Ref.

Ares(2015)4173087 - 08/10/2015

The land use change impact of

biofuels consumed in the EU
Quantification of area and
greenhouse gas impacts
The land use change impact of
biofuels consumed in the EU
Quantification of area and greenhouse gas impacts

By:
Hugo Valin (IIASA), Daan Peters (Ecofys), Maarten van den Berg (E4tech), Stefan Frank,
Petr Havlik, Nicklas Forsell (IIASA) and Carlo Hamelinck (Ecofys), with further
contributions from: Johannes Pirker, Aline Mosnier, Juraj Balkovic, Erwin Schmid, Martina
Dürauer and Fulvio di Fulvio (all IIASA)

Date: 27 August 2015

Project number: BIENL13120

Reviewers: Ausilio Bauen (E4tech), Michael Obersteiner (IIASA) and the Scientific Advisory Committee: - Prem
Bindraban, Don O’Connor, Robert Edwards, Jacinto Fabiosa, David Laborde, Chris Malins, André
Nassar, Koen Overmars and Richard Plevin
Project coordination: Michèle Koper (Ecofys)

This study has been commissioned and funded by the European Commission.

A cooperation of Ecofys, IIASA and E4tech

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 ii
Acknowledgements

The study consortium is grateful for the useful information and the many comments, questions and
recommendations we received throughout the preparation of the study between September 2013 and
June 2015 from a wide variety of stakeholders, including representatives from industry associations,
NGOs and scientific experts, either by attending one of the eight stakeholder meetings we organised
during the study process or by corresponding with us per email. We also thank the European
Commission Steering Committee for their input and reviews. In particular, we would like to thank our
Advisory Committee members, who agreed to review the study approach, progress and draft results
on a voluntary basis: Prem Bindraban, Don O’Connor, Robert Edwards, Jacinto Fabiosa, David
Laborde, Chris Malins, Andre Nasser, Koen Overmars and Richard Plevin. We would also like to thank
Alexandre Gohin for his critical comments on our initial study approach.

The input received from many sides improved the quality of our study, while at the same time the
study consortium bears the full responsibility for this study.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 iii
Executive Summary

Introduction

Biofuels are promoted as an option to reduce climate emissions from the transport sector. As most
biofuels are currently produced from land based crops, there is a concern that the increased
consumption of biofuels requires agricultural expansion at a global scale, leading to additional carbon
emissions. This effect is called Indirect Land Use Change, or ILUC. The EU Renewable Energy Directive
(2009/28/EC) directed the European Commission to develop a methodology to account for the ILUC
effect.

The current study serves to provide new insights to the European Commission and other stakeholders
about these indirect carbon and land impacts from biofuels consumed in the EU, with more details on
production processes and representation of individual feedstocks than was done before. ILUC cannot
be observed or measured in reality, because it is entangled with a large number of other changes in
agricultural markets at both global and local levels. The effect can only be estimated through the use
of models. The current study is part of a continuous effort to improve the understanding and
representation of ILUC.

Background

Most biofuels today use feedstock grown on land that is suitable for food, feed or material production.
An increase in biofuel consumption could therefore lead to cropland expansion in one of two ways:
 Directly, when new cropland is created for the production of biofuel feedstocks. This is called
direct land use change, or DLUC;
 Indirectly, when existing cropland is used for biofuel feedstock production, forcing food, feed and
materials to be produced on new cropland elsewhere. This expansion is called indirect land use
change, or ILUC.

Direct and indirect land use change are intertwined in reality. They can lead to changes in carbon
stocks on land, most notably through loss of above and below ground living biomass and soil organic
carbon, which leads to an increase of greenhouse gases in the atmosphere. However, the uptake of
carbon by crops and the effective use of co-products from biofuel production can partly compensate
these emissions. The outcome of emission quantification studies present the net result.

This study aims to quantify emissions resulting from the existing EU biofuel policy up to 2020. The
study therefore enables policy makers to assess the complete climate impacts associated with biofuel
policies. Biofuel policies aim to mitigate climate change, but high emissions could compromise biofuels’

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 iv
mitigation potential. Insights of this study can assist policy makers in designing future EU biofuel
policy in such way that land use change impacts are effectively addressed.

ILUC modelling

Because ILUC occurs through global market mechanisms with many direct and indirect effects, it can
only be modelled, not measured. Direct measurement will only provide partial accounting of the total
effects. Previous studies have tried to quantify ILUC related emissions, to understand whether the use
of biofuels really avoids greenhouse gas emissions on a global scale and by how much. The current
study focuses on biofuels consumed in the EU. Note that it does not discuss whether biofuel producers
should be held accountable for effects that are indirectly induced by their actions but which take place
outside their control. Nor does it answer the question regarding how it can be ensured that biofuels
actually reduce greenhouse gasses emissions compared to fossil fuels, within a certain timeframe. The
aim of this study is only to model biofuel induced land use change and its greenhouse gas emission
consequences, as consistently as possible, using a tailored version of the GLOBIOM model. Whilst this
is not the first study that quantifies land use change impacts of EU biofuels – it follows a study
published by the International Food Policy Research Institute (IFPRI) in 2011 (Laborde, 2011) – the
current study quantifies for the first time land use change emissions from advanced biofuel feedstocks
as well as several ‘alternative scenarios’, as further explained below. The study is relevant for the
discussion on the 2030 EU policy framework for energy and climate change.

The study follows the general principles of ILUC modelling used in earlier studies, in which a “world
with additional biofuels” (the policy scenario) is compared to the same world “as it would have
developed without the additional biofuels” (the baseline). In this study, the policy scenarios are based
on the European Union Renewable Energy Directive1 (commonly known as ‘the RES directive’ or ‘the
RED’). The computed ILUC impact of the additional biofuels follows from the difference between
emissions in the policy scenarios and those of the baseline. This difference is then attributed to the
additional biofuel demand in the policy scenarios.

The results of this study, commonly referred to as ‘ILUC values’ (or ‘factors’), are in fact the sum of
direct and indirect emission effects. When comparing a policy scenario with a baseline, it is certain
that the difference in quantity of land conversion and its greenhouse gas impact results from the
difference between scenario and baseline: the additional biofuel demand. The modelling does not show
to what extent the land conversion is caused directly or indirectly. For this reason, this study speaks
about ‘LUC values’ rather than ‘ILUC values’ and about ‘land use change’ rather than ‘direct or indirect
land use change’. Even the term ‘land use change emissions’ does not fully cover the different sources
of emissions included in the final results, as some of the emissions are related directly to the change
in crop or plantation type, which impacts carbon stock in biomass and soil. These emission savings are

1
Directive 2009/28/EC

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 v
deducted from the land use change related emissions, leading to the LUC values. For each modelled
scenario we provide a precise breakdown of the result into various contributing factors.

This study includes various emission sources and sinks linked to related to biomass and soil carbon
stocks. This includes direct soil carbon emissions resulting from the removal of forestry residues from
forests2. Not included are emissions directly related to the biofuel production chain, including
emissions related to feedstock cultivation and processing, biofuel production, transport and
distribution. Box 1 gives an overview of emission sources included in this study.

Box 1: Overview of emissions included in this study and emissions not included

Emission sources included in this study

Peatland oxidation: emissions caused by peatland drainage due to oil palm plantation expansion.
Soil organic carbon: changes in carbon stored in soils.
Natural vegetation reversion (foregone sequestration): avoided emission savings due to reduced
afforestation or reduced return of cropland to other natural land due to increased use of cropland. This effect
takes place in particular in Europe where a trend exists of cropland abandonment.
Natural vegetation conversion emissions: release of carbon stored in forest biomass or natural biomass,
at the moment the land use change occurs.
Agricultural biomass: changes in carbon stored in agricultural crops. These can either be biofuel feedstocks
cultivated as a direct consequence of increased biofuel demand, or other crop cultivation, triggered indirectly
by increased biofuel demand.

Some of these emission sources can be both positive and negative, even within the same scenario.
Soil organic carbon emissions, for example, are positive emissions when carbon stored in soils is released, e.g.
when forests or other natural biomass are converted and tilled for farming. The emissions are also positive
when the build-up of soil organic carbon is avoided (relative to the baseline), e.g. when the collection of forest
residues is increased. These emissions can result directly from increased cultivation of specific biofuel
feedstocks, or result from the increased cultivation of other crops triggered by increased biofuel demand. At
the same time, soil organic carbon emissions can be negative when carbon is stored in soils or crops, due to a
switch of crop cultivation methods.

Emissions not included in this study

Agricultural production and chain emissions (direct and indirect): emissions resulting directly from the
cultivation of crops (fertiliser production and use, machinery, etc.), conversion into biofuels, and product
transport and distribution.

2
These could also be accounted for in the direct GHG emissions of biofuels, but that is not the case in the methology specified in the RES
Directive.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 vi
Scenarios

Earlier studies have shown that the LUC impact differs per crop and supply chain. In the current study,
14 crop-specific scenarios for the main conventional and advanced biofuel crops are modelled, as well
as separate scenarios for the cereal, starch and oilseed crop groups. Also, a central aggregated
scenario is modelled for the EU 2020 biofuel mix, with 8.6% conventional biofuel consumption and
0.8% advanced biofuels3 (in line with National Renewable Energy Action Plans).4 In addition, an EU
2020 biofuel mix scenario with a maximum cap on the consumption of conventional biofuels of 7% is
modelled, based on the same feedstocks, with 6,7% conventional biofuel consumption and 1,7%
advanced biofuels (by volume). The division between conventional and advanced biofuels and the
chosen feedstock mix have an important influence on the results of the aggregated scenarios. The
division between conventional and advanced biofuels in the EU 2020 biofuel mix scenario is based on
the National Renewable Energy Action Plans (NREAPs) submitted by Member States to the European
Commission to allow comparibility with the previous LUC study (IFPRI 2011). The chosen feedstock
mix only includes feedstocks which have been selected to be part of the study scope. Not all
feedstocks that are part of the actual EU biofuels feedstock mix have been selected, most notably
used cooking oil and animal fats are not included. The mix of conventional feedstocks is based on EU
FAS Posts (USDA 2014), the mix of advanced feedstocks is determined by the model cost
minimisation. While these choices are based on best available consistent information on the EU overall
biofuel feedstock division, it is clear that the resulting feedstock mix does not necessarily reflect the
actual situation by 2020. This means that the resulting LUC values for the aggregated scenarios
should be treated with caution.

In addition to the EU 2020 biofuel mix scenario and the 7% cap scenario, several explorative scenarios
are modelled to understand how the results would change, if more abandoned land in the EU was to
be used for the biofuels feedstock production; if worldwide deforestation was to either increase or
decrease; or if there were a global ban on peatland drainage. In total, 28 scenarios have been
modelled (incl. four for straw), as presented in the figure below.

3
Before double counting.
4
As submitted by Member States to the European Commission in 2010-11.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 vii
Figure 1: Overview of assessed crops, crop groups, policy scenarios, explorative scenarios and contextual variations

Both baseline and biofuel scenarios are modelled for the years 2020 and 2030. However, in the
absence of a biofuel target for 2030, we assume that no further increase in biofuel consumption would
occur after 2020, and that the feedstock composition would not change. For this reason, all our biofuel
policy shocks are assessed for the year 2020.

Land use change emissions that result from the modelling are distributed over a 20-year period. For
foregone sequestration emissions, avoided carbon stock accumulation is included for the first 20
years. Peatland emissions, which emit carbon dioxide every year, are also accounted over a 20 year
period. A distribution of emissions over 20 years is common practice in land use change modelling,
since most LUC emissions take place shortly after the conversion of previously non-agricultural land to
agricultural land and it makes little sense to allocate all emissions to the first year after the conversion
and to have zero LUC emissions in year two. The twenty-year period is in line with the period used for
the allocation of direct land use change emissions in the greenhouse gas calculation methodology as
laid down in the EU-RED. If a longer allocation period were chosen, for example 30 or 50 years, LUC
emission values would be lower for some sources, since the total land use change emissions
associated with a certain quantity of biofuels would be divided over larger number of years. However,
annual flows from peatland and future foregone sequestration would not be reduced before 50 to 100
years (time for peat to be fully oxidised or forest to be fully regrown). Given the significant
contribution from continued peatland oxidation, the LUC emissions from 50 year perspective is overall
significant higher than from the 20 year perspective. Annual LUC emissions would however decrease,
as shown under ‘main findings’ below.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 viii
Results

The total LUC emissions results are presented in Figure 2, expressed in grams of CO2 equivalent per
megajoule of biofuels (gCO2eq/MJ). More detailed modelling results are provided in Chapter 4.

This study has two types of outcomes: quantities of land conversion caused by additional biofuel
demand and, based on this land conversion, greenhouse gas emission impacts for each of the
modelled scenarios. The total land use change caused by the EU 2020 biofuel mandate is 8.8 Mha
(million hectares), of which 8 Mha is new cropland and the remaining 0.8Mha consists of short rotation
plantations on existing cropland. From the 8.8 Mha, 2.9 Mha of conversion takes place in Europe by
less land abandonment and 2.1 Mha of land is converted in Southeast Asia under pressure from oil
palm plantation expansion, half of which occurs at the expense of tropical forest and peatland. The
abovementioned 8.8 Mha is 0.6% of the total global crop area in 2012 of 1,395 Mha (FAO). This is
around 4% of the total land area of Indonesia, or equal to the total land area of Austria.

Figure 2 below shows the LUC emission values for each of the modelled scenarios and their break-
down between various emission sources (see also Box 1 above). The part of each bar above zero on
the y-axis represents positive emissions, while the part of the bar below zero represents negative
emissions that are being deducted from the emissions. The resulting net LUC emission value is
represented by the small triangle in each bar and by the number on top of each bar.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 ix
Figure 2: Overview of modelling results: LUC emissions per scenario. Source: GLOBIOM

Main findings

Feedstock-specific scenarios
1 Conventional biodiesel feedstocks have high LUC effects compared to the direct emissions
resulting from the biofuel production process, with very high emissions for palm oil (231 grams of
CO2e per megajoule of biofuel consumed – gCO2e/MJ), high emissions for soybean oil (150
gCO2e/MJ) and 63 and 65 gCO2e/MJ for sunflower and rapeseed respectively;
2 Drainage of peatlands in Indonesia and Malaysia plays a large role in LUC emissions for vegetable
oils. This is especially the case for palm oil: 69% of gross LUC emissions for palm oil is caused by
such peatland oxidation after land conversion;

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 x
3 The large and local emission source of peatland oxidation has an impact on the LUC values of
other vegetable oils through the substitution effect, with vegetable oils interchangeable to a
certain extent. Based on empirical data, we conjecture a relatively limited substitution effect,
hence the large difference in LUC values for palm oil – the most cost competitive vegetable oil –
and other more costly vegetable oils. Still, substitution plays a role and transfers some of the
peatland emissions from palm oil to other vegetable oils;
4 The conventional ethanol feedstocks – sugar and starch – have much lower LUC emission
impacts, at 14 and 34 gCO2e/MJ biofuel consumedfor maize and wheat, 17 gCO2e/MJfor
sugarcane and 15 gCO2e/MJ for sugarbeet. These feedstocks lead to a much lesser extent to
peatland oxidation and deforestation compared to vegetable oils;
5 In general, crops with higher energy yield per hectare have lower indirect impacts on land use
change and greenhouse gas emissions. A notable exception is palm oil, a high yielding crop
whose performance is strongly impacted by emissions from deforestation and peatland
conversion, as explained above;
6 Advanced biofuels have negative LUC emissions if produced from short rotation crops (-29
gCO2e/MJ biofuel consumed) or perennials (-12 gCO2e/MJ ), mainly because of the increase in the
carbon stock on the land that is converted to produce these higher carbon stock crops;
7 Advanced biodiesel (Fischer-Tropsch) from forestry residues leads to a significant LUC emission
value of 17 gCO2e/MJ biofuel consumed, despite the fact that no land use change takes place per
se when harvesting forestry residues. The emissions result instead from a lower build-up of soil
organic carbon. It is therefore more appropriate to speak about a ‘soil organic carbon (SOC)
emission value’ for forestry residues, instead of a ‘LUC emission value’. Note that, according to
the Renewable Energy Directive, the emissions associated with collecting wood residues from the
forest floor have to be included in the direct emissions (since it is the point of collection5).
However, the impact on soil organic carbon associated with the same collection of residues is not
included in the direct emissions, which is why it is accounted here;
8 Ethanol from cereal straw can lead to a LUC value of 16 gCO2e/MJ biofuel consumed, caused by a
slight reduction in yields of the main commodity (i.e. the cereal) in cases of overharvesting in
areas where already high volumes of straw are harvested for purposes such as animal feed and
bedding. This overharvesting leads to soil carbon depletion, and a small yield loss. If straw
harvesting is limited to a sustainable removal rate of 33-50% (Ecofys 2013), no yield effect
occurs and therefore no land use change effect is observed. Based on four different scenarios
modelled for cereal straw, it can be concluded that the LUC value of 16 gCO2e/MJ biofuel
consumed would become 0 gCO2e/MJ if a sustainable straw removal rate was introduced limiting
the straw removal to once every two to three years or 33-50%.

5
Directive 2009/30/EC, Annex V, Part C, point 18.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xi
Aggregated EU 2020 biofuel mix scenarios
9 The central ‘EU 2020 biofuel mix’ scenario gives a high LUC impact of 97 gCO2e/MJ biofuel
consumed. This high number is largely due to the fact that palm oil constitutes 16% of the
feedstock of additional biofuels in 2020;
10 Applying a maximum percentage (‘cap’) on the consumption of conventional biofuels reduces the
overall LUC emission effect from 97 gCO2e/MJ biofuel consumed to 74 gCO2e/MJ with a 7% cap
on conventional biofuels, mainly because the share of advanced biofuels with low or negative
emissions increases compared to a situation without a cap;
11 If total LUC emissions would be amortised over 50 years instead of 20 years, annual emissions
would amount to 79 gCO2e/MJ in the EU 2020 biofuel mix scenario.

Explorative scenarios
12 A scenario in which more abandoned land in the EU is used for biofuel production reduces LUC
emissions of the EU 2020 biofuel mix from 97 gCO2e/MJ biofuel consumedto 52 gCO2e/MJ. Part of
this reduction results directly from using abandoned land, while partly it results from a reduced
share of palm oil in the total feedstock mix. Using abandoned land can be a good policy option,
particularly if the land is degraded and soil carbon stocks are restored though use;
13 Global efforts to stop deforestation and peatland drainage could effectively reduce LUC emissions.
The very low deforestation scenario shows that a substantial global incentive to leave forests
intact, created in our modelling by charging a price of USD 50/t CO2 emissions from
deforestation, could reduce deforestation to a level that would result in overall LUC emissions for
the EU 2020 biofuel mix of 48 gCO2e/MJ, instead of the central scenario impact of 97 gCO2e/MJ
biofuel consumed. If such a low deforestation scenario were to be combined with an effective ban
on peatland drainage, the overall LUC emission effect of EU biofuel policy would further decrease
to just 4 gCO2e/MJ. A more moderate incentive to reduce deforestation of USD 10/t CO2 would
have more modest results in reducing deforestation and would mean that the LUC emissions of
EU biofuel policy would remain at a relatively high level of 87 gCO2e/MJ biofuel consumed.

The very large LUC emissions resulting from increased palm oil use as a biofuel feedstock will likely
lead to the question of how the existing EU sustainability criteria for biofuels are factored into this
study. These criteria prohibit expansion into forests, expansion into areas with high biodiversity levels
and peat land drainage. While these restrictions have a positive impact on the direct sustainability of
biofuel production, unsustainable land conversion can still take place. The ban on ‘unsustainable land
conversion’ causes biofuel feedstocks to be sourced mainly from existing farms and plantations,
resulting indirectly in increased unsustainable land conversion to meet demand for food, feed and
materials, or to supply other markets than the European Union. Only if sustainability criteria that offer
a similar level of protection are extended to the food, feed and materials sectors and if these are
applied and effectively enforced globally, then these unsustainable practices may be effectively
tackled.

Whereas a global approach could be effective to tackle unsustainable land use change, this study
shows that one of the major contributors to LUC emissions, peat land drainage, is a relatively local

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xii
problem. If peatland drainage in Indonesia and Malaysia were stopped, the negative greenhouse gas
impact of land use change would reduce dramatically. This requires an effort either from the
Indonesian and Malaysian governments, all palm oil using sectors (food, personal care products,
biofuel) or, best of all, a combination of both. Whether by global action to stop unsustainable land
conversion, or by local action to stop peatland drainage, our study shows that LUC values can be
reduced by effective policies.

A modest (for most feedstocks) but interesting emission source is foregone sequestration, which is the
effect that, without demand for biofuels, cropland area might decrease and partly revert into grassland
or forest. Using more cropland to produce biofuel feedstocks in Europe slows down this process of land
abandonment. This has a negative carbon impact, because it implies that carbon accumulation
through natural vegetation and young forest regrowth does not take place. If such “foregone
sequestration” is indeed considered a business-as-usual development included in the baseline, it will
have an impact on LUC emissions. In this study, most foregone sequestration takes place in the EU
and more intensive cropland usage in Europe prevents reversion from taking place. We acknowledge
that this topic can be debated, as the extent to which the effect occurs in reality is not well
documented. Cropland which is abandoned due to agricultural market dynamics does not always
automatically revert to forest, due to, for example, annual mowing by farmers in order to receive CAP
money, occasional mowing by local smallholders, or extensive grazing. Foregone sequestration was
largely left out of the IFPRI study: forest regrowth on abandoned land was not included although some
afforestation was included in the IFPRI baseline. Because of the uncertainty concerning foregone
sequestration and in order be able to better compare the results of the present study with the results
of the IFPRI study, we present Figure 3 below with LUC values both with and without foregone
sequestration. In the scenario result sheets in Section 4.2, results are also presented both with and
without foregone sequestration. Excluding foregone sequestration has a large impact on ethanol
feedstocks; the LUC value for wheat for example drops from 34 to 22 gCO2e/MJ biofuel consumed and
for maize from 14 to 9 gCO2e/MJ. The EU 2020 biofuel mix scenario result drops from 97 gCO2e/MJ to
90 gCO2e/MJ without foregone sequestration.

In addition to the values with and without foregone sequestration, Figure 3 also shows the estimated
ranges of uncertainty for each scenario for which a Monte Carlo sensitivity analysis has been
performed. Important uncertainties remain, as will always be the case in modelling exercises. They
are related to variability around biophysical values that cannot be reduced and uncertainty around
causalities assumed by the modelling approach. However, a significant number of uncertainties can be
explored within the modelling framework. The most important ones analysed in this study are varying
levels of market and producer responses (related to demand, trade, vegetable oil substitution,
intensification and land expansion), and some biophysical characteristics (water availability, co-
product protein content, the soil carbon and yield impact of straw removal, and the peat land emission
factor). The sensitivity analysis shows that, in some cases, LUC emissions of conventional biofuels
could be negative; it could however also lead to much higher results per scenario. It is important to
keep in mind that the uncertainties are large and often considerable ranges of modelling results exist.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xiii
Figure 3: Overview of modelling results: LUC emissions per scenario with and without foregone sequestration and

with uncertainty ranges (bars indicate the range within the first and the last decile). Source: GLOBIOM

Some important parallels exist between this study and the previous LUC quantification study focusing
on EU biofuels that was published by the International Food Policy Research Institute IFPRI in 2011.
Both studies show that sugar and cereal feedstocks perform better than vegetable oils. Both studies
show the large influence co-product use and yield increase have on lowering LUC effects. Both studies
also show that peat land drainage for oil palm plantation expansion plays a large role in LUC emission
values for palm oil and other vegetable oils. An important difference with the IFPRI study is the very
high LUC impact for palm oil and soybean oil in the current study, arising from the high share of new
oil palm plantations that are being developed on peatland and the higher peatland emission factor
assumed, based on the latest available literature. Another important difference is the resulting total

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xiv
land use change measured in hectares. As described above, the aggregated ‘EU 2020 biofuel mix’
scenario in the present study leads to 8.8 Mha of LUC.

In the IFPRI study (Laborde, 2011), however, the total EU biofuel demand shock results in 1.7 Mha of
LUC, four times less than the area result in the present study. In line with this large difference in LUC
area, the total estimated LUC emissions of the present study are also considerably higher than those
estimated by IFPRI: 1,495 MtCO2e in our central ‘EU 2020 biofuel mix’ scenario, compared to 495–516
MtCO2e in IFPRI depending on the chosen central scenario. Whereas we estimate the area effect to be
more than four times larger than IFPRI, the emission effect is only three times larger. Looking,
however, to individual crop-specific scenario results, LUC emission values in the present study are
approximately similar to those in the IFPRI study, although palm oil and soybean oil are striking
exceptions.

There has been an important debate on whether or not LUC emission factors should be used in biofuel
policy. Our results show that LUC emissions are likely to be substantial, but some inherent uncertainty
cannot be avoided in the estimation of such emissions and many parameters and assumptions
influence the results. From this perspective, only a few feedstocks can be designated as having high or
low LUC emissions with a high degree of confidence, with advanced feedstocks having low LUC
emissions, or soil organic carbon but no LUC emissions, while palm oil and soybean oil clearly have
substantial LUC emissions. However, our work also identifies some clear chains of effects and
highlights impact patterns that can vary significantly between feedstocks. If, for example,
deforestation and peatland drainage in Indonesia and Malaysia could be avoided by introducing
appropriate environmental safeguard systems, LUC emissions for palm oil, soybean oil and other
vegetable oils would strongly decrease. These effects should be kept in mind when discussing the
emission impacts of current biofuel policy.

For this work, our consortium gathered the best available datasets and built upon the most recent
literature published up to early 2014. Stakeholders have been consulted in 2013/14 to obtain inputs
and feedback, as further described in the Introduction (Section 0). Following suggestions from
stakeholders, the GLOBIOM model was also improved substantially for a number of topics during the
course of 2014, followed by the actual LUC modelling. A Scientific Advisory Committee provided
valuable comments on our approach that we took into account to the largest possible extent.
Notwithstanding these efforts, many particular aspects will still require future research and LUC
quantification will always remain the reflection of our understanding of agricultural market behaviour.
While modelling can be improved with better datasets and better understanding of certain dynamics
and interlinkages, uncertainties cannot be avoided. The main uncertainties are described and tested in
Annex V.

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xv
Table of contents

Acknowledgements iii

Executive Summary iv
Introduction iv
Background iv
ILUC modelling v
Scenarios vii
Results viii
Main findings ix

1 Introduction 1
1.1 Study context and aim 1
1.2 ILUC: a brief background 1
1.3 Modelling approach 3
1.4 Using GLOBIOM to model LUC 3
1.5 Modelling baseline, scenarios, feedstocks and sensitivity analyses 4
1.6 Interactive project approach leading to transparent results 6
1.7 How the current study follows previous land use change studies 7

2 Description of the modelling baseline 10

2.1 Baseline assumptions 11
2.2 Baseline results 18

3 Description of scenarios and sensitivity analysis 29

3.1 Introduction 29
3.2 Crop-specific scenarios 30
3.3 EU 2020 biofuel mix scenario without and with 7% 31
3.4 Explorative scenarios: abandoned land and lower or higher deforestation 34
3.5 Sensitivity analysis 35

4 Modelling results 37
4.1 Summary of modelling results 37
4.2 Detailed results by feedstock 47
4.3 Wheat ethanol 49
4.4 Maize ethanol 51
4.5 Barley ethanol 53
4.6 Sugar Beet ethanol 55
4.7 Sugar Cane ethanol 57

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xvi
 4.8 Silage Maize biogas 59
4.9 Sunflower oil biodiesel 61
4.10 Palm oil biodiesel 63
4.11 Rapeseed oil biodiesel 65
4.12 Soybean oil biodiesel 67
4.13 Cereal straw ethanol produced in the EU 69
4.14 Miscanthus and switchgrass FT biodiesel produced in the EU 72
4.15 Short Rotation Plantation FT biodiesel produced in the EU 74
4.16 Forest residues FT biodiesel produced in the EU 76
4.17 Starchy crops group 78
4.18 Sugar crops group 80
4.19 Vegetable oil group 82
4.20 EU 2020 biofuel mix scenario (all feedstocks) 84
4.21 EU 2020 biofuel mix scenario with 7% cap on conventional biofuels 86
4.22 Abandoned land in the EU 88
4.23 Lower deforestation 90
4.24 Very low deforestation with no peatland drainage 92
4.25 Higher deforestation 93
4.26 Comparison of results with previous LUC assessments 94

5 References 97

6 Glossary 100

Annex I Description of GLOBIOM and comparison with MIRAGE-BioF (IFPRI) 102

I.1 Summary of differences between GLOBIOM and MIRAGE-BioF 103
I.2 Representation of agriculture and yield development 106
I.3 Representation of woody biofuel feedstocks and forestry 110
I.4 Overview of feedstock processing and biofuel production 112
I.5 Processing activities and bioenergy pathways 113
I.6 Capturing the world markets and the global economy 117
I.7 Modelling land use change and associated GHG emissions 120
I.8 Modelling changes in food consumption 126
I.9 GLOBIOM and MIRAGE-BioF characteristics – technical summary 127
I.10 Technical comparison table GLOBIOM versus MIRAGE-BioF 130
I.11 References 134

Annex II Building an improved version of GLOBIOM 139

II.1 Improve the representation of cereal straw 140
II.2 Include carbon sequestered in annual and perennial crops 146
II.3 Update peat land emission factors 148
II.4 Represent expansion of oil palm plantations into peat land 161

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xvii
 II.5 Expand the inclusion of soil organic carbon (SOC) worldwide 168
II.6 Include forest regrowth and reversion time on unmanaged land 170
II.7 Refine co-product substitution 173
II.8 Represent multi-cropping 178
II.9 Represent imperfect substitution between vegetable oils 180
II.10 Separate representation of Argentina, Indonesia, Malaysia and Ukraine 185
II.11 Represent unused agricultural land in Europe 187
II.12 Refine biofuel feedstock processing coefficients 189
II.13 References 189

Annex III Technical background of modelling 198

III.1 Calculation of sustainable potential 198
III.2 Supply cost calculations 199
III.3 Soil carbon losses 199
III.4 Amortisation of emissions over 50 instead of 20 years 200
III.5 References 203

Annex IV Data used in the GLOBIOM model 205

IV.1 Parameters 205
IV.2 Land cover data 205
IV.3 Carbon stocks 207
IV.4 Crop yields 207
IV.5 Bioenergy transformation pathways 209
IV.6 Co-product replacement coefficients 214
IV.7 Biofuel feedstock demand 215
IV.8 Demand elasticitities 219

Annex V Sensitivity and uncertainty analyses 222

V.1 Most important uncertainties in LUC modelling 222
V.2 Detailed results per scenario 225

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
Chamber of Commerce 30161191 xviii
1 Introduction

1.1 Study context and aim

In order to fulfil its commitment to mitigate greenhouse gas emissions, the European Union (EU)
engaged in an ambitious programme to develop renewable energy sources by 2020. The 2009
Renewable Energy Sources (RES) Directive (2009/28/EC), or ‘RED’, includes a target of 10%
renewable energy in transport. The majority of this renewable energy comes, and is expected to
come, from biofuels. The EU introduced mandatory sustainability criteria for biofuels in the RED.
These criteria ensure that feedstock production does not cause unsustainable land conversion, i.e.
conversion of land with high biodiversity values or carbon stocks. However, when feedstock is
(sustainably) sourced from existing farms or plantations, this could still lead to expansion of
agricultural land elsewhere, causing indirect land use change, refered to as ILUC. The carbon impact
of ILUC can temporally reduce or undo the carbon benefits of biofuels. ILUC is a sensitive topic, with
widely varying opinions on whether the effect can be quantified in a robust way and how ILUC
modelling results should, or should not, be used in EU biofuel policy. In October 2012, the European
Commission published a legislative proposal6 to introduce measures aimed at addressing ILUC. The
European Council and Parliament reached agreement on an amended version of this proposal in
2015, which means that measures to address ILUC will be included in the Renewable Energy
Directive7.

This study aims to quantify land use change emissions resulting from the existing EU biofuel policy
up to 2020 and assesses also the land use change impacts of this policy in 2030. The study enables
policy makers to assess the complete climate impacts from biofuels policies. Biofuel policies have
been designed to mitigate climate change, and high land use change emissions can compromise
biofuels’ mitigation potential. More insights into land use change emissions resulting from biofuel
production can help policy makers to find the best way to design the future EU biofuel policy in such
way that land use change is effectively addressed.

1.2 ILUC: a brief background

When demand for biofuels increases and food and feed crops are starting to be used for biofuels, the
shortage in food production may be compensated by new food production on previously non-
agricultural areas elsewhere, such as forests or grasslands. Alternatively, land remains in agricultural
production that would otherwise be abandoned. This has a climate impact, because conversion of
forest or grassland to agricultural land can lead to significant releases of CO2 to the atmosphere.
ILUC takes place outside the biofuel production and supply chain, but can be linked to biofuel
production due to the international nature of agricultural commodity markets. The effect cannot be
measured, only modelled with large and complex economic models.

6
COM(2012)595
7
As well as in the Fuel Quality Directive (FQD – 2009/30/EC).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 1

How does the ILUC effect work in practice? At present, biofuels are mainly produced from agricultural
crops that are also used for food, such as rapeseed, maize or palm oil. If more biofuels are produced
to fulfil renewable energy targets, demand for these crops rises as well. Following the basic law of
supply and demand, increased demand compared to supply leads to a price increase of the crop. The
market can respond to this price increase in several ways:
 Reduce consumption;
 Increase supply by creating additional cropland (somewhere);
 Improved agricultural productivity.

Firstly, increased crop prices will cause some decline in food consumption, both because people will
eat less and because food waste in the supply chain will be reduced. Secondly, farmers will invest in
increasing their yield by improving their agricultural methods, because they can get a better price for
their crops. Thirdly, to a certain extent, previously non-agricultural land will be converted to
agricultural land to compensate for the crop that was taken from the market. Because of the open
and global nature of agricultural commodity markets, this conversion of land can take place
anywhere in the world. This effect can be even more indirect, since an increase in demand for crop x
can cause this crop to expand at the expense of crop y, which in turn can drive the conversion of
forest or grassland elsewhere. This makes ILUC a cross-border effect, acting internationally and also
across crops. Agricultural commodities are partly interchangeable, depending on their function,
location and price levels. For example, palm oil can be used by the food sector to compensate for an
increased use of other vegetable oils, such as rapeseed by the biofuels sector. This means that if
palm oil is cheaper than rapeseed oil, increased consumption of rapeseed for biofuels in Germany at
the expense of rapeseed previously used in the food sector may lead to an increased interest in palm
oil and hence to deforestation in Indonesia. Note that the EU RED does not allow deforestation and
expansion into peatland for biofuel feedstock. It should also be noted that Indirect Land Use Change
(ILUC) is not exclusively related to biofuel production, but that other land using sectors cause land
use change. This study focuses on LUC effects from biofuels, since it is relevant for policy makers to
assess how to ensure a policy that is designed to mitigate climate change can indeed serve its
purpose.

The results of this study, commonly referred to as ‘ILUC values’, are in fact a mix of direct and
indirect emission effects. When comparing a policy scenario with a baseline, it is certain that the
differences in quantity of land conversion and its greenhouse gas (GHG) impact results from the
difference between scenario and baseline: the additional biofuel demand. The modelling does not
show to what extent the land conversion is caused directly or indirectly. For this reason, this study
speaks about ‘LUC values’ rather than ‘ILUC values’ and about land use change rather than direct or
indirect land use change.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 2

1.3 Modelling approach

This study follows the method of LUC modelling used in most studies, in which the world with an
increase in biofuels, called “the policy scenario”, is compared to exactly the same situation but now
without the biofuels mandate, “the modelling baseline”. In the current study, we focus on the impact
of the EU RED. Therefore, we compare a world with increasing EU biofuels consumption to a world in
which this consumption is fixed at 2008 levels and does not further increase. The computed LUC
impact is the difference between the baseline and the policy scenarios. Differences in LUC emissions
between the two scenarios will provide emissions that are attributable to the increase in biofuel
demand between the baseline and applied policy scenarios.

According to the methodology for calculating DLUC emissions laid out in the EU RED, and also used
in the October 2012 LUC proposal Impact Assessment by the European Commission, the LUC factor
is obtained by dividing CO2 emissions from land use change by an amortisation period of 20 years to
provide a final estimate in grams of CO2-equivalent per megajoule (gCO2e/MJ). More details on the
model and modelling approach are provided in the sections below.

1.4 Using GLOBIOM to model LUC

For the purpose of this study we use the GLOBIOM (Global Biosphere Management Model)8,
developed by IIASA (see Havlik et al. 2011, 2014). The model effectively represents the world’s
agricultural and forestry sectors and most relevant economic and demographic indicators and trade
relations. GLOBIOM is an equilibrium model, meaning that the supply and demand sides of the
agricultural and forestry sectors are represented, with supply and demand being equal at a certain
price level. During the modelling, a biofuel demand ‘shock’ is applied and compared to the ‘baseline’
situation. This means that a certain quantity of biomass demand increase is assumed, leading to an
increase in prices. The model calculates the supply side changes and feedback-loops that this shock
causes. This iteration or adjustment stops when a new equilibrium between supply and demand sides
is found at a new price level.

GLOBIOM is a global recursive dynamic partial equilibrium model with a bottom-up representation of
agricultural, forestry and bioenergy sectors. The model is global because it covers 57 countries and
regions worldwide (EU28 plus 27 countries and regions in rest of world). GLOBIOM is recursive
dynamic instead of static, and is thus able to model changes over periods of time. The model is a
partial equilibrium, as opposed to general equilibrium, because it covers the most relevant sectors
(agriculture and forestry) in great detail while information from other sectors is kept external to the
model. Finally, GLOBIOM is bottom up, because the supply side of the model is built up from bottom
(land cover, land use, management systems) to top (production/markets). GLOBIOM is developed
since 2007 and a EU dedicated version has been set-up over the past four years (Frank et al., 2013).

8
www.globiom.org

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 3

The model computes the global agricultural and forest market equilibrium by choosing land use and
processing activities to maximise the sum of producer and consumer surplus, subject to resource,
technological and policy constraints. The level of production in a given area is determined by the
agricultural or forestry profitability in that area (dependant on suitability and management), market
prices (reflecting the level of demand) and the conditions and costs associating with conversion of
the land, expansion of production and, where relevant, to international market access. Trade is
modelled following the spatial equilibrium approach, which means that the trade flows are balanced
out between different specific geographical regions. This allows tracing of bilateral trade flows
between individual regions.

By including the bioenergy sector, forestry, cropland and grassland management, and livestock
management, the model allows for a full account of all agriculture and forestry GHG sources.
GLOBIOM accounts for ten sources of GHG emissions, including crop cultivation N2O emissions from
fertiliser use, CH4 from rice cultivation, livestock CH4 emissions, CH4 and N2O emissions from manure
management, N2O from manure applied on grassland, above and below ground biomass CO2
emissions from biomass removal after converting forest and natural land to cropland, and CO2
emissions from soil carbon, including cultivated organic soil (drained peat land, at country level).
These emissions inventories are consistent with IPCC accounting guidelines.

A more detailed description of the GLOBIOM model and how the model is used to quantify LUC is
provided in Annex I.

1.5 Modelling baseline, scenarios, feedstocks and sensitivity analyses

This study models a number of scenarios by comparing them with a modelling baseline. This baseline
describes the evolution of relevant sectors between the base-year 2010 – the year which the EU RED
entered into force – and the year 2020 – for which the ‘biofuels shock’ is modelled.

The baseline includes biofuel consumption outside the EU plus the level of EU biofuel consumption
(3.2%), as also used in the IFPRI study. The baseline excludes the implementation of the Renewable
Energy and Fuel Quality directives, assuming that EU biofuels will remain at 3.2% in the baseline up
to 2020. The baseline assumptions are presented in Chapter 2.

Several policy scenarios are compared with the baseline. Selected scenarios are listed in Table 3.
First, feedstock-specific scenarios are modelled, looking at the effect of increasing the incorporation
level of one biofuel feedstock only (the list of feedstocks is presented in Table 7 in Chapter 3).
Scenarios on the total EU biofuel mix in 2020 were also modelled. In addition, alternative scenarios
are developed that assess the impact of using abandoned farmland for biofuel crop production in the
EU and lower or higher deforestation. There is large recognition of the sensitivity of LUC impacts to
behavioural parameters in economic models. For that reason, sensitivity analyses are performed to
explore uncertainty ranges around the results of these scenarios. These highlight different
developments of the model variables from the same baseline. For instance, changing the elasticity of
endogenous yield response can lead, for the same future food consumption patterns, to different
land use changes.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 4

The sensitivity analysis is performed through Monte-Carlo simulations, i.e. the GLOBIOM model ran a
large number of times, drawing random values for parameters in a plausible distribution, to produce
an estimate of the results distribution. The Monte Carlo simulations, modelling parameters used in
the simulations and outcomes of the simulations are further described in Section 3.5 and Annex IV,
while summary graphs are included in the modelling result sheets presented in Section 3.2.

The table below provides an overview of scenarios modelled in this study. A more in-depth
description of each of the scenarios is provided in Chapter 3.

Table 1: List of scenarios in this study

# Baseline and scenarios Nr. Sensitivity analysis

Baseline

A0 Baseline: global trends between 2000 and 2030 YES

Feedstock scenarios

A “Marginal feedstock”: A0 +1% biofuel consumption per feedstock 13 YES

“Marginal feedstock for cereal straw”: A0 + 1% shock of straw

A1 4 YES
ethanol for EU and for three selected Member States
“Marginal feedstock groups”: as A, but with crop groups (ILUC
A2 3 YES
proposal)
Policy scenarios

B “EU biofuel mix in 2020”: A0 + biofuel consumption forecasts from MS

1 YES
NREAPs

“EU biofuel mix in 2020 with 7% cap”: B + maximum of 7%

B1 1 NO
conventional biofuels
Explorative scenarios

C “Biofuels + increased use of abandoned land in EU”: incentivised

1 NO
land expansion into EU abandoned land in the baseline + Scenario B
“Biofuels + low deforestation ”: assumed lower deforestation (two
C1 levels) worldwide and halting of peatland conversion in the baseline 3 NO
compared to recent trends + Scenario B

“Biofuels + high deforestation”: assumed higher deforestation

C2 1 NO
worldwide in the baseline compared to recent trends + Scenario B
TOTAL NUMBER OF SCENARIOS 27

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 5

1.6 Interactive project approach leading to transparent results

Stakeholder involvement is essential in improving the understanding of LUC impacts and to create
maximum transparency in the modelling exercise. The following actions have been taken for this
purpose:
 We have provided a detailed description of the modelling approach and differences in
comparison to the previous IFPRI study;
 We invested in stakeholder outreach;
 We established a scientific advisory committee;
 In this report, inputs are discussed in detail;
 Modeling results are decomposed to increase the understanding of various ‘LUC dampening
effects’ such as yield increase and demand reduction, and the role of various sources of
emissions as part of the total results.

Stakeholders received ample opportunities to provide input to the study. Two stakeholder
consultations were organised, ongoing exchange with stakeholders took place via the project mailbox
([email protected]) and an Advisory Committee was formed with nine international experts on ILUC
quantification and agriculture. The first stakeholder consultation took place in November-December
2013 and aimed to make stakeholders familiar with the GLOBIOM model and identify possibilities to
improve the model during the course of the study project. To this end, separate stakeholder
meetings with conventional ethanol supply chain, conventional biodiesel supply chain, advanced
biofuel producers and non-government organisations (NGOs) were organised in Brussels. Prior to
these meetings, the consortium circulated a brief description of GLOBIOM, plus a more detailed
description of the GLOBIOM and comparison with MIRAGE-BioF model (IFPRI) to around 200
stakeholders in the ILUC debate, with the invitation to provide comments, suggestions or questions
to the project mailbox. This consultation resulted in a long-list of 47 possible improvements to
GLOBIOM and suggestions for feedstocks and scenarios to be modelled. This long-list was discussed
with the Advisory Committee and the European Commission’s steering committee

The second consultation took place in February-March 2014, this time to discuss proposals for a
number of improvements to be made to the GLOBIOM model, proposals for a modelling baseline,
scenarios and feedstock choice, and to outline the planned sensitivity and uncertainty analysis. The
consortium circulated relevant documentation to around 200 stakeholders and published the
consultation documents on the project website.9 Four stakeholder meetings with the above-
mentioned stakeholder groups were organised and stakeholders were invited to submit comments
via the project mailbox. Comments and suggestions obtained in this second consultation were
assessed by the consortium and discussed with the European Commission’s steering committee. This
resulted in a final selection of changes to be made to the GLOBIOM model, a set-up for the modelling
baseline and scenarios, and the feedstocks to be modelled.

9
www.globiom-iluc.eu

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 6

Following the second consultation, the consortium started to implement the selected changes to
GLOBIOM (see Chapter 3 for a more detailed description). In parallel, the consortium prepared a
document with modelling parameters, mainly focusing on biofuel production pathways, which were
shared for comments with selected biofuel industry associations and subsequently published on the
project website.

Some stakeholders requested our consortium to obtain access to the model. It was not possible to
fulfil this wish within the scope of our study, as the model in itself is not ‘open source’ and is
proprietary owned by IIASA who invested significantly in developing and fine-tuning the model and
datasets used. It is clear that the model, like any equilibrium model, is a highly complex tool, as it
represents the entire global agricultural and forestry sectors and the most important global economic
drivers and trade relations, with thousands of lines of modelling code. This means it can only be
effectively operated by modelling experts. IIASA works with other research groups in several joint
research projects, during which those research groups are being trained to use the model and
subsequently have access to the model. IIASA is open to collaborate with research group(s) who
would like to perform a research project, which could take the form of a peer review of the current
study.

The study consortium had several meetings with the scientific advisory committee, whose role was to
critically assess our proposed modelling approach, suggest improvements to the GLOBIOM model
and assess draft modelling results. The committee was not involved in the actual modelling but was
able to obtain a good overview of the way in which IIASA performed the modelling.

1.7 How the current study follows previous land use change studies

Land use change quantification started in the United States. In 2008, Searchinger and colleagues
were the first to publish estimates of indirect land use change impacts associated with US biofuel
consumption, by means of a modelling framework. They looked at different alternative feedstocks
used to produce ethanol using the FAPRI-CARD model. They calculated that greenhouse gas
emissions from indirect land use change would represent 104 gCO2e/MJ for corn ethanol alone if
amortised on a 30 years period. They calculated that, in order to achieve 20% emission savings from
corn ethanol relative to fossil fuel, the corn ethanol would need to be produced from the same land
for over 167 years to repay the ILUC emissions. Looking at some other feedstocks, the authors were
pessimistic: growing miscanthus instead of corn in fertile areas would still generate 111 gCO2e/MJ in
impacts and need 52 years to repay (thanks to a better LCA direct saving coefficient) and Brazilian
sugar cane ethanol would need four years to repay if expansion occured into grassland, but 45 years
if tropical forest was converted.

US researchers Keeney and Hertel (2009) strongly criticised Searchinger’s paper, arguing that the
role of endogenous yield response to price change had not been adequately addressed in the
analysis. They argued that endogenous yield response could be higher than in Searchinger’s
alternative scenario, in which 20% of additional demand could be met by increased corn yields. They
presented simulations with a variant of the GTAP model where a third of the additional demand could
be met through crop yield increases. Their model was further used to provide more comprehensive

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 7

analysis of US biofuel mandates. They found a 30-year LUC value (LUC emissions per unit of biofuel
averaged on a 30-year period) of 27 gCO2e/MJ. Although this value is a quarter of the value initially
calculated by Searchinger, this result is still too high to allow climate change mitigation benefits from
using corn ethanol.

The GTAP model has also been used in a wider set of LUC impact estimations led by the California Air
Resource Board (CARB) in the context of the Low Carbon Fuel Standards regulation. LUC impacts
used by CARB are 30 gCO2e/MJ for corn ethanol, 46 gCO2e/MJ for sugarcane ethanol, and 62
gCO2e/MJ for soybean biodiesel (CARB, 2009).

In parallel, a more comprehensive assessment of impacts of different US biofuel feedstock is the

Regulatory Impact Analysis performed by US EPA and released in 2010. Using a wide set of models
(FAPRI, GREET, FASOM), the exercise computed ILUC factors for many existing and advanced
biofuels. The ILUC factor for corn ethanol from EPA is identical to the CARB estimate, at 30
gCO2e/MJ, but it is lower for soybean biodiesel, at 40 gCO2e/MJ for 30 years (EPA, 2010). Sugar cane
ethanol has the lowest ILUC factor at 4 gCO2e/MJ, whereas switchgrass ethanol is attributed 14
gCO2e/MJ.

In 2010, the first large LUC quantification for EU biofuels was published. Al Riffai and colleagues
estimated a 20-year LUC factor of 18-20 gCO2/MJ for EU biofuel policy, with scenarios relying
significantly on sugar cane imports (with the range reflecting different trade assumptions), using the
MIRAGE-BioF model. The model was also used to look at the respective impact of each feedstock by
testing the effect of some marginal shocks. They found that biodiesel feedstocks typically result in
higher LUC impacts per unit of energy than bioethanol ones. This IFPRI study by Laborde (2011),
assessing the impact of the NREAPs with the same model, has been used by the European
Commission as the scientific basis for its Impact Assessment10 that accompanied the ‘ILUC proposal’
referred to in Section 1.1 above. Other computable general equilibrium (CGE) models have found
similar results: Britz and Hertel (2011) used the GTAP model to explore rapeseed related LUC
impacts in Europe and estimated a LUC value of 42 gCO2/MJ, confirming the higher LUC emissions
from biodiesel feedstock are mostly due to lower yields and the typical replacement by palm oil
causing expansion in high carbon stock land.

Most modelling exercises that have been performed so far were based either on general equilibrium
approaches (models such as GTAP, EPPA or MIRAGE), or economic model linkages (EPA design).
Both techniques suffer from notable limitations:
 CGEs have a clear lack of sectorial detail, robust supply side description and lack of tractability
of the biophysical variables. These models are mainly based on social accounting matrixes and
rarely incorporate a precise account of input-output physical constraints and process
technologies;
 Model linkages incorporate greater detail thanks to refined national models but can suffer from
inconsistencies. For example, the 2010 EPA model could not reproduce similar production and
export levels for some commodities as the two FASOM and FAPRI models.

10
SWD(2012)343

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 8

While these models have been improved, questions on the uncertainty around LUC impacts have
been raised more often and more strongly in recent years.

In 2010, Plevin and colleagues assessed the uncertainty in LUC models through a simplified model.
They showed that the 95% confidence interval on carbon stock, model behaviour, or amortization
period would result in range of LUC impacts from 21 to 142 gCO2e/MJ/y. More strikingly, they found
an upper bound of 340 gCO2/MJ, much higher than all previous estimates, whereas their lower
estimate would be only about 10 gCO2e/MJ.

In order to support the scientific foundation for its legislative proposal on ILUC, the European
Commission commissioned IFPRI to improve and refine their MIRAGE-Biof model and estimate LUC
values for EU biofuels. In October 2011, the IFPRI report “Assessing the Land Use Change
Consequences of European Biofuel Policies” was published, which to date is the most referred to
source of quantitative information on LUC GHG effects of EU biofuel consumption.

It can be concluded that a wealth of analysis has been undertaken on LUC impacts, but significant
uncertainties remain in part due to shortcomings in the modelling approaches. From the previous
studies, the IFPRI-MIRAGE study in particular is relevant, as it focuses on EU biofuels like this study.
For this reason we compare GLOBIOM with IFPRI-MIRAGE in our detailed description of GLOBIOM in
Annex II.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 9

2 Description of the modelling baseline

In this study, the LUC impacts of the European biofuel policy are assessed by comparing different
biofuel demand scenarios with a baseline scenario (see Figure 11 in Chapter 3). Basically, we
compare a world without the EU RED and FQD directives to a world with the European biofuel
incentives under various scenarios. The baseline represents the way the world develops between the
model base year, 2000, and 2030, without European biofuel incentives. The model is calibrated in
the year 2000 because some important spatially explicit datasets are not available every year and
2000 is the most commonly studied reference point.11 However, because more recent statistics are
available on market data, some more recent parameters, such as evolution of GDP, population, fossil
fuel prices, exchange rates, average yield and consumption patterns, have been used to better
model recent developments, permitting comparison with the modelled results for the period 2000-
2010, as illustrated by this section.

The baseline uses the level of biofuel consumption in the EU in 2008 just before the RED and FQD
Directives were passed by EU legislators. This EU demand equals 9.8 Mtoe of conventional biofuels
(equivalent to 3.2% of the total liquid fuel demand) and is kept constant throughout the modelled
time span. However, biofuel policies that have been, or will be, introduced between 2008 and 2030
in other regions of the world are included in the baseline.

Aside from biofuel demand, there are many variables that influence demand for land-based products
and thereby land use. Examples include population growth, GDP and dietary patterns. Equally, there
are variables on the supply side that influence the acreage needed to meet demand for land-based
products, such as crop yield and livestock productivity. Furthermore, the development of the energy
demand in the European transport sector will play a role in determining the amount of biofuel that is
needed to meet the 10% EU RED target and hence influences the total LUC impacts from the EU
biofuel policy.

In the establishment of the modelling baseline, assumptions have been taken regarding the
development of the variables mentioned above. Data and sources for the most important exogenous
parameters used in the baseline are presented in Section 2.1. The results for the most important
endogenous parameters are presented in Section 2.2. Further information on the input data used in
this modelling study is provided in Annex V.

11
The JRC global land cover dataset (GLC 2000) has been released for the year 2000 only. For the EU, the Corine Land Cover dataset is
available for the years 2000, 2006 and 2009. The crop allocation model from IFPRI (SPAM) provides data for two years, 2000 and 2005.
The Gridded Livestock of the World (GLW) dataset on livestock distribution is available for the years 2000 and 2005 only. The global
biomass carbon map from Ruesch and Gibbs (2008) relies on land cover for the year 2000 and has not been yet updated to later years.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 10

2.1 Baseline assumptions

This section presents the most important assumptions that are used in the baseline scenario.

2.1.1 Macroeconomics

Driver Assumption Data source

“Middle of the Road” pathway (SSP2 scenario) in which the
Population SSP Database: (IIASA,2015)
world population reaches 7.6 and 8.3 billion in 2020 and
growth
2030 respectively.

The Shared Socio-economic Pathways (SSPs) are consistent and harmonised prospective scenarios
developed and widely used by the scientific community in the framework of research on climate
change. The “Middle of the Road” pathway (SSP2) used in the baseline assumes the continuation of
currently observed trends in population growth with 7.6 billion people globally in 2020 and 8.3 billion
by 2030.

Driver Assumption Data source

“Middle of the Road” pathway (SSP2 scenario) in which the
SSP Database: IIASA (2015)
GDP growth global per capita GDP increases from USD 6,700 in 2005 to
USD 8,800 and USD 10,900 in 2020 and 2030 respectively.

Data from the same (Middle of the Road) Socio-economic Pathway is used to ensure consistency of
GDP projections with population assumptions. In SSP2, the trend of fast growth in emerging regions
continues. Per capita GDP is projected to increase by 125% for China and 170% for India between
2010 and 2030.

2.1.2 Energy

Driver Assumption Data source

Fuel
Total liquid fuel demand in the EU-28 transport sector EU Energy, Transport and GHG
demand in
decreases from 12,947 PJ in 2010 to 12,294 and 11,955 PJ emissions Trends to 2050
EU
in 2020 and 2030 respectively. (European Commission, 2013)
transport

Fuel consumption in the transportation sector has been declining in Europe since the peak of oil
prices in 2007-2008. We follow the Reference 2013 scenario of DG Energy for our projections of
future fossil fuel demand in the transportation sector, which anticipates continuation of this trend.
Under this scenario, total EU demand for transportation fuel is expected to decrease further by about
8% between 2010 and 2030, also as a consequence of accelerating energy efficiency improvements.
The share of diesel in total diesel and gasoline consumption increases from 68% in 2010 to 82% in
2030 in total transport fuel demand and from 42% to 61% in passenger car fuel demand.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 11

 Driver Assumption Data source
Kept constant at 2008 levels:
Biofuel
demand in 1G: 3.2% (408 PJ) of total EU transport fuel demand Laborde (2011)
EU 2G: 0%

The EU biofuel demand in the baseline is kept constant at 2008 levels, the year before the 10%
renewable energy target for the transport sector was enforced in the EU RED and the emissions
reduction target was revised to 6% in the FQD. This equates to 408 PJ, equivalent to 3.2% of the
total fuel demand in the EU transport sector. Of this 408 PJ biofuel demand in 2008, 83% is biodiesel
and 17% is ethanol (consistent with the assumption used in the IFPRI 2011 study), all produced from
‘first generation’ (1G) feedstocks. Hence zero ‘second generation’ (2G) biofuel demand is assumed in
the baseline. The EU biofuel demand is kept constant until 2030 at 2008 levels to assess the LUC
impact of the EU mandate. At the same time this assumption allows for comparison with the 2020
LUC values reported by IFPRI (2010), in which the same approach is taken.

Driver Assumption Data source

Main biofuel mandates incorporated, summing up to:
Biofuel
Values based on Lotze-Campen
demand in 1G: 338 PJ (2000), 1,717 PJ (2010), 2,406 PJ (2020) and
et al. (2014) adapted for lower
rest of the 2,828 PJ (2030)
biofuel demand in US and Brazil
world 2G: 0 (2000 – 2010), 16 PJ (2020) and 21 PJ (2030)

1G biofuel demand in the rest of the world is based on the US Information Energy Administration for
USA and on AgMIP 1G scenario (Lotze-Campen et al., 2014) for the rest of the world.
This latter set of projections has been developed by a consortium of modellers working on global
agricultural scenarios. The demand for biofuel outside the EU comes mainly (but not exclusively)
from the following countries as a consequence of national biofuel commitments:

USA: Partial implementation of the 2,871 PJ (36 billion gallon) Renewable Fuel Standards mandate
by 2022; 1,166 PJ (14.5 billion gallon) from maize ethanol in 2020 and 259 PJ (3.6 billion gallon)
from advanced non-cellulosic biofuels (70% biodiesel and 30% sugar cane based). Cellulosic ethanol
development remains marginal with only 16 PJ deployed by 2020 (0.2 billion gallon).
Brazil: Stable ethanol incorporation and assumption of prolongued transportation fuel demand over
the next decades in line with the 2000-2010 increase, rising from 467 PJ in 2010 to 731 PJ and 994
PJ in 2020 and 2030, respectively. Biodiesel incorporation triples during the period, from 64 PJ in
2010 to 219 PJ in 2030.
Argentina: Incorporation of 10% biodiesel in diesel fuel by 2020 at 64 PJ.
China: Stable ethanol incorporation rate, but increases of 8% per year in fuel transport demand, to
reach 27 PJ by 2030.
Canada: Incorporation of 5% ethanol in gasoline by 2020 (27 PJ); no biodiesel demand considered.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 12

Indonesia: Consumption of 0.9 Mt of palm oil biodiesel in 2013 (USDA). Indonesia has introduced a
biodiesel mandate that sets strong targets in coming years, up to 20% in 2020. However, biodiesel
consumption in reality lagged far behind the mandate quantity, so it remains unsure what the
expected biodiesel consumption will be in 2020. For this reason, we chose to include double the
quantity of the real biodiesel quantity consumed in 2013 in our study baseline.

Biofuel produced in the regions above is not all freely traded. Indeed, some restrictions are currently
in force, such as EU anti-dumping duties on biodiesel imports from Argentina, US and Indonesia and
US corn ethanol. In our modelling, we therefore consider that soybean biodiesel exports from
Argentina to the EU, and corn ethanol exports from the US to the EU, are impossible. However, we
do not put restriction on palm based biofuels from Indonesia, due to the potential to produce
hydrogenated vegetable oil from palm and to export it to the EU market. Duties on biodiesel are
relatively inefficient for limiting the flow of palm oil use from these different regions, due to the
possibility to directly ship the raw feedstock to another country or to the EU directly to produce the
biodiesel.

Biofuel feedstocks are transformed into various types of liquid fuels through different transformation
processes, whose conversion efficiencies are provided in Appendix IV.5. It should be kept in mind
that no specific assumption is made in the baseline about variation in conversion efficiencies over
time.

Driver Assumption Data source

Solid Global solid biomass demand continues to grow from its

World Energy Outlook, 2010
biomass 2010 level (43,800 PJ) until 2030, but at a decreasing
(IEA, 2010); “Current policies”
demand for pace, reaching 47,200 PJ in 2020 and 48,500 PJ in 2030
scenario.
energy (final energy).

The model assumptions on solid biomass demand levels are fitted to historical data from 2000
(38,500 PJ) and 2010 (43,800 PJ) using data from International Energy Agency. Electricity
generation from woody biomass is assumed to strongly increase from 4,000 PJ/y in 2010 to 10,200
PJ/y in 2030 at a global level. Traditional use of biomass remains significant in developing countries
and increases by 4.5% up until 2030 before decreasing. In particular, demand growth remains high
in Sub-Saharan Africa (+18% 2010–2030) whereas it stabilises in India (+3%) and decreases in
China (-30%). This use still represents 76% of the total solid biomass consumption by 2030.

Driver Assumption Data source

Crude oil price is considered stable at a high level over the
Fossil fuel World Energy Outlook 2013
2010-2030 and is assumed at USD 121 for 2020–2030 (in 12
prices (IEA, 2013)
real terms).

12
The consortium used the most up-to-date price assumptions from the World Energy Outlook at the time of the modelling (IEA 2013), but
relied on some older edition for the long term solid biomass projections, as all the detailed datasets from that year were already available to
the consortium. Energy consumption projections, that depend heavily on past investments, are subject to more ineria than energy price
projections, that are more influenced by conjonctural developments.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 13

The model captures the increased price of fossil fuel since 2000 with a four-fold increase in the price
of crude oil from USD 25-30 per barrel in 2000 to USD 109 in 2012. This shock is implemented in the
model for the year 2010 and impacts the price of fertilisers, and therefore farm gate prices, in large
producers of agricultural products (see crop prices in Section 2.2.9).

2.1.3 Food and agriculture

Driver Assumption Data source

GLOBIOM with SSP2
In the 30 year modelling period, per capita food
macroeconomic assumptions
consumption increases across the world by 11.6% from
Diet patterns and diet preference changes
an average of 2,729 kcal/capita/day in 2000 to 3,045
from Alexandratos and Bruinsma
kcal/capita/day in 2030 (see Figure 4).
(2012)

Food demand in GLOBIOM depends on two main factors: i) an exogenous component, depending on
evolution of income per capita and food pattern changes as anticipated by FAO; ii) an endogenous
response, depending on change in price level in the model. Projections for food demand in the model
baseline are illustrated in Figure 4. Total level of consumption increases in our baseline scenario from
2729 kcal/capita/day to 3,045.13 These projections are in line with those from FAO (Alexandratos and
Bruisma, 2012), which project 2960 kcal/capita/day by 2030, but are slightly larger due to higher
GDP growth.14 For the year 2010, however, our projections are lower by 3.5% compared to FAOSTAT
data, due to the response in the model of consumers from developing countries to price increases in
the period 2000-2010.15 Per capita meat consumption increases in developing regions by 34%, from
27 kg/person/year in 2000, to 36 kg/person/year in 2030. In the same period, the meat
consumption per capita increase in the developed regions is smaller (12%), mainly due to the
decrease of consumption of ruminant meat. Compared to developing countries, however, the
absolute per capita meat consumption in 2030 (97 kg/person/year) is still three times higher. The
global average per capita meat consumption increases by 21%, from 37 kg/person/year to 45
kg/person/year. In developed regions, a small substitution occurs from bovine meat to pig and
poultry meat consumption. Similar patterns are reported for milk and other dairy consumption.

13
Our definition for food consumption following the one from FAO, this variable accounts here for food effectively ingested and for
household waste. Therefore the values reported excess by far recommended daily intakes from usual dietary guidelines.
14
See Valin et al. (2014) for more details on how SSP2 macroeconomic growth rate changes FAO projections compared to FAO assumed
growth rate.
15
It is important to note that FAOSTAT data on food consumption – or literally “food supply quantity per capita” – correspond to calculated
data by difference between production, trade and other uses and is not reported data. Comparison with such statistics should therefore
been done with care.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 14

Figure 4: Per capita food consumption projections in the baseline. Source: GLOBIOM and FAOSTAT

Driver Assumption Data source

- Direct payments under CAP stay constant
Common
- No subsidy on energy crops
agricultural Eurostat (2014)
- No further impact of Ecological Focus Area policy on
policies
level of set-aside land in EU agricultural production

The level of direct payments (financial support directly granted to farmers to ensure a stable income
in volatile market) is assumed to stay constant throughout the modelling timeframe and energy
crops are not financially incentivised. The Ecological Focus Area (EFA) policy, which obliges farmers
to appoint 5% of their land as an EFA (which covers a range of land types including fallow land,
hedges and areas with nitrogen fixing crops) – a value that might go up after review in 2017 – is not
assumed to have any further impact on the EU agricultural production and the level of set-aside land
is considered here constant. Recent policy developments on the EU Common Agricultural Policy, such
as the sugar reform, is not taken into account.

Driver Assumption Data source

- Status quo of trade policies accounted for, with the
World Trade Organization
exception of the WTO accession of China and Russia (in
Trade www.wto.org
2001 and 2012 respectively)
policies MacMap-HS6 tariffs database
- Free Trade Area of the Americas (FTAA), currently
www.cepii.fr
under negotiation, is not accounted for

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 15

Tariff information in GLOBIOM is based on MAcMap-HS6 2001, which provides details on the applied
bilateral protection at the level of product tariff lines. This tariff information is used to calibrate the
model in the year 2000 and tariff changes are then applied in the baseline where relevant. China’s
accession to WTO in 2001 is considered to have had a major impact on the imports of soybeans from
Latin America, and in the case of Russia, accession to WTO in 2012 prevents any form of price
regulation by import or exports tariff adjustments. With respect to trade of biofuels, we assume that
imports to the EU of corn ethanol from US and biodiesel from Latin America and Southeast Asia are
restricted by anti-dumping measures. The corresponding feedstocks, however, can be traded.

2.1.4 Biomass demand in other sectors

Driver Assumption Data source

Uses other than food and feed are assumed to follow the
pattern of food demand, except for some particular
commodities for which outstanding trends have emerged
over the past decade. In particular, we took into account
the expansion of palm oil use in Asia and in North
Other
America, and the expansion of cotton in South Asia.
biomass FAOSTAT
At the EU level, biomaterial (biopolymers, bitumen) and
demand
biochemical (surfactants, solvents, lubricants) have been
increasingly used over the past decade, but scenarios on
their prospects diverge. As a defaut assumption, we do
not consider here any further increase in the incorporated
share of biomaterials.

Biomass demand grows in all sectors throughout the modelling timeframe and total annual biomass
demand increases to 4.5 Gt in 2030 (a 70% increase compared to 2000). Of the total additional
biomass demand in 2030 (compared to 2000), the feed sector takes the largest share (35%),
followed by the food sector (28%), other sectors (23%) and the biofuel sector outside the EU (14%).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 16

Figure 5: Global biomass projected demand by sector in the baseline. Source: GLOBIOM

2.1.5 Land use

Driver Assumption Data source

Land World database on protected
Protected areas based on WDPA
protection areas: IUCN and UNEP (2014)

The World Database on Protected Areas (WDPA) is the most comprehensive global dataset on
terrestrial and marine protected areas and is maintained at the UNEP World Conservation Monitoring
Centre (UNEP-WCMC). This dataset is introduced in the model here to define ‘no expansion’ areas
corresponding to protected areas into IUCN categories Ia (Nature Reserve), Ib (Wilderness area) and
II (National Park). Land conversion is excluded from these areas in the model and no agricultural or
forestry activities are allowed.

Driver Assumption Data source

Global Forest Resource
Deforestation No particular deviation from deforestation trend
Assessment 2010: FAO (2010)
policy observed in the period 2006–2012
and Hansen et al. (2015)

The reference period for deforestation is 2005–2012. This means that the effects of policies put in
place between 2000 and 2010 are only considered to the extent that their effects have been visible
by 2012, the most recent date available for land cover change data at the time of the start of the
modelling. Evidence for the effectiveness of policies to avoid deforestation includes, for example,
Brazil, where deforestation decreases from an average 3.4 Mha forest losses/y in the period 2001–
2005, to 2.4 Mha in the period 2006–2009, and 2.3 Mha in the period 2010–2012, following
substantial efforts for a greater enforcement of protection and monitoring policies.
In contrast, in spite of several initiatives to better control pressures on forests, clearings in Indonesia
has increased over time from 0.9 Mha in 2001–2005, to 1.6 Mha in the period 2006–2012. The year
2012 has been marked by the highest deforestation rate registered to date, with more than 2 Mha of
forest cleared. Therefore, no inflexion in deforestation trend is considered in that region in the model
baseline.16 The effects of climate change on the forest cover and the carbon uptake rate are not
modelled.

Deforestation as modelled in GLOBIOM only captures a part of the historic deforestation, because the
model only represents expansion of land for agricultural activitiesand not illegal logging and forest
degradation for fuel wood or other purposes. In Brazil, for the period 2010–2020, the model

16
At the time of redaction of this report, newly released statistics for the year 2013 by Global Forest Watch indicate for that year a much
lower deforestation rate in Indonesia than on the previous years, at 1 Mha. This information could not be used by the consortium, but would
have remained anyway too isolated to conclude that the trend has been reversed in this region. In order to explore potential implications
that such favourable development would have on the results of this study, we refer the reader to the low deforestation scenarios presented
in the next section.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 17

calculates 0.6 Mha of forest clearing per year.17 In Southeast Asia, deforestation is projected at 0.5
Mha per year, and in Sub-saharan Africa, at 1 Mha per year.18

2.2 Baseline results

2.2.1 Biofuel production and feedstock quantities used

Biofuel production increased from 25 PJ to 350 PJ between 2000 and 2008 in Europe (with an
additional contribution of silage corn, whose production amounted to 155 PJ, both for biofuels, heat
and power combined). This can be observed in Table 2, which shows biofuel production levels from
the model between 2000 and 2030. Both 2000 and 2010 data are sourced from GLOBIOM but
calibrated on external statistics from EurObserv’ER and estimates by EU FAS posts (USDA Foreign
Agriculture Service) for the EU and based on US Energy Information Administration for global
statistics. During the same period, the production of biofuel in the rest of the world increases from
338 PJ in 2000 to 2,873 PJ in 2030, following our assumptions. The strongest growth between 2000
and 2010 corresponds to the development of corn ethanol in the US, whose volume has been
multiplied by eight as a result of the Renewable Fuel Standards program. After 2010, we consider
that this development stalls and the highest growth (5.8% per year) is observed in ethanol from
sugar cane, stimulated by Brazilian demand (incorporation policy is assumed unchanged) and
exports to North America (Table 2).

Wheat straw, short rotation coppice, forestry residues and grassy crops are not reported in this table,
because no demand for these biofuels is assumed in the baseline. Silage corn, however, is reported
in this table, because it is used in the baseline for both cogeneration and combustion in transport
(the two different uses could however not be distinguished).

17
Note that this version differs from the regional version of GLOBIOM dedicated to the study of deforestation in Brazil, and that represents
all the deforestation drivers in that region (see www.redd-pac.org).
18
This version differs from the regional version of GLOBIOM applied to deforestation in the Congo Basin (Mosnier et al., 2012).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 18

Table 2: Biofuel production per feedstock

Biofuel and biogas production from crops (PJ/year)

Feedstock 2000 2010 2020 2030
EU RoW EU RoW EU RoW EU RoW
Corn ethanol 130 10 1,055 9 1,166 9 1,183

Palm oil biodiesel 2 29 95 94

Rapeseed oil biodiesel 20 212 202 198

Soybean oil biodiesel 3 0 93 166 88 388 86 509

Sugar beet ethanol 19 18 17

Sugar cane ethanol 204 488 826 1,093

Sunflower oil biodiesel 0 5 4 4

Wheat ethanol 4 12 28 11 44 11 61

Silage maizea 155 155 155

a
Both uses in transportation and in heat and electricity sector are accounted here.

2.2.2 Livestock productivity

Driver Assumption Data source

Livestock feed Livestock feed conversion efficiencies increase in

conversion developing regions by up to 30-50% by 2030 for SSP2 Animal Change: IIASA 2011
efficiency but grow only slowly in Europe (below 5% increase).

Livestock feed efficiency is increasing in all parts of the world and is driven in GLOBIOM by
technological change (exogenous trend from the Animal Change project; see IIASA, 2011) and
livestock system transitions, with the model explicitly representing different livestock management
systems. In the EU, livestock productivity for meat measured per unit of land increases by 8% for
ruminants between 2000 and 2030. If we measure this productivity in terms of feed conversion
efficiency, the increase is even lower, at 7%. For pigs and poultry, however, feed conversion
efficiency is considered close to its maximum and no significant change is observed in the period.
It is in the rest of the world that productivity gains are the most impressive in the period 2000-2030.
It increases by 63% for ruminant meat on per ha basis, and 33% for feed conversion efficiency. On
the pigs and poultry side, productivity is also increasing with the transition from smallholders to
more industrial systems, but in terms of resulting conversion efficiency, this results in a decrease in
the model, because smallholders systems rely heavily on scavenging, which progressively disappears
with livestock sector industrialisation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 19

 Table 3: Meat

productivity in
kg protein/ha grassland kg protein/t dm feed
the EU and the

rest of the world

2000 2010 2020 2030 2000 2010 2020 2030

Bovine meat 64.4 66 67.6 67.9 9.9 10.3 10.7 11

Sheep and goat meat 27.2 28.6 29.7 30.9 5.5 5.7 6 6.3
EU28
Pig meat 16.7 16.8 16.8 16.8
Poultry meat 32.9 32.8 32.2 32.1

Bovine meat 6.8 8.2 9.7 11.1 3.5 4.1 4.6 5.2

Rest of the Sheep and goat meat 5.1 6.2 7.4 8.7 3.1 3.4 3.7 3.9
world Pig meat 18.2 17.4 16 15.7

Poultry meat 26.3 25.7 24.7 24.5

2.2.3 Crop yield

Driver Assumption Data source

AgLINK-COSIMO baseline 2010-2030 for the EU28 and
Crop yield extrapolation of yield change on period 1998-2012 for the FAOSTAT
rest of the world

Between 2010 and 2030, the global crop productivity of biofuel crops increases on average
1.0%/year (0.6%/year in the EU28 and 1.1%/year in the rest of the world). Strongest productivity
growth is expected in wheat production between 2010 and 2020, with a strong catching up of yield in
Latin America – still at two third of EU average yield in 2010, but steadily increasing. Maize
productivity is assumed to continue with a significant yield increase globally, whereas sugar crops
follow a more moderate yield increase. In particular, the strong increase in yield productivity
observed historically for sugar beet in the period 2000-2010, and related to the sector restructuring,
is no longer reflected in the model projections for 2010-2030, following the AgLINK-COSIMO
assumptions.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 20

Figure 6: Crop yield projections in the baseline for some biofuel feedstocks. Source: GLOBIOM

Note that the yields from FAO have been further adjusted in GLOBIOM to better reflect the effect of
multi-cropping, the practice of harvesting two or more crops successively from the same cropland in
one year. Country specific shifts in cropping intensities are obtained from the literature as exogenous
variables, although it should be kept in mind that no specific endogenous response is associated with
this feature (see Annex II.8 for more details on the multi-crop modelling assumptions taken in
GLOBIOM).

2.2.4 Total crop production

In the baseline, in response to food demand, feed needs for livestock, biofuel and fibre demand, total
crop production increases by 70% from 2.60 Gt/year in 2000, to 4.52 Gt/year in 2030. The crop
production increase between 2000 and 2030 is only 5% in Europe, which is consistent with current
observations, as cereal production has been observed to grow by only 3.5% in Europe between 2000
and 2010, and oilseeds expansion (+40%) has been historically associated to the biofuel demand,
which is maintained stable in our modelling after 2008. In contrast, production is projected to
increase by 82% in the rest of the world. This can be compared with a 22% increase in cereal
production between 2000 and 2010, and 56% increase for oilseeds (accounted in primary
equivalent).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 21

Figure 7: EU and rest of the world crop production in the baseline. Source: GLOBIOM. Line for ‘all crops’ includes

non-cereal and non-oilseed crops

2.2.5 Cultivated area

The global cultivated area for crops included in GLOBIOM (about 70% of total cultivated areas)
increases by 11.1% between 2000 (954 Mha) and 2030 (1,060 Mha), after a modest growth of 0.6%
between 2000 and 2010 (960 Mha) and a much more rapid increase by 2020 (1,013 Mha).
Cultivated area corresponds to areas used for crops, and differs from harvested areas that add to
this number all the multi-cropped areas. According to FAOSTAT, arable land would have increased by
20 Mha between 2000 and 2010, therefore extrapolating this trend linearly would lead to 60 Mha of
expansion by 2030. The increase in cultivated areas between 2000 and 2010 is more limited in the
model due to the assumption about multi-cropping development, in particular in India and China,
and also due to the slightly decreased demand in the model from the response to food price changes
over the period 2000-2010 (see food demand assumptions). These areas however increase in the
period 2010-2030 when prices are more stable and demand is increasing more steadily with the
increasing population and economic growth. Throughout the modelling timeframe, the strongest
growth is expected in Sub-Saharan Africa (53% in absolute terms), also, followed by Central & South
America (42%), North Africa (30%) and Southeast Asia (18%). Total cultivated area declines in
Oceania (-20%), Eastern Asia & Pacific (-12%), the EU (-13%) and South Asia (-4%).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 22

Figure 8: Total cultivated area per region in the baseline projections. Source GLOBIOM

2.2.6 Absolute crop area change

Global harvested areas have significantly increased over the past years, with an additional 100 Mha
harvested between the years 2000 and 2010, of which the GLOBIOM crops account for around 80
Mha. The absolute area change per crop projected between 2010 and 2030 amounts in GLOBIOM to
159 Mha, which corresponds to the same rate of expansion per decade as 2000–2010. Expansion
patterns across crops differ significantly between regions (see Figure 9). Changes in the EU are
relatively small and, in the absence of any biofuel policy and following the slowing down of meat
demand, total crop area declines, mainly through wheat harvested area (-2.5 Mha) and maize (-1.2
Mha). Changes are more pronounced in the Americas, Eastern Asia & Pacific and the Africa with the
largest absolute crop area increase in the Americas and Africa. Significant changes at a crop level
include maize area expansion at the expense of wheat, soybean area expansion in Central & South
America (20 Mha) and oil palm area expansion in Southeast Asia (6 Mha) and Sub-Saharan Africa
(2.5 Mha).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 23

Figure 9: Crop area change in the baseline between 2010 and 2030. Source: GLOBIOM

2.2.7 Land use change

Absolute LUC by land use type between 2010 and 2030 shows a similar pattern across regions:
forest and other natural vegetation land is lost to grassland and cropland. Cropland expands in
particular in Central & South America and Sub-Saharan Africa at the expense of forest and other
natural vegetation, which is well in line with historic observations (see Section 2.2.5 on cultivated
area).19 The EU is an exception and sees an increase in forest and abandoned land. Without biofuel
policies, traditional cropland decreases between 2010 and 2030 by 9 Mha, to the benefit of energy
plantations for solid biomass (5 Mha) and afforestation (7 Mha), which also expand into other natural
vegetation. Around 4 Mha are additionally abandoned over the period in regions where cropland
reduction is not followed by any other uses. Abandoned land can also occur in some regions other
than the EU, but does not expand between 2010 and 2030. On the contrary, past abandoned land is
observed to decrease in Eastern Asia and in South Asia due to the demand increase for agricultural
products.

19
It should be noted that outside of Europe, no afforestation policies are implemented in these scenarios. Therefore, some regions, like
Eastern Asia, which is notoriously characterised by a trend of reforestation, do not show this pattern in our projections.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 24

Figure 10: Land use change baseline projections between 2010 and 2030. Source: GLOBIOM

2.2.8 Greenhouse gas emissions from agriculture and forestry

Greenhouse gas emissions from different sources in agricultural and forestry sectors (including crop
cultivation and livestock) increase continuously along the baseline. Agricultural emissions in
GLOBIOM increase from 3,710 MtCO2 to 4,440 MtCO2 between 2010 and 2030, due to the increase in
livestock population, the increased use of fertiliser and the expansion of rice cultivation.20 This
corresponds to an increase of 20% over the 20 year period, to be compared with the historical
growth rate from FAO of 13% between 2000 and 2010. Absolute levels for the year 2010 differ
between FAO and GLOBIOM due to differences in the GHG sources accounted for and the emission
factors used,21 as well as different production levels related to the food prices impact (see Section
2.1.3).

20
Note that there are uncertainties on emission factors for each sources and different approaches are found in the literature. FAO accounted
for the same sources and production levels as GLOBIOM 1,170 MtCO2-eq in 2000 versus 1,110 MtCO2-eq.
21
See Herrero et al. (2013) for differences in accounting on the livestock sector and Valin et al. (2013) for the crop sector.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 25

Figure 11: Greenhouse gas emissions from agriculture and forestry in the baseline. Source: GLOBIOM

2.2.9 Crop price

Prices in GLOBIOM develop in the baseline under the pressure of change in exchange rates, increases
in oil prices, and changes in demand for food, feed and biofuels. Due to changes in these drivers in
the period 2000-2010, initial prices in GLOBIOM, calibrated to average historical values of these
drivers in 2000, are shifted up when projecting towards 2030. Table 4 shows how GLOBIOM
projected prices compare to recently observed ones. Crop prices all reproduce an increase between
2000 and 2010 in line with recent trends for cereals, sugar cane and oilseeds. In the case of corn,
the model projection is higher than in historical record because the US biofuel shock is fully applied
in the baseline in the year 2010 (decadal time step). Sugar beet prices are higher in the model than
in the historical record, because some sources of production costs decrease, steaming from change
in farm structure, are not captured in the model. After 2010, because oil prices and exchange rates
are considered to stabilise, all prices follow a slight downward trend until 2030, supported by
productivity increases.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 26

Table 4: Main crop prices in the baseline. Source: GLOBIOM

Hist. average GLOBIOM

World prices Calibration
2009-2011 2010 2020 2030
2000
(€/wet tonne)
Maize (EU) 164 155 232 211 197
Wheat (EU) 160 103 152 141 133
Barley (EU) 143 93 150 126 109
Sugar beet (EU) 32 40 54 56 55
Sugar cane (Brazil) 18 10 14 14 13
Rapeseed (EU) 363 176 398 270 240
Soybeans (Brazil) 281 181 247 233 224
Sunflower (EU) 371 211 295 277 259
Note: Historical producer prices in current EUR (source FAOSTAT, reference France). GLOBIOM prices correspond to average producer prices

in the EU or Brazil, expressed in real terms. Exchange rate fluctuations between EUR and USD since 2000 are accounted for.

2.2.10 Livestock prices

Livestock prices are primarily influenced by the price of feed and, for internationally traded products,
by exchange rate. These determinants are represented in our model and fluctuations of livestock
product prices reflects the variation in crop prices. For ruminant products, the increase in the price of
grain and oilseeds is only partially reflected in product prices, due to the presence of other costs for
feeding and animal management. For monogastric products however, the link to crop prices is more
direct (main cost source) and the model reproduces the upward trend in prices observed in the
historical period between the calibration prices from 2000, and the calculated price from 2010. From
2010 to 2030, the model then projects a stabilisation and then slight decline of prices until 2030.

Table 5: Main livestock product prices in the EU in the baseline. Source: GLOBIOM

Hist. average GLOBIOM

Calibration
World prices 2009-2011 2010 2020 2030
2000
(€/tonne carcass weight)
Bovine meat 3491 2481 3331 3313 3295
Sheep and goat meat 5187 4544 5757 5418 5006
Milk 321 258 358 357 356
Pig meat 1258 1178 1590 1445 1381
Poultry meat 1434 1268 1667 1575 1507
Poultry eggs 1036 900 1218 1189 1165

Note: Historical producer prices in current EUR (source FAOSTAT, reference France). GLOBIOM prices corresponds to average producer

prices in the EU or Brazil, expressed in real terms. Exchange rate fluctuations between EUR and USD since 2000 are accounted for.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 27

2.2.11 Biofuel prices

Prices of biofuels in the EU28 are strongly correlated with feedstock prices. Ethanol price is calculated
in the model as around EUR 0.50/l over the period 2010–2030. Biodiesel prices have been strongly
fluctuating over the past decade due to the high price volatility of vegetable oil in the period 2000-
2010, which leads to a spike in the price of biodiesel that later stabilises around a value of EUR
0.80/l.

Table 6: EU28 average biofuel prices in the baseline. Source: GLOBIOM

Historical
Calibration
EU28 biofuel prices prices 2009- 2010 2020 2030
2000
2011
(€/liter)
Ethanol 0.45-0.70a – b
0.54 0.53 0.51
c
Biodiesel 0.55-1.10 0.49 1.13 0.89 0.8
a
Ethanol prics varied within a 0.45-0.70 EUR/l range over the period, with average yearly value of 0.5 EUR/l in 2009, 0.55 EUR/l in 2010,

0.65 EUR/l in 2011 (source: Platts, Ethanol T2 Rotterdam).

b
No bioethanol consumption for EU28 in 2000 in GLOBIOM.
c
Biodiesel producer price significantly varied over the period within a 0.55-1.10 EUR/liter range. Yearly average were observed around 0.6

EUR/l in 2009 and 1 EUR/liter in 2011 (source: UFOP).

Note: Historical producer prices in current EUR (source FAOSTAT, reference France). GLOBIOM prices corresponds to average producer

prices in the EU or Brazil, expressed in real terms. Exchange rate fluctuations between EUR and USD since 2000 are accounted for.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 28

3 Description of scenarios and sensitivity analysis

3.1 Introduction

Following the construction of the modelling baseline as described in the previous chapter, this
chapter describes the various scenarios that are modelled against the baseline. A large number of
scenarios are modelled. Firstly, a series of crop-specific scenarios for the main conventional and
advanced biofuel crops as well as the deployment of separate cereal, starch and oilseed crop groups
are modelled. Also, aggregated scenarios of 9.4% EU biofuel consumption following the National
Renewable Energy Action Plans (NREAPs)22 (8.6% conventional plus 0.8% advanced biofuels), as
well as a scenario that includes a maximum cap on the consumption of conventional biofuels of 7% ,
are modelled. In the latter scenario, a total of 9.4% biofuels is modelled of which 7% consists of
conventional biofuels and the rest of advanced biofuels, taking account of the EU RED double
counting provision.23 In addition to this, some explorative scenarios are modelled: increased use of
abandoned land in the EU, lower than expected worldwide deforestation plus a ban on peatland
drainage, and higher than expected worldwide deforestation.

In the baseline, we assume a biofuel consumption of 3.2%, which equals the consumption level in
2008. Feedstock-specific scenarios are compared with the baseline by modelling separately for each
feedstock an increased consumption of 1% biofuels as a share of total road transport fuels, or 3
Mtoe. The NREAP scenarios and explorative scenarios are modelled by applying a ‘shock’ of 6.2%
additional biofuel consumption as compared to the baseline biofuel volume of 3.2%. An overview of
the various scenarios as compared to the baseline is provided in Figure 12 below.

22
As submitted by Member States to the European Commission in 2010-12. http://ec.europa.eu/energy/en/topics/renewable-
energy/national-action-plans
23
EU RED Article 21(2) states that biofuels produced from wastes, residues, lignocellulose and non-food cellulose material count twice
towards national targets for renewable energy in transport.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 29

Figure 12: Scenario setting for the modelling of biofuel policies between 2000 and 2030. Two scenario types are

considered: feedstock specific shocks (+1%) and policy shocks (full mandate). Plain dots indicate years for which

the model generates results

3.2 Crop-specific scenarios

A total of 14 feedstock-specific scenarios are modelled in which a shock of 1% biofuel consumption

from each feedstock (123 PJ) as part of total road transport fuels in the EU in 2020is compared to
the baseline. The feedstock-specific scenarios are modelled for the following biofuel feedstocks:

Table 7: Overview of feedstock-specific scenarios

Conventional biofuels Advanced biofuels

Wheat ethanol Miscanthus biodiesel

Maize ethanol Short rotation plantation biodiesel

Barley ethanol Forest residue biodiesel

Sugarbeet ethanol Straw ethanol

Sugarcane ethanol

Silage maize biogas

Sunflower oil biodiesel

Palm oil biodiesel

Rapeseed oil biodiesel

Soybean oil biodiesel

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 30

For straw ethanol, an alternative approach is used because, due to relatively high transport costs, we
do not assume an EU-wide market for straw, and straw trade with countries outside the EU is
negligible. The modelling of straw ethanol takes into account this fragmented market situation. Straw
removal potential is assessed in three regions with different straw availability: Hungary, Great Britain
(excluding Northern Ireland) and Northern France around Paris. The 1% shock is applied for these
regions24 and results are subsequently aggregated at EU level. For all other regions in the world,
level of biofuel demand is kept constant. Therefore, no change in biofuel consumption level can serve
as a buffer to divert more biofuel to the EU market. The approach taken for straw is described in
more detail in Section II.1 of Annex II.

3.3 EU 2020 biofuel mix scenario without and with 7%

EU 2020 biofuel mix scenario assumes that the 10% target on renewable energy in transport is
fulfilled with 9.4% biofuels (before double counting) following the National Renewable Energy Action
Plans (NREAPs) that were submitted to the European Commission by EU Member States in 2010–11.
While it is generally recognised that many of the NREAPs are outdated, no other official projections
on biofuel consumption in 2020 for each EU Member State is available.

The NREAPs provide an overall forecast on the level of biofuel consumption in 2020 and a split
between conventional and advanced biofuels. According to article 21(2) of the EU directive on
Renewable Energy Sources, biofuels produced from wastes, residues and25 cellulosic material count
twice towards national targets. This lowers the overall quantity of biofuels required to meet the
target. The NREAPs assume a very limited uptake of advanced biofuels, including UCOME (biodiesel
from used cooking oil), TME (biodiesel from animal fats) and other double counting biofuels of 0.8%.
This means that the projected 9.4 % biofuels in the NREAPs represent an actual food crop based
biofuel consumption in volume of 8.6% of EU transport fuels.

The NREAPs do not provide an estimated split in biofuel feedstocks used. In fact, it is difficult to
obtain a reliable picture of the EU biofuel feedstock mix, since the biofuel industry generally does not
share information on their feedstock mix and most Member States (except the UK, Germany and the
Netherlands) do not publish the feedstock mix of consumed biofuels. The consortium invited the
industry to provide this information, but in the end had to rely on estimates by EU FAS posts (USDA
2014). More transparency on this would certainly help to improve the estimate of land use change
emissions of the total EU biofuel mix in 2020 and beyond. This study bases the assumed feedstock
mix on USDA estimates for 2013 and keeps this constant up to 2030. The shares and mix of
advanced biofuel feedstocks, are determined endogenously by the model based on least cost
optimisation. Based on the above, the following EU biofuel consumption level and and feedstock mix
are assumed:

24
In the case of Central France, the 1% shock is applied to the entire country of France. This has little impact on modelling results as
abundant straw is only available in Central France. This is further explained in Section II.1 in Annex II.
25
The estimates in this USDA report is collected by USDA Foreign Service Officers stationed in EU Member States (EU FAS posts). The
method of information collection is not known, but we assume that it is based on public information, such as press releases, magazines,
combined with interviews.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 31

 10%

9%
Grassy crops FT Diesel
8%
Short rotation coppice FT
Diesel
7% Forest residues FT Diesel
Incorporation level

6% Palm oil FAME

5% Soya FAME

Sunflower FAME
4%
Rape FAME
3%
Sugar Cane Ethanol
2%
Sugar Beet Ethanol
1% Barley Ethanol

0% Corn Ethanol
Baseline EU 2020 Mix EU 2020 Mix
+ Cap 7%

Figure 13: Feedstock composition in the baseline and EU 2020 biofuel mix scenarios

The default EU2020 biofuel mix scenario without constraints placed on the consumption of
conventional biofuels is characterised by a more than marginal share of palm oil in 2020 (16% of
total biofuel mix), used both for biodiesel (FAME) and drop-in renewable diesel (Hydrotreated
Vegetable Oil or HVO), as can also be seen in Figure 14 below. Indeed, we assume that one third of
additional vegetable oil used in the mandate comes from palm oil, based on USDA observation on
recent change in composition mix. The rather substantial share of palm oil found by USDA is rougly
equal to the quantity found in a biofuel sample analysis study performed by UFOP in Germany, which
estimated that around 14% palm oil was used in German biodiesel consumption in 2013 or around
12% in the total biofuel mix.26 However, as stated above, no better data on the EU-wide feedstock
mix are available than the USDA data.

26 Union zur Förderung von Oel- und Proteinpflanzen e.V, Rohstoffbasis der Biodieselanteile in Dieselkraftstoffen (2014). This study
estimated that, based on samples taken at fuel stationsin 2013, around 14% of palm oil was used in biodiesel in Germany. German
government agency BLE however reports that 26.316TJ of palm oil was used for biofuels consumed in Germany in 2013, which equals 21%
of total German biodiesel consumption. .

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 32

Figure 14: Feedstock composition as share of energy consumed in the EU 2020 biofuel mix (baseline + shock)

We also consider one alternative policy scenario corresponding to the political agreement on ILUC
between the Council and the European Parliament reached in April 2014 modifying biofuel
incorporationby limiting the contribution of conventional biofuels to 7%.

The introduction of maximum levels (cap of 7%) of incorporation of conventional biofuels modifies
the fuel mixes initially prescribed by NREAPs, potentially increasing the share of advanced biofuels to
some extent. It is likely that the introduction of new - and the increase of existing- multipliers when
counting the use of renewable electricity in road and rail may reduce significantly or even remove the
1.6 % gap between the “cap” and the NREAP scenario for conventional biofuels. The 7% cap is
introduced in each Member State. As some Member States did not plan in their NREAP to reach 7%
from conventional biofuels (e.g. Belgium, the Netherlands, Denmark, etc.), the aggregated share of
conventional biofuels equals 6.7% at the EU27 level under this scenario, instead of 8.6%. Because
advanced biofuels and renewable electricity are subject to multiple counting, the total level of biofuel
consumption in real energy terms decreases in the 7% cap scenario, as the incorporation limit
becomes more restrictive for conventional biofuels: It drops from 9.4% to 8.4% (figure 13).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 33

3.4 Explorative scenarios: abandoned land and lower or higher
deforestation

In order to better understand the role of some of the contextual elements on the results of the
assessment, some explorative scenarios are now presented that complement the previous
calculations. These scenarios look at some particular elements for which uncertainty is high and that
can influence the response of agricultural production and land use emissions in response to a biofuel
policy incentive: i) restoration of agriculture on abandoned land in the European continent ii)
different developments of deforestation policies and their enforcement level.

To test to what extent abandoned land in the European continent can help buffer some impacts of
biofuel deployment, we construct the C1 scenario on abandoned land restoration. This scenario is
implemented in the context of the ‘EU 2020 biofuel mix scenario’, because restoration of abandoned
land with free entry of low price feedstocks on the EU market leads to very limited effect on
abandoned land in the model: imports of feedstocks remain in that case the preferred option, even
with some form of support on EU feedstocks. The scenario tests a dedicated incentive in EU biofuels
policy for increased use of abandoned land in the EU. We represent this scenario by a combination of
abandoned land restoration with restrictions on feedstock use: biofuel based on palm oil, soybean oil
and sugar cane can no longer be used in the feedstock mix beyond the baseline levels, which
increases the reliance on feedstocks produced in the EU. This restriction is complemented by a
decrease in conversion costs to other natural land and a subsidy on conversion of new land with low
carbon stocks. This scenario is then compared to the baseline (A0).

To study the impact of deforestation context, we follow a quite different approach. The cost of
conversion of forest into agricultural land are decreased (High deforestation scenario) or increased
(Low deforestation scenario, Very low deforestation scenario), with all scenarios being based on the
central ‘EU 2020 biofuel mix scenario’ including the same assumed feedstock mix. We consider
additionally for the case of low deforestation two levels of carbon prices as an incentive against
deforestation: USD 10 /tCO2 (Low deforestation) and USD 50/tCO2 (Very low deforestation). As a
consequence, land expansion in these scenarios is more likely to occur into other natural vegetation
rather than into forest. The deforestation rates associated with these modified baselines are
displayed in the table below. Note that these rates corresponds only to the net deforestation, ie.
forest natural regrowth is accounted for in the tropics, and represents only the share attributed in
GLOBIOM to agricultural drivers.

Table 8: Worldwide deforestation patterns in different baselines and shocks depending on deforestation context

Region A0 (baseline) A0 + High A0 + Low A0 + Very low

deforestation deforestation deforestation
Baseline 2010- -62,500 -88,900 -10,200 -1,400
2030 (kha)
Impact of the -1,132 -1,920 -1,210 -170
NREAP shock by
2020 (kha)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 34

We additionally modelled the Very low deforestation scenario with a ban on peatland drainage in
Indonesia and Malaysia, both for current and new concessions for oil palm plantations expansion.

3.5 Sensitivity analysis

In order to test how models depend on the different assumptions in the modelling, it is useful to
explore further simulation results where the model assumptions are varied. Such extensive sets of
simulations are called sensitivity analyses. Different sensitivity analysis techniques can be used. For
instance, it is possible to vary each crucial parameter around its central value, one after the other, to
test first order effect responses. Another, slightly more resource consuming, approach, consists of
approximating each of the parameter distributions by a Gaussian curve and running a few points of
the Gaussian. We followed here an even more comprehensive approach, called Monte-Carlo analysis.
For this approach, a large number of initial simulations are run repeatedly with randomly varied
parameters. In the present case, 300 runs have been performed for each of the feedstock specific
scenarios A, A1, A2 and the EU 2020 biofuel mix scenario. To perform this analysis, 11 parameters
were varied along the specifications reported in Table 9.

The first set of parameters to be varied relate to the modelled behavioral responses. These
responses or parameters depend on how the model functions. Elasticities were varied for demand
response, trade response, expansion reponse, vegetable oil substitution and impact of the biofuel
policy on the feedstock yield. These elasticities determine how much land use change occurs and in
what regions.

A second set of parameters concerns biophysical characteristics. These parameters are direct model
inputs on some resource or product properties. Co-product protein content is the first important one,
as it determines the extent of substitution of co-products with other oilseed meals. Additional testing
was applied on the impact of removing yield residues on yield and soil organic carbon. Degree of
water availability to expand irrigated systems was also varied. Finally, the emission factors for peat
land, as well as the share of (palm oil) plantation expanding into peat land, were varied for Indonesia
and Malaysia.

In the Monte Carlo analysis, the chosen parameters are randomly varied, but this still involves a pre-
defined distribution shape27. Some parameters are varied between -50% and +100%. For
parameters that are known with more accuracy, the range and shape of variation is pre-set in line
with the data used in the modelling, as described in Annex IV and in the model improvement
descriptions provided in Annex II.

27
Most values are varied along a loguniform distribution, because the central value is not necessarily more plausible than other points in the
distribution. Biophysical parameters were varied along different distribution shapes, either uniform when no better information was known,
or along the distribution determined in the Improvement document (Annex II).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 35

Table 9: Parameter variation used for the Monte-Carlo analysis

Parameter Value range Motivation for parameter selection

Minimum Maximum
Behavioral parameters
Demand elasticity - 33% +50% Determines the degree of food consumption
adjustment
Trade elasticity -50% +100% Determines trade response patterns
Vegetable oil -50% +100% Determines the degree of substitution between
substitution elasticity different vegetable oils
Land expansion -50% +100% Determines ease of expansion into the different
elasticity land use types
Yield response on Elasticity Elasticity Determines the degree of adjustment of yield to
feedstock model – 0.05 model + 0.2 prices
Expansion response of 12% 54% Determines the degree of expansion of palm
palm into peat land plantation into peatland in Indonesia and
Malaysia
Biophysical
parameters
Co-product protein -10% +10% Determines the degree of substitution of co-
content products
Soil carbon impact -10% 0% Determines the impact of straw removal beyond
straw sustainable levels on soil organic carbon
Yield impact straw -4% 0% Determines the impact of straw removal beyond
(mean value -2%) sustainable levels on soil organic carbon
Peat land emissions 27 tCO2 ha-1 113 tCO2 ha-1 Determines the level of peatland emissions in
factor yr-1 yr-1 Indonesia and Malaysia
Water availability -50% +100% Determines the possibility of intensification
through more irrigation

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 36

4 Modelling results

4.1 Summary of modelling results

This section presents the modelling results of all 27 modelled scenarios, both in terms of land use
change area effect and the resulting greenhouse gas impact.

The results of this study, commonly referred to as ‘ILUC values’, are in fact a mix of direct and
indirect emission effects, in the sense of traditional life cycle analysis. The modelling does not show
to what extent the land conversion is caused directly or indirectly. For this reason, this study speaks
about ‘LUC values’ rather than ‘ILUC values’ and about land use change rather than direct or indirect
land use change. Indeed, when looking at the impact of a policy against a counterfactual, the notion
of direct and indirect effect only depends on the subjective choice of where the boundaries are set on
the system of analysis. In life cycle analysis, the systems analysed are usually limited to the sites of
production, transformation and consumption of the products, and GHG emission accounting is limited
to sources identified within these boundaries. In a global economic model, no geographical
boundaries apply, because all emissions from countries around the globe are simultaneously
accounted for. From this perspective, our approach can be considered comprehensive. However,
some limitations still apply in the scope of the analysis along two dimensions: GHG emission
accounts, and temporal horizon. Regarding GHG emission accounting, it is important to note that the
term ‘land use change emissions’ covers a certain number of emissions sources that go beyond the
most usual source of living biomass carbon emissions. In our final results are also included some
other stocks of carbon, such as those in agricultural biomass and in mineral or organic soils.
Agricultural biomass carbon sequestration is deducted from the land use change related emissions,
to obtain the LUC values. For each modelled scenario, we provide a precise breakdown of the result
into various contributing factors. Soil organic carbon emissions from crop management can also be
indirectly related to feedstock cultivation. Accounting for these emissions leads to higher emissions
for expansion of new cropland in Europe, as well as more foregone sequestration as discussed below.
A more comprehensive presentation of system boundaries used for this study and a comparison with
the MIRAGE model used by IFPRI is provided in Box 2.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 37

Box 2: GHG emission accounting scheme in this study and system boundaries

Five categories of emissions are accounted for in this report, which correspond to a subset of GHG
emission sources directly associated to the Agriculture, Forestry and Land Use Change sector
(AFOLU) of the IPCC guidelines. The selection of these sources was discussed and agreed with the
European Commission’s steering committee to be in line, and better comparable with the past
literature on ILUC.

GHG fluxes from these sources are related to the following carbon pools:
 Carbon from living biomass above and below ground
 Carbon in dead wood and litter in forest
 Organic carbon in mineral soil
 Carbon in organic soils

These fluxes are analysed in this report according to five categories:

1 “Natural vegetation conversion”: fluxes associated with the emission or sequestration of carbon
in living biomass, dead wood and litter when land use is converted (can be positive in case of
deforestation or negative if managed grassland expands into fallow land);
2 “Natural vegetation reversion”: fluxes associated to the natural regrowth of living biomass on
previously abandoned land. This source is accounted for separately to factor in the uncertainty
of C growth rate and the importance of timing;
3 “Agricultural biomass”: fluxes associated to the sequestration in agricultural living biomass
(cropland) when agricultural activities expand. This flux can be negative (sequestration) if palm
tree expand into grassland, or positive (emissions) if sugar cane is replaced by soybeans.
4 “Soil organic carbon”: this source accounts for all change in soil organic carbon stocks from
agricultural land, forest and other natural vegetation. The methodology for such accounting is
explained in more details in Appendix II.5;
5 “Peatland oxidation” corresponds to release of mineral carbon in organic soil associated with
drainage of peatland. In this study, these emissions are only accounted for in Indonesia and
Malaysia. They are subject to significant uncertainties and studied in more details in Appendices
II.3 and II.4.

Our accounting is performed as follows: categories 1, 2 and 3 are accounted for based on emissions
calculated by the model on the basis of land cover changes for the period of the simulation. Carbon
stock variation is then divided by 20 years to obtain annual emission flows. Categories 4 and 5 are
directly accounted for as annual emission flows, on the basis of difference between drained peatland
areas in the scenario and in the baseline, or between cultivated areas under different tillage for soil
organic carbon (IPCC tier 1 assumes release of soil organic carbon on a 20 year period).

We do not account in this report for sources other than those listed above due to the focus on land.
In particular, non-CO2 GHG emission associated to agriculture or to LUC are not reported (for
instance emissions from fertiliser related to intensification of agricultural production). A part of them
are covered in the direct accounting of the EU renewable energy directive.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 38

The LUC values provided in this chapter are single, central emission value estimates which should be
considered with greatest prudence when they are considered separately. The central emission values
are complemented with a much more comprehensive distribution of results that stem from the
sensitivity analysis on model parameters, as explained in Section 3.5 and Annex V.

Figure 15, below, shows the LUC emission values for each of the modelled scenarios and their break-
down between various emission sources. The part of each bar above zero on the y-axis represents a
quantity of emissions, while the part of the bar below zero represents negative emissions that are
deducted from the emissions. The resulting net LUC-emission value is represented by the small
triangle in each bar and by the number on top of each bar. Positive emission sources are commonly
peatland oxidation (by peatland drainage), soil organic carbon emissions (carbon stored in soils),
forest reversion (foregone sequestration, see below) and natural vegetation conversion (removal of
above and below ground living biomass in converted land). Negative emission sources are commonly
agricultural biomass (carbon stored directly in cultivated feedstocks) and soil organic carbon
sequestered during feedstock cultivation.

Figure 15: Overview of modelling results: LUC emissions per scenario. Source: GLOBIOM

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 39

4.1.1 Results for crop-specific scenarios

 Conventional biodiesel feedstocks have high LUC effects compared to the direct emissions
resulting from the biofuel production process, with very high emissions for palm oil (231
gCO2e/MJ biofuel), high emissions for soybean oil (150 gCO2e/MJ) and 63 and 65 gCO2e/MJ for
sunflower and rapeseed oil respectively. For all oil crops and in particular for palm oil, peatland
conversion is a large contributor to LUC emissions. Due to substitution effects, the use of other
oil crops also leads to some conversion of peatland and related LUC emissions. However, the
more limited substitution assumed among vegetable oils leads to higher differentiation of results
for the different oil types;
 Even with a moderate substitution of palm oil in the model on the demand side, the link to oil
markets operates by replacement of soybean meals. Less soya is then cultivated and the
missing soya oil is then replaced by cheaper palm oil. A slight increase in plantation acreage
immediately leverages considerable quantities of peatland emissions under a mid-range
assumption on peat expansion patterns and emission factors; this emphasises the ambiguous
effect of DDGS28 return on the protein meal market, which can on one side save land for
soybean producer countries, but at the same time can generate additional emissions in other
parts of the world and possibly degrade the environmental balance further;
 Conventional ethanol feedstocks, sugar and starch, have much lower LUC emission impacts,
with 14, 34 and 38 gCO2e/MJ biofuel for maize, wheat and barley, 17 gCO2e/MJ for sugarcane
and 15 gCO2e/MJ for sugarbeet;
 The LUC value for maize is lower than for wheat or barley due to its higher average yield and
higher protein substitution of wheat co-products with other protein sources, leading to some
small oil palm plantation expansion. Additionally, maize transformed into ethanol performs
better than maize silage transformed into biogas, partly through the coproduct effect and
agricultural biomass contribution; however, the ambiguous effect of coproducts leads to a larger
spread of results as for maize silage;
 Advanced biofuel crop production leads to low, or even negative, LUC emissions, because
emissions are compensated by carbon credits related to carbon sequestered in the new land
covers, or in carbon sequestered in soil resulting from no-till practices for feedstocks. This is,
however, under the condition that feedstocks do not expand into carbon-dense areas, as occurs
for some sensitivity analysis cases for sugar cane. Additionally, because the foregone
sequestration accounting is aligned on the biofuel reference period of 20 years, we do not
consider the full regrowth of forest as a counterfactual, which masks significant opportunity
costs for long term sequestration. Forestry residues have relatively high soil organic carbon
emissions of 17 gCO2e/MJ biofuel, but no LUC emissions. The LUC value for straw is 16
gCO2e/MJ with unsustainable straw removal and zero gCO2e/MJ when not more than 33-50% of
straw is removed;
 In general, crops with higher energy yield per hectare (corn, sugar cane, sugar beet) have lower
indirect impacts on LUC and GHG emissions. However, a notable exception is palm oil, a high
yielding crop whose performance is strongly impacted by GHG emissions from deforestation and,
in particular, peatland conversion;

28
Distiller's dried grains with solubles.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 40

  In Europe, a trend exists of cropland being abandoned, with abandoned land partly turning into
forests and partly into grasslands. An increase of cultivation of biofuel crops in Europe leads to a
slower pace of land abandonment. This has a carbon impact, because it implies that carbon
sequestration through natural vegetation and young forest regrowth does not take place. This
“foregone sequestration” on abandoned land in the EU contributes LUC emissions. It should be
noted however that some strong uncertainties exist on the extent to which this effect occurs in
reality, as discussed in the Executive Summary.
 Modelling results are distributed over a 20 year period. Most LUC emissions take place shortly
after the conversion of previously non-agricultural land to agricultural land and it makes little
sense to allocate all emissions to the first year after the conversion and to have much lower LUC
emissions in year two. If a longer allocation period were chosen, for example 30 or 50 years,
LUC emission values would be lower for some sources, since the total land use change emissions
associated with a certain quantity of biofuels would be divided over larger number of years.
However, annual flows from peatland and future foregone sequestration would not be reduced
before 50 to 100 years (time for peat to be fully oxidised or forest to be fully regrown). Given
the significant contribution from continued peatland oxidation, the LUC emissions from 50 year
perspective is overall significant higher than from the 20 year perspective. If total LUC emissions
would be amortised over 50 instead of 20 years, annual emissions would amount to 79
gCO2e/MJ in the EU 2020 biofuel mix scenario. More background is provided in Annex III.

4.1.2 Results for aggregated policy and explorative scenarios

 The EU 2020 biofuel mix scenario assumes that the 10% target on renewable energy in
transport is fulfilled with 9.4% biofuels, of which 8.6% are conventional biofuels. The assumed
increase in each feedstock (see also Section 3.3) leads to a total LUC of 8.8 Mha, of which 8 Mha
are devoted to new crop and perennial cultivation and the remaining 0.8Mha consists of short
rotation plantations on existing cropland. From the 8.8 Mha, 2.9 Mha of conversion takes place
in Europe through less land abandonement and 2.1 Mha takes place in Southeast Asia under
pressure from oil palm plantations, the latter for 50% at the expense of tropical forest and
peatland. 8.8 Mha is equivalent to 0.6% of the total global crop area in 2012 of 1,395 Mha
(FAO). The 8 Mha new cropland for biofuels to be used in the EU compares to a total 64 Mha of
new cropland for all uses of agricultural biomass due to overall increased demand during the
same time span, i.e. 2008-2020.29 The resulting LUC emissions mainly relate to deforestation
and peat land conversion for palm plantations, which provide vegetable oils for direct
replacement of volume consumed by biofuels and compensate for the soybean production
decrease due to co-products substitution. The LUC emission value of the scenario is 97
gCO2e/MJ;
 The introduction of a maximum percentage (‘cap’) for conventional biofuel consumption in the
EU could significantly lower the LUC emissions. A cap of 7% results in an LUC value of 74
gCO2e/MJ biofuel, with the same assumed feedstock mix for conventional biofuels as in the EU
2020 biofuel mix scenario. The decrease is mainly caused by an increase in the share of
advanced biofuels, which have lower or even negative LUC emissions. To assess the overall ILUC
mitigation by the ILUC Directive agreed upon in 2015, one has to add the fact that not only the

29
Extrapolated from the 53 Mha projected by the model on 2010-2020.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 41

 LUC intensity of the biofuels used in the EU declines but that also less biofuels are needed in
quantitive terms to achieve the 10% target for renewable energy in transport (see section 3.3
and figure 13);
 Abandoned land can provide acreage for biofuel expansion in the EU. This is the case already in
the ‘EU 2020 biofuel mix’ scenario, especially in the explorative ‘Abandoned land in the EU’
scenario, which considers a larger use of abandoned land in the EU for biofuel feedstock
cultivation. This scenario reduces LUC emissions from 97 gCO2e/MJ biofuel to 52 gCO2e/MJ. This
reduction is caused both by increased use of abandoned land, and a restriction on feedstock
use: biofuel based on palm oil, soybean oil and sugar cane can no longer be used in the
feedstock mix beyond the baseline levels, which increases the reliance on feedstocks produced
in the EU. However, 52 gCO2e/MJ remains large compared to emissions from the biofuel
production process. Partly this is due to the fact that greater use of land in the EU still removes
carbon stock from natural vegetation – present or future in case of abandoned land – and soil, in
part because the substitution of co-products with other protein meals still lead to some
displacement of land in other parts of the world. It should be noted that, in reality, land is often
not completely unused but rather underused, used extensively by local smallholders who cut
grass for their animals, or the land is not used for crop cultivation but ploughed once every year
in order to obtain CAP subsidies. This means that in many cases, abandoned land will not refer
to forest or grassland and can be used for crop production with limited emission effects. The use
of abandoned land can be an effective LUC mitigation strategy in cases where it is currently
‘underused’, for example as grassland with relatively low carbon stocks where reversion to
forest is being prevented. Our abandoned land scenario focuses on improved access to
abandoned land in Europe. This means we take a conservative approach on the potential
positive effect of restoring abandoned land for agriculture expansion;
 A serious effort to limit deforestation and expansion into peat land leads to a significant lowering
of LUC effects. The ‘Very low deforestation’ scenario leads to LUC emissions of the EU 2020
biofuel mix of 47 gCO2e/MJ biofuel instead of 97 gCO2e/MJ. On the other hand, in a high
deforestation scenario, biofuels induced expansion into forest and peat land also increases. The
LUC impact therefore increases alongsideincreasing from 97 gCO2e/MJ to 110 gCO2e/MJ, an
increase of 15%. The intermediate ‘low deforestation’ scenario leads to a LUC value of 90
gCO2e/MJ. As explained in Section 3.4, the low deforestation scenario assumes a carbon price of
USD 10/t as an incentive to stop deforestation. This incentive is shown by the modelling to be
insufficient to reduce deforestation rates drastically, mainy because commodities such as palm
oil are so profitable that converting the land remains economically beneficial. (See also: Deiniger
et al. in 2010). The USD 50 incentive in the very low deforestation scenario is much more
efficient. LUC emissions of the EU 2020 biofuel mix can be reduced dramatically to just 5
gCO2e/MJif the USD 50 incentive in the very low deforestation scenario is combined with an
effective ban on peatland drainage.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 42

4.1.3 Results comparison on distribution of response to the shock

Figure 16: Overview of distribution of effects between demand side and supply side adjustments for all crops

modelled in GLOBIOM (aggregated by ton dry matter) and for all scenarios

Figure 16 provides an overview of how the agricultural system reacts to the shocks from the different
scenarios. The demand side reacts mostly on feed, because food demand is more inelastic than feed
demand, especially for cereals. In the case of oilseeds, feed response is lower and sometimes even
negative (soybean), because the yield of protein meals per unit of fuel is stronger than the yield of
DDGS for cereals based ethanol and boosts the consumption of other feedstuffs. Co-products
themselves are accounted for in a different category (yellow bars) and are directly related to the
technical coefficients in the crushing or biofuel supply chain. In many cases, yield is found to be an
important contributor. However, some crops do not respond in yield in the model response, in
particular sugar crops, because marginal yields are found to be lower than average yields for these
crops. For cereal straw, yield response corresponds to use of unsustainable removal rates where
straw was previously harvested within sustainable rate limits. Perennials and short rotation crops
mainly provide the extra production though area increase, because only one management type is
considered for these in the model.

As stated above, yield increases induced by additional biofuel demand are significant. How does the
estimated yield on ‘new’ arable land compare with the average yield on the baseline arable area? The
table below shows the marginal yield effect for each of the feedstock specific scenarios.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 43

Table 10. Marginal production increase divided by marginal harvested area increase for the different feedstocks in

each of the feedstock specific scenarios [t fresh matter per ha]. Only regions with production increases are shown.

For perennial crops, newly planted areas are also accounted.

Maiz Maize Sugar Sugar Rape Soyb Sunfl Palm Perennial

Wheat Barley
e silage beet cane seed eans ower oil grasses

Latin America 4.7 5.9 69.5 3.2 2.0

South Asia 3.4 46.6 1.0

North America 6.6 4.7 3.3 4.2 3.1

EU28 4.9 8.2 4.9 44.3 64.1 3.8 2.3 1.9 9.2

East Asia 3.4

Southeast Asia 5.6 2.3 13.1

Russia and neighbouring

countries formerly part of 3.6 2.5 1.5

the USSR

Sub-Saharan Africa 3.9

Oceania 1.4

Middle-East North Africa 2.6 2.5 2.7

Eastern Europe 4.7 2.8 2.4

World 4.9 7.5 4.6 44.3 64.1 67.4 3.2 3.7 1.9 12.8 9.2

When coming to policy and explorative scenarios, all the different effects above balance across
feedstocks and contribute to the mitigation of the shock.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 44

4.1.4 Results comparison on land use change

Figure 17: Overview of land use changes in the model in 2020 in feedstock scenarios (A) and policy and explorative

scenarios (B & C) compared to the baseline

Figure 17 shows the differential amount of land use expansion in each cover type between the
scenario and the baseline in 2020. Cropland reports all crop cultivation increases, but do not account
for short rotation coppices that are singled out as energy plantations. This figure illustrates well the
contrast in land use impact between high yielding crops (sugar beet, sugar cane, maize) and lower
yielding crops (soybean, barley, sunflower, and (imported) rapeseed). It is noteworthy that forests
are little affected for feedstock specific scenarios, with the exception of the palm oil scenario.
Deforestation is, however, more impacted by the large land use changes associated with the policy
and the explorative scenarios, except when some particular assumptions are made to limit its extent
(abandoned land scenario and low deforestation scenarios for example).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 45

Figure 18: Overview of land use converted in different locations by 2020 in all scenarios compared to the baseline.

Numbers on the top of the nar indicate the extent of total land use conversion in Mha

Figure 18 provides more detail about the location of LUC. Contribution of EU abandoned land is
significant in most scenarios, with the exception of palm oil, sugar cane and soybean oil, for which
the feedstocks are not grown in the EU. Among most converted land cover types outside of the EU,
forest in Southeast Asia comes first, followed by other natural land and grassland in Latin America,
other natural land and grassland in Southeast Asia, and only later comes changes in Oceania, Sub-
Saharan Africa, or North America.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 46

4.2 Detailed results by feedstock

The following pages present LUC emission results for the different conventional and advanced biofuel
feedstocks in the form of factsheets. Some pages also look at feedstock group results (sugar crops,
starch, vegetable oils) and policy scenario results (full NREAP, NREAP with a 7% incorporation limit
for conventional biofuels). Each feedstock factsheet first provides information on the energy
efficiency performance of the feedstock, usually a good indicator of the land use requirements. This
efficiency corresponds to the amount of biofuel that can be produced from one hectare of land,
assuming for each feedstock the average yield in the EU, according to the modelled yield projections,
by 2020. The model results are then presented under the form of two indicators: land
requirement/TJ (requirement of cropland and requirement of total agricultural land, i.e. cropland +
grassland) and LUC emissions, which includes both “direct” and “indirect” land use emissions.

The impact of the shock is decomposed in each factsheet through a series of three graphs. The first
graph (left hand-side) looks at the contribution of different channels that can buffer the initial area
requirement for the additional biofuel feedstocks: i) demand change ii) co-product feedback on
supply iii) yield response iv) area response. The first three elements are ‘ILUC dampening effects’,
which limit the ILUC-effect. They are calculated directly on the basis of the model results on how
much demand changes in response to the shock, and how much co-product is generated. The last
effect, area response, leads to increased cultivation of feedstocks leading to changes in LUC and to
GHG emissions (see paragraphs below). This decomposition is conducted for the feedstock demand,
yield and co-product responses for the targetgroup alone (left bar) and for all crops contributions
(right bar). The different elements of the decomposition always sum to 100%.

The second graph (central bar chart) represents the extent of LUC associated with the shock. This
decomposition is obtained by summarising the land that disappears in each world region to be
allocated to new agriculture or forestry uses. The total LUC is reported above the bar; it corresponds
to a gross change, not to the net change (at the global level, land disappearance in one region, e.g.
grassland in Brazil can be compensated by land expansion in another one, e.g. grassland in Europe).
Since several types of land cover can expand in the model (cropland, grassland, plantations,
abandoned land, etc.), the total land conversion should not be confused with cropland or even
agricultural land expansion.

The third graph (right hand-side bar chart) corresponds to the cumulative change in terms of GHG
emissions over the period of the shock in the different land use carbon stocks scrutinised in the
study. Four carbon pools are monitored: i) living biomass in natural vegetation (forest, other natural
vegetation, grassland) – this one is decomposed between emissions from land conversion and
sequestration from vegetation reversion, ii) living biomass in crops and perennials, iii) soil organic
carbon, iv) mineral carbon in organic soils (peat lands). The net cumulative emissions (including
negative flows) are reported on the top of the bar. In addition, foregone sequestration is taken into
account.

The descriptive text below the graphs further details the dynamics at play, firstly at the level of the
feedstock market and subsequently for all feedstock markets and for global LUC.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 47

Explanation of scenario result sheet
Biofuel type

Energy productivity 2020: xx GJ ethanol/ha

Land use impact indicators
Cropland and agricultural land displacement: xx ha/TJ and xx ha/TJ

Land use emissions: xx gCO2/MJ

Land newly Total LUC-related Land use change results

Response to shock
put in GHG emissions with sensitivity analysis
for target feedstock
production (incl. agricultural
and for all crops
grassland biomass)

1) New feedstock requirements and

Explanatory text
decomposition of the response to the shock
Additional demand of 1% ethanol (123 PJ) is produced from…
… 2) Feedstock new land requirement

Additional feedstock production is located in… 3) Adjustments in production and

… demand on global markets

Overall agricultural production is affected by…

4) Global land use change impacts
…

Land expansion requires…

5) Global land use emissions
…

Land use emissions are mainly associated to…

6) Land use change emission factor
…
for a reference period of 20 years

Total land use emissions of 123 PJ additional wheat ethanol are…

…

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 48

4.3 Wheat ethanol

Energy productivity 2020 42 GJ ethanol/ha (equals 2000 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 12.8 ha/TJ and 10.4 ha/TJ

Land use emissions: 34 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) is produced from 16 million tons (Mt) of wheat, with
57.5% of additional production taking place in the EU. This shock leads to a price increase of wheat
of 12% in the EU and 1.8% at the global level.

Adjustments to the shock

The additional feedstock supply is achieved 15% through a decrease in feed and, to a lesser extent,
food demand; 26% by displacement of purpose-grown feed by co-product of ethanol; and 59% by
extra production, with yield increase contributing 10% and area 49%. At the total cropland level, due
in part to the impact on the livestock sector, feed demand declines even further (26%) and only 46%
additional supply is required.

Additional feedstock production is located in the EU (7.2 Mt), North America (2.3 Mt), Latin
America (1.5 Mt), and Russia, Ukraine and rest of Europe30 (1.2 Mt). This new production requires an
acreage of 1.5 million ha (Mha) in the EU, 340 kha in North America, 310 kha in Latin America and
310 kha in Russia, Ukraine and rest of Europe.

30
Rest of Europe designates here European Union neighbours, to the exception of Ukraine and states from the Commonwealth of
Independent States (in particular Russia, Belarus, Moldova…), represented as two separate regions.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 49

Overall agricultural production is affected by the expansion of wheat demand and total demand
for cereals decreases by 3.6 Mt due to higher prices. Demand for protein meals (incl. DDGS)
increases by 2.8 Mt as a result of the extra supply of biofuel co-products on the market.

Land Use Change effect

Land expansion requires conversion of 1.7 Mha of land globally, of which 1.6 Mha becomes new
cropland. In the EU, cropland expands by 1.2 Mha, of which 490 kha is sourced from abandoned land
by 2020 and 750 kha from other natural vegetation. North America and Latin America, extra wheat is
produced on the current cropland, whereas in Ukraine and the rest of Europe, cropland expands at
the expense of other natural vegetation (-100 kha). Oil palm plantation expands globally, because
when DDGS displaces protein meals, it also decreases the production of their vegetable oil co-
products and this triggers an increase in palm oil production.In Indonesia and Malaysia, new
plantations represent 34 kha.

Land use change emissions

Land use emissions are mainly associated to soil organic carbon emissions with 54 MtCO2e. Foregone
carbon sequestration of abandoned land in the EU also increases by 29 MtCO2e following the shock
due to expansion of cropland while additional carbon sequestration in crop biomass decreases
emissions by 15 MtCO2e.

Total land use emissions are 123 PJ. Additional wheat ethanol emissions are found to be 83 MtCO2e.
With an assumed 20 year amortisation this results in an LUC emissions factor of 34 gCO2e/MJ. If no
natural vegetation were to regrow on abandoned agricultural land in absence of biofuels, the total
land use emissions would be lower, at 54 MtCO2e, and the LUC emission factor would be 22
gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 50

4.4 Maize ethanol

Energy productivity 2020: 64 GJ ethanol/ha (equals 3030 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 7.7 ha/TJ and 6.4 ha/TJ

Land use emissions: 14 gCO2/MJj

Additional demand of 1% ethanol (123 PJ) is produced from 14.2 Mt of corn, with 82% of
additional production taking place in Europe. The shock leads to a price increase of 4% in the EU
and 0.4% at the global level.

Adjustments to the shock

The additional feedstock supply is achieved 18% through a decrease in feed, with food demand
hardly impacted, 26% by displacement of feed by co-products of ethanol, and 56% by extra
production, where yield increases account for 11%.

Additional feedstock production is located in Europe (9.6 Mt), and Latin America (1 Mt). This new
production requires acreage of 1.2 Mha in the EU, and 130 kha in Latin America.

Overall agricultural production is affected by maize acreage expansion and grain demand
decreases of 2.8 Mt, while demand for protein meals (including DDGS) increases by 3.1 Mt.

Land use change effect

Land expansion requires 950 kha of additional land globally for cropland, most of it coming from the
EU. In the EU, cropland expands 700 kha into other natural vegetation, whereas 250 kha are sourced
from abandoned land. In Latin America, extra corn production substitutes soybean production, which
is substituted by corn DDGS, and no cropland expansion is necessary.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 51

In North America, production of soybean meal is also decreased and the decreased price of protein
meals leads to more substitution for grain-based production systems, and 110 kha of grassland is
returned to other natural vegetation.

Palm oil production increases to replace displaced soybean oil due to protein meal substitution and
palm plantations expand globally by 10 kha.

Land use change emissions

Land use emissions are mainly associated with soil carbon changes on cropland (26 MtCO2), most of
it taking place in the EU, and emissions from foregone sequestration (14 MtCO2). Carbon
sequestration in agricultural crops decreases emissions by 10 MtCO2.

Total land use emissions of maize ethanol are found to be 35 MtCO2e. With an assumed 20 year
amortisation this results in an LUC emissions factor of 14 gCO2e/MJ. One can note that this emission
factor is much lower than for wheat. This can be explained by several reasons: first, corn yield in the
EU28 is 40% higher per hectare than wheat yield in terms of energy productivity. This is the result of
more heterogeneous yields for wheat across the EU. Additionally, maize DDGS contains less protein
than wheat DDGS, so additional maize implies less soybean substitution (than in the case of wheat)
and hence less decrease in soybean oil production. This in turn leads to less replacement by palm oil
and thus less palm oil expansion. If no natural vegetation were to regrow on abandoned agricultural
land in absence of biofuels, the total land use emissions would be lower at 21 MtCO2e and the LUC
emission factor would be 9 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 52

4.5 Barley ethanol

Energy productivity 2020: 38 GJ ethanol/ha (equals 1770 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 15.0 ha/TJ and 11.5 ha/TJ

Land use emissions: 36 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) is produced from 16 Mt of barley, with 82% of
additional production taking place in the EU. This shock leads to a price increase of barley of 18% in
the EU and 6% at the global level.

Adjustments to the shock

Extra production from barley comes from 7% from demand, mainly feed, 25% from co-products and
69% from production, where yield increase account for 21% and area expansion for 47%.

Additional feedstock production is located in the EU (12.7 Mt), in North America (1 Mt) and in
Ukraine and rest of Europe (0.4 Mt). This new production requires acreage of 2.6 Mha in the EU, 220
kha in North America and 150 kha in Ukraine and rest of Europe.

Overall agricultural production is also affected by decrease in demand for grains of 4.2 Mt
globally and 4.9 Mt considering the EU alone. Demand for protein meals and DDGS increases by 2.7
Mt.

Land use change effect

Land expansion requires 1.9 million ha of additional land, which comes predominantly from
conversion of other natural vegetation and grasslands to cropland. In the EU, cropland expands by
1.5 kha, of which 950 kha are sourced from other natural vegetation, 460 kha from abandoned land
and 130 kha from grassland. Cropland increases in Ukraine and rest of Europe expand by 110 kha

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 53

but global forest area decreases by only 60 kha overall. Oil palm plantations expand as a result of
co-product substitution, with an increase of 40 kha in South East Asia.

Land use change emissions

Land use emissions are mainly associated with emission from soil organic carbon (65 MtCO2e),
emissions from foregone sequestration on abandoned land (27 MtCO2), and peat land emissions (13
MtCO2e). Agricultural crops sequester additional 18 MtCO2e.

Total land use emissions of barley ethanol are found to be 94 MtCO2e. With an assumed 20 year
amortisation this results in an LUC emissions factor of 38 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower, at 67 MtCO2e, and the LUC emission factor would be 27 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 54

4.6 Sugar Beet ethanol

Energy productivity2020: 145 GJ ethanol/ha (equals 6840 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 2.6 ha/TJ and 2.2 ha/TJ

Land use emissions: 15 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) is produced from 58 Mt of sugar beet, with 100% of
additional production taking place in the EU. This shock leads to a price increase of 7.4% at the
European level.

Adjustments to the shock

The additional feedstock is achieved 4% through a decrease in food and feed demand, 19% by
displacement of feed by co-product of ethanol and 77% by extra production, which occurs fully
through expansion. Slightly lower yield in the newly producing land contributes negatively to the
adjustments (-6%).

Additional feedstock production is exclusively located in Europe (55 Mt) and requires acreage of
860 kha.

Overall agricultural production is also affected by the additional sugar beet demand and global
demand for grains. Beside the decrease in sugar crop demand of 3.0 Mt, cereals demand decreases
by 3.2 Mt. Protein meals and DDGS increase by 3.0 Mt, while vegetable oil demand is barely
impacted (-0.1 Mt).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 55

Land use change effect
Land expansion leads to 320 kha of additional cropland globally, which expands mostly into
abandoned land. Cropland expands by 220 kha, of which 200 kha are sourced from abandoned
landand other natural vegetation and 20 kha is sourced from grassland.

Land use change emissions

Land use emissions are mainly associated with soil carbon changes in cropland (26 MtCO2).
Reversion in natural vegetation accounts for 11 MtCO2 and carbon sequestration in agricultural crops
decreases emissions by 6 MtCO2.

Total LUC emissions of sugar beet ethanol are found to be 38 MtCO2e. With an assumed 20 year
amortisation this results in an LUC emissions factor of 15 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower at 27 MtCO2e and the LUC emission factor would be 11 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 56

4.7 Sugar Cane ethanol

Energy productivity 2020: 118 GJ ethanol/ha (equals 5,570 litres/ha)

Cropland and agricultural land displacement: 5.1 ha/TJ and 5.0 ha/TJ

Land use emissions: 17 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) is produced from 69 Mt of sugar cane, taking place
mainly in Latin America. This shock leads to a price increase of 0.8% in Brazil, but has negligable
impact at the global level.

Adjustments to the shock

The additional feedstock is mostly achieved through extra production (97%), with a slight decrease
in yield due to lower suitability of land used by marginal expansion (-14%).

Additional feedstock production is predominantly located in Latin America (63 Mt), which requires
acreage of 900 kha, and 4 Mt in South Asia (90 kha).

Overall agricultural production is also affected by a decrease in sugar crop demand of 0.7 Mt in
Latin America and 0.8 Mt in South Asia. At the global level, sugar crop demand decreases by 2.1 Mt
and grains demand decreases by 500 kt in response to some slight price increases in Brazil and its
trade partners.

Land use change effect

Land expansion totals 0.6 Mha of net land expansion globally. The majority of this total expansion
takes place in Latin America, where cropland expands 320 kha at the expense of other natural
vegetation. In South Asia, where sugar cane yield are much lower, cropland expands by 70 kha, also
at the expense of other natural vegetation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 57

Land use change emissions
Land use emissions are mainly associated with the conversion of other natural vegetation and
primary forests (109 MtCO2). Soil organic carbon stock increase by 31 MtCO2e globally, driven by the
expansion of sugar cane, and 37 MtCO2 are sequestered in biomass.

Total land use emissions of sugar cane ethanol are found to be 43 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 17 g CO2e/MJ. Changing assumptions on the
natural vegetation regrowth on abandoned agricultural land in absence of biofuels does not affect the
final result here.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 58

4.8 Silage Maize biogas

Energy productivity 2020: 123 GJ biogas/ha (EU28 avg., before accounting for co-
products)

Cropland and agricultural land displacement: 4.8 ha/TJ and 5.6 ha/TJ

Land use emissions: 21 gCO2/MJ

Additional demand of 1% biogas (123 PJ) is produced from 41.4 Mt of silage maize (fresh matter),
with 100% of additional production taking place in the EU. This shock leads to a price change of
29.5% at the EU level.

Adjustments to the shock

The additional feedstock is achieved 20% through a decrease in feed demand, 74% by area
expansion and 6% through yield increase.

Additional feedstock production is located in the EU (33.2 million tonne). This new production
requires acreage of 750 kha in Europe.

Overall agricultural production in the EU is also affected by a decrease in demand of 570 kt of

grains, 171 kt of sugar crops and 240 kt of protein meals.

Land use change effect

Land expansion leads to conversion of 0.7 Mha globally, if which 590 kha are needed for additional
cropland and 100 kha for grassland globally. In the EU, cropland expands by 560 kha and grassland
by 70 kha, of which 210 kha are sourced from abandoned land and 420 from other natural
vegetation. Land use in regions outside the EU is hardly affected.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 59

Land use change emissions
Land use emissions are mainly associated with soil organic carbon,with 31 MtCO2e emitted globally
following the shock. Foregone carbon sequestration in the EU also comes as a significant emission
source (16 MtCO2e).

Total land use emissions of silage maize biogas are found to be 52 MtCO2e. With an assumed 20 year
amortisation this results in an LUC emissions factor of 21 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower and the LUC emission factor would be 15 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 60

4.9 Sunflower oil biodiesel

Energy productivity 2020: 24.5 GJ biodiesel/ha (equals 740 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 11.6 ha/TJ and 7.3 ha/TJ

Land use emissions: 63 gCO2/MJ

Additional demand of 1% FAME (123 PJ) is produced from 3.5 Mt of sunflower oil, with 28% of
additional production taking place in the EU. This shock leads to a price increase of 8.1% on
sunflower seeds, and 16.7% on sunflower oil in Europe. At the global level, the price impacts are
5.2% and 8.3% for sunflower seed and oil, respectively.

Adjustments to the shock

The additional feedstock is achieved 12% through a decrease in food and feed demand, 32% by
displacement of feed by sunflower meal and 57% by extra production, of which 52% is from area
expansion and 4% from yield increase.

Additional feedstock production requires 6.1 Mt of sunflower globally, located in the EU (1.7 Mt),
Ukraine and rest of Europe (1.7 Mt), in Russia and its neighbouring countries formerly part of the
USSR (1.3 Mt) and in Latin America (0.9 Mt). This new production requires acreage of 870 kha in the
EU, 660 kha in Ukraine and rest of Europe, 860 kha in Russia and its neighbors and 450 kha in Latin
America.

Overall agricultural production is also affected globally by an increase in consumption of 2.1 Mt of

protein meals and the displacement of 530 kt of vegetable oils on the demand side. Extra availability
of protein meals leads to increased feed consumption. Meat production increases by 130 kt globally
and milk by 120 kt.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 61

Land use change effect
Land expansion leads to 1.5 Mha of additional land conversion globally, mainly for cropland. In the
EU, cropland expands by 625 kha of which 290 kha are sourced from abandoned land and 290 kha
from other natural vegetation. In Ukraine and rest of Europe, cropland expands by 270 kha mainly
into other natural vegetation. Global grassland also decreases by 530 kha as protein meals
availability favors grain-based production systems instead of grass-based ones, in particular in Latin
America (-140 kha). At the same time, palm oil plantation expands by 160 kha in Southeast Asia,
which leads to 50 kha of extra deforestation in the region. Deforestation however decreases in Latin
America by 100 kha, due to lower expansion of grassland.

Land use change emissions

Land use emissions are mainly associated to soil carbon changes in cropland (53 MtCO2), natural
vegetation conversion emissions (61 MtCO2e) and peatland emissions (56 MtCO2e). Carbon
sequestration in agricultural crops decreases emissions by 32 MtCO2.

Total land use emissions of sunflower FAME are found to be 155 MtCO2e. With an assumed 20 year
amortisation this results in a resulting LUC emissions factor of 63 gCO2e/MJ. If no natural vegetation
were to regrow on abandoned agricultural land in absence of biofuels, the total land use emissions
would be lower, at 137 MtCO2e, and the LUC emission factor would be 56 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 62

4.10 Palm oil biodiesel

Energy productivity 2020: 88 GJ biodiesel/ha (equals 2660 litres/ha)

Cropland and agricultural land displacement: 8.5 ha/TJ and 6.3 ha/TJ

Land use emissions: 231 gCO2/MJ

Additional demand of 1% FAME (123 PJ) is produced from 3.5 Mt of palm oil. This shock leads to a
price increase of 3.1% on palm fruit in Southeast Asia and 4.2% on the price of palm oil regionally.
At the global level, impact on palm oil price is 2.1%.

Adjustments to the shock

The additional feedstock is achieved 42% through a decrease in food (2%) and other uses (40%)
demand, 14% through co-product substitution (palm kernel) and 44% by extra production.

Additional feedstock production is mainly located in South East Asia (3.1 Mt) and requires
acreage of 1.2 Mha palm oil plantations.

Overall agricultural production is also affected globally by a decrease in demand of 0.6 Mt of

sugar crops and the displacement of 210 kt of grains and 270 kt of vegetable oils on the demand
side.

Land use change effect

Land expansion leads to 1.0 Mha of additional land conversion globally, with new cropland at the
expense of grassland, other natural vegetation and forest area. In Southeast Asia, cropland expands
by 930 kha, of which 290 kha are sourced from grassland, 80 kha from other natural vegetation and
570 kha from primary forest. Increase in palm oil plantations remains more limited in other regions,
with only 40 kha in Sub-Saharan Africa.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 63

Land use change emissions
Land use emissions are mainly associated to natural vegetation conversion emissions (362
MtCO2e) and peat land emissions (450 MtCO2e). Carbon sequestration in biomass (-224 MtCO2e) and
soil carbon sequestration (-19 MtCO2e) decrease emissions.

Total land use emissions of palm oil FAME are found to be 569 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 231 gCO2e/MJ. Changing assumption on the
natural vegetation regrowth on abandoned agricultural land in absence of biofuels does not affect the
final result here.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 64

4.11 Rapeseed oil biodiesel

Energy productivity 2020: 52 GJ biodiesel/ha (equals 1570 litre/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 15.5 ha/TJ and 11.9 ha/TJ

Land use emissions: 65 gCO2/MJ

Additional demand of 1% FAME (123 PJ) is produced from 3.5 Mt of rapeseed oil, with 41% of
additional production taking place inside the EU. This shock leads to a price increase of 25% for the
price of rapeseed and 28% for the price of rapeseed oil in the EU. At the global level, impacts on
seed and oil prices are 5.3% and 7%, respectively.

Adjustments to the shock

The additional feedstock is achieved 13% through a decrease in food and feed demand, 32% by
displacement of feed by co-product of biodiesel and 54% by extra production.

Additional feedstock production corresponds to 6.2 Mt of rapeseed, mainly located in the EU (3.0
Mt), North America (2.6 Mt) and Oceania (0.5 Mt). This requires acreage of 790 kha in the EU, 780
kha in North America and 350 kha in Oceania.

Overall agricultural production is also affected globally by an increase in consumption of protein

meal of 2.3 Mt and the decrease of 720 kt in vegetable oil demand. Grain demand increases by 1.1
Mt to serve as feed complement to newly consumed protein meals, whereassugar crops demand
decreases by 1.7 Mt. The livestock sector benefits from the extra feed production and meat and milk
production increase globally by 130 kt and 330 kt, respectively.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 65

Land use change effect
Land expansion requires 1.9 Mha of additional cropland globally. In the EU, cropland expands by
1.1 Mha, of which 630 kha is into abandoned land and 470 kha is into other natural vegetation.
Global grassland decreases by 440 kha as protein meal availability favors grain-based production
systems over grass-based ones, in particular in Latin America (-140 kha) and North America (-180
kha). At the same time, palm oil plantation expands by 110 kha in Southeast Asia, which leads to 50
kha of extra deforestation in the region. Deforestation, however, decreases in Latin America by 80
kha, due to lower expansion of grassland.

Land use change emissions

Land use emissions are mainly associated with soil carbon changes (72 MtCO2e), peatland
emissions (36 MtCO2e) and foregone sequestration (36 MtCO2e). Carbon sequestration in palm
plantations decreases emissions by 31 MtCO2e.

Total land use emissions of rapeseed FAME is found to be 160 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 65 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower, at 124 MtCO2e, and the LUC emission factor would be 50 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 66

4.12 Soybean oil biodiesel

Energy productivity 2020: 17 GJ biodiesel/ha (equals 530 litres/ha. EU28 avg., before
accounting for co-products)

Cropland and agricultural land displacement: 14.8 ha/TJ and 12.6 ha/TJ

Land use emissions: 150 gCO2/MJ

Additional demand of 1% FAME (123 PJ) is produced from 3.5 Mt of soybean oil, with 9.8% of
additional production taking place inside the EU. This shock leads to a price increase of 2.3% for
soybean and 10.8% for soybean oil at the global level.

Adjustments to the shock

The additional feedstock is achieved 18% through a decrease in food and feed demand, 38% by
displacement of feed by soybean meal and 44% by extra production, in which area expansion
accounts for 36% and yield increase for 8%.

Additional feedstock production requires 7.3 Mt of extra soybeans locally, mainly located in North
America (4 Mt) and Latin America (2.7 Mt). This requires an area of 960 kha in North America and
860 kha in Latin America. Inside the EU, soybean production increases by 570 kt, which corresponds
to 250 kha.

Overall agricultural production is also affected globally by an increase in consumption of protein

meal by 5.9 Mt and the decrease of 1.4 Mt in demand for vegetable oils. Grain demand increases by
1 Mt to serve as feed complement to newly consumed protein meals, whereas sugar crops demand
decreases by 0.9 Mt. The livestock sector benefits notably from the extra feed production and meat
and milk production increase globally by 620 kt and 1,280 kt, respectively.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 67

Land use change effect
Land expansion leads to 2.0 Mha of land conversion globally, 1.8 Mha of which corresponds to
additional cropland. In Latin America, cropland expands (500 kha) mainly into other natural
vegetation (420 kha), whereas grassland decrease (-190 kha) due to protein meal availability, which
favors grain-based production systems instead of grass-based ones. As a consequence, deforestation
decreases by 120 kha. The same effect is observed North America, where cropland expansion (190
kha) partly benefits from grassland decrease (-100 kha) and expands into other natural vegetation
for only 90 kha. At the same time, palm oil plantation expands by 240 kha in Southeast Asia and
cereal production also grows to provide more animal feed. This leads to 560 kha of cropland
expansion in the region, replacing 160 kha of grassland, 150 kha of primary forest and 260 kha of
other natural vegetation.

Land use change emissions

Land use emissions are mainly associated with LUC emissions (244 MtCO2e), soil carbon changes
(105 MtCO2e) and peatland emissions (78 MtCO2e). Carbon sequestration in biomass decreases
emissions by 60 MtCO2e.

Total land use emissions of soybean FAME are found to be 368 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 149 gCO2e/MJ. Changing assumption on the
natural vegetation regrowth on abandoned agricultural land in absence of biofuels does not affect the
final result here.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 68

4.13 Cereal straw ethanol produced in the EU

Full EU28 1% shock

Energy productivity 2020: 15 GJ ethanol/ha (EU28 avg., equals 710 litres/ha)

Cropland and agricultural land displacement: 2.0 ha/TJ and 2.3 ha/TJ

Land use emissions: 16 gCO2/MJ

Additional demand of 1% straw ethanol (123 PJ) in the EU is produced from 20.5 Mt of cereal
straw, with 100% of additional production taking place in Europe.

Adjustments to the shock

Most of the new demand is fulfilled through extra production (93%), although 7% comes from a
decrease in animal demand and other uses. Expansion of area where residues are harvested
contributes 64% of the biofuel demand through sustainable systems and 29% from intensification of
removal in countries where sustainable removal is not sufficient to meet demand.

Additional feedstock production amounts to 19.1 Mt and requires residues to be removed from
an additional 7.3 Mha in the EU. Harvesting management also varies, to accommodate the new
demand. In total, 5.0 Mha are newly harvested under sustainable management (see Appendix II.1
for the definition of this type of management in each region). In addition, 2.3 Mha corresponds to
removal rates of more than 33-50%, with corresponding yield and soil organic carbon impacts. This
unsustainable removal mainly occurs in the Netherlands, Belgium, Greece, Hungary, Ireland, Italy,
Poland and the UK. In the Netherlands, Lithuania and Portugal, supply even reaches its limit, leading
to substitution of straw in the livestock sector.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 69

Overall agricultural production is only marginally affected by the additional cereal straw demand.
Grain production is affected by some yield decrease in cases where more than 33-50% of straw is
removed, but the additional market value for cereal production coming from residues also leads to a
decrease in price of wheat (-1.3%) in Europe, and an overall increase in grain production of 0.3 Mt.

Land use change effect

Land use conversion amounts to 0.4 Mha. This only concerns cropland expansion in the EU for extra
cereals, two thirds of which isinto abandoned land and one third in other natural land.

Land use change emissions

Land use emissions result from this conversion of land for new cropland and from the soil organic
carbon impact of removing residues in case of unsustainable management. Soil organic carbon
emissions account for 42 MtCO2e and foregone sequestration from forest reversion for 13 MtCO2e.

Total land use emissions are found to be 51 MtCO2e. With an assumed 20 year amortisation, this
results in an LUC emissions factor of 16 gCO2e/MJ. If no natural vegetation were to regrow on
abandoned agricultural land in absence of biofuels, the total land use emissions would be lower, at
38 MtCO2e, and the LUC emission factor would be 11 gCO2e/MJ.

Country specific results

Due to the heterogeneity of situations in straw market across Europe, we present below some
additional results showing contrasting situations in different EU countries. We look at the impact of
removing residues for three specific regions: Hungary, where the market is particularly tight and the
sustainability threshold is met; the UK, where straw is largely used and additional production could
push the production beyond the sustainability threshold, and France, where a large reserve of
residue supply exists. The impacts for these three regions are provided in Figure 19 below.

In Hungary, the shock of straw ethanol demand leads to some large effects, as most of the additional
straw comes from unsustainable management and impacts soil organic carbon and yields. The span
of LUC impact ranges for the full distribution, from -75 to 150 gCO2/MJ, with an average impact of 60
gCO2/MJ. Most values are in the range 13-113 gCO2/MJ (without first and last decile).

In the UK, 40% of the supply can be obtained by increase removal of straw under sustainable
management, but 60% is sourced from removal above 40%. As a result, impacts on yield and soil
organic carbon occur and the range of impact goes from -30–80 gCO2/MJ, with most of the
distribution in the range 4–39 gCO2/MJ. The average value for the UK is 20 gCO2/MJ, which is close
to the average EU effect presented above.

In contrast, in France, all the additional straw supply occurs through increase of straw removal well
below the sustainable removal rate. Impacts are then invisible, with an average value of 0 gCO2/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 70

Figure 19: Impact on LUC emissions of 1% incorporation of straw ethanol in national transportation targets of

Hungary, UK and France

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 71

4.14 Miscanthus and switchgrass FT biodiesel produced in the EU

Energy productivity 2000-2030: 90 GJ biodiesel/ha (EU28 avg., without accounting for co-
products)

Cropland and agricultural land displacement: 7.5 ha/TJ and 7.7 ha/TJ

Land use emissions: -12 gCO2/MJ

Additional demand of 1% FT diesel (123 PJ) is produced from 13.1 Mt DM of annual grassy crops
(miscanthus, switchgrass), with 100% of additional production taking place in the EU.

Adjustments to the shock

All the grassy crop expansion comes from acreage increase but other crops adjust through demand
(8%). Yield contributions are slightly negative (-3 %).

Additional feedstock production is located in the EU (13.1 Mt). This new production requires
acreage of 1.4 Mha in the EU. Production is based mainly on switchgrass (9.3 Mt) in Western, South
and Central EU, and miscanthus (3.9 Mt) in Northern EU. Average yield for miscanthus/switchgrass is
assumed to be 10.5 t dry matter/ha in Northern Europe, 9.2 t dm/ha in Central Europe and 9 t
dm/ha in Western Europe. Switchgrass is assumed yield of 7.4 t dm/ha in Southern Europe, 8.7 t
dm/ha in Eastern Europe and 10.1 t dm/ha in Western Europe.

Overall agricultural production is also affected by additional feedstock demand and grain demand
decreases of 360 kt globally. Moreover, 100 kt of sugar crops, 40 kt of protein meals, 70 kt of meat
and 170 kt of milkare displaced on the demand side.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 72

Land use change effect
Land expansion requires 920 kha of additional cropland and 20 kha of grassland globally. In
Europe, cropland expands by 890 kha, of which 300 kha is sourced from abandoned land and 580
kha from other natural vegetation. Outside Europe, only a small LUC takes place.

Land use change emissions

Land use emissions are mainly associated with the forgone carbon sequestration in Europe (19
MtCO2e) and LUC emissions (7 MtCO2e). Soil organic carbon stocks increase by 30 MtCO2e following
the shock. Additional carbon sequestration in agricultural crops increases by 27 MtCO2e.

Total land use emissions of grassy FT biodiesel are found to be --29MtCO2e. With an assumed 20
year amortisation, this results in an LUC emissions factor of -12 gCO2e/MJ. If no natural vegetation
were to regrow on abandoned agricultural land in absence of biofuels, the total land use emissions
would be lower, at -48 MtCO2e, and the LUC emission factor would be -20 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 73

4.15 Short Rotation Plantation FT biodiesel produced in the EU

Energy productivity 2000-2030: 97 GJ biodiesel/ha (EU28 avg)

Cropland and agricultural land displacement: -2.85 ha/TJ (excluding SRC) and 6.4 ha/TJ (including
SRC).

Land use emissions: -29 gCO2/MJ

Additional demand of 1% FT diesel (123 PJ) is produced from 13.1 Mt of woody biomass (short
rotation coppice), with 100% of additional production taking place in the EU.

Adjustments to the shock

Extra supply occurs 91% through increase in planted areas and 9% through increase in yield of
coppice produced in the EU.

Additional feedstock production is located in the EU and requires acreage of 1.2 Mha in the EU.

Overall agricultural production is also affected by additional decreases of 240 kt in demand for
grains inside the EU. Moreover, 160 kt of sugar crops, 50 kt of meat products and 180 kt of milk are
displaced on the demand side, which shows the effect of competition between energy plantations and
cropland.

Land use change effect

Land expansion requires 1.2 Mha of additional land globally, of which 1.2 Mha areplantations in the
EU. Cropland decreases by 390 kha, grassland by 60 kha, other natural vegetation by 510 kha and
abandoned land by 220 kha. Inside the EU, plantations mainly displace green fodder and fallow (-190
kha), wheat (120 kha) and oats (40 kha). Outside EU, only small LUC takes place (below 10 kha).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 74

Land use change emissions
Land use emissions are mainly associated to carbon sequestration through afforestation (-82
MtCO2e) and soil carbon change (-12 MtCO2e), while emissions from forgone carbon sequestration on
abandoned land increase by 14 MtCO2e globally.

Total land use emissions of short rotation coppice FT biodiesel are found to be -71 MtCO2e. With an
assumed 20 year amortisation, this results in an LUC emissions factor of -29 gCO2e/MJ. If no natural
vegetation were to regrow on abandoned agricultural land in absence of biofuels, the total land use
emissions would be lower, at -89 MtCO2e, and the LUC emission factor would be -35 gCO2e/MJ.
However, it should be kept in mind that short rotation plantations can only sequester more carbon
than a regrowing forest if short period of time are considered. If accounting for forest regrowth for
time periods beyond 20 years, the short rotation plantation would quickly become less efficient from
a sequestration standpoint than a forest plantation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 75

4.16 Forest residues FT biodiesel produced in the EU

Energy productivity 2020: 5.8 GJ diesel/ha/y (EU28 avg.)

Cropland and agricultural land displacement: 0 ha/TJ and 0 ha/TJ

Soil organic carbon emissions: 17 gCO2/MJ

Note: parameters tested in the sensitivity analysis do not change the results of this shock.

Residue availability from forest under sustainable management practice are estimated to be about
14.4 million m3 in the EU28, which corresponds to about 7.2 million tons. Therefore, we apply a
shock of only 0.5% for forestry residues (62 PJ), which corresponds to 6.6 Mt of forestry residues
demand.

Forest residues considered here include: i) losses from the harvesting of roundwood at the
harvesting site (i.e. rotten wood, piece of wood unsuitable for roundwood), excluding the bark (bark
is assumed to be harvested and delivered to industry with roundwood); ii) all branches attached to
the stem; iii) tops when dimensionally unsuitable for production of roundwood (threshold adjusted by
country, e.g. 7-10 cm of diameter). No stumps are included. The amount of available residues we
assume in our modelling is rather low because residues that are already in use for existing purposes
are not taken into account. These sustainably harvested residues with no current uses can be
processed into 68 PJ of biodiesel, which equals just over 0.5% of total EU transport fuel consumption
in 2020. See Appendix III.3 for background surveys and literature on the various impacts of
changing management.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 76

Adjustments to the shock
Additional demand of 0.5% FT diesel (62 PJ) is produced from 13.1 Mm3 of forest residues, being
entirely sourced from Europe forests. For cases where countries demand more than their country can
provide, adjustments through trade takes place at the EU level from countries with excess of
residues to countries with residue deficits.

Additional feedstock production is located in managed forest area in the EU (9.4 Mha). Main
producers of forest residue FT biodiesel are Northern Europe (28 PJ), Western Europe (14 PJ),
Southern Europe (11 PJ).

Overall agricultural production is not impacted by the additional forest residue harvesting with no
observed land expansion. Therefore, for a shock of 0.5%, production costs remain below competitive
feedstocks and only direct effects are recorded. Under such practices, however, the soil organic
carbon is impacted by removal of residues.

Land use change effect

No change in land cover is observed in this scenario, but land management is affected in managed
forest.

Soil organic carbon emissions

No LUC emissions take place. However, collecting forestry residues does lead to a soil carbon change
which amounts to 20.4 MtCO2e over a 20 year period. This soil organic carbon loss results from the
removed residue no longer decomposing in the soil. We consider here only the removal of residues
for areas that are not ecologically vulnerable. In these areas, 60-70% of available residues are
removed, depending on the harvesting method. As we do not know if the soil carbon response is
linear with the extraction rate, we cannot determine if the emission level per quantity of residue
would be lower or higher for a more limited extraction rate. See Annex III.3 for further details.

Total soil organic carbon emissions of forestry residues FT biodiesel are found to be 20.4
MtCO2e. With an assumed 20 year amortisation, this results in an soil organic carbon emissions
factor of 17 gCO2e/MJ. Changing assumption on the natural vegetation regrowth on abandoned
agricultural land does not affect the final result here.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 77

4.17 Starchy crops group

Energy productivity 2000-2030: 51 GJ ethanol/ha (EU28 avg., before accounting for co-
products)

Cropland and Grassland displacement: 11.3 ha/TJ and 9.6 ha/TJ

Land use emissions: 29 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) is here distributed across the feedstocks proportionally
to their historical proportion. We do not consider substitution here, which would then go fully for
adoption of rye. Feedstocks used are here maize ( 7.9 Mt), wheat (5.2 Mt) and rye (1.9 Mt), coming
for most part from the EU. The shock has limited price impact on cereals (2% in the EU and 0.2%
globally) and food prices are unaffected.

Adjustments to the shock

The additional feedstock is achieved 20% through a decrease in food and feed demand, 34% by
displacement of feed by co-product of ethanol, and 46% by extra production, of which yield accounts
for 17% and area expansion 29%.

Additional feedstock production is mainly located in the EU (9.2 Mha), followed by Latin America
(1.3 Mha) and North America (0.8 Mha).

Overall agricultural production is also affected by a decrease in demand for cereals of 2.7 Mt
globally and 3.8 Mt considering the EU only. On the other side, the displacement leads to an increase
of 3.1 Mt in protein meal for cattle. As a result, meat production increase globally by 100 kt and milk
production by 290 kt.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 78

Land use change effect
Land expansion requires 1.5 Mha of additional land globally, mainly cropland. In Europe, cropland
increases by 1.3 Mha, which leads to 790 kha of forest losses and 440 kha of abandoned land
recovery.

Land use change emissions

Land use emissions are increasing very slightly (15 MtCO2e) through LUC, but soil carbon
emissions play a larger role in this scenario, although of small absolute magnitude (50 MtCO2e).
Emissions from peatland coming from co-product substitution in the oilseed market amount to 7
MtCO2e.

Total land use emissions of starchy crops are found to be 72 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 29 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower, at 46 MtCO2e, and the LUC emission factor would be 19 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 79

4.18 Sugar crops group

Energy productivity 2020: 135 GJ ethanol/ha (EU28 avg., before accounting for co-
products)

Cropland and agricultural land displacement: 2.3 ha/TJ and 2.0 ha/TJ

Land use emissions: 11 gCO2/MJ

Additional demand of 1% ethanol (123 PJ) from sugar crops is here distributed across the
feedstocks proportionally to their historical proportion. As for starch, we do not consider substitution
here, which would imply that only sugar cane remains. Feedstocks uses would be two-thirds sugar
beet (40 Mt) and one third sugar cane (21 Mt), mostly imported. The shock has no significant impact
on sugar products globally, with the exception of the EU where sugar beet price increases by 4.7%.
Food prices are found to be unaffected.

Adjustments to the shock

The additional feedstock is achieved 16% through a decrease in food and feed demand, 13% by
displacement of beet pulp and 71% by extra production, with a negative contribution from yield of
5%, due to lower marginal yields than the average yields.

Additional feedstock production is mainly located in the EU (0.6 Mha), followed by Latin America
(0.3 Mha). Occupied areas are considerably smaller than for some other feedstock scenarios.

Overall agricultural production is also affected by a decrease in demand globally of 2.4 Mt for
cereals and 1.9 Mt for sugar crops. Cattle benefit from the shock through higher inputs of co-product
(2.1 Mt).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 80

Land use change effect
Land expansion is relatively limited for this scenario, with high yielding feedstocks. Cropland
expands by 0.3 Mha globally. In Europe, expansion is notably limited, with only 125 kha of cropland
increase.

Land use change emissions

Land use emissions increase moderately in this scenario (37 MtCO2e) through LUC. Soil carbon
emissions in the shock for EU amount to about 12 MtCO2e, but are compensated by sequestration in
soil carbon in Latin America (-22 MtCO2e).

Total land use emissions of sugar crops are found to be 28 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 11 gCO2e/MJ. If no natural vegetation were to
regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would be
lower, at 20 MtCO2e, and the LUC emission factor would be 8 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 81

4.19 Vegetable oil group

Energy productivity 2020: 37 GJ biodiesel/ha (EU28 avg., before accounting for co-
products)

Cropland and agricultural land displacement: 12.6 ha/TJ and 8.6 ha/TJ

Land use emissions: 101 gCO2/MJ

Additional demand of 1% biodiesel (123 PJ) from oilseeds is here again distributed across the
feedstocks proportionally to their historical proportion. We do not consider substitution, which would
then imply that only palm oil would be used in this group. Feedstocks used under this scenario are
50% rapeseed oil, 24% soybean oil, 23% palm oil and 3% sunflower oil, for a total of 3.5 Mt. The
shock increases vegetable oil prices by 2.9% globally and 12% in the EU. Global crop prices increase
by 0.2% and the world food price index by 0.1%.

Adjustments to the shock

The additional feedstock provision is achieved 12% through a decrease in food and feed demand,
25% by displacement of feed by protein meals and 63% by extra production.

Additional feedstock production is mainly located in the EU (0.5 Mha), followed by North America
(0.3 Mha) and Southeast Asia (0.4 Mha).

Overall agricultural production is affected globally by this shock, in particular through the
provision of the protein meals (+2.2 Mt). The boost to the livestock sector also drives an increased
demand for grains (1.2 Mt), in particular in Latin America, where they serve as a complement to
protein meals excess. Meat consumption increases by 200 kt globally and milk by 460 kt.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 82

Land use change effect
Land expansion under this scenario leads to 1.6 Mha of land conversion globally, most of it (1.6
Mha) for additional cropland. In the EU, expansion is 780 kha, whereas Southeast Asia expands land
by 390 kha, most of it through expansion of oil palm plantations. At the same time, grassland
decrease globally by 500 kha, mainly in Latin America and Southeast Asia. Tropical forests increase
in Latin America by 100 kha compared to the basline, but decrease in Southeast Asia by 90 kha.

Land use change emissions

Land use emissions are significant in this scenario for Southeast Asia (72 MtCO2e) due to LUC and
reach 117 MtCO2e globally. Peatland emissions amount to about 131 MtCO2e and soil organic carbon
drives additional emissions of 55 MtCO2e globally.

Total land use emissions of vegetable oils are found to be 249 MtCO2e. With an assumed 20 year
amortisation, this results in an LUC emissions factor of 101 gCO2e/MJ. If no natural vegetation were
to regrow on abandoned agricultural land in absence of biofuels, the total land use emissions would
be lower, at 219 MtCO2e, and the LUC emission factor would be 89 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 83

4.20 EU 2020 biofuel mix scenario (all feedstocks)

Energy productivity 2020: 48 GJ ethanol/ha (EU28 avg., without accounting for co-
products)

Cropland agricultural land displacement: 10.4 ha/TJ / 9.6 ha/TJ

Land use emissions: 97 gCO2/MJ

Additional demand of biofuels for a full EU 2020 biofuel mix is coming from a mix of 11.1 Mt of
corn, 6.6 Mt of wheat, 15.4 Mt of sugar beet, 5.8 Mt of rapeseed oil, 4.5 Mt of palm oil, 2.9 Mt of
soybean oil and more minor uses of sunflower oil, sugar cane andbarley. We distributed feedstocks
proportionally to their historical shares. To reach their commitments, Member States deploy some
second generation biofuels, and 7.8 Mt of short rotation coppice and 2.1 Mt of grassy crops are
established for biofuel production in this scenario. Due to the large contribution of palm oil and
soybeans, only 37% of total additional feedstocks under this scenario are produced in the EU. This
shock leads to a price change globally of 9.3% for vegetable oil and 0.8% for cereals, while protein
meal prices decrease by 12%. In the EU, the prices of these products change by 38%, 4.6% and -
24% respectively. Overall, crop prices increase by 0.5% globally and food prices by 0.3%.

Adjustments to the shock

The additional feedstock is achieved 22% through a decrease in food and feed demand, 26% by
displacement of feed by co-products and 52% by extra production, in which yield accounts for 13%.

Additional feedstock production is located in the EU (2.9 Mha), Southeast Asia (2.2 Mha), Latin
America (1.3 Mha) and to a more limited extent in Ukraine and rest of Europe, North America, Sub-
saharan Africa, and Oceania (0.4 Mha for each).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 84

Overall agricultural production is also affected by a decrease in demand of 6.5 Mt for cereals in
Europe and 9.2 Mt globally. Additionally, 2.4 Mt vegetable oils are no longer consumed globally.
Protein meal consumption increasesmassively by 13 Mt, which boosts milk production by 1 Mt and
meat production by 0.2 Mt.

Land use change effect

Land expansion requires an additional 8.8 Mha of land conversion globally compared to the
reference case – 8 Mha for additional cropland and the rest of short rotation plantations for advanced
biofuels. In Europe, less cropland is abandoned and crop area increase by 2.9 Mha. In Southeast
Asia, cropland expands by 2.1 Mha under the pressure of palm plantation, 50% of this is at the
expense of tropical forest. At the same time, grassland decreases by 1.4 Mha globally, mainly in
Latin America, Southeast Asia and the EU.

Land use change emissions

Land use emissions are mainly associated with LUC emissions (855 MtCO2e) and peatland
emissions (880 MtCO2e). Carbon sequestration in biomass decreases emissions by 480 MtCO2e,
however, through new palm trees, mainly in Southeast Asia. Soil organic carbon release 228 MtCO2e
over the period and foregone sequestration in natural vegetation accounts for 110 MtCO2e.

Total LUC emissions for the EU 2020 biofuel mix scenario reach a total of 1,493 MtCO2e over the full
20 year period, which corresponds to an LUC emissions factor of 97 gCO2e/MJ. If no natural
vegetation were to regrow on abandoned agricultural land in absence of biofuels, the total LUC
emissions would be lower, at 1,385 MtCO2e, and the LUC emission factor would be 90 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 85

4.21 EU 2020 biofuel mix scenario with 7% cap on conventional biofuels

Energy productivity 2000-2030: 60 GJ ethanol/ha (EU28 avg., before accounting for co-
products)

Cropland and Grassland displacement: 8.0 ha/TJ and 9.0 ha/TJ

Land use emissions: 74 gCO2/MJ

Additional demand when implementing a EU 2020 mix with 7% cap on first generation comes from
a more important share of second generation bioenergy. Under this scenario, 17 Mt of woody
biomass is sourced from short rotation coppice and 6.1 Mt from grassy crops. The largest feedstocks
used after these are sugar beet (10 Mt), maize (8.4 Mt) and wheat (4.8 Mt), with palm oil
importance decreased at 3.3 Mt. This shock lowers the impact of the policy on the feedstock markets
(1.1%) and no food price increase is observed on average.

Adjustments to the shock

Additional feedstocks are provided 17% through a decrease in food and feed demand, 22% by
displacement of feed by co-products and 61 % by extra production, in which yield accounts for 17%.

Additional feedstock production is located in the EU (2.3 Mha), Southeast Asia (1.8 Mha), Latin
America (0.6 Mha for each), and Ukraine and the rest of Europe, Oceania and Sub-saharan Africa
(0.3 Mha each).

Overall agricultural production is also affected globally by a decrease of 2.1 Mt in demand for
cereals and 2.8 Mt for sugar crops. Protein meal demand increases by 6.5 Mt, whereas demand for
vegetable oil is less affected (-1.5Mt). Meat and milk demand increase globally by 0.2 and 0.6 Mt,
respectively.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 86

Land use change effect
Land expansion leads to 6.7 Mha of land conversion globally, of which 5.2 Mha are used for
additional cropland and 1.5 Mha for short rotation coppice. In the EU, cropland expands by only 1.8
Mha, half at the expense of abandoned land and half through other natural vegetation. Southeast
Asia (1.6 Mha) and Latin America (0.6 Mha) are the two other regions where most cropland changes
are taking place, although to a lower extent than in some other scenarios. At the same time,
grassland decreases by 0.9 Mha globally, mainly in Latin America, Southeast Asia and in the EU.

Land use change emissions

Land use emissions from living biomass conversions are more contained in this scenario at 617
MtCO2. The EU is almost neutral on this source (6 MtCO2) through expansion of woody biomass
plantations. However, peatland emissions are still high (684 MtCO2e), although plantations sequester
517 MtCO2e. Some foregone sequestration on other feedstock still amounts to 89 MtCO2e globally
and soil carbon losses reach 81 MtCO2e globally.

Total land use emissions of the EU 2020 scenario with 7% cap are found to be 952 MtCO2e,
therefore resulting in an LUC emissions factor of 74 gCO2e/MJ for the EU policy. If no natural
vegetation were to regrow on abandoned agricultural land in absence of biofuels, the total LUC
emissions would be lower, at 865 MtCO2e, and the LUC emission factor would be 67 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 87

4.22 Abandoned land in the EU

Energy productivity 2020: 51 GJ ethanol/ha (EU28 avg., before accounting for co-
products)

Cropland and and agricultural land displacement: 10.6 ha/TJ and 10.0 ha/TJ

Land use emissions: 52 gCO2/MJ

In order to assess the impact of better use of land, we explore the effect of an alternative setting on
the policy scenario EU 2020 biofuel mix, now with restrictions on the import of feedstocks other than
those traditionally grown in the EU and with further possibilities for expanding cropland into
abandoned land and other natural vegetation in Eastern Europe and Ukraine.

Additional demand when implementing the EU 2020 scenario with more reliance on EU feedstocks
leads to a larger part of bioenergy sourced from rapeseed, sugar beet and cereals. The feedstocks
the most used are then rapeseed oil (422 PJ), maize (82 PJ), wheat (61 PJ), short rotation coppice
(66 PJ) and sugar beet (38 PJ). This shock leads to an impact of about 1% on prices of these
feestocks.

Adjustments to the shock

Additional feedstock are provided 11% through a decrease in food and feed demand, 33% by
displacement of feed by co-products and 56% through extra production, composed of 22% of yield
increase and 35% of extra area.

Additional feedstock production is located to a larger extent in the EU (4.0 Mha compared to 2.9
Mha in the EU2020 scenario). Other regions where feedstocks are grown are Ukraine and rest of
Europe (1.7 Mha), Southeast Asia (1.2 Mha), Latin America (0.8 Mha) and Russia and its
neighbouring countries (0.8 Mha).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 88

Overall agricultural production is also affected globally by a decrease of 8.0 Mt in demand for
cereals and 6.5 Mt for sugar crops. Protein meal demand increases by 14.9 Mt, whereas demand for
vegetable oil decreases by 2.1 Mt.

Land use change effect

Land expansion leads to 8.9 Mha of land use conversion globally, with 8.2 Mha used as additional
cropland, mainly at the expense of abandoned land (2.8 Mha), grassland (1.2 Mha) and other natural
vegetation (4.5 Mha). In the EU, cropland increases by 4.0 Mha and abandoned land is reduced by
2.1 Mha. Ukraine and rest of Europe increase their cropland at the expense of other natural
vegetation (former abandoned land) by 0.8 Mha. At the same time, some deforestation still occurs (-
0.4 Mha) due to the effect of co-products on the oilseed market. .

Land use change emissions

Land use emissions from living biomass conversion amount in this scenario to 333 MtCO2
additional emissions compared to the baseline. The EU is a significant emitter of soil carbon
emissions (204 MtCO2) and foregone sequestration in this region totals 145 MtCO2e. Peatland
emissions still constitute the largest source globally (359 MtCO2e, whereas plantations sequester 321
MtCO2e), but this source is still reduced by more than half compared to the full EU 2020 biofuel mix
peatland emissions (880 MtCO2). However, it is noteworthy that emissions in the EU increase by one
third (145 MtCO2e for soil organic carbon and 110 MtCO2e for natural vegetation reversion in the EU
2020 mix scenario).

Total land use emissions of the abandoned land restoration scenario are found to be 802 MtCO2e,
therefore resulting in an LUC emissions factor of 52 gCO2e/MJ for the EU policy. The scenario slightly
improves the overall emissions level through avoided expansion in carbon rich areas (Latin America),
and reduces the induced deforestation in Southeast Asia, although leaked emissions remain large.
The LUC of 52gCO2/MJ associated to this shock is to be compared to 97 gCO2/MJ for the EU 2020
biofuel mix alone. If no natural vegetation were to regrow on abandoned agricultural land in absence
of biofuels, the LUC emission factor would be 42 gCO2e/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 89

4.23 Lower deforestation

Energy productivity 2020: 49 GJ biofuel/ha (EU28 avg., before accounting for co-
products)

Cropland and agricultural land displacement: 9.2-9.9 ha/TJ and 7.7-8.2 ha/TJ

Land use emissions: 87 gCO2/MJ (low) and 48 gCO2/MJ (very low)

Very low deforestation scenario

We observe that at a very high level of protection of forests (USD 50/tCO2), most deforestation is
seriously halted and the LUC effect is drastically reduced (to 48 gCO2/MJ), even if some significant
emissions from peatland drainage can still occur. These latter emissions account for 689 MtCO2e.
Indeed, peatland is not limited to forested areas and some additional protection measures are
necessary to limit its drainage.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 90

Low deforestation scenario

The low deforestation scenario includes a more moderate carbon price to disincentivise deforestation
of USD 10/tCO2. This moderate incentive is sufficient to limit a part of deforestation seen in the
baseline, but is less effective in preventing deforestation associated with palm oil expansion, as this
type of agricultural activity can remain more profiable than the conservation alternative. The
obtained LUC emission value is 87 gCO2/MJ.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 91

4.24 Very low deforestation with no peatland drainage

Energy productivity 2020: 49 GJ biofuel/ha (EU28 avg., before accounting for co-
products)

Cropland and agricultural land displacement: 9.2 ha/TJ and 7.7 ha/TJ

Land use emissions: 4 gCO2/MJ

For the case where peat land emissions are not considered and deforestation is very low, the overall
effect of a shock has a quite different outcome. The total LUC effect is drastically reduced (4
gCO2/MJ), because carbon sequestered in palm trees compensates for a part of the LUC emissions,
soil organic carbon and forest reversion carbon that is emitted as a result of the shock. Only 55
MtCO2e are emitted under low deforestation and peatland protection, much less than the baseline
situation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 92

4.25 Higher deforestation

Energy productivity 2000-2030: 51 GJ biofuel/ha (EU28 avg., beforeaccounting for co-

products)

Cropland and Grassland displacement: 9.9 ha/TJ and 8.8 ha/TJ

Land use emissions: 110 gCO2/MJ

When considering a baseline with higher deforestation, the level of encroachment into forest is nearly
doubled, with 1.9 Mha of forest disappearance as a result of the full mandate shock. This
deforestation takes place in particular in Latin America (-0.9 Mha) and Southeast Asia (-1 Mha). Palm
plantations in Southeast Asia increase in such scenario by 2.6 Mha, which means that land use types
other than tropical forest are also converted to palm plantations. This scenario leads to LUC emission
value of 110 gCO2/MJ biofuel consumed in the EU.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 93

4.26 Comparison of results with previous LUC assessments

This is not the first study which assesses the LUC impacts of biofuel feedstocks. Some previous
assessments have been performed in the EU, and more often in the US, where the literature is
prolific. An overview is presented below of the way that the results of of the current study compare
to previous assessments published by environmental agencies or in peer-reviewed journals. We
provide some external sources for a more exhaustive overview of LUC model results (e.g. De Cara et
al., 2012 covering 485 estimates from various studies).

Table 11: Selected values from the literature on LUC emission factors from biofuel feedstocks

Amortisation
Study Location Feedstock LUC result LUC 20 years
period
Searchinger et
USA Corn 104 gCO2e/MJ 30 years 156 gCO2e/MJ
al. (2008)
Keeney and
USA Corn 27 gCO2e/MJ 30 years 40 gCO2e/MJ
Hertel (2009)
CARB (2009) USA Corn 30 gCO2e/MJ 30 years 45 gCO2e/MJ
CARB (2009) USA Soybean 62 gCO2e/MJ 30 years 93 gCO2e/MJ
US EPA (2010) USA Corn 30 gCO2e/MJ 30 years 45 gCO2e/MJ
US EPA (2010) USA Soybean 40 gCO2e/MJ 30 years 60 gCO2e/MJ
Britz and Hertel
EU Rapeseed 42 gCO2e/MJ 30 years 63 gCO2e/MJ
(2011)
Plevin et al. a
US Corn 21-142 gCO2e/MJ 15-45 years 47-106 gCO2e/MJ
(2012)
Taheripour and
US Corn 13.3 gCO2e/MJ 30 years 20 gCO2e/MJ
Tyner (2013)

The only existing study that focuses on multiple conventional biofuel feedstocks for biofuels
consumed in the EU is the study from Laborde (2011). The emission values resulting from this study
are compared with results from the current study in the figure below.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 94

 250
Emission factor (gCO2/MJ)
200

150

100

50

Laborde et al., 2011 This study

Figure 20: Comparison of results from Laborde et al. (2011) with central estimates from this study

The following conclusion can be drawn based on this comparison:

1 There is a consensus that vegetable oil feedstocks have higher associated LUC emissions
compared with cereals or sugar feedstocks. This is due to the following reasons:
 Vegetable oils are more directly connected to expansion of palm oil in Southeast Asia;
 Increases in vegetable oil transformation do not proportionally increase the production of
protein meals (extra supply comes in part from more crushing but also from vegetable oil
diversion), whereas increase in cereals transformation into ethanol generates proportional
increase in DDGS.

2 There are strong similarities on the results of some feedstocks:

 Sugar beet, sugar cane central values are between 7 and 17 gCO2/MJ for both studies;
 Cereal central values (wheat and maize) are between 10 and 34 gCO2/MJ for both studies;
in both studies, LUC from wheat is higher than LUC from maize due to lower energy
efficiency yield. In the case of GLOBIOM, DDGS protein content and forest reversion
patterns also explain some of the differences – see next paragraph;
 Rapeseed and sunflower oils are between 53 and 65 gCO2/MJ in both studies

3 Some significant differences appear on two vegetable oil feedstocks, palm oil and soybean oil,
which are associated much higher emission factors in our study. This is explained by the lower
elasticity of substitution assumed for our central case in the EU28 following the dedicated
improvement (Annex II.9). As a consequence, demand for palm and soybean based oils leads to
more import of these type of oils in our study than in Laborde (2011). Additionally, it should be
kept in mind when comparing results with other sources from the US literature that the EU
soybean oil can have a very dfferent impact from the US one, because the former is imported,
for a part from Latin America, whereas the US benefits from a large domestic production. This
being said, it should be reminded that emission factors for soybeans were also found to be large
in the US studies, in particular the one from CARB in 2009, which obtained LUC emissions of 62
gCO2e on a 30 year basis, equating to 93 gCO2e on a 20 year basis.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 95

4 Strong similarities exist in what sources contribute to emissions when looking at natural
vegetation biomass, organic carbon in mineral soil and peatland emissions. However, emissions
sources are accounted for in our study with inclusion of reversion carbon and agriculture
biomass. This contributes to some differences, for instance in the case of wheat, where
reversion has a relatively high contribution.

5 Our central EU2020 scenario (97 gCO2e/MJ) results differ significantly from the results in Laborde
(2011), who finds 38 gCO2e/MJ for the biofuel mix. There are several reasons for this:
 Some feedstocks have higher emission factors in the current study than in Laborde (2011);
 More palm oil and soybean oil are used in our scenarios than in Laborde (2011);
 No sugar cane is used in our scenario, whereas Laborde (2011) uses it as a large source to
fill the full mandate;
 Emissions factors are non-linear in GLOBIOM response and the EU2020 emissions increase
with the size of the mandate. The weighted sum of single feedstock LUC for the EU2020 mix
leads to 88 gCO2/MJ, instead of 97 gCO2/MJ, because two times more deforestation occurs
in the EU2020 scenario than would have been anticipated based on the marginal 1%
scenario as EU abandoned land is being used (see Table 12).

Table 12: Comparison of EU 2020 land use change and ILUC effect with weighted sum of feedstock specific scenarios

Land use change compared to baseline (Mha)

LUC
Short
Natural Abandoned emissions
Scenario Cropland Grassland rotation Forest
land land (gCO2/MJ)
plantation
EU2020 Mix
8,038 -1,380 -4,381 -1,906 761 -1,133 97
results
Feedstoc specific
results (weighted
7,856 -1,714 -4,226 -1,954 657 -619 88
avg. with EU2020
shares)

6 Much larger uncertainty ranges are identified in our study than in Laborde (2011). In particular,
some possible negative emission factors are found for cereals, due to the role of co-products,
and some ambivalent impacts appear for sugar cane (higher values are possible if leakage
occurs to Amazon), soybean oil (high uncertainty on the expansion pattern on deforestation
frontier) or palm oil (possibly low emission factors if low expansion in peat and forest).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 96

5 References

Alexandratos and Bruinsma (2012) World agriculture towards 2030/2050, the 2012 Revision [Online]
Available from: http://bit.ly/1lIkiWx [Accessed 4 February 2015]

Al-Riffai, P.; Dimaranan, B. & Laborde, D. (2010), 'Global Trade and Environmental Impact Study of
the EU Biofuels Mandate', Final Report for the Directorate General for Trade of the European
Commission, International Food Policy Research Institute. [Online] Available from:
http://bit.ly/1Cvlw4O [Accessed 4 February 2015]

Britz, W. & Hertel, T. W. (2011), 'Impacts of EU biofuels directives on global markets and EU
environmental quality: An integrated PE, global CGE analysis', Agriculture, Ecosystems &
Environment 142(1--2), 102 - 109.

CARB (2009), 'California's Low Carbon Fuel Standard Final Statement Of Reasons', Technical report,
California Air Resource Board.

De Cara S, Goussebaile A, Grateau R, Levert F, Quemener J, Vermont B (2012). Revue critique des
études évaluant l'effet des changements d'affectation des sols sur les bilans environnementaux des
biocarburants (Critical review of studies assessing the impact of LUC on the environmental balance of
biofuels).

EurObserv’ER, Biofuels Barometers [online] Available from: http://www.eurobserv-er.org/?s=biofuels

European Commission (2013) EU Energy, Transport and GHG Emissions

Trends to 2050 [Online] Available from: http://bit.ly/1C0kpYc [Accessed 4 February 2015]

Eurostat (2014) Farm structure statistics [Online] Available from: http://bit.ly/1zPsrE7 [Accessed 5
February 2015]

FAO (2010) Global Forest Resources Assessment 2010 [Online] Available from: http://bit.ly/1c6wbEA
[Accessed 5 February 2015]

Frank, S.; Böttcher, H.; Havlík, P.; Valin, H.; Mosnier, A.; Obersteiner, M.; Schmid, E. & Elbersen, B.
(2013), 'How effective are the sustainability criteria accompanying the European Union 2020 biofuel
targets?', GCB Bioenergy 5(3), 306--314.

Havlík, P.; Valin, H.; Herrero, M.; Obersteiner, M.; Schmid, E.; Rufino, M. C.; Mosnier, A.; Thornton,
P. K.; Böttcher, H.; Conant, R. T.; Frank, S.; Fritz, S.; Fuss, S.; Kraxner, F. & Notenbaert, A. (2014),
'Climate change mitigation through livestock system transitions', Proceedings of the National
Academy of Sciences 111(10), 3709--3714.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 97

Havlík, P.; Schneider, U. A.; Schmid, E.; Böttcher, H.; Fritz, S.; Skalský, R.; Aoki, K.; Cara, S. D.;
Kindermann, G.; Kraxner, F.; Leduc, S.; McCallum, I.; Mosnier, A.; Sauer, T. & Obersteiner, M.
(2011), 'Global land-use implications of first and second generation biofuel targets', Energy Policy
39(10), 5690-5702.

IEA (2010), World Energy Outlook 2010, OECD/IEA.

IEA (2013), World Energy Outlook 2013, OECD/IEA.

IIASA (2015) SSP Database, version 0.9.3 [Online] Available from: http://bit.ly/1F4ZmVe [Accessed
4 February 2015]

IIASA (2011) Animal Change - Theme 2: food, agriculture and fisheries, and biotechnologies [Online]
Available from: http://bit.ly/1vusW6t [Accessed 5 February 2015]

IPCC (2006), Volume 4: Agriculture, Forestry and Other Land Use, in S. Eggleston; L. Buendia; K.
Miwa; T. Ngara & K. Tanabe, ed., 'IPCC guidelines for national greenhouse gas inventories', Institute
for Global Environmental Strategies, Hayama, Japan, .

IUCN and UNEP (2014) World Database on Protected Areas (WDPA) [Online] Available from
http://bit.ly/1zCttkZ [Accessed 5 February 2015]

Keeney, R. & Hertel, T. W. (2009), 'The Indirect Land Use Impacts of United States Biofuel Policies:
The Importance of Acreage, Yield, and Bilateral Trade Responses', American Journal of Agricultural
Economics 91(4), 895--909.

Laborde, D. (2011), 'Assessing the Land Use Change Consequences of European Biofuel Policies',
Final report, International Food Policy Research Institute, ATLASS Consortium, October 2011.

Lotze-Campen et al. (2014) Impacts of increased bioenergy demand on global food markets: an
AgMIP economic model intercomparison Agricultural Economics 45 (2014) 103–116

Mosnier, A.; Havlík, P.; Obersteiner, M.; Aoki, K.; Schmid, E.; Fritz, S.; McCallum, I. & Leduc, S.
(2012), 'Modeling Impact of Development Trajectories and a Global Agreement on Reducing
Emissions from Deforestation on Congo Basin Forests by 2030', Environmental and Resource
Economics, 1-21.

Plevin, R. J.; O'Hare, M.; Jones, A. D.; Torn, M. S. & Gibbs, H. K. (2010), 'Greenhouse Gas
Emissions from Biofuels Indirect Land Use Change Are Uncertain but May Be Much Greater than
Previously Estimated', Environmental Science & Technology 44(21), 8015-8021.

Ruesch, A. S. & Gibbs, H. K. (2008), 'New IPCC Tier-1 Global Biomass Carbon Map For the Year
2000', Technical report, Oak Ridge National Laboratory, Oak Ridge, Tennessee., Available at
http://cdiac.ornl.gov.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 98

Searchinger, T.; Heimlich, R.; Houghton, R. A.; Dong, F.; Elobeid, A.; Fabiosa, J.; Tokgoz, S.; Hayes,
D. J. & Yu, T.-H. (2008), 'Use of U.S. Croplands for Biofuels Increases Greenhouse Gases Through
Emissions from Land-Use Change', Science 319(5867), 1238-1240.

Taheripour, F. & Tyner, W. E. (2013), 'Biofuels and Land Use Change: Applying Recent Evidence to
Model Estimates', Applied Sciences 3(1), 14--38.

Union zur Förderung von Oel- und Proteinpflanzen e.V (2014), Rohstoffbasis der Biodieselanteile in
Dieselkraftstoffen.

USDA (2014), Foreign Agricultural Service GAIN report, EU Biofuels Annual 2014 (NL4025)

US EPA (2010), 'Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis'(EPA-420-R-
10-006), Technical report, US Environmental Protection Agency, Office of Transportation and Air
Quality, Assessment and Standards Division.

Valin, H.; Sands, R. D.; van der Mensbrugghe, D.; Nelson, G. C.; Ahammad, H.; Blanc, E.; Bodirsky,
B.; Fujimori, S.; Hasegawa, T.; Havlík, P.; Heyhoe, E.; Kyle, P.; Manson-D'Croz, D.; Paltsev, S.;
Rolinski, S.; Tabeau, A.; van Meijl, H.; von Lampe, M. & Willenbockel, D. (2014), 'The Future of Food
Demand: Understanding Differences in Global Economic Models', Agricultural Economics 45(1), 51-
67.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 99

6 Glossary

This Glossary contains brief descriptions of terms used in the study report. More technical terms used
in the Annexes are not included here.

Abandoned land: land reported as abandoned in GLOBIOM results corresponds to land used as
cropland or grassland in the year 2000, but no longer used in the baseline due to decrease in
harvested areas. Land already abandoned in 2000 is represented in the other natural land category.
For more details, see appendix II.11.

Behavioural parameters: parameters that are independent from the model used

C-growth rate: rate of accumulation of carbon in living biomass above and below ground due to
vegetation regrowth. See appendix II.6.

Cropland: we designate by cropland the cultivated areas for annual crops and permanent crops such
as palm plantations or semi-perenials (sugar cane, miscanthus, etc.). An important exception is short
rotation coppice that are cultivated for energy purpose and accounted as energy plantation. Because
GLOBIOM only model 18 crops globally, it should be kept in mind that GLOBIOM only represents a
part of the cropland (about 70%).

Cultivated area: physical area on which one or several crops can be cropped in one year,
subsequently (multi-cropping) or simultaneously (inter-cropping).

Equilibrium model: A macro-economic model in which supply and demand sides of certain sectors
in the economy are represented, with supply and demand being equal at a certain price level.

Endogenous: the value of a parameter is determined by other parameters in the model. E.g. Food
demand is endogenous in GLOBIOM and depends on population size, gross domestic product (GDP)
and product prices.

Exogenous: the value of a parameter comes from a dataset and is fixed, not impacted by the
model.

Foregone sequestration: carbon sequestration that occurs in the baseline, but not in the scenario.

Gaussian Curve: a bell shaped distribution typical for input parameters and results.

General Equilibrium Model: An equilibrium model (see above) that considers the whole global
economy.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 100

Grassland: grassland (or pasture) is used in this report to designate áreas occupied a part of the
year by grawing animals. In our approach, grassland accounts for 1.6 Gha and therefore does not
comprehend some open áreas classified as cropland in the FAO definition.

Harvested area: Total area of land that has been harvested in one year. In case of multi-cropping,
harvested area is greater than cultivated area.

Other natural land: land not classified as cropland, grassland or forest in the initial land cover data
of GLOBIOM. This includes abandoned agricultural land in the dataset (i.e. in year 2000), but this
does not include land voluntarily set-aside (included in cropland). Land abandoned in the model
projections (2000-2030) is accounted separately and not mixed with other natural land. For more
details, see appendix II.11.

Partial Equilibrium Model: An equilibrium model that focuses on specific sectors of the economy,
with more detail then general equilibrium models.

Monte Carlo Analysis: uncertainty analysis, through random testing of a large range of input
values.

Substitution effects: the effect that co-products of a biofuel conversion process can substitute
other commodities elsewhere in the global economy.

World prices: in this study world prices for agricultural commodity corresponds to world consumer
prices, and not to world market prices. World consumer prices are calculated as consumption-
weighted prices, which covers parts of the consumption whose prices are not directly correlated to
world market prices.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 101

Annex I Description of GLOBIOM and comparison
with MIRAGE-BioF (IFPRI)

This annex provides a detailed comparison of main features of the GLOBIOM and compares the
model with the MIRAGE-BioF model used by IFPRI for their study to LUC emissions associated with
EU, their respective strengths and limitations, written from the perspective of the GLOBIOM model.
The version of GLOBIOM used for the comparison is the model as it stands at the end of this
project.31 We compare this with the version of the MIRAGE-BioF model as used in the IFPRI 2011
study for the European Commission.32 The purpose of this comparison is not to argue that one of the
two models is better than the other, but merely to give an insight into how GLOBIOM works, partly
by comparing it with MIRAGE-BioF.

The main features of both models are presented, focusing on data and mechanisms that play an
important role in the assessment of LUC impacts of biofuels. The following aspects are discussed in
detail: the representation of bioenergy processing chains and their co-products, the dynamics of LUC
and the response of agricultural yield or food demand to change in domestic and international
market prices. These aspects are taken into consideration when presenting GLOBIOM and comparing
it to MIRAGE-BioF in this annex (Figure 21).

31
Note that this version features a detailed representation of the European Union and may differ from some earlier standard versions used
for global level assessments. The description presented in this appendix also features the improvements developed in this project and is
updated compared to the report “Description of the GLOBIOM (IIASA) model and comparison with the MIRAGE-BioF (IFPRI) model”
published in October 2013 on www.globiom-iluc.eu.
32
The model may have been changed in the course of other projects in the meanwhile but no documentation was accessible on these
changes at the time of writing this document.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 102

Figure 21: Main steps in the biofuel supply chain, how they relate to the model description in this report and what

their associated LUC issues are

The following questions will be addressed in this annex:

1 How does the model represent the important elements of the ILUC debate?
2 What advantages can GLOBIOM bring?
3 What are the current shortcomings and how can they be addressed?

Models are complex and the use of technical terms is inevitable at times. This annex aims to clearly
describe the GLOBIOM model, outline the main differences between this and MIRAGE-BioF as well as
note any implications to interested stakeholders. Experienced readers can find more technical
information in the final sections I.8 and I.9.

I.1 Summary of differences between GLOBIOM and MIRAGE-BioF

GLOBIOM and MIRAGE-BioF belong to two different families of economic models. None of the two
models is a priori superior to the other, but depending on the topic addressed, some characteristics
can be important. GLOBIOM is a model designed to address various land use related topics
(bioenergy policy impacts, deforestation dynamics, climate change adaptation and mitigation from
agriculture, long term agricultural prospect). MIRAGE-BioF, besides its use for biofuel policies, is
generally used to assess trade policy impacts and impacts of agricultural policies on income and
poverty in developing countries. The main differences between GLOBIOM and MIRAGE-BioF are
summarised in Table 13. More technical descriptions can be found in Section I.9.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 103

Table 13: Main differences between GLOBIOM and MIRAGE-BioF
GLOBIOM* MIRAGE-BioF*

Bottom-up, starts from land and Macroeconomic, starts from national

Model framework
technology accounting relations

Detailed focus on agriculture

All economic sectors represented with
(including livestock), forestry and
Sector coverage agricultural sector disaggregated
bioenergy
(General equilibrium)
(Partial equilibrium)

Global Global
Regional coverage
(28 EU Member states + 29 regions) (1 EU region + 10 world regions)

Resolution on Detailed grid-cell level Regional level, with land split into up
production side (>10,000 units worldwide) to 18 agro-ecological zones
Time frame 2000-2030 (ten year time step) 2004-2020 (one year time step)
GTAP economic accounts, harmonized
Market data source EUROSTAT and FAOSTAT
with FAOSTAT
Factor of production More detailed on natural resources More detailed on economic resources
explicitly modelled (land, water) (labour, capital, land)
Aggregated representation.
Geographically explicit. Substitution of land use at regional
Land use change Land conversion possibilities allocated and agro-ecological zone level.
mechanisms to grid-cells taking into account Allocation of agriculture and forest
suitability, protected areas. land expansion across other land
covers using historical patterns
Detailed biophysical model estimates Input-output coefficient from GTAP
for agriculture and forestry with database or national statistics at
Representation of
several management systems regional level.
technology
Literature reviews for biofuel Literature reviews for biofuel
processing processing
One representative agent per region
One representative consumer per
Demand side adjusting its consumption between
region and per good, reacting to the
representation different goods depending on prices
price of this good.
and level of income
12 sources of GHG emissions covering
Only land use change emissions.
crop cultivation, livestock, land use
Deforestation and soil organic carbon
change, soil organic carbon based on
calculated with default IPCC emissions
advanced accounting framework.
GHG accounting factors.
Peatland IPCC emissions values
Peatland IPCC emission values revised
revised upward based on exhaustive
upward based on Edwards et al.
recent literature review (see Appendix
(2010).
II.3).

* GLOBIOM version with disaggregated EU as at the start of this project. MIRAGE-BioF as in Laborde (2011).

As a model specialised in land use based activities, GLOBIOM benefits from a more detailed sectoral
coverage, backed by a solid representation of production technologies and a geographically explicit
representation of land use33 and associated greenhouse gas emission flows (see Figure 22).
GLOBIOM is a partial equilibrium model, meaning that the only economic sectors represented in
detail are agriculture (including livestock), forestry and bioenergy. In MIRAGE-BioF, all sectors of the
economy are represented but with a more limited level of detail on the supply side representation
due to the use of socioeconomic accounting matrices.

33
By geographically explicit, we refer to the fact that the model makes allocation of production based on precise geographical data on land
characteristics (> 10,000 spatial units).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 104

Many of the modelling issues raised during the previous LUC assessment can be more easily and
accurately addressed in GLOBIOM. Some GLOBIOM characteristics that differ from MIRAGE-BioF
include:
 A more precise representation of LUC dynamics
 The robustness of biophysical relations for production, conversion and substitution processes
 The level of detail in the description of available technologies
 The representation of non-linear responses on land (for instance, fallow land can be used, but
only up to a certain maximum level).

Figure 22: Overview of the GLOBIOM model structure

While GLOBIOM can address many modelling limitations raised in the ILUC debate, a limitation of
GLOBIOM is the fact that it does not have some of the mechanisms that are present in MIRAGE-BioF,
such as the macroeconomic effect of bioenergy policy on the fuel market or population income.
These effects are discussed in more detail in Sections I.8 and I.9. We argue that these should remain
of second order when compared with drivers of indirect land use change.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 105

Beyond differences in modelling frameworks and their capacity to describe mechanism at play, a
main challenge in ILUC modelling remains the treatment of uncertainties on parameters that are
independent from the model used (behavioural parameters, emission factors, future technologies).
These uncertainties receive particular attention in this study. A range of confidence is established for
emission factors on the basis of literature review and other available information. Sensitivity analyses
are performed on the behavioural parameters considered as the most critical for the final results (for
instance, yield response or land conversion costs).

I.2 Representation of agriculture and yield development

As a model specialised on land use issues, GLOBIOM benefits from a greater level of detail in its
representation of agriculture, with a larger number of crops and livestock systems represented than
in MIRAGE-BioF. This increases the number of biofuel feedstocks that can be modelled, and allows for
a more precise description of crop and livestock interaction, including co-products utilization. Like in
MIRAGE-BioF, yields are sensitive to prices and farmers can intensify their production in response to
market signals.

I.2.1 Crops

GLOBIOM represents 18 crops globally and 27 crops for the European Union. The full list of crops
covered is detailed in Section I.9. Harvested areas are based on FAOSTAT statistics but are spatially
allocated using data from the Spatial Production Allocation Model (SPAM).34 In the case of the EU,
crops are allocated across NUTS2 regions using data from EUROSTAT. This setting provides a very
detailed framework compared to the previous modelling with MIRAGE-BioF, where Europe was
represented as a single region and the number of crops more limited. MIRAGE-BioF relies on a
modified version of the GTAP 7 database35 that only contains 8 crop aggregates. IFPRI extended the
number of these crops to 11 by disaggregating oilseeds and singling out corn (see full list in Section
I.9).

The aggregated approach of MIRAGE-BioF is too coarse to trace all single crop substitutions, but
allows a mapping of the total global harvested area. In GLOBIOM, the crop level approach is more
precise but, as all crops are not represented, a small fraction of harvested areas is not explicitly
modelled. Cultivated area currently represents in GLOBIOM around 84% of the total harvested area
in the world. Harvested area for the non-covered crops is kept constant.36 Global harvested area
amounts to 78% of land classified by FAO as “Arable land and permanent crop” category, which
shows the importance of abandoned land, idle land and temporary meadows in the definition of this
category. The advantage of the GLOBIOM approach is that this “not harvested” arable land is also

34
See You and Wood (2008) and http://mapspam.info/
35
The Global Trade Analysis Project (GTAP) database is a large database describing the world economy and compiled using national
statistics and global trade datasets. This database is formatted to satisfy certain properties of consistency on economic accounts in order to
be used by computable general equilibrium models, a class of macro-economic models to which MIRAGE-BioF belongs. For more details see
Narayanan et al. (2012) and the GTAP website www.gtap.org .
36
The five most harvested crops in FAOSTAT nomenclature subject to this assumption in GLOBIOM are in decreasing order: other fresh
vegetable, coconuts, olive, coffee, natural rubber.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 106

explicitly represented in the model. The standard assumption for model projections is to keep this
area constant.

However, as explained in Section II.11 in Annex II, the representation in GLOBIOM of expansion into
abandoned land has been improved during the study project, in order to represent more complex
dynamics (for instance, decrease in fallow land).

In GLOBIOM, yields for all locations and crops are determined in a geographically explicit framework
by the Environmental Policy Integrated Climate Model (EPIC). The yields are distinguished by crop
management system and land characteristics by spatial unit.37 They are however rescaled by a same
factor to fit FAOSTAT average yield at the regional level, in order to catch other managements
parameters not supplied to EPIC or other cause of yield mismatch. This approach with differentiated
yields is different from the one in MIRAGE-BioF that assumes a homogenous yield within a region
and agro-ecological zone.

Different crop management systems are distinguished in GLOBIOM. At the world level, four
technologies can be used (subsistence, low input rainfed, high input rainfed and high input irrigated).
In Europe, a larger set of options is available, with two different levels of fertilizer input, two levels of
irrigation, and three different levels of tillage. EPIC has additionally been run for a large combination
of different rotation systems for all NUTS2 regions.38 This therefore allowed a more precise
simulation of the yield achieved through optimisation of rotations, a practice well observed in Europe.
Input requirements for each system and location are determined by EPIC (quantity of nitrogen and
phosphorus, irrigated water). At the base year, production cost for these systems (i.e. all input costs
plus the farmer margins) are calibrated using FAOSTAT producer price data at the national level,
assuming that all units within the country supply the market at this price.
The representation of crop production is therefore much more detailed than the one used in MIRAGE-
BioF, which is also consistent with FAOSTAT but has a more aggregated description at the regional
level for output and at the agro-ecological level (AEZ) for land distribution.39 The production
structure in MIRAGE-BioF relies on a single aggregate production function at regional level,
describing how output can be obtained from various production factors and intermediate
consumption interactions (see Box 3 below). The description of the link between output and land is
therefore not based on any biophysical model and relies on a simplified relation of substitution
between inputs. A specific treatment of fertilizer input has however been added to better represent
the saturation effect of yield in case of excessive addition of fertilizer.40

Additionally to production of grains or fibres, GLOBIOM also represents the production of straw for
some of the major crops (barley, wheat) and corn stover. Only a part of the residues produced is
considered available because of the role of residues for soil fertilisation. The residues removed are

37
EPIC is run over a large number of spatial units covering the global land cover (over 200,000) that are then aggregated for model runs
into around 10,000 larger units. See for more details.
38
NUTS (Nomenclature of Units for Territorial Statistics) is the standardized format for administrative divisions in the European Union. The
level 2 of NUTS (NUTS2) corresponds to 271 regions in Europe.
For more information see http://epp.eurostat.ec.europa.eu/portal/page/portal/nuts_nomenclature/introduction
39
More information on the GTAP-AEZ framework can be found in Lee et al. (2007). This framework relies on aggregation of input data from
Ramankutty et al. (2008) database.
40
See appendix 4 of Al Riffai et al. (2010).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 107

used for the livestock sector and the industrial and energy uses. The impact of using agricultural
residues as biofuel feedstock has been more precisely looked at in a specific improvement to the
model (Annex II.1).

Several rates of residue removal are now considered and the effect of changing this rate on yield and
carbon sequestration is now analysed using the EPIC model. Agricultural residues were not
represented in the version of MIRAGE-BioF used in Laborde (2011).

I.2.2 Livestock

GLOBIOM is one of the most refined global models in its representation of the livestock sector. It
includes in its dataset all relevant information from the Gridded Livestock of the World database41
and represents eight animal types spatially distributed, and producing seven animal products (see
for the list of animal and products). This allows for a more precise representation of the links
between livestock production, feed requirements and the link to land through grazing needs.
Livestock productivity for ruminants (buffalos, cows, sheep, goats) is estimated in GLOBIOM on the
basis of animal feed ration using RUMINANT, a digestibility model.42 The use of this model ensures
consistency between the livestock sector input (grass, grains, stover, etc.) and output under different
management systems.43 For monogastric animals (pigs, poultry) the same consistency has been
achieved using the results of a literature review to identify feed conversion efficiencies under two
management systems (industrial and smallholder). Production costs for these systems are all based
on FAOSTAT producer prices for product output and for grains input.

Grazing needs of ruminants depend on the rearing management system. For instance, cattle under
mixed temperate systems spend a longer period of the year in stables and have lower grazing needs
than cattle under extensive grazing management. A grassland map indicating levels of biomass
production in the different regions is used to determine possible stocking densities of animals. The
link between animals and land is therefore fully consistent, allowing the need for additional land in
response to changes in the livestock sector to be traced.

This level of detail and consistency is an important asset when compared with the more simplified
representation of the livestock sector in MIRAGE-BioF, where only two types of animals are
distinguished: cattle and other animals (derived from the three sectors present in GTAP: cattle dairy,
cattle other and other animals). A caveat of the GTAP approach is that animal numbers are not
explicitly represented which makes the calculation of the feed requirement more complex.44 Feed
intake and conversion efficiencies are derived from the input/output relations observed in the
economic statistics of the sector as a whole. Each sector is only represented through one aggregated
production function, similar to the approach for crops.

41
See Wint and Robinson (2007)
42
See Herrero et al. (2013)
43
Eight production systems are used that are based on the classification from Seré and Steinfeld (1996)
44
Animals in GTAP are assimilated to capital for these sectors.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 108

I.2.3 Dedicated energy crops

In addition to the crops mentioned in Section I.2.1, GLOBIOM also contains yield information from
EPIC to simulate deployment of dedicated energy crops, such as switchgrass and miscanthus.
Because these crops are not cultivated at large scale in the base year, only their production potential
is represented in the model. The use of the EPIC model to estimate the biophysical characteristics of
the crops provides information on the suitability of land in different locations, as well as the fertilizer
and water requirements.

Woody biomass can also be supplied on agricultural land using short rotation coppice. In the current
version of the model, all short rotation woody biomass production is described through a single
sector of short rotation plantation (Section I.3.1) that can be deployed on agricultural land or on
other types of land.45 These dedicated energy crops and woody biomass sectors in agriculture are not
represented in MIRAGE-BioF.

I.2.4 Yield responses and intensification

The response of agricultural yield to market signals has been an important point of debate in the
assessment of ILUC.46 In both GLOBIOM and MIRAGE-BioF assumptions are made on technological
changes that allow yields to increase over time independently from other economic assumptions
(e.g. breeding, introduction of new varieties, technology diffusion, etc.) In addition both models
represent yield responses to prices, although in a different way.

In GLOBIOM, crops and livestock have different management systems with their own productivity
and cost. The distribution of crops, animals and their management types across spatial units
determines the average yield at the regional level. Developed regions rely for most of their
production on high input farming systems whereas developing countries have a significant share of
low input systems and even, in the case of smallholders' subsistence farming with no fertilizer at all.
Farmers can adjust their management systems and the production locations following changes in
prices, which impact the average yields in different ways:
 shifts between rainfed management types (subsistence, low input and high input) and change in
rotation practices;47
 investment in irrigated systems. This development is controlled through a simplified
representation of the regional water supply potential;
 change in allocation across spatial units with different suitability (climate and soil conditions).

In MIRAGE-BioF, yield response to prices is described in a much more simple manner due to the
aggregated production function that does not differentiate land suitability or management systems
(see Box 3). When the relative price for land increases, yield can increase too by adding additional
fertilizer, capital and labour. Hence, additional demand can be met while keeping land requirement
constant. As explained in Section I.2.1, the production function has been modified to avoid

45
See Havlik et al. (2011)
46
See for instance Keeney and Hertel (2009) or CARB (2011).
47
Change in tillage practice can also intervene. However, the impact on yield is second order, this management most significant impact on
the level of carbon stocked in the soil.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 109

unrealistic responses of yield in case of strong fertilizer input increase. The yield response however
remains based on a simplified representation without explicit link to the real biophysical potentials.

Box 3: Production functions in MIRAGE-BioF and in GLOBIOM

Production in GLOBIOM and MIRAGE-BioF follow two different settings, due to differing theoretical approaches.
GLOBIOM, as a bottom-up mathematical programming model, relies on a detailed representation of technology
for each sector with different management systems and production locations. Each management option has its
own input requirements, production cost, and production efficiency. For instance, in the case of crops, the level of
fertilizer and water requirements is precisely known depending on the level of intensity of the management (low,
high input, irrigated). The model computes for a given demand, what the most cost-efficient systems are under a
constraint of land availability and cost of resources. At the level of a region, the production pattern is then
obtained by the sum of all production systems and locations used. This representation provides non-linear supply
functions, whose slope patterns directly depend on the distribution of cost-efficiency across management systems
and locations. The advantage of this approach is the explicit link between technological options and the
production potentials. The shape of the supply function, however, cannot be simply inferred ex-ante and requires
simulation experiments to be calculated.

MIRAGE-BioF, as a computable general equilibrium model (CGE), relies on a more aggregated representation of
production, directly calculated at the regional level. Input and production factor requirements (land, capital,
labour) are set for the base year at regional level, as observed in statistics. When prices of these inputs or factors
change, their level of consumption and level of output changes as well, following a simplified formula designed to
capture the aggregated effect directly. For this function, MIRAGE-BioF, like many applied CGEs, relies extensively
on the Constant Elasticity of Substitution form (CES) that defines the easiness of substitution between all factors
(labour, capital, land) from a specific parameter, the elasticity of substitution (see Box 4 for more details on the
CES). MIRAGE-BioF uses this design at the regional level for all its sectors, but relies for several levels on CES
nesting, with different elasticity values that depend on inputs and factors.48 Such stylized representations are very
convenient for macro-economic approaches (trade policies, budgetary policies, etc.) when estimation of the
different level of substitution around an equilibrium point is the main interest. They however lose the link to
underlying technological relations, and generally display a smooth supply profile and lower sensitivity to
biophysical constraints due to input substitution possibilities.

I.3 Representation of woody biofuel feedstocks and forestry

In addition to crop feedstocks, the GLOBIOM model also provides potential for woody biomass
feedstock extraction that can be used for bioelectricity and second generation biofuels. This is based
on a detailed representation of plantation deployment potentials, as well as a refined description of
the forestry sector. This combination of a detailed agriculture and forestry sector in one modelling
framework is a strong asset of GLOBIOM. The description of forestry in MIRAGE-BioF is limited to a
single sector, without biofuel feedstocks, whereas GLOBIOM explicitly models extraction of five
primary wood products and distinguishes between short rotation plantations and managed forests.

48
See Bouet et al. (2010) for a description of most CES nesting in the different production functions.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 110

I.3.1 Short rotation plantations

Besides energy crop, woody biomass can be supplied in GLOBIOM through short-rotation plantations,
a sector that covers very short rotation periods (short rotation coppice i. e. 2 to 5 years) but also
longer rotation periods (short rotation forestry, closer to 10 years).49

Suitable areas for this sector are determined by using a geographic information system (GIS) that
analyses temperature, precipitation (rain), altitude, and population density. The productivity of
plantations is based on estimates from the Potsdam Net Primary Productivity50 Model Inter-
comparison, and production costs are calculated based on literature sources.51 Several deployment
potentials can be considered depending on the assumption used for plantation type (cropland,
grassland, other natural vegetation). These data are also used to update the model with the amount
of carbon that is sequestered.

I.3.2 Woody biomass from managed forests

GLOBIOM relies on information from the global forestry model G4M52 for its representation of forestry
productivity. Locations of forests are supplied to GLOBIOM at a half degree resolution (see Figure
22). Harvest potentials of stemwood are determined based on net primary productivity (NPP) maps
and combined with maps of forest biomass stock such as the Global Forest Resources Assessment
provided by FAO.

The information on forestry harvest potential from G4M allows four main primary woody resources to
be represented in GLOBIOM: industrial roundwood, non-commercial roundwood, harvest losses and
branches and stumps. Harvesting costs include logging and timber extraction and depend on
harvesting equipment, labour costs and terrain conditions. Primary resources, once extracted, are
separated into five primary woody products: sawn wood biomass, pulp wood biomass, energy wood
biomass (biofuels, heat and electricity), traditional use biomass (fuel, cooking) directly collected in
the forest (no processing chain) and other non-energetic use biomass. Primary forest residues are
included (branches and stumps) and can be used for second generation biofuels, electricity and
heating. All harvested primary woody products are sent to processing activities which can lead to
other types of bioenergy feedstocks (secondary residues such as saw dust and cutter shavings, black
liquor, bark). MIRAGE-BioF only contains one aggregated forestry sector (see Box 3) which does not
supply feedstock for biofuels and contains no specific information on forest biomass productivity.

49
Weih (2004)
50
Net primary productivity is the measure of the net carbon flow from the atmosphere to the terrestrial biomass, ie the amount of biomass
that is growing in a given period of time, a year in our approach. See Cramer et al., 1999.
51
See Havlik et al (2011) for full details.
52
See Kindermann et al. (2008).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 111

I.4 Overview of feedstock processing and biofuel production

GLOBIOM expands the number of feedstocks and processing pathways that have been explored so
far with MIRAGE-BioF. It includes second generation technologies and offers a flexible framework
that can be further developed to describe additional biofuel pathways, present or future, with their
expected production costs and conversion coefficients.

I.4.1 Sector coverage and role of supply chain

At the level of primary sectors, GLOBIOM represents, in total, 27 crops, 7 animal products and 5
primary wood products (see details in ). These products can then be directly sent to markets to
satisfy the demand of households and various industries and services (food industry, seeds, cosmetic
industry, etc. which are not explicitly represented in the model53). Part of the commodities can also
be used as animal feed in the livestock sector, which is the case for a significant share of many
crops. Some other products are transformed explicitly in the model into intermediate or final
products, before being sent to the market. This is the case for oilseeds, wood primary products and
products used as bioenergy feedstocks. For these products, all processing industries are explicitly
represented in the model, with their transformation coefficients, their co-products and processing
costs. The role of processing industries in the supply chain is illustrated in Figure 23.
The representation of market flows in GLOBIOM is based on information from FAOSTAT that provides
details on the quantities of biomass which is processed, directly purchased by final consumers, used
as animal feed, or allocated to seeds or other industrial users. The accounting of this distribution
across potential users is important to assess the competition between food, energy and other uses.

Figure 23: Supply chain in GLOBIOM and role of processing industries

53
Industrial uses are captured in the FAOSTAT database in the category “Other uses” of the Supply Utilisation Accounts.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 112

In comparison, MIRAGE-BioF relies on a more comprehensive representation of economic flows,
because it contains a complete mapping of all economic sectors with their demand for raw products.
For instance, the food industry is represented explicitly in the model, as well as the chemical
industry. However, the number of primary sectors is more limited, with only 11 crops, two animal
products, and one forestry product. The volume of input in each sectors relies on aggregated regional
accounts in monetary units, less precise than in the FAOSTAT utilization accounts expressed in
quantities. For crops of interest for biofuels however, some bottom-up data reconstructions were
performed in MIRAGE-BioF to refine the initial data and make them more consistent with FAO
statistics.

Because the supply chain is long and complex and commodities are aggregated according to their
economic value, it is often difficult to trace the flow of the raw commodities ‘from field to plate’. For
instance, the raw output of an aggregated sector such as fruit and vegetables can be purchased by
sectors as diverse as (in decreasing order) food processing (other), vegetable and fruits54, beverage
and tobacco, trade margins, textile, other general services, vegetable oil, chemical rubber plastic,
etc.55 The outputs of these sectors can in turn be purchased by many other sectors.

I.5 Processing activities and bioenergy pathways

The main processing industries currently represented in GLOBIOM are the oilseed crushing industry,
forestry industry, and a certain number of bioenergy industries.
Table 14 provides a detailed overview of processing activities and indicates if they are represented in
MIRAGE-BioF. Conversion coefficients from input to final products, quantities of co-products
generated and processing costs of pathways are currently sourced from literature. They can be
updated if better information would become available.

Table 14: List of current processing activities in GLOBIOM and availability in MIRAGE

Processing Output IFPRI-

Input product GLOBIOM*
category product MIRAGE*

Oilseed crushing

Rape oil
Rapeseed crushing Rapeseed  
Rape meal
Sunflower oil
Sunflower crushing Sunflower Sunflower  
meal
Soybean oil
Soybean crushing Soybeans  
Soybean meal
Palm fruit Palm oil
Palm fruit ** 
processing Palm fruit fiber
Wood processing

54
Auto-consumption in the GTAP database is frequent and is simply the result of aggregation across sectors having input-output flows.
55
Based on the GTAP8 database for the year 2007 at world level. All sectors listed purchase more than 2% of the output consumed as
intermediate consumption (40% of the total production).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 113

 Processing Output IFPRI-
Input product GLOBIOM*
category product MIRAGE*
Sawn wood
Sawn wood Saw dust
Sawmill 
biomass Saw chips
Bark
Pulp wood Mechanical
Mechanical pulping biomass pulp 
Saw chips Bark
Pulp wood Chemical pulp
Chemical pulping biomass Black Liquor 
Saw chips Bark
Plywood
Sawn wood Sawdust
Plywood production 
biomass Saw chips
Bark
Pulp wood
Fiberboard biomass
Fiberboards 
production Saw chips
Sawdust
Bioenergy
Energy biomass
Sawdust
Electricity
Combustion Saw chips 
Heat
Black Liquor
Bark
Traditional
Cooking Stove energy 
biomass
Ethanol
Biofuel corn based Corn  
DDGS
Ethanol
Biofuel wheat based Wheat  
DDGS
Biofuel sugar based Sugar cane Ethanol  
Ethanol
Sugar beet  
Sugar pulp
Biodiesel
Biofuel FAME Vegetable oil  
(FAME)
Woody biomass
Cellulosic ethanol Grassy crops Ethanol 
Cereal straw
Fischer-Tropsch Woody biomass Biodiesel

biodiesel Grassy crops (drop-in)
Biogas fermentation Corn silage Biogas 
* GLOBIOM version with disaggregated EU as used for this project. MIRAGE-BioF as in Laborde (2011).
** Palm fruit fibers are not represented in the current version of GLOBIOM.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 114

The flexibility of GLOBIOM with respect to modelling supply chains can be used to improve the
processing description at the level of detail deemed relevant for an accurate assessment. It is
possible for instance to:
 Disaggregate the pathway to represent more precisely the underlying technologies currently
used;
 Refine the type of inputs and co-products associated to a processing pathway (e.g. the use of
ethanol or methanol during the transesterification for biodiesel production and production of
glycerol);
 Account for the quality of co-products generated depending on the supply chain, for example the
protein content of dried distiller grain solubles (DDGS).

In Laborde (2011), the number of bioenergy sectors in MIRAGE-BioF is limited to conventional (first
generation) biofuels (see Table 14). Main crushing sectors and bioenergy production sectors have
been carved out in the GTAP database and their processing costs and conversion coefficients are
derived from literature.

However, this process is time-consuming because any change made to the model database requires
a full rebalancing of all economic flows in the model. In a computable general equilibrium model like
MIRAGE-BioF, the total country income, households and industry purchases, must remain consistent
with the national accounts, also after addition of new sectors. For this reason, the number of sectors
that can be added is more limited.

Another issue in MIRAGE-BioF is related to the unit of substitution. A product can only have one rate
of substitution with other products within the same nest. By default, the unit of substitution is the
economic value in the base year, when the model is calibrated. For instance, one dollar of vegetable
and fruit imported in the base year from one region can be replaced by one dollar of vegetable and
fruit from a different region. In the case of MIRAGE-BioF, the most important flows of homogenous
products (wheat, corn, sugar, ethanol, vegetable oil, biodiesel) were reconstructed with the same
prices per tonne to make the substitution equally consistent on a quantity basis. But this remains an
imperfect adjustment because some products are substituted differently in reality, depending on
their final use (for instance, calories or proteins).

GLOBIOM allows for taking into account quality aspects relevant per type of use (food, feed,
bioenergy feedstock). For instance, it is possible to substitute bioenergy feedstock on the basis of
their biofuel yield when used in a bio refinery and at the same time express the quantity in
kilocalories (or other nutrition metric) for the final consumer. Similarly, in the case of feed, the
protein and energy content are both important for the calculation of livestock rations, i.e. the bundle
of feed given to the animals, as we will see in the next section.

Another interesting feature of GLOBIOM is its capability to model discontinuities in substitution

patterns. For instance, it is possible to represent substitutions between several types of biodiesel
sourced from different vegetable oils, but to restrict this substitution to some maximum incorporation
constraint (for instance the amount of soybean oil or palm oil, following quality standards). In that
case, a competitive price plays a role only when substitution is possible. Once the maximum

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 115

incorporation level is reached, the more expensive feedstocks have to be used to satisfy the extra
demand.

I.5.1 Dealing with co-products

The role of co-products of biofuel feedstocks has been discussed intensively in the ILUC debate.
There is consensus about the fact that the cogeneration of products can limit the land footprint of
bioenergy production but evaluations find varying estimates for this effect. The assessment of this
effect is in particular related to the representation of feed intake by the livestock sector.
The feed representation of GLOBIOM provides detailed information on animal requirements. Rations
of animal feed are calculated based on a digestibility model, which ensures consistency between
what animals eat and what they produce, and rations are specific to each management system.
When the price of a crop changes, the price of the feed ration changes as well, causing a change in
profitability of each livestock management system. Switching between management systems allows
for representing changes in the feed composition of the livestock sector.

Oilseed meals are explicitly modelled in GLOBIOM as a part of the rations represented in the
livestock sector. If availability of one type of meal increase (e.g. rape), it can replace another type of
oilseed meal (soybean) or increase the number of animals relying on a higher share of protein
complement in their diet. The substitution of feed is handled under a double constraint of constant
protein and energy requirement. For instance, it is possible to represent the fact that DDGS can be
incorporated in high quantities to substitute some oilseed meals on a protein content basis, but that
beyond a certain level of incorporation, this generates a deficit in energy needs that requires other
feed items to be added in the ration. Other co-products such as corn and wheat DDGS are modelled
in a simpler way and are just considered to replace some crop groups with a substitution ratio that is
determined exogenously (see Annex II.7). Some constraints on the maximum level of incorporation
of co-products in the livestock sector are also represented.

Contrary to GLOBIOM, the representation of feed in MIRAGE-BioF is based on a top-down

decomposition of inputs based on economic statistics and FAOSTAT information and not established
on the basis of a biophysical model. For that reason, feed quantity and composition are not explicitly
linked to production levels for the livestock sectors (and based on aggregated statistics). Grazing
input is not determined on the basis of animal feed needs but on the amount of land classified as
grassland in the model. Therefore, increasing production requires increasing grassland,
independently from the cattle density on land. This can overestimate the response of grassland area
to change in livestock production level.

Co-products are well represented in MIRAGE-BioF and they also substitute in the livestock production
functions associated to feed. Feed substitution is managed at two levels: the first level deals with
substitution of different types of grains and the protein complement aggregate; the second level
disaggregates the protein complement category to represent an easier substitution between oilseed
meals and DDGS. However, the substitution ratio remains determined by the economic value
associated to the different meals, which are highly correlated with protein contents in the case of
protein meals. Although these substitution patterns have been compared and found consistent with
literature, the flexibility to fit a specific substitution patterns is limited by the model design. In

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 116

particular, it is not possible to exactly match a substitution ratio to multiple crops, for instance, one
tonne of wheat DDGS replacing 0.5 tonne of soybean and 0.66 tonne of wheat.56

I.6 Capturing the world markets and the global economy

Both GLOBIOM and MIRAGE-BioF have a full representation of world markets but GLOBIOM trades all
single goods as perfect substitute57 in tonnes, whereas MIRAGE-BioF represents imperfect
substitution between trade flows measured in monetary terms. GLOBIOM is therefore more suitable
to account for replacements between specific goods on international markets, between the sectors
covered. In GLOBIOM the description of the economy is however limited to main land based sectors
(it is a partial equilibrium model whereas MIRAGE-BioF is a general equilibrium model).

MIRAGE-BioF has a greater understanding of interactions between all sectors of the economy but is
coarser for sectors highly relevant to the ILUC debate such as agriculture and forestry. The GLOBIOM
model might miss certain interactions (fuel market feedback, income impact), a caveat that can be
addressed by calculating separately the magnitude of these effects.

I.6.1 International markets and trade

Both GLOBIOM and MIRAGE-BioF represent international markets for the various products that are
traded between regions. They both rely on international trade statistics for trade flows and tariffs.58
Their trade specifications however differ, as explained in the section below.

Trade in GLOBIOM follows a representation where products are all expressed in tonnes across the 53
economic regions and are considered as identical goods. Products are always sourced from the region
with the least expensive production costs, adjusted by international transportation costs and tariffs.
An increasing cost of trade prevents that all exports are provided by the same region. The advantage
of such an approach is that it allows to trace precisely all substitutions of traded goods on a
quantitative basis. Some patterns of trade creation are also possible, i.e. if increase of population
requires or if a change in production costs makes it more profitable, two countries can start to trade
in the future even if they were not trading partners before. This is not possible in the MIRAGE-BioF
model.

In MIRAGE-BioF, all products are traded based on their economic value in the base year and
consequently all substitution relations are by default measured in base year US dollars. To allow the
substitution of agricultural goods and biofuel feedstocks to be on a 1 to 1 basis in quantitative terms,
trade values are adjusted in the MIRAGE-BioF database. But this rate of substitution is difficult to

56
Estimate from CE Delft in the Ghallager review (2008).
57
A perfect substitution means that an importing country will always decide to import from the country which hasthe lowest cost. This is
different from an imperfect substitution representation where some stickiness in trade flows is assumed, meaning that trade patterns are
not immediately impacted by small changes in price because it takes some effort to switch to a different supplier. Note that in GLOBIOM, as
explained in this section, transportation costs increase with the size of trade flows, which also introduce some stickiness.
58
GLOBIOM relies for its trade on FAOSTAT net export and reallocates trade bilaterally using COMTRADE. MIRAGE-BioF use data from the
GTAP database that is built on COMTRADE statistics. Both models use the tariffs information from the MAcMap-HS6 database (Bouët et al.,
2008).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 117

maintain in case of large change in trade flows, due to the function of substitution used (see Box 4
for a detailed discussion). Additionally, trade patterns can only evolve in MIRAGE-BioF around the
base year trade flows, and no new trade flow can appear. Products are not necessarily sourced from
the cheapest region, because consumers are assumed to differentiate them on other criteria (quality
or sanitary measures for instance).

This can help in reproducing some trade patterns in high value products (limited change in meat
trade) but is sometimes a constraint to replicate rapid changes in the trade balance of bulk
commodities (for instance, the change in rapeseed trade direction in Europe in the 2000s).

I.6.2 Including the economy partly or entirely: PE versus CGE

GLOBIOM is a partial equilibrium (PE) model, this means that the relevant sectors (agriculture,
forestry and bioenergy) are represented in detail, which makes it suitable for modelling LUC effects.
Other economic sectors however are not included or only included in a very coarse way. GLOBIOM
assumes that the economy outside land using sectors evolves independently from the policies
assessed in the model, following a ceteris paribus approach59.

In the MIRAGE-BioF model, all sectors of the economy are simultaneously and immediately
interconnected, for this reason the model is classified as a computable general equilibrium (CGE)
model. It works as a giant water bed – if you press on one side of the economy, it moves
everywhere. This is because all relations in the economy are described through equations that take
the trickle-down effect to all other sectors already into account. The relation between all sectors is
described on the basis of base year economic flows (national accounting perspective).

As stated above, GLOBIOM models only agriculture, forestry and bioenergy and focuses on
understanding the land use impact of these activities. The impacts on the rest of the economy are
assumed to have a second order effect and are not accounted for the modelling of LUC. Using the
GLOBIOM model instead of the MIRAGE-BioF model implies that some sector interactions are
missing. These interactions are however predictable in the case of biofuel policies, in which an
increase in biofuel demand leads to more demand for biofuel crops. In particular, two interactions
with sectors not covered in GLOBIOM can have a feedback effect on the land use 1) the effect biofuel
policies have on the fuel market and its feedback on agriculture and forestry via fossil fuel and
fertilizer prices and 2) the increased regional income in developing countries associated to the
development of bioenergy production, that lead to higher consumption of land based products. These
interactions are described in more detail in Box 4. Where necessary, effects not currently covered by
GLOBIOM can be calculated ex-ante (before the event) and added to the simulation of the GLOBIOM
model. For example, the change in oil price associated to the biofuel policy can be calculated based
on a literature review or a simplified model. If required, it is also possible to introduce a simplified
representation of the fuel market in GLOBIOM to represent its relation to the bioenergy market.

59
Ceteris paribus is an assumption widely used in economics where the effect of changing a parameter in the economic system is analysed,
while considering that all other parameters influencing the economy are kept unchanged.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 118

Box 4: General equilibrium effect from biofuel policies not captured in GLOBIOM

A certain number of general equilibrium effects are not captured in GLOBIOM, for instance across sectors or
economic agents. Depending on additional information coming available during the present study (from literature,
from stakeholders), we can decide to improve the description of interaction between the increasing biofuels
volume and the feedback effects on other land based sectors.

 Fuel market leakage: Biofuels can lead to a decreased demand for fossil fuel and therefore somewhat reduce
prices of fossil fuels. In response, cheaper prices can lead to an additional consumption of fossil fuel. This
means that the replacement of fossil fuels by biofuels may not be 1 to 1.60 This leakage is of different nature
than the ILUC leakage but is important for the final GHG balance of biofuels.
 Impact on fuel prices and feedback on agriculture: Al-Riffai et al. (2010) report (using MIRAGE)-BioF that the
EU biofuel mandate will lead to a fall in oil price of about 0.8% and a price reduction in the EU of about 0.3%
of conventional petroleum based fuel at the pump. This could have a feedback effect on the input side of the
agricultural sector and forestry sector. As this impact is usually small, this feedback effect should remain
limited.
 Impact on fertilizer prices: fertilizer prices can be influenced by the price of fossil energy as well as by the
change in production level required by the expansion of the agricultural sector. Furthermore, changing crop
prices change the specific intensity of input use, increasing demand for biofuels thus increases the use of
fertilizer on the existing cropland. As mentioned in the previous bullet, oil prices are expected to change in
response to biofuel policy shocks, which could impact the price of fertilizer. However, as this effect is
expected to remain small, the magnitude of this impact will remain limited, as long as quantities of extra
fertilizer are low compared to overall agricultural needs.
 Change in consumer income. Impact on food prices is captured by a model like GLOBIOM. But in some
developing regions, the development of a biofuel sector can have a significant impact on national income.
Additionally, a change in fuel price can also lead to an increase or decrease of purchasing power and
consequently a higher or lower consumption of other products.
 Change in exchange rates, wages, cost of capital, service input prices, etc.: many other interlinkages are
described in a general equilibrium framework such as impacts on the labour market, capital market and
currency market. However, as the biofuel sector remains of limited size compared to the rest of the
economy, its macroeconomic impact usually remain limited. Bouet et al., 2010 find that the welfare impact of
the EU biofuel mandate is close to zero (-0.01%) for the EU and regions in the world that are more notably
affected are the least advanced countries, due to change in commodity prices, captured in GLOBIOM.

60
Rajagopal et al., 2011; 2013; de Gorter et al., 2011

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 119

I.7 Modelling land use change and associated GHG emissions

The modelling of LUC is a strength of GLOBIOM, as land is the starting unit to all production
processes. GLOBIOM uses a flexible framework that represents all major land use substitution
possibilities and takes into account the heterogeneity of production across locations. This offers a
solution to the limitations observed with the simplified representation of land substitution in MIRAGE-
BioF. The link between LUC and GHG emissions is more precise in GLOBIOM because it relies on a
more refined approach than MIRAGE-BioF, which uses the default IPCC coefficients, and an updated
estimate for peatlands. Additionally, GLOBIOM contains sources of non-CO2 emissions from
agriculture that can complement the understanding of the full GHG effect of bioenergy policies.
MIRAGE-BioF however can inform policy on the emissions from the rest of the economy (industry and
services).

I.7.1 Land allocation for crops

Trade in GLOBIOM is handled at the level of its 57 economic regions (EU28 + 29 world regions). The
supply side of the model optimises the location of crop cultivation at a much finer resolution in so-
called Supply Units: geographical areas of similar topographic, climatic and soil conditions of which
more than 10,000 are distinguished in GLOBIOM. Depending on the potential yield and cost in each
Supply Unit the model determines which crops will be allocated in that unit and in what quantity.61
Each supply unit contains information (derived from the biophysical model EPIC) on the productivity
of each crop. Therefore the quality of land is not an absolute characteristic of a Supply Unit, but is
crop specific. Additionally for the EU region, more precise data could be fed into the model to
represent crop rotations in GLOBIOM and substitutions occurs between these rotations (defined as
group of crops including rotations) instead of between single crops.

This representation is more detailed than in MIRAGE-BioF, which relies on an aggregated approach at
the level of the region and agro-ecological zone. Land substitution in MIRAGE-BioF is managed
through a Constant Elasticity of Substitution (CES) nested structure that allows different levels of
substitutability between crop productions to be distinguished (see Box 5). Two elasticities of
substitution determine which crops are grown (see Figure 24). For instance, corn and wheat are
highly substitutable. If the price of wheat increases, a significant share of corn harvested area is
reallocated to wheat. Rice is placed at a lower level of substitution. Therefore, for the same wheat
price increase, the increase of rice acreage will be much smaller. This simplified approach has the
advantage of representing all the relevant substitution mechanism at the aggregated level. However,
it does not use the full biophysical information useful to know the relative crop profitability in each
location and may neglect non-linearities in the system. For instance, in the previous example, it is
possible that corn remains very profitable in many suitable regions, initially limiting substitutions.
When the wheat price hits a record, making it more profitable everywhere, substitutions will occur
more massively.

61
This process of allocation of land between crops can be assimilated as a perfect substitution. In practice, to avoid the model to reallocate
too abruptly across production systems, a flexibility constraint is implemented, often a lower or upper limit to the share of harvested area
that the crop can use in the given location. In the EU, crop rotations also play this role of flexibility contraint.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 120

Figure 24: Land substitution nesting structure for crops in MIRAGE-BioF

Box 5: CES and CET functional forms, the bricks of MIRAGE-BioF

The Constant Elasticity of Substitution (CES) function is a production function widely used in applied economics to
define how the level of output of one sector depends on a certain number of inputs or production factors (labour,
capital, land, etc.). In a CES, when a quantity Q of output is obtained using two inputs in quantity q 1 and q2, a
mathematical relation defines how q1 and q2 can substitute each when their relative prices change. The central
parameter of the CES is the elasticity of substitution σ that defines the easiness of substitution. For instance, an
elasticity of substitution of 0.1 between labour q1 and capital q2 means that if price of labour increases by 10%
relative to capital, more capital will be used such as capital purchase over labour purchase (q2/q1) increases by
1% (10% x 0.1).

The Constant Elasticity of Transformation (CET) function is used to define allocation of production factors
(typically land or labour) where the returns are the highest. It works exactly as a CES but in the other direction:
quantity is increased for the good that gets a higher relative price. For instance, a land owner puts more land in
production in the sector that has higher market prices. Mathematically, the two functional forms are the same but
the sign of the elasticity of substitution becomes negative for the CET.

CES and CET are used as elementary bricks in many CGE models and are extensively used in MIRAGE-BioF. One
level of substitution (one elasticity only and all products at the same level) is usually too coarse to represent the
complex substitution patterns observed at the aggregated level., In an attempt to approach reality and model
capacities, modellers usually increase the number of levels (nested CES or CET) to differentiate different levels of
substitution.

Although this approach allows for controlling the ease of substitution with different elasticity values for the
different nests, three limitations are however to be noted. First the number of degrees of freedom for the
calibration is only one per nest. So for instance, using one CET level for land means that forest, crop land and
pasture is the same.62 Moreover, the substitution patterns are by construction symmetrical, i.e. it is possible to
reverse the land conversion with the same easiness. In other words, if prices come back to their initial values,
land use comes back to its initial distribution.63 In a conversion cost approach as in GLOBIOM, costs can be
different for changing from land type A to B and for the reverse relation.

62
This specification is for instance used in GTAP-BIO (Golub and Hertel, 2012). See discussion in CARB (2011) and Laborde and Valin
(2012).
63
In MIRAGE-BioF, this symetry is also observed. However, if natural forest disappears, and is later replaced again by forest, it is assumed
that the level of carbon is lower, equivalent to a managed forest.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 121

A second issue is that the substitution around the equilibrium is performed on the basis of input values (one USD
versus one USD).

To obtain a substitution in a different metric (for instance, one tonne for one tonne), it is necessary to reconstruct
all the input values, using a same price per unit of substitution (in our example, same price per tonne). In
MIRAGE-BioF, this was indeed required for a large number of agricultural goods, to ensure consistent substitution
patterns.

A last drawback for the CES and CET is that the sum of volumes is not conserved by the substitution. This is a
critical issue in the case of land use substitution and in the case of MIRAGE-BioF, it has been corrected for land by
applying a correction factor.64 However, it remains a limitation for the many other CES functional forms in the
model, when moving away far from the initial equilibrium point.

I.7.2 Cropland, grassland and agricultural land expansion

Another important difference between GLOBIOM and MIRAGE-BioF is the way in which land use is
represented between land cover types.
In MIRAGE-BioF, several representations have been tested. The design used for Laborde (2011) is as
represented in Figure 25. Land expansion is managed at two levels:
 First level: Land expansion within agricultural and managed forest area (i.e. economic use area).
For these cases, the substitution between cropland, grassland and forest is managed through a
CET functional form (see Box 5).
 Second level: Land expansion in other natural area is managed through a separated elasticity of
total managed land expansion.

Figure 25: Land use type substitution nesting and expansion in MIRAGE-BioF

The limitation of this approach is the proper evaluation of land rents associated to grassland and
managed forest (see

64
See Golub and Hertel (2012) for an illustration of how this correction is made. Elasticities of transformation are however no longer
constant in that case, which imposes some recalibration when far from the initial point.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 122

Box 6). The aggregated representation of MIRAGE-BioF does not allow for capturing the value of land
in the different locations adequately. Moreover, the quality of land is often related to the type of
cultivation, livestock activity, or forest plantation that land owners can choose, depending on output
and inputs prices.

A very suitable land for wheat is not necessarily as suitable for corn, or for cotton, as illustrated by
the regional specialization observed all over the world.

In GLOBIOM, the productivity of land for each type of crop is specific to the grid cell, also for land not
currently used as cropland. Therefore, it is possible to consider conversion of other land to cropland
on the basis of the expected profitability associated to productivity of new locations. A similar
approach is used for grassland and grass productivity. This allows for direct calculation of the value
of the marginal productivity of land in the model (a parameter often discussed in the ILUC debate).
This value is estimated on the basis of real land use productivity estimates from EPIC (see Section
I.2.1) instead of using an ad-hoc coefficient like in MIRAGE-BioF. By default, a yield value equivalent
to 75% of average yield in the region was applied in MIRAGE-BioF in case of land expansion (with an
interval for sensitivity analysis of 50%-100%).

Land expansion in GLOBIOM is described at the level of each spatial unit. Instead of substituting use
with an aggregated function at a regional level, as for crop substitution in MIRAGE-BioF, land
conversion is performed at the local level, on a one to one hectare basis, to allocate the new
production to the spatial unit. A matrix of land use conversion between land use types defines which
land use conversions are possible and what the associated costs are (Figure 26).
The land transition matrix has the great advantage of offering a flexible representation of land
conversion patterns that has close resemblance with the real world. Conversion costs are not the
same and vary between land types. For instance, it can be less costly to expand into natural
vegetation than into forest (although less economically rewarding if the timber can be valued). This
conversion cost approach in particular allows for a more flexible representation of the main drivers of
LUC and deforestation observed in the different regions of the world.65

Peatland is one of the land covers that are under scrutiny in the biofuel debate. No spatially explicit
information on peatland is currently available in GLOBIOM. Therefore, as in MIRAGE-BioF, drainage
of peatlands drainage is currently accounted ex-post (with hindsight) in the model and based on
other indicators, in particular cropland expansion in areas already containing drained peatlands. This
representation can be improved if more information becomes available in the course of the project.

65
All land use changes in GLOBIOM are driven by expansion of agriculture and forestry. Hosonuma et al. (2012) estimate that 80% of
deforestation is driven by agriculture.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 123

Box 6: Land rent and land areas

One of the main challenges of the CGE approach used by MIRAGE-BioF is the mapping between the value of land
represented in the production function, and the effective land area observed in the statistics. In the GTAP
database, land use is represented as land rent, because production functions account for purchases of the
different production factors (labour, capital, land). High value products (e.g. vegetables, fruits, cash crops) are
therefore allocated a higher land rent, but only for a limited cultivated area. This can become a problem when
starting to reallocate land input from one sector to another. For example, in GTAP, cereals have generally lower
value added and therefore lower land rent per hectare. Transferring all land rents from vegetable and fruits to the
cereal sector provides a lot of virtual land because, even though the value of land rent is very high (providing
great expansion possibilities for cereals), the real biophysical area transferred is in reality small.
In MIRAGE-BioF, this anomaly has been fixed in the crop sector by reconstructing all land rents and assuming the
same rent per hectare for all crops in a given region. However, the issue remains in the mapping of other land
use. In particular, it is not possible to assume the same land rent per hectare for grassland, managed forest and
cropland, as the areas considered are too vast. Consequently, representation of cropland expansion remains
delicate when managed through a CET function (see Box 3). Several modelling options are proposed in Al Riffai et
al. (2010) and Laborde and Valin (2012).
The methodological difficulties above are avoided in the bottom-up approach taken with GLOBIOM by relying on
an explicit gridded representation of land, based on detailed information from remote sensing and data
downscaling. This approach however does not remove the need for specification and calibration efforts when
defining land conversion costs associated to the different transitions allowed.

Figure 26: Land cover representation in GLOBIOM with land use distribution in each model gridcell (left hand-side)

and land transition matrix defining in each gridcell the conversion allowed (arrows, right hand-side)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 124

I.7.3 GHG emissions of agriculture and land use change

A dozen different GHG emissions sources related to agriculture and LUC are represented in
GLOBIOM. Agricultural emission sources covered represent 94% of total agricultural emissions
according to FAOSTAT, and LUC emissions are consistent with recent reporting, although slightly
lower66(Valin et al., 2013). All GHG emission calculations in GLOBIOM are based on IPCC guidelines
for GHG accounting (IPCC, 2006). These guidelines specify different levels of detail for the
calculations. Tier 1 is the standard calculation method with default coefficients, whereas Tier 2
requires local statistics and Tier 3 onsite estimations. Seven out of eleven GHG sources in GLOBIOM
are estimated through Tier 2 or Tier 3 approaches.

Table 15: GHG emission sources in GLOBIOM

Sector Source GHG Reference Tier

Crops Rice methane CH4 Average value per ha from FAO 1

Crops Synthetic fertilizers N2O EPIC runs output/IFA + IPCC EF 1

Crops Organic fertilizers N2O RUMINANT model + Livestock systems 2

Carbon from cultivated organic

Crops CO2 FAOSTAT 1
soil (peatlands)

Livestock Enteric fermentation CH4 RUMINANT model 3

Livestock Manure management CH4 RUMINANT model + Literature review 2

Livestock Manure management N2O RUMINANT model + Literature review 2

Livestock Manure grassland N2O RUMINANT model + Literature review 2

Land use change Deforestation CO2 IIASA G4M Model emission factors 2

Land use change Other natural land conversion CO2 Ruesch and Gibbs (2008) 1

Harmonised World Soil Database

Land use change Soil organic carbon CO2 1/2
JRC for the EU28

For specific cases of LUC emissions, four different sources are particularly relevant:
 Deforestation: only changes in above and below ground living biomass are accounted for. G4M
provides estimates that are consistent with Forest Resource Assessment (FAO, 2010). When
forest is converted to a non-forest land cover, forest C stock is lost and replaced by the carbon
stock from the new land cover (see next bullet);
 Natural land conversion: for other land cover than forest, above and below living biomass is
accounted for based on the Ruesch and Gibbs (2008) database. This applies to grassland, other
natural land and short rotation plantations;

66
This is due to the fact that the model only represents land use change emissions from agricultural activities and not from other activities
such as illegal logging, mining, etc. Current observations however show decreasing patterns of deforestation in some regions with
significant deforestation in the past, in particular Brazil.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 125

  Soil organic carbon: following improvement performed in this project, soil organic carbon (SOC)
is accounted for the world using information from the Harmonised World Soil Database. In
Europe more precise data from JRC are used. SOC is influenced by crop management practices,
in particular tillage. See Annex II.5 for more details;
 Organic soil cultivation: this concerns peatlands that are taken into cultivation and emit GHG
emissions over multiple years. The associated emissions have been precisely studies in this
study, as described in Annex II.3.

In comparison to GLOBIOM, MIRAGE-BioF LUC GHG accounts are based on more generic calculations
as they often rely on Tier 1 approach from IPCC. Non-CO2 emissions from agriculture were not used
to avoid some double counting with the direct emissions coefficients from biofuel life cycle analyses.

From the four types of emission sources listed above, only three sources are represented in MIRAGE-
BioF, natural land carbon stocks in living biomass are not represented. MIRAGE-BioF models the
other three as follows:
 Carbon stock in forests is based on IPCC Tier 1 emission factors applied to the different AEZ in
the 15 regions (Laborde and Valin, 2012). Forest coefficients correspond to above and below
ground living biomass and a distinction is made between primary forest, managed forest, and in
the case of EU, afforested areas. However, the carbon stocks are not spatially allocated like in
G4M;
 Soil organic carbon is estimated for all regions in the world using IPCC Tier 1 emission factors;
 IPCC coefficients are not applied for peatland, but instead a higher value of 55 tCO2 ha-1 yr-1 is
used, sourced from more recent literature estimates. Based on historical observations, a share
of 33% of oil plantations is assumed to expand into peatland (based on Edwards et al., 2010).

I.8 Modelling changes in food consumption

Both GLOBIOM and MIRAGE-BioF represent a response of food demand that increases the price of
agricultural products. MIRAGE-BioF features the most sophisticated approach to modelling food
demand having a full representation of household substitution patterns. Consumers in GLOBIOM do
not substitute across products, but the impact of their change in food intake can be estimated in a
more tangible way, using statistics on kcal per capita per day provided by FAO.

Food demand is endogenous in GLOBIOM and depends on population size, gross domestic product
(GDP) and product prices. When population and GDP increase over time, food demand also
increases, putting pressure on the agricultural system. Change in income per capita in the baseline
drives a change in the food diet, associated to changing preferences. Current trends in China for
example show that per capita rice consumption decreases, whereas pig consumption increases and
milk consumption grows even faster. Food prices are another driver for a change in food
consumption patterns. When the price of a product increases in GLOBIOM, the level of consumption
of this product decreases by a value determined by the price elasticity associated to this product in
the region considered. The price elasticity indicates by how much the relative change in consumption
is affected with respect to relative change in price.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 126

For instance, an elasticity of -0.1 means that if the price of the product increases by 10%, the
consumption of this product decreases by 1% (10x-0.1). The values of these elasticities in GLOBIOM
are sourced from the USDA demand elasticity database67. In this database, price elasticities of
demand are lower for developing countries than for developed countries and lower for cereals than
for meat products. This is consistent with observations.

The representation of demand in MIRAGE-BioF is more comprehensive because the model

incorporates a full representation of the consumer budget covering consumption responses to
changes in household income and to the different product prices at the same time.
In particular MIRAGE-BioF allows for representation of cross-price effects. This means that when the
price of wheat increases the consumption of wheat decreases (own-price effect) whereas the
consumption of corn increases to compensate for wheat loss (cross-price effect). GLOBIOM models
the own-price effect but does not account for the cross-price effect. Therefore its assessment of food
demand change cannot account for substitution, which may underestimate the transmission of
effects across agricultural markets.68 An additional feature of GLOBIOM compared to MIRAGE-BioF is
that it accounts for kcal per capita supplied per day by using FAOSTAT data. The impact of food
prices on food demand can therefore be assessed as a change in kcal per capita per day for each of
the products.

I.9 GLOBIOM and MIRAGE-BioF characteristics – technical summary

GLOBIOM and MIRAGE-BioF are two very different types of economic models: GLOBIOM is a detailed
multi-sector multi-region mathematical programming model focused on agriculture and forest
activities, and therefore follows a partial equilibrium approach; MIRAGE-BioF is a multi-sector multi-
region computable general equilibrium model (CGE), based primarily on the Global Trade Analysis
Project database (GTAP). Although these approaches differ in several important points, they are both
grounded in microeconomic traditions and based on the same assumptions of optimizing behaviours
of the agents they focus on, producers for GLOBIOM, and producers and consumers for MIRAGE-
BioF. Prices play a central role in these models to shape decisions of agents.

I.9.1 GLOBIOM, a partial equilibrium mathematical programming model

GLOBIOM is a multi-sectoral model developed at the International Institute for Applied Systems
Analysis (IIASA) since 2007. The model is grounded in the mathematical programming tradition
(McCarl and Spreen, 1980). This type of model is derived from aggregation of more simplified linear
programming models of production used in microeconomics (Day, 1963). This type of approach has
been long used in economics for many sectoral problems, in particular in agricultural economics
(Takayama and Judge, 1964; 1971).

67
This database provides demand elasticities for 144 regions and eight food product groups. See Muhammad et al. (2011).
68
Market interactions however also occur through the supply side with land use competition.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 127

Development of recent computation capacities allowed application of this framework to large scale
problems with a high level of details, for example to US policies affecting agriculture and forestry
sectors (Schneider et al., 2007; US EPA, 2010) and GLOBIOM has common roots with the US-
FASOM69 model. Sectors covered by GLOBIOM are currently agriculture, forestry and bioenergy, with
their supply side production functions, their markets and the demand side. The model is therefore a
partial equilibrium model, because not all goods, factors or agents are represented in this approach.
It is therefore designed to address issues affecting land use based sectors, and consider that
situation in the rest of the economy is unchanged (ceteris paribus).

The economic formulation problem in GLOBIOM is expressed as follows: the model optimises an
objective function defined as the sum of producer and consumer surplus associated to the sector
represented, under a certain number of constraints. Producer surplus is determined by the difference
between market prices and the cost of the different production factors (labour, land, capital) and
purchased inputs. International transportation costs are also taken into account in the producer
costs. On the consumer side, surplus is determined by the level of consumption on each market: the
lower a price is, and the higher this consumption level can be, as well as the consumer surplus.
Technically, this is achieved by integrating the difference between the demand function of the good
on its market and the market price level. Constraints in the model are related to various dimensions:
technologies available, biophysical resources availability (land, water), capacity constraints, etc.
In this type of approach, the supply side can be very detailed, in particular benefiting from the
possibility of linearizing the non-linear elements of the objective function, the model can be solved as
a linear programming (LP) model, allowing a large quantity of data to be used for production
characteristics. The GLOBIOM model for instance can optimize the production for each sector on a
large number of geographic units (maximum resolution is 212,000 units but typically the model is
run at a more aggregated level of around 10,000 units). Additionally, many technologies and
transformation pathways can be defined for the different sectors. This detailed representation on the
production side however induces a trade-off on the demand side. Because of the linear optimization
structure, demand is represented through separated demand functions, without a representation of
total households budget and the associated substitution effects (McCarl and Spreen, 1980).

I.9.2 MIRAGE-BioF, a computable general equilibrium based on the GTAP framework

MIRAGE-BioF is a computable general equilibrium model (CGE) dedicated to biofuel impact analysis,
derived from the trade policy analysis model MIRAGE developed at CEPII (Bchir et al, 2002; Decreux
and Valin, 2007) and based on the GTAP database (Narayanan et al., 2012).
CGE models have their basis grounded in microeconomic theory, but operate in a macroeconomic
framework, with a complete coverage of economic flows circulating in the economy for purchase of
goods, remuneration of production factors. The father of the general equilibrium theory is Leon
Walras who defined this framework in 1871, emphasizing the importance of interactions across the
different component of the economy, ie sectors and regions, but also factor market, government
expenditure, households savings and investment, current accounts disbalances, etc.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 128

Kennet Arrow and Gérard Debreu implemented these principles in the 1950s in a more systematic
formalized framework. To the difference of partial equilibrium models, all prices in CGEs are
endogenous determined through equations to other economic trade flows, including real wages,
return on capital, or exchanges rates (only one single price needs to be fixed to serve as a reference,
called numeraire). These models are calibrated to a pre-existent state of the economy, considered in
equilibrium. Prices all vary around this initial equilibrium in response to a shock (change in tax level,
tariff, level of quota). Data on the pre-existing state is supplied by extensive datasets, called Social
Accounting Matrices (SAM), usually produced by national statistical agencies.

The big advantage of CGEs is their full theoretical consistency as no ceteris paribus assumption is
necessary with all sectors of the economy simultaneously interconnected. This however comes at the
expense of details because SAMs are often more limited in their sectoral representations, due to their
macroeconomic perspective and they only provide economic flows in monetary terms.
Even if some countries produce precise datasets tracking all economic interdependencies, with high
level of representation of sectors, households and factor markets, many others rely on coarser
information, and must rely on construction assumption and allocation rules to build up a complete
and consistent SAM. These models were used until the end of the 80s mainly to assess the effect of
taxation policies and trade policies (Shoven and Whaley, 1984), but they have been progressively
extended to other applications such a climate change impact, carbon trading policies or bioenergy
policies.

In the case of global CGEs, the GTAP database is very often used as the source of data, as it
represents a unique effort of reconciling information from the SAMs of the different countries around
the world. The process is however delicate as SAMs from various countries are usually not consistent
with each another, due to differences in accounting method but also to the year in which the SAM
has been constructed (SAMs are rarely available for every year). The GTAP consortium performs this
reconciliation process and succeeded to put together an increasing number of SAMs over the years
(96 for GTAP6 with base year 2001, 112 for GTAP7 with base year 2004, 134 for GTAP8 with base
year 2007).

The GTAP database currently uses a nomenclature of 57 sectors, including 12 for raw agricultural
products, and 1 for forestry. This often makes the data too coarse for a precise assessment of
bioenergy. For instance, ethanol and biodiesel are missing but also fossil fuel. Oilseeds are
aggregated and vegetable oil and their co-products are in the same sectors. For that reason, IFPRI
has developed an extended database used with the MIRAGE-BioF model (82 sectors) for the different
biofuels assessments, in which the most important missing sectors have been singled out.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 129

I.10 Technical comparison table GLOBIOM versus MIRAGE-BioF

GLOBIOM MIRAGE-BioF
Land use resolution
Simulation units (SimU) architecture (Skalsky et al., 2008)

 Global-SimU
= Countries boundaries
x HRU* at 5' resolution
x Grid layer with 30' resolution
GTAP Land database (GTAP-AEZ)

Total number of Global-SimU (incl. EU): 212,707

1 spatial unit = 18 agro-ecological zones
Usual aggregation for global runs (2°x2°): 10,893
x GTAP7 countries (112)

Max number of Global-SimU for Brazil: 11,003

Typical aggregation, world:
Usual aggregation for global runs, Brazil: 443
155 units but of unequal importance
 EU-SimU (Laborde and Valin, 2012)
= NUTS2 spatial unit
x HRU at 1 km resolution
Typical aggregation, EU27:
10 units (87% of rent in 2 AEZ)
Max number of EU SimU: 379,220
Typical aggregation, Brazil:
Usual aggregation: 648
9 units (98% of rent in 4 AEZ)
(NUTS2 x AEZ regions)

*HRU = Region of same altitude, soil type, slope and other

characteristics (Balkovic et al., 2010)

Land cover types

World: Global Land Cover 2000 (JRC, 5’x5’) FAOSTAT database
EU: CORINE Land Cover 2000 (EEA, 1 x 1 km)
Land cover types imported into MIRAGE:
Land cover types imported into GLOBIOM:  Arable land
 Meadows and permanent pasture
 Cropland
 Permanent crops
 Other agricultural land
 Forest
 Grassland
 Other
 Forest
 Wetlands
 Other natural land
Land available for expansion:
 Not relevant
GAEZ (IIASA and FAO, 2002)

Improvements performed in the model

 Split managed/unmanaged forest (G4M data) Improvements performed in the model:
 Grassland match to grazing requirements  Split managed/unmanaged forest (GTAP-AEZ
 Short rotation plantation land cover data)
Crop production
World: 18 crops: World: 11 crops aggregates from an extended
 cereals: barley, corn, millet, rice, sorghum, wheat,
GTAP database:
 oilseeds: groundnut, rapeseed, soybeans, sunflower, palm
 sugar cane (Laborde and Valin, 2012)
 roots/tubers/vegetables: cassava, chick peas, dry beans,  Wheat
potatoes, sweet potatoes  Maize (built by IFPRI)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 130

 GLOBIOM MIRAGE-BioF
 cotton  Sugar crops
 Soybeans (built by IFPRI)
 Sunflower (built by IFPRI)
EU: 9 additional crops:  Rapeseed (built by IFPRI)
 cereals: soft wheat, durum wheat, rye, oat  PalmFruit (built by IFPRI)
 sugar beet  Rice
 peas  OthCrop (aggregates of GTAP other crops,
 green fodder: corn silage, other green fodder plant fibers and other coarse grains)
 fallow  Other oil seeds
 Vegetable and fruits

Harvested area: Harvested area:

World: FAOSTAT with spatial allocation from Spatial Production FAOSTAT distributed by AEZ according to the

Allocation Model (IFPRI) M3 database (Ramankutty et al., 2008)

EU28: EUROSTAT NUTS2 statistics

Yield:
Yield: FAOSTAT, only at regional level.

World: EPIC model on SimU grid for the 18 crops.

Yield values adjusted to fit FAOSTAT country level Spatial and
management system differentiation.
EU28: EPIC run for combination of different rotation systems for
all NUTS2 regions. Production:
FAOSTAT, only at the regional level.
Production:
At SimU level. Consistent with FAOSTAT & EUROSTAT
aggregates. Production cost:
GTAP database
Production costs:
FAOSTAT producer prices. Technology: 1 aggregated nested CES
function
Technology:
Substitution between Leontieff technologies
World : 4 technologies estimated by EPIC
- Subsistence
- Low input, rainfed
- High input, rainfed
- High input, irritaged
EU28: large set of technologies
- 2 different levels of fertilizer
x 2 different levels of irrigation
x 3 different level of tillage

Livestock sector
Eight Animal types (and seven associated products): Two livestock sectors:
 Bovine dairy (bovine milk and meat)  cattle
 Bovine other (bovine meat)  other animals

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 131

 GLOBIOM MIRAGE-BioF
 Sheep and goat dairy (small ruminant milk and meat)
(three sectors in GTAP: cattle dairy, cattle
 Sheep and goat other (small ruminant meat)
 Pigs (pig meat) other, other animals)
 Poultry hens (eggs)
 Poultry broilers (poultry meat)
 Poultry mixed (poultry meat and eggs)

Animal number: Animal number:

ILRI/FAO Gridded Livestock of the World (GLW) animal number Not available in GTAP or MIRAGE-BioF.
and distribution at the 3’x3’ resolution. (Assimilated to capital)

Yield:
Estimated using RUMINANT, a digestibility model. Ensures Yield:
perfect consistency between feed input (grass, grains, stover…) Input/Output coefficient from the SAM
and output.
For monogastric, based on a literature review.

Production: Production:
Seven products: GTAP production value. Can be matched ex
 Bovine meat (from bovine dairy and bovine other) post with FAO quantities.
 Bovine milk
 Sheep and goat meat (from sheep and goat dairy and sheep
and goat other)
 Sheep and goat milk
 Pig meat
 Poultry meat (from broiler and poultry mixed) Production cost:
 Poultry eggs (from hens and poultry mixed)
GTAP database

Production cost:
Technology: 1 aggregated nested CES
FAOSTAT producer prices and grains input.
function

Technology:
Substitution between Leontieff technologies
Ten systems (Seré and Steifeld classifications)
8 systems for ruminant:
 Grassfed arid • Mixed arid
 Grassfed humid • Mixed humid
 Grassfed temperate • Mixed temp.
 Urban • Other
2 systems for monogastrics
 Industrial
 Smallholders
Forestry sector
Forest area: based on G4M model (0.5°x0.5°) Forestry as one single sector.

Harvest yield: Stemwood harvest potential determined from

net primary productivity (NPP) maps, combined with maps on
forest biomass stock (Global Forest Resources Assessment,
FAO)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 132

 GLOBIOM MIRAGE-BioF
Forest primary products:
4 forest resources.
 Industrial roundwood
 Non-commercial roundwood
 Harvest losses
 Branches and stumps
Separated into 5 primary woody products:
 Sawn wood biomass
 Pulp wood biomass
 Energy wood biomass (biofuels, heat and electricity)
 Traditional use biomass (fuel, cooking)
 Other use biomass

Forest secondary products:

Secondary forestry residues from forest industries and milling Production cost:

activities: GTAP database

 Saw chips
 Sawdust
 Bark
 Black liquor

Production costs: Harvesting costs including logging and Technology:

timber extraction account for: 1 aggregated nested CES function
 Unit cost of harvesting equipment and labour
 A slope factor accounting for terrain conditions
 A regional adjustment of labour cost by the ratio of mean PPP
(purchasing power parity over GDP).

Technology:
Substitution between Leontieff technologies
Technologies with yield estimated for:
 Sawmills
 Mechanical pulp mills
 Chemical pulp mills
 Fiberboard production
 Plywood production

Conversion technologies in agriculture, forestry and

bioenergy
List of sectors/processes:

List of sectors/processes:
 Agriculture
o Rapeseed crushing
o Sunflower crushing  Agriculture
o Soybean crushing o Rapeseed crushing
 Forestry o Sunflower crushing
o Sawmill o Soybean crushing
o Mechanical pulping o Palm fruit processing
o Chemical pulping  Bioenergy
o Plywood production o 1st gen biofuel corn
o Fiberboard production o 1st gen biofuel wheat
 Bioenergy o 1st gen biofuel sugar cane
o Combustion o 1st gen biofuel sugar beet
o Cooking o 1st gen biofuel FAME
o 1st gen biofuel corn
o 1st gen biofuel wheat

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 133

 GLOBIOM MIRAGE-BioF
o 1st gen biofuel sugar
o 1st gen biofuel FAME
o 2nd gen biofuel fermentation
o 2nd gen biofuel gasification

Conversion coefficients and costs:

Conversion coefficients and costs: Based on the GTAP modified database.
Based on FAOSTAT and literature reviews. Changing in technology representation
Can be expanded or updated more easily as a CGE. technical due to modification to report in the
SAMs
GHG Emission sources
Eleven emission sources from agriculture and land use change:

 Rice methane CH4 CO2 Industrial and service emissions

 Synthetic fertilizers N2O
+ Three emission sources from land use
 Organic fertilizers N2O
 Enteric fermentation CH4 change
 Manure management CH4  Deforestation CO2
 Manure management N2O  Soil organic carbon CO2
 Manure grassland N2O  Cultivated organic soil CO2
 Deforestation CO2
 Other natural land conversion CO2
 Soil organic carbon CO2
 Cultivated organic soil CO2

I.11 References

Al-Riffai, P., Dimaranan, B. & Laborde, D. (2010). Global Trade and Environmental Impact Study of
the EU Biofuels Mandate (Report to the European Commission). International Food Policy Research
Institute.

Balkovic, J., van der Velde, M., Schmid, E., Skalský, R., Khabarov, N., Obersteiner, M., Stürmer, B. &
Xiong, W. (2013). Pan-European crop modelling with EPIC: Implementation, up-scaling and regional
crop yield validation . Agricultural Systems 120 (0), 61 - 75.
Retrieved from http://www.sciencedirect.com/science/article/pii/S0308521X1300070X.

Bchir, H., Decreux, Y., Guerin, J.-L. & Jean, S. (2002). MIRAGE, a Computable General Equilibrium
Model for Trade Policy Analysis (Working Paper 2002-17). CEPII. Retrieved from
http://www.cepii.fr/anglaisgraph/workpap/summaries/2002/wp02-17.htm.

Bouët, A., Decreux, Y., Fontagné, L., Jean, S. & Laborde, D. (2008). Assessing Applied Protection
across the World. Review of International Economics 16 (5), 850--863.

Bouët, A., Dimaranan, B. V. & Valin, H. (2010). Modeling the global trade and environmental impacts
of biofuel policies (IFPRI Discussion Paper 01018). International Food Policy Research Institute
(IFPRI).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 134

Britz, W. & Hertel, T. W. (2011). Impacts of EU biofuels directives on global markets and EU
environmental quality: An integrated PE, global CGE analysis. Agriculture, Ecosystems &
Environment 142 (1-2), 102 - 109. (Doi: 10.1016/j.agee.2009.11.003.) Retrieved from
http://www.sciencedirect.com/science/article/pii/S016788090900334X.

CARB (2011). Final Recommendations From The Elasticity Values Subgroup (ARB LCFS Expert
Workgroup ). California Air Resource Board.

Chen, X., Huang, H. & Khanna, M. (2012). Land-use and Greenhouse Gas Implications of Biofuels:
Role of Technology and Policy. Climate Change Economics 03 (03), 1250013. (Doi:
10.1142/S2010007812500133.) Retrieved from
http://www.worldscientific.com/doi/abs/10.1142/S2010007812500133.

Cramer, W., Kicklighter, D. W., Bondeau, A., Iii, B. M., Churkina, G., Nemry, B., Ruimy, A., Schloss,
A. L. & Intercomparison, T. P. O. T. P. N. M. (1999). Comparing global models of terrestrial net
primary productivity (NPP): overview and key results. Global Change Biology 5 (S1), 1--15. (Doi:
10.1046/j.1365-2486.1999.00009.x.) Retrieved from http://dx.doi.org/10.1046/j.1365-
2486.1999.00009.x.

Day, R. H. (1963). On Aggregating Linear Programming Models of Production. Journal of Farm

Economics 45, 797-813.

Decreux, Y. & Valin, H. (2007). MIRAGE, Updated Version of the Model for Trade Policy Analysis
Focus on Agriculture and Dynamics (Working Papers 2007-15). CEPII research center. Retrieved
from http://www.cepii.fr/anglaisgraph/workpap/summaries/2007/wp07-15.htm.

EC (2012). Impact Assessment accompanying the document Proposal for a Directive of the European
Parliament and of the Council amending Directive 98/70/CE relating to the quality of petrol and
diesel fuels and amending Directive 2009/28/EC on the promotion of the use of energy from
renewable sources. (Commission Staff Working Document). European Commission.

Edwards, R., Mulligan, D. & Marelli, L. (2010). Indirect Land Use Change from Increased Biofuels
Demand: Comparison of Models and Results for Marginal Biofuels Production from Different
Feedstocks. Joint Research Center - European Commission.

EPA (2010). Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis ( EPA-420-R-10-
006). U.S. Environmental Protection Agency, Office of Transportation and Air Quality, Assessment
and Standards Division. Retrieved from
http://www.epa.gov/otaq/fuels/renewablefuels/regulations.htm.

FAO (2010). Global Forest Resources Assessment ( ). Food and Agriculture Organization of the
United Nations. Retrieved from http://www.fao.org/forestry/fra/fra2010/en/.

Gallagher, E. (2008). The Gallagher review of the indirect effects of biofuels production ( ).
Renewable Fuel Agency.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 135

(http://webarchive.nationalarchives.gov.uk/20110407094507/http://renewablefuelsagency.gov.uk/si
tes/renewablefuelsagency.gov.uk/files/_documents/Report_of_the_Gallagher_review.pdf)

de Gorter, H. & Drabik, D. (2011). Components of carbon leakage in the fuel market due to biofuel
policies. Biofuels 2 (2), 119--121.

Havlík, P., Schneider, U. A., Schmid, E., Böttcher, H., Fritz, S., Skalský, R., Aoki, K., Cara, S. D.,
Kindermann, G., Kraxner, F., Leduc, S., McCallum, I., Mosnier, A., Sauer, T. & Obersteiner, M.
(2011). Global land-use implications of first and second generation biofuel targets. Energy Policy 39
(10), 5690 - 5702. (Doi: 10.1016/j.enpol.2010.03.030.) Retrieved from
http://www.sciencedirect.com/science/article/pii/S030142151000193X.

Herrero, M., Havlik, P., Valin, H., Rufino, M., Notenbaert, A., Thornton, P., Blummel, M., Weiss, F. &
Obertsteiner, M. (2013). High resolution livestock productivity, biomass use, non-CO2 greenhouse
gas emissions and excretion maps for global change and sustainability research. PNAS.

Hertel, T., Hummels, D., Ivanic, M. & Keeney, R. (2007). How confident can we be of CGE-based
assessments of free trade agreements?. Economic Modelling 24 (4), 611--635.

Hosonuma, N., Herold, M., Sy, V. D., Fries, R. S. D., Brockhaus, M., Verchot, L., Angelsen, A. &
Romijn, E. (2012). An assessment of deforestation and forest degradation drivers in developing
countries. Environmental Research Letters 7 (4), 044009. Retrieved from http://stacks.iop.org/1748-
9326/7/i=4/a=044009.

IPCC (2006). IPCC guidelines for national greenhouse gas inventories. In S. Eggleston, L. Buendia, K.
Miwa, T. Ngara & K. Tanabe (ed.),. Institute for Global Environmental Strategies, Hayama, Japan.

Keeney, R. & Hertel, T. (2009). The Indirect Land Use Impacts of United States Biofuel Policies: The
Importance of Acreage, Yield, and Bilateral Trade Responses. American Journal of Agricultural
Economics 91 (4), 895--909.

Kindermann, G., Obersteiner, M., Sohngen, B., Sathaye, J., Andrasko, K., Rametsteiner, E.,
Schlamadinger, B., Wunder, S. & Beach, R. (2008). Global cost estimates of reducing carbon
emissions through avoided deforestation. Proceedings of the National Academy of Sciences 105 (30),
10302.

Laborde, D. (2011). Assessing the Land Use Change Consequences of European Biofuel Policies
(ATLASS Consortium, Final report ). IFPRI. Retrieved from
http://www.ifpri.org/publication/assessing-land-use-change-consequences-european-biofuel-policies.

Laborde, D. & Valin, H. (2012). Modeling land-use changes in a global CGE: Assessing the EU biofuel
mandates with the Mirage-BioF Model. Climate Change Economics 03 (03), 1250017. (Doi:
10.1142/S2010007812500170.) Retrieved from
http://www.worldscientific.com/doi/abs/10.1142/S2010007812500170.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 136

McCarl, B. A. & Spreen, T. H. (1980). Price Endogenous Mathematical Programming as a Tool for
Sector Analysis. American Journal of Agricultural Economics 62 (1), 87--102. Retrieved from
http://www.jstor.org/stable/1239475.

Muhammad, A., Seale, J., Meade, B. & Regmi, A. (2011). International Evidence on Food
Consumption Patterns: An Update Using 2005 International Comparison Program Data (Technical
Bulletin 1929). USDA-ERS.

Narayanan, B. G., Aguiar, A. & McDougall, R. (2012). Global Trade, Assistance, and Production: The
GTAP 8 Data Base. Center for Global Trade Analysis, Purdue University.

Rajagopal, D. (2013). The fuel market effects of biofuel policies and implications for regulations
based on lifecycle emissions. Environmental Research Letters 8 (2), 024013. Retrieved from
http://stacks.iop.org/1748-9326/8/i=2/a=024013.

Rajagopal, D., Hochman, G. & Zilberman, D. (2011). Indirect fuel use change (IFUC) and the
lifecycle environmental impact of biofuel policies. Energy Policy 39 (1), 228 - 233. (Doi:
10.1016/j.enpol.2010.09.035.) Retrieved from
http://www.sciencedirect.com/science/article/pii/S0301421510007214.

Ramankutty, N., Evan, A., Monfreda, C. & Foley, J. (2008). Farming the planet: 1. Geographic
distribution of global agricultural lands in the year 2000. Global Biogeochemical Cycles 22 (1), 1--19.

Ruesch, Aaron & Gibbs, H. K. (2008). New IPCC Tier-1 Global Biomass Carbon Map For the Year
2000. (Available online from the Carbon Dioxide Information Analysis Center [http://cdiac.ornl.gov]
). Oak Ridge National Laboratory. Retrieved from
http://cdiac.ornl.gov/epubs/ndp/global_carbon/carbon_documentation.html.

Schneider, U. A., McCarl, B. A. & Schmid, E. (2007). Agricultural sector analysis on greenhouse gas
mitigation in US agriculture and forestry. Agricultural Systems 94 (2), 128 - 140. (Doi: DOI:
10.1016/j.agsy.2006.08.001.) Retrieved from
http://www.sciencedirect.com/science/article/pii/S0308521X06001028.

Seré, C. & Steinfeld, H. (1996). World livestock production systems: current status, issues and
trends (Animal and Health Paper 127). Food and Agriculture Organisation. Retrieved from
http://www.fao.org/WAIRDOCS/LEAD/X6101E/X6101E00.HTM.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 137

Shoven, J. B. & Whalley, J. (1984). Applied General-Equilibrium Models of Taxation and International
Trade: An Introduction and Survey. Journal of Economic Literature 22 (3), 1007--1051. Retrieved
from http://links.jstor.org/sici?sici=0022-
0515%28198409%2922%3A3%3C1007%3AAGMOTA%3E2.0.CO%3B2-Y.

Takayama, T. & Judge, G. (1971). Spatial and temporal price and allocation models. North-Holland
Amsterdam.

Takayama, T. & Judge, G. G. (1964). An Interregional Activity Analysis Model for the Agricultural
Sector. Journal of Farm Economics 46 (2), pp. 349-365. Retrieved from
http://www.jstor.org/stable/1236540.

Valin, H., Havlík, P., Mosnier, A., Herrero, M., Schmid, E. & Obersteiner, M. (2013). Agricultural
productivity and greenhouse gas emissions: trade-offs or synergies between mitigation and food
security?. Environmental Research Letters 8 (3), 035019. Retrieved from http://stacks.iop.org/1748-
9326/8/i=3/a=035019.

Weih, M. (2004). Intensive short rotation forestry in boreal climates: present and future
perspectives. Canadian Journal of Forest Research 34 (7), 1369-1378. (Doi: 10.1139/x04-090.)
Retrieved from http://www.nrcresearchpress.com/doi/abs/10.1139/x04-090.

Wint, W. & Robinson, T. (2007). Gridded livestock of the world 2007. FAO.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 138

Annex II Building an improved version of
GLOBIOM

The GLOBIOM model is well suited for the purpose of modeling LUC. During the projest the
consortium worked to further improve the model for this specific purpose. The consortium invited
stakeholders to provide input on possible improvements. This Annex describes the improvements
which the consortium implemented during the study. Selected Improvements are taken from a
longer list of improvements based on stakeholder recommendations received during the first
stakeholder consultation in November-December 2013. The consortium selected in co-operation with
the project Advisory Committee the most important improvements, which were discussed with the
European Commission in January 2014, with stakeholders during the second stakeholder consultation
in February and again with the Commission in March 2014.

Nr Improvement
Improve the representation of cereal straw to enable the modelling of possible LUC effects of removing
1.
cereal straw from fields in 3 selected EU Member States
Include carbon sequestered in annual and perennial crops as this carbon influences the carbon flows
2.
(emissions and sequestration) due to land use change.
3. Update peat land emission factors based on best available literature.
4. represent the expansion of plantations into peat land based on literature review.
Inclusion of soil organic carbon (SOC) to rest of the world as currently only included for the EU. Tillage
5.
data are included at regional level where available.
Include forest regrowth and reversion time on unmanaged land based on IPPC estimates. These effects
6. increase the opportunity costs of using abandoned farmland in areas where forest regrowth is likely to
occur.
Improve protein and energy content representation to refine co-product substitution. Co-product
substitution was already represented in GLOBIOM, accounting for both protein and energy content.
7.
Substitution patterns are a highly debated topic, therefore, fine-tuning of this mechanism is important to
produce reliable substitution effects.
Represent multi-cropping. Multi-cropping increases production per hectare and reduces agricultural land
expansion. It can therefore influence the LUC response. As it is challenging to model multi-cropping
8.
response to price, multi-cropping will be included in the modelling baseline (with yield projections) rather
than in the model itself.
Represent imperfect substitution of vegetable oils. Previously, GLOBIOM could only represent perfect
substitution or fixed composition of vegetable oils. Market leakage across vegetable oil markets is crucial
9.
to the LUC impact of biodiesel but is characterised by imperfect substitution. Hence, it was important to
introduce this in GLOBIOM. Substitution elasticities are based on previous modelling exercises

Separate representation of Argentina, Indonesia, Malaysia and Ukraine. The first three are important
players in the production of biofuels and Ukraine has the potential to become an important supplier of
10. agricultural products to Europe in the future. Previously, the countries were not represented separately in
GLOBIOM but were represented as part of larger regional areas.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 139

Nr Improvement
Represent unused agricultural land in Europe. Large amounts of unused land, mostly abandoned
farmland, exist in Europe that could be potentially be used to produce additional quantities of biofuel
11.
feedstock. Using this land would limit expansion in other regions of the world and thus reduce LUC.
Biofuels produced on this land would be low ILUC risk.
Refine biofuel feedstock processing coefficients (oilseed crushing, ethanol production coefficients).
12. Crushing rates are important to determine the final land use impact of biofuels. Coefficients used in
GLOBIOM are fine-tuned.

II.1 Improve the representation of cereal straw

Motivation for improvements

Agricultural residues were so far represented in GLOBIOM only on the supply side, without
consideration of competitive uses and sustainability removal threshold. As agricultural residues
constitute one of the feedstocks studied in this assessment, it was decided to improve their
representation on the supply and demand side to overcome the current loopholes. The
representation of agricultural residues focuses in this study on wheat and other cereals straw.70
Three regions were selected as illustrative case studies. They were chosen taking into account data
availability on straw production and uses, a geographic coverage consistent with straw market size,71
but also contrasting situations with respect to sustainability of additional residue removals, if a 1%
biofuels shock from residues is implemented at the national level. The three selected regions are : i)
Hungary, as an example of limited availability of straw, as more than half of cereal straw potential is
currently being used for feed and animal bedding, which is considered beyond the sustainability
removal level, ii) Great Britain as a region with greater availability,72 but where sustainable potential
would be hit if 1% of transportation fuel was supplied from straw biofuels, and iii) Center of France73
where supply is relatively larger and sustainable potential should not be fully exhausted if 1% of
French transportation fuel was supplied from straw biofuels. Statistics on characteristics of the
different countries are summarized in Table 16 and the contrasting situations of these three regions
detailed in Table 17.

Methodological approach
Supply of residues: Three different management systems were distinguished to reflect different
levels of residue removal: i) no residue removal; ii) sustainable residue removal (around 33-50%
depending on the region); iii) high residue removal (greater than sustainable removal). The first and
second systems are assumed to have the same biophysical characteristics in terms of crop
production (yields, soil organic carbon stocks), but production costs in the second system are higher
due to collection of residues.74

70
We will represent cereal straw market in GLOBIOM with straw from wheat, barley, oat and rye, which are found to supply most of straw in
Europe (Ecofys, 2013).
71
Typical transportation distances are reported to be below 500 km.
72
Excluding Northern Ireland
73
Defined as NUTS1 regions FR1 (Ile de France) and FR2 (Bassin parisien).
74
The cost for residue removal or residue incorporation is based on the data Standarddeckungsbeitragskatalog 2008, from the Austrian
Ministry of Agriculture. We follow an assumption of 74 EUR/ha for full residues collection with balling.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 140

The third management system also has a different collection cost, and modified characteristics for
yield and soil organic carbon (SOC). Depending on the management, in particular the degree of
fertiliser application that can compensate the yield losses, these effects can have more or less
impact.

Impact of residue removal on yield is occurring through multiple channels, such as change in soil
temperature and moisture, nutrient content, soil texture and sensitivity to water and wind erosion
(Blanco-Canqui & Lal, 2007; Johnson & Barbour, 2010). To assess its impact, we EPIC simulations on
EU data assuming a linear decrease of yield between sustainable removal rate and yield loss
observed when 90% of residues are removed. EPIC simulations only capture a part of the drivers
cited above and provide effects on yield of around -2% for median value after 20 years, with first
quartile at -4.8% and third quartile at 0% (see Figure 27). Some other authors find greater impacts
on some crops but this is highly dependent on soil type (no impact in two types of soil or up to -15%
one type for corn stover in Blanco-Canqui & Lal, 2007; around -10% in Wilhelm, Doran & Power,
1986). Our simulations lead in particular to some positive feedback in case of low input system when
residues are removed.75 We also analyzed with the EPIC model the relative change of soil organic
carbon associated with straw removal of 90% (Figure 28). At such rate of removal, under full tillage,
SOC decreases after 20 years by 8 ± 3% with some significant differences across locations. For our
sensitivity analysis in the Monte-Carlo simulation, we vary the full range of possible value of impact
around representative values based on these findings. Our range of value acknowledges the
uncertainty related to the farmer management response and soil quality implications. For yield
impact, we assume an impact ranging from no impact (0%) to high impact (-4%) with a median
value at -2%. For soil organic carbon impact, we assume an impact ranging from no impact (0%) to
full impact (-10%) with a median value at -5%. The two ranges of impact are considered correlated.

Demand for residues: Several sectors are represented in the model that can compete for residues.
First, the livestock sector uses straw as bedding, and to a lesser extent as feed. A generic substitute
to straw has been represented in the model for bedding, which allows straw to be replaced by some
other materials above a certain price.76 Animal needs are implemented, with requirements based on
Scarlat, Martinov & Dallemand (2010): straw use for cattle is 1.5 kg/day/head for 25% of population,
sheep is 0.1 kg/day/head, pigs is 0.5 kg/day/head for 12.5% of population adjusted to (Ecofys,
2013) data when available. Straw used as feed can also substitute with other feedstuff in the
livestock sector, with some implications on land use. Additional uses are also considered for energy
and horticulture (mushrooms, strawberry, vegetables etc.) as well as industry (material use, pulp
and paper).

75
Producing sufficient and timely quantities of crop residues is expected to increase soil organic carbon and overall soil quality. Incorporated
crop residues also support recycling of essential nutrients in the soil and, from long-term perspective, improve soil fertility and have positive
impact on yields. However, mineralization is a complex process driven by weather, soil mixing efficiency, soil moisture and nutrients
available for microorganisms, and also by the ratio of C (mostly lignin) to N and P in incoming litter. Therefore, EPIC provides quite variable
results as these major drivers vary in time and space. Most importantly, high quantities of soil-available N and P are used by
microorganisms during plant residue decay (immobilization) which may also negatively impact yields in the following year as nutrients are
then lacking for plants. These processes are explicitly included in C, N and P routines in EPIC. Repeated and intensive straw ploughing may
therefore have negative effects on yields under management with generally low nutrient inputs. Moreover, other processes including
leaching, erosion, runoff, or nitrification/denitrification determine fate of crop residue nutrients and introduce variability into our results.
76
We currently assume substitute material for bedding available at 22-32 Euro per m3 of wood chips (0.42 m3/t wood chips) depending on
country according to the Finnish Forest Research Institute (Asikainen, Liiri, Peltola, Karjalainen & Laitila, 2008). One ton of straw requires
1.5 tonne of wood chip in the substitution due to different absorption rate.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 141

These latter uses are rather small at the EU level (5.5%, 4.8% and 1.5%, respectively, of total
residue demand) and are considered as fixed in the model. In total cereal straw uses amounts in the
EU28 to 63 Mt per year.

Sustainable straw potentials after removal of other uses (Table 16 and Table 17) are consistent with
estimates from the Biomass Futures project (BIOMASS FUTURES, 2012), as illustrated in Figure 29
below. The few available sources disagree however on the straw collection potential and the amount
of residues already used. For instance, for the three regions of interest here, HGCA (2014) reports
significantly lower availability of residues for the UK (9.5 Mt) than the estimate we rely on. LUC value
associated with 1% extra demand could then be underestimated for this region. For France, ADEME
(2002) also reports smaller residue potential (25 Mt) and higher uses (17 Mt) compared to ECOFYS
(2013). However, for the latter, a sufficiently large potential remains, and results would not be
expected to change in France would the potential be reduced accordingly.

Regional markets: Straw is usually not being traded on long distance (usually transported within
500 km maximum). Therefore, we base our representation of local markets on NUTS1 region supply
as a general rule, because their size corresponds approximately to this order of magnitude. However,
for regions where NUTS1 are of relatively smaller size (Germany, UK, the Netherlands), larger units
were considered (for instance Great Britain as a whole for the UK). For regions with straw deficit
(Netherlands), import needs were added to the demand of neighbour countries.

Implications for model results

With this representation of cereals residues, it will be possible to look at land use implications of
increasing straw removal. A shock in straw demand at the NUTS1 level will lead to cereal production
cost increase in each of the three countries of focus. Residue prices will increase, but will be capped
by the price of the substitution materials. Grains, as a joint product with straw, will be affected by
the extra demand for residues. Primarily on its price, related to the change in price of residues. But
also in production level, as food and feed demand reacts to prices. Beyond the sustainability
threshold, soil organic carbon stock will also be impacted. Effect on cereal yields will also be looked
at in the case of a sensitivity analysis, for cases where the suitability threshold is reached. The focus
on different regions will help to understand the regional nature of results for straw removal due to
the limited extent of trade for this material.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 142

Table 16: Cereals straw balance in 2000 for EU Member States, and impact of a 1% demand shock of bioenergy from

straw (1000 tonnes)

Gap
Gap
compared 1%
Sustainable after
Technical Supply - to national Share
Country Demand Potential 1%
Potential demand sustainable demand wheat
(40%) national
straw straw req
supply
available
AT 382 2,535 1,126 744 66% 25% 461 50%
BE 556 1,476 656 100 15% -73% 577 78%
BG 289 4,741 2,107 1,818 86% 79% 148 82%
CZ 228 5,158 2,292 2,064 90% 73% 399 64%
DE 6,010 32,718 12,360 6,350 51% 23% 3,489 51%
DK 3,635 4,799 2,133 -1,502 -70% -85% 305 47%
EE 46 638 284 238 84% 65% 54 24%
ES 2,949 16,823 7,477 4,528 61% 28% 2,413 41%
FI 179 3,818 1,697 1,518 89% 72% 298 13%
FR 11,382 36,756 20,420 9,038 44% 30% 2,945 76%
GR 795 2,617 1,163 368 32% -9% 476 84%
HU 3,522 5,201 1,907 -1,615 -85% -100% 300 74%
IE 1,191 1,702 756 -435 -58% -99% 314 31%
IT 1,664 9,094 4,042 2,378 59% -7% 2,653 83%
LT 148 1,952 868 720 83% 73% 84 49%
LU 30 77 34 4 12% -446% 156 26%
LV 64 958 426 362 85% 67% 75 46%
NL 1,311 1,087 483 -828 -171% -336% 797 71%
PL 18,427 24,179 10,746 -7,681 -71% -79% 758 40%
PT 396 563 250 -146 -58% -229% 427 66%
RO 1,240 7,595 3,376 2,136 63% 55% 287 79%
SE 289 4,699 2,089 1,800 86% 62% 505 40%
SI 77 198 88 11 13% -108% 106 76%
SK 112 2,138 950 838 88% 76% 119 65%
UK 7,740 17,698 7,866 1,230 16% -21% 2,917 66%

Not reported: Malta, Cyprus, Croatia. Demand data for year 2000 were used applying Scarlat et al. (2010) coefficients on

livestock number, or when Ecofys data for recent years were available.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 143

Table 17: Selected three case studies for the marginal 1% shock at national level (1000 tonnes)

Remaing
Remaining sustainab Straw
Sustainable Sust.
Technical sustainable le straw required Share
Country Demand Potential Supply -
Potential straw after 1% for 1% wheat
(40%) demand
available biofuel biofuel
demand
Centre
3,103 20,253 11,252 8,149 72% 50% 2,509 80%
France*
Great
7,034 17,210 7,649 615 8% -24% 2,485 66%
Britain
Hungary 3,522 5,201 1,907 -1,615 -85% -98% 255 74%

* This value excludes imports demand from the Benelux.

Figure 27: Crop yield relative change (%) in the EU when removing 90% of residues compared to 40%. Ten

representative years are shown with their representative climate, after 20 years of removal. Estimates are sourced

from the EPIC crop model simulations in all cropland location in the EU. Boxes indicate the first and third quartile of

values and whiskers the 5%-95% range

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 144

Figure 28: Soil organic carbon (SOC) relative change (%)in the EU when removing 90% of residues compared to

40%. Ten representative years are shown with their representative climate, after 20 years of removal. Estimates are

sourced from the EPIC crop model simulations in all cropland location in the EU. Boxes indicate the first and third

quartile of values and whiskers the 5%-95% range

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 145

Figure 29: Comparison of straw potential for bioenergy use with data from BIOMASS FUTURES (2012). BIOMASS

FUTURES data has been converted from ktoe to Mt wet matter using a LHV of 18 GJ/t and a dry matter content of

85%. IIASA data corresponds to the year 2000 while BIOMASS FUTURES refers to 2004

II.2 Include carbon sequestered in annual and perennial crops

Motivation for improvements

LUC drives GHG emissions in particular due to changes in carbon stocks from different biomes. So
far, in GLOBIOM, carbon stocks were represented for forest, following statistics from the Forest
Resource Assessment 2010 (FAO, 2010) and for grassland, and other natural vegetation, based on
the Ruesch & Gibbs (2008) database. However, cropland was not covered. This could lead to some
bias in the emissions associated with cropland expansion because some carbon can be sequestered
in crops during the harvest cycle, and this for several years in the case of perennial crops.

Methodological approach
Different sources were applied to collect carbon stock values associated to each crops. In the case of
annual crops, we used the EPIC crop model information, directly related to the management of crop
in each simulation unit. EPIC provides crop yield but also dry matter living biomass produced per ha.
The IPCC (2006) default value of 0.47 tonne C per tonne dry matter biomass recommended for
herbaceous biomass was applied. The carbon stock of annual crops was then multiplied by the
fraction of the year during which crops are grown and divided by 2 on the basis of the assumption of
a linear growth.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 146

In the case of semi-perennials or perennial crops (sugar cane, miscanthus), the crop calendar was
applied over the number of years of the plantation cycle. Results for EU crops are provided in
Table 18.

A particular case is palm plantations that is modelled as a crop activity in the model but contains
carbon stocks that are those of tree plantations. Carbon stock default value for palm oil tree from
IPCC (2006) is 68 tonnes C per ha for a mature plantation (136 tonnes dry matter above ground
biomass x 0.5 tonne carbon per dry matter tonne as for woody biomass; see IPCC AFOLU Guidelines
Chap 5, Table 5.3). However, a typical rotation period for palm oil is 25 years and palm trees are
continuously growing on this time period. Therefore, the IPCC value needs to be corrected to account
for the growing stock. Khasanah et al. (2012) consider based on several site studies an average of
40 tC/ha on the life-cycle on a plantation, using growth profiles that are consistent with IPCC values
for a mature plantation. Therefore, the estimate of 40 tC/ha appears appropriate for above biomass
of palm plantations in GLOBIOM. For calculation of below-ground biomass, we also rely on IPCC
below to above biomass ration of 0.2 (subtropical humid forest; above biomass lower to 125 t dry
matter per ha).

Implications for model results

Accounting for cropland carbon stocks should lead to a more comprehensive calculation of CO2
emissions resulting for LUC, when cropland expands in another land use type or is converted to
another use.

Table 18: Above- and below-ground average carbon stock in living biomass for GLOBIOM crops, and annualized

stock values. Crops stocks are aggregated at the global level based on their location in 2000

Annualized carbon stock

Carbon stock at
Crop in living biomass
harvest (tC/ha)
(tC/ha)
Barley 5.8 1.5
Dry beans 3.4 0.9
Cassava 2.4 0.6
Chick peas 2.5 0.6
Maize 9.6 2.4
Cotton 6.3 1.6
Groundnuts 8.6 2.2
Millet 4.2 1.1
Potato 3.3 0.8
Rapeseed 2.5 0.6
Rice 8.3 2.1
Soybean 6.3 1.6
Sorghum 3.5 0.9
Sunflower 6.4 1.6
Sweet potato 3.1 0.8
Wheat 5.6 1.4
Flax* 3.0 0.8
Peas* 3.2 0.8

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 147

 Annualized carbon stock
Carbon stock at
Crop in living biomass
harvest (tC/ha)
(tC/ha)
Sugar beet* 5.5 1.4
Fallow* 4.2 1.1
Fodder* 3.8 1.0
Maize silage* 2.4 0.6
Oats* 8.5 2.1
Rye* 4.1 1.0
Grassy crop** 6.2 6.2
Sugar cane*** 13.4 13.4
Oil palm -- 48.0

* for EU only, ** On the basis of miscanthus annual stocking rate for a yield of 10 dry matter ton per year *** Carbon stock after

one year, harvest after two years

II.3 Update peat land emission factors

Motivation for improvement

Past studies on LUC have found biodiesel consumption to impact palm oil production, directly or
indirectly. Expansion of palm plantations in Indonesia and Malaysia, which represents 80% of global
production (FAO, 2014), has occurred for a significant share on tropical peat soils (Gunarso, Hartoyo,
Agus, & Killeen, 2013). As a consequence of soil being drained, the peat starts to slowly decompose
and can emit greenhouse gas (GHG) for several decades.77 Although the number of studies that
estimate these peat emissions has increased tremendously during the last decade, the scientific
community has not yet reached a consensus on an appropriate range of emission factors. For
instance, revision of IPCC guidelines for wetlands has been hotly debated (IPCC, 2013).

So far, GLOBIOM could only take into account this type of emissions by mapping cropland area to the
organic soil emissions reported in FAOSTAT at the national level. The objective of this improvement
is to define, on the basis of the existing literature, a more specific range of emission factors for peat
drainage78 for the LUC simulations. This range of values will then be used in the Monte-Carlo
simulation, i.e. an iterative approach to take account of a range of plausible emission values.
Moreover, this section also aims at providing an overview of the most prominent drivers of GHG
emissions from peat and explains the most critical methodological issues that may explain the
scientific disagreement observed between the various author groups.

77
CO2 is by far the most prominent GHG, accounting for about 98% of all peat-related GHG’s (Hergoualc’h and Verchot 2013; Schrier-Uijl et
al. 2013). CH4 and N2O, the later mainly upon application of mineral fertilizers, constitute the remainder of the total GHG emissions. In the
present literature review we thus focus on the role of CO2.
78
Due to methodological uncertainties, immediate emissions from peat fires and emissions from peat drainage in forests adjacent to
plantations will not be considered in the LUC assessment.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 148

Methodological approach
Our analysis of potential emission factors builds here upon a wide examination of past literature.
A number of determinants were identified that drive the pace and the magnitude of GHG emissions
from peat:
 The level of the water table (drainage depth) directly determines peat decomposition rate.
Several scholars provide estimates of GHG emissions per additional centimeter of drained soil
(Hirano et al. 2012; Hooijer et al. 2006; Wösten et al. 1997).
 Natural respiration variability and timing of measurement: Both intra-annual changes
(e.g. temperature and rainfall distribution over the year) and inter-annual changes, such as the
el Niño phenomenon can explain a significant part of the variability observed in measurement.
Additionally, peat respiration curves show a tendency to peak over the 5-10 years after drainage
followed by a flattening of the emission curve (Page et al., 2011), which needs to be taken into
account to provide reasonable emission factors.
 Current and past land use and management: land use and land management affect the
level of peat oxidation and thus the measured emission flow. For instance, fertilization practices
stimulates microbial soil activity and can increase peat emissions; furthermore, different types
of land use imply different drainage depths (Dariah et al., 2013).
 Peat bulk density (BD) and the fraction of carbon in soil influence peat decomposition
rates. BD values vary throughout the soil profile and need to be sampled with care as they feed
directly in the formula for emissions in the case of studies based on measurement of the soil
subsidence (Melling and Henson 2011; see Box 1).
 The measurement method used to estimate fluxes of GHG from peat to the atmosphere,
namely measurements of soil subsidence, of direct flux measurement through closed chambers,
and measurements by Eddy Covariance techniques (see Box 7).

Table 19 and Table 20 provide an overview of 12 studies based on subsidence and closed chambers,
respectively, and lists some of the determinants mentioned above. No Eddy covariance studies were
found for oil palm plantations.
Subsidence studies find the highest potential emissions, due to the full accounting of the emission
cycle along the exploitation process of plantation. The method strength relies on the explicit
representation of peat oxidation process but due to the long period of study required, only a few
estimates are available. Estimates critically depend on the subsidence rate. Hooijer et al. (2012) find
the highest estimates as they also account for the initial subsidence in the few years following the
drainage, whereas other studies look at emission fluxes for a period after 5 years of drainage.
Close chambers studies are more numerous but the range of their results is highly variables. Earlier
studies were flawed by methodological problems, such as interference of root respiration,79 too short
periods of measurements and bias due to the time of measurement in the day. Figure 30 shows that
closed chamber estimates tend to increase over the past recent years and the most extreme points
corresponds to non-peer-reviewed results (Melling et al., 2007; Agus et al., 2010; Comeau et al.
2013). The lowest published value is from Dariah et al. (2013) with measurements at 34.1 and 38.2
MtCO2-eq ha-1 yr-1 and the highest to Husnain et al. (2014) with 66 MtCO2-eq ha-1 yr-1 and
Jauhiainen et al. (2012) with 80 MtCO2-eq ha-1 yr-1.

79
Trees on the plantation site emits CO2 through root respiration (autotrophic respiration) which is also captured by closed chambers

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 149

This last estimate is however sourced from several acacia sites, and authors disagree on whether
such flux chamber measurements are directly transposable to the case of palm oil plantations.80
The three to four research groups publishing actively on peat land emissions also authored a number
of literature reviews (Table 21). Their usually reported estimates and recommendations vary. Page et
al. (2011) repeatedly find high emission rates and recommended a value of 95 t CO2-eq ha-1 yr-1
(based on Hooijer et al. 2012) while the group around Agus usually reports emissions of 43 t CO2-eq
ha-1 yr-1 (Agus et al., 2013; A. Hooijer et al., 2010, 2012). IPCC (2013) chose a Tier 1 emission
factor of 40 tCO2-eq ha-1 yr-1.

Sources of uncertainty are too large to lead to a narrow estimate of peat emissions in South-East
Asia given these emission estimates. To derive our final range of values, we proceed in two steps:
 Filtering of studies: all studies analysed in this review are not equal in terms of level of
details, robustness of the methodology and validation of the results. To improve the quality of
our reference values, we consider here as relevant only the values produced by studies
respecting two criterias: i) peer-reviewed and ii) in the case of closed chambers, we only
consider studies separating autotrophic (root respiration) and heterotrophic (peat oxidation)
calculations, a bias that can play a significant role around trees (Dariah et al., 2013). As a
consequence, three field studies are removed from our sample: Melling et al. (2007), Agus et al.
(2010), Comeau et al. (2013).81 This particularly leads to removal of the lowest and highest
values in our closed chamber range for palm oil. In addition, although we kept in the sample the
measurements on acacia, we displayed them separately due to the on-going debate about
differentiated impact of peat drainage for palm oil and for acacia.82
 Distribution of emissions: if we follow the subsidence method, emissions depends on different
uncertain multiplicative drivers, among which oxidation rate, peat bulk density and subsidence
rate (related to water table level). There is no large scale dataset on the distribution of these
factors over the regions of interest for our study. If we assume such values are symmetrically
distributed and independent, the resulting distribution should be log-normal shaped.83 This
profile is confirmed by observation with flux chambers (see for instance records from Dariah et
al. (2013). Based on some distribution of oxidation rates in the range 40-92% (Page et al.,
2011; Hooijer et al., 2012), 0.06-0.12 g cm-3for peat bulk density at 55% C (Jauhiainen et al.,
2012) and a water table of 0.6-0.85 m (Page et al., 2011), we can reproduce a distribution
profile consistent with the literature. The mean of the distribution is 61 ± 22 tCO2 ha-1 yr-1.

80
The question whether acacia and palm oil should be considered similar is still unresolved. IPCC (2013) rejected comments from the US
government to consider acacia and palm oil plantations equivalent pointing four reasons that could justify differences: i) shorter rotation
time of 6 years versus 25 years leading to higher soil disturbance, ii) difference in fertilization and nitrogen cycle, iii) larger depth of
drainage for acacia iv) different regions of plantations. However, they also acknowledge that studies could have reported different values for
the two types of plantations due to some different sites being looked at. Some more recent studies (Couwenberg and Hooijer, 2013 on
subsidence; Husnain et al., 2014 on closed chambers) suggest that differences might not be as high as for the currently proposed Tier 1
emission factor from IPCC (11 tC ha-1 yr-1 for palm oil and 20 tC ha-1 yr-1 for acacia) and can be of similar magnitude for a same site
(Husnain et al., 2014).
81
Some of the sources above appear in particular very little documented. Melling et al. (2007) is only three pages of explanations, not peer-
reviewed; Agus et al. (2010) is the same level of details, with a very succinct results section and no peer-review. Comeau et al. (2013) is
more developed, but do not distinguish the effect of root respiration (autotrophic respiration).
82
See footnote 4 above.
83
A log normal distribution is the distribution of a random variable whose log value is normally distributed. It is typically characterized by a
longer right tail and a mean value higher than the median value. Log-normal distributions are usually observed when evenly distributed
random variables are multiplied together. For illustration of the role of log-normal distribution in science, see (Limpert, Stahel, & Abbt,
2001)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 150

 The median value is 58 tCO2 ha-1 yr-1 and the confidence interval at 95% in the range 27--112
tCO2 ha-1 yr-1.
The first quartile of the distribution is at 44 tCO2 ha-1 yr-1, which is in the magnitude of the Tier 1
value of IPCC (2013). The third quartile is 74 tCO2 ha-1 yr-1, which is above most closed chamber
measurements, but below measurements on acacia plantations (Jauhiainen et al., 2012).

Limitations to this approach: Average range of subsidence assumed here is 5cm/yr, with a
confidence interval of 3.4cm yr-1 to 7 cm yr-1. This is in line with most records from subsidence at
steady state, but this does not account for the emission peak of the first five years described by
Hooijer et al. (2012) on the observation of an acacia plantation. However, we did not find any study
quantifying the effect of such a peak on a palm plantation. The peak effect can be partly represented
through high bounds of peat bulk density, typical of higher layer of peat, and the higher values of our
subsidence rate, but might be underestimated compared to Page et al. (2011) for instance. More
comprehensive information on subsidence rate distribution could help overcome this caveat but is not
yet available. Another limitation comes from the assumption that some variables are independent,
such as oxidation rate and age of the plantation (reflected through subsidence rate). The most recent
publications suggest that the oxidation rate should increase with the age of plantation and
compensate the decrease of subsidence rate with the plantation aging (Hooijer et al. (2012),
Couwenberg & Hooijer (2013)). However, information to properly quantify this relation on a
systematic basis is not yet available.

As more sites will be monitored over time, we can progressively expect better information on key
drivers distributions on oxidation rates, subsidence rate and peat bulk densities. For the time being,
we base our final range on the simple subsidence relation above, which covers the current
observations for an average water table level of 0.6-0.85 m with a mean value of 61 ± 22
tCO2 ha-1 yr-1 and a 95% confidence interval of 27--112 tCO2 ha-1 yr-1. As a more
comprehensive set of records from the literature will be recorded, this method will be able to evolve
and the range be refined.

Implications for model results

This range of peatland emission factors will be used with the expansion pattern of improvement 29 to
account for impact of palm plantation expansion. These emissions being highly uncertain, they will be
clearly identified in the total accounting of LUC emissions.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 151

Box 7: Measurement methods used for peatland emissions

Subsidence: Drainage of peat results in the increase in the oxidation of carbon and the transfer of carbon to the
atmosphere. The removal of carbon from peat results in its shrinkage and the increase in its bulk density, thereby
resulting in peat subsidence. CO2 emission estimates may be based on peat subsidence as they are related to one
another. Carbon loss is calculated using the formula:

Where St is the surface height loss, DBD1 is the dry peat bulk density and Cdw is the volumetric carbon density of
peat below the water table (the product of carbon bulk density and carbon concentration in the peat). Estimation
of the contribution of the oxidative component to overall subsidence is critical in order to infer soil CO2 emissions.
Assumed or calculated oxidation rate is a source of uncertainty across studies with values in the range 40-90%
(Wösten et al. 1997; Hooijer et al. 2012; Couwenberg et al. 2009). The subsidence-based CO2 estimates are cost-
efficient, allow for high spatial resolution in the sampling process, and emissions from peat (heterotrophic
emissions) can clearly be distinguished through this method from vegetation emissions (autotrophic emissions).
The technique yields results which are comparable to techniques that measure emissions directly (Page et al.,
2011). Yet, emission estimates are limited to CO2 (no CH4 or N2O) and they critically depend on the estimation of
the oxidative fraction of peat subsidence which is subject to a high uncertainty (Dariah et al. 2013).
Closed chambers: they provide direct measurements of gas fluxes from the soil at discretionary spatial and
temporal resolution, the results of which can be up-scaled to obtain emission factors for given site conditions.
Rigorous measurement scheme allows for reliable measurements of heterotrophic soil emissions. Homogenous
experiment set-up (e.g. chamber location in the micro-relieve and varying chamber sizes) is imperative in order
to obtain reliable results. The advantage of this method is the direct measurement of emissions, and the number
of samples collected so far. However, a certain number of technical challenge weaken the reliability of
measurements, due to the high variability of results, depending on the location and the moment in the year or in
the day where experiment are conducted, and the risk of measurement bias related to the distance to root of
planted trees that can create interference between heterotrophic respiration from peat oxidation and autotrophic
respiration from roots (Page et al., 2011).
Eddy Covariance (EC): method to measure gas fluxes on towers reaching above the top of the vegetation
cover. It is suitable to capture the total GHG balance of larger sites with trees, but it is limited by high costs, low
portability and low spatial resolution. EC studies on peat were presented by Hirano et al. (2007; 2012) but not for
oil palm, thus we did not consider them further in this review84.

84
Similar observations were made on EC by IPCC (2013).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 152

Table 19: Summary of available data from studies based on the subsidence method on plantations on peat (Source: authors’ compilation)

Location Determinants Mean

Average estimated Associated

Study Drainage
Study Affiliation observed Time Observed Oxidation emission range85
Number peer depth Duration of
(year) / funder subsidence after peat bulk rate factor [t CO2- ha-1
reviewed Land use Location [m below estimation
[cm/year] drainage density observed [t CO2-eq yr-1]
ground [months]
[years] [g cm3] or applied ha -1 -1
yr ]
surface]

Wageningen
Wösten et
/ Malaysian Sarawak,
1 al. Yes Oil palm 4.686 0.7 14-28 275 0.1 60% 6187 30-9188
Ministry of Malaysia
(1997)
Agriculture

Singapore-
Hooijer et Sumatra, 0.07–
2 Yes Delft Water Oil palm 4.3-6.589 0.5 – 1.0690 14 - 1991 24 92%93 10994 47-11995
al. (2012) Indonesia 0.0992
Alliance

Couwenber Singapore-
Sumatra, 0.08 –
3 g & Hooijer Yes Delft Water Oil palm 3.2-4.497 0.4 – 0.9 5 – 20 36 ~80%99 62.4 51-75100
Indonesia 0.1398
96
(2013) Alliance

85
Information based on interpretation by authors of this note under assumptions below. Not provided by authors of the studies.
86
The study reports the profile of subsidence for a long period with 4.6 cm/year for 14 to 28 years of age and 2 cm/year beyond. We apply here the subsidence rate of the earlier period.
87
Assume 60% decomposition rate and subsidence rate of 4.6 cm/year. The initial published value in Wösten et al. (1997) is 26.5 tons CO2-eq ha-1 yr-1, on the basis of subsidence of 2 cm/year but correspond to a
more aged plantation (>28 years).
88
Replicating the sensitivity analysis of the authors on peat bulk density.
89
Value observed for plantation older than 5 years old: 5.4 cm/yr on average with standard deviation of 1.1 cm/yr. Very large subsidence rate was observed in the case of acacia in the 5 years after drainage and
used for the calculation on palm oil (142 cm in 5 years).
90
73 cm drainage measured in particularly wet year – usually WT is lower
91
Emissions from first years of plantation are also accounted for on the basis of acacia measurements.
92
Assumption of homogenous BD and carbon content over soil profile
93
Oxidation rate is here directly inferred from bulk density measurements and subsidence rate, assuming steady state in the subsidence process.
94
Estimate is an annualized value over 25 years rotation time – taking into account the first five years of a very high emission level (178 tCO2-eq/yr) and then 73 tCO2-eq/yr. The high initial subsidence rate was
measured on acacia plantation. Assume 70 – 92% decomposition rate.
95
Using the subsidence range on a period of 25 years, with the 5 first year collapse of peat observed in acacia plantation and without it.
96
The study looks at three sites, in same provinces as in Hooijer et al. (2012). Calculation methods differ and the study only looks here at emissions after peat consolidation (over 5 years).
97
Site of young plantation showed subsidence of 3.7 ± 0.5 cm/yr and old plantation 3.9 ± 0.5 cm/yr.
98
Only lower layer value is used in the steady state calculation.
99
Due to the carbon loss difference method chosen, oxidation rate is directly derived from peat density and subsidence measurements.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 153

Table 20: Summary of available data from peat carbon emission studies on plantations on peat

Contextual information Determinants Mean Stdev. or range

Study estimated of estimated

Affiliation / Drainage Separation
Study Study peer emission emission
funder depth Time after Duration of auto- and
number (year) review- Number of [t CO2-eq [t CO2-eq
Land use Location [m below drainage estimation hetero-
ed sites ha-1 yr-1] ha-1 yr-1]
ground [years] [months] trophic

surface] respiration

Tropical Peat

Melling et Research Lab. / Sarawak, 0.6

1 Yes Oil palm 1 7 12101 No 60.6 15-107
al. (2005) Malaysian ministry Malaysia (variable)

of Science

Tropical Peat

Melling et Research Lab. / Sarawak,

2 No Oil palm 1 NA 5 12 Yes 33.6/40.1102 NA
al. (2007) Malaysian ministry Malaysia

of Science

Agus et al. Indonesian Soil Sumatra,

3 No Oil palm 3 0.7 – 1.5 1-10 2103 Yes 19.5 ±13.2
(2010) Research Institute Indonesia

Jauhiainen SDWA / Grant of

Sumatra, 0.45 –
4 et al. Yes Academy of Acacia104 8 7 24 Yes 80105 ±15106
Indonesia 1.39
(2012) Finland

Indonesian Soil
Dariah et Sumatra, 0.52 –
5 Yes Research Institute Oil palm 2 8 10 Yes 34.1/38.2107 ±9.5/15.9
al. (2013) Indonesia 0.58
/ EU FP7 program

100
Using the confidence interval on subsidence rate.
101
Short daily measurement period (two hours) – unclear if measurements are representative
102
Melling and Henson (2011) report 33.6 MtCO2 which corresponds to microbial respiration, Marwanto & Agus (2013) also report other soil emissions not associated to roots.
103
Measurement period of 2 months only.
104
This study is looking at acacia palm but is retained here because it has been largely cited and also discusses application of findings to palm oil plantations.
105
The authors of the study reduce the daytime measurement of 94 tCO2-eq ha-1 yr-1 by 14.5% to account for night temperature correction.
106
After applying the same correction same correction on standard deviation as for the mean.
107
Lower value for a plantation aged of 15 years, higher value for a six-year-old plantation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 154

 Marwanto & Indonesian Soil
Sumatra, 0.59 –
6 Agus Yes Research Institute Oil palm 1 15 12 Yes108 46 ±30
Indonesia 1.27
(2013) / EU FP7 program

CIFOR / Australia,
Comeau et Sumatra, 0.65 –
7 No Norway and EU Oil palm 1 10 9 No 104 ±4
al. (2013) Indonesia 1.05
FP7 program

Tropical Peat

Research Lab. /

Melling et Malaysian ministry Sarawak, 0.56 –

8 Yes Oil palm 3109 1-7 24 Yes 60.1110 ±3
al. (2014) of Science Malaysia 0.66

Husnain et Indonesian Soil

Sumatra,
9 al. Yes Research Institute Oil palm 1 0.2 – 1.4 7 7 - 13 Yes 66 ±25
Indonesia
(2014)111 / EU FP7 program

Source: authors’ compilation

108
No explicit distinction is performed but measurements were performed sufficiently far from the palm tree according to authors.
109
No separation of auto- and heterotrophic emissions
110
Median value for a 5 year-old palm plantation. Authors report 54 and 68 tCO2-eq ha-1 yr-1 for a one year and seven-year old plantation, respectively.
111
The study uses results from two other studies already listed here: Marwanto & Agus (2013) and Dariah et al. (2013). To avoid double-counting we only report here the specific site added by the paper in the
Riau province.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 155

Table 21: Overview of reviews and meta-studies on peatland emissions (Source: authors’ compilation)

Peer- Final range of Recommended

Number of Common
Study (year) reviewed Affiliation / funder estimates [t emission factor [t Comment
studies assumption
study CO2-eq ha-1 yr-1] CO2-eq ha-1 yr-1]
Verwer, Meer,
 Take 60-80 cm 10 per 10cm  Quantitative estimates based mostly
and Nabuurs Alterra, Wageningen -- drainage on Hooijer et al. (2006)
drainage depth
(2008)
 Estimated values (drainage depth,
emission factors) based on studies of
 Average
Melling and the Hokkaido University
Uryu et al. (2008) WWF Indonesia 2 drainage depth 5 - 165 85  Point out the large variations of
of 53 cm
drainage depth as a function of the
weather (e.g. El Niño)
 Per 10cm of
drainage depth
 For 50 –
Univ. Greifswald / 100cm
Couwenberg ≥ 9 per 10cm
x Wetlands -- drainage --  Based on Couwenberg et al. (2009a)
(2009b) depth, 40% of drainage depth
International subsidence
caused by
oxidation
 Drainage depth
7 studies
 0.95 m (0.80 –
Hooijer et al. reporting 1.10 m)  Relation of WT-depth and emissions
x Deltares, SDWA  0.91 t CO2-eq 73 - 100 86 based on Hooijer et al. (2006)
(2010) water table
ha-1 yr-1 per
depth cm of drainage
 Recommended value based on Hooijer
Univ. Leicester /
et al. (2012)
International council  Drainage depth  Exclusion of some studies for
Page et al. (2011) 12  0.6 – 0.85 m 54 - 115 95 methodological flaws
of Clean
 Annualized value over 30 years
Transportation (ICCT) rotation time
11  Meta-model based on sample of
 Drainage depth
Hergoualc’h and CIFOR/ Grants from studies (input-output method)
x (2 for oil 0.60 m (0.55 – 24.1 – 44.1 34.1  Includes CH4 and N2O (ca. 2% of
Verchot (2011) Australia and Finland 0.65)
palm) total emissions)

Tropical Peat
Melling and
x Research Laboratory 19 (8) -- 33.6—89.8 --  Review also CH4 and N2O
Henson (2011)
Unit, Malaysia

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 156

 Peer- Final range of Recommended
Number of Common
Study (year) reviewed Affiliation / funder estimates [t emission factor [t Comment
studies assumption -1 -1 -1 -1
study CO2-eq ha yr ] CO2-eq ha yr ]
Indonesian Soil  Recommended value based on
recalculation by Agus et al. (2013) of
Research Institute / results by Hooijer et al., (2010) and
Agus et al.  Drainage depth
Round Table on 14 20--95 43 proposed correction factor for root-
0.5 – 0.7 m
(2013) related respiration by Jauhiainen et
Sustainable Palm Oil al. (2012).--> Approach challenged
(RSPO) by Schrier-Uijl and Anshari (2013)
CIFOR / Grants from  Meta-model based on sample of
Hergoualc’h and studies
x Australia, Norway 28 0.65 +/- 0.05 m 35.2-50 45.1  Includes CH4 and N2O (ca. 2% of
Verchot (2013)
and EU FP7 program total emissions)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 157

Table 22: Subsidence parameters selected for our distribution of peatland emission factors and results (Source: authors’ calculations)

Range Distri- Source

Parameter Notation Unit Comment
Min Max bution Min Max
Typical drainage for oil palm cultivation is supposed to be 0.7 m for
Hergoualc’h
Page et al. oil palm and recommended depth is 60-80 cm (Verwer et al. 2008a;
and Verchot
(2011); Mutert et al. 1999). Lower depth can be observed (Agus et al., 2013
Water table WT cm 60 85 Uniform (2013);
Hooijer et al. report low bound at 0.5 m) but can also be much higher in
Page et al.
(2010) industrial plantations (Hooijer et al., 2010). We follow Page et al.
(2011)
(2011) that cover the most common values also found in Table 1-3.
Wösten et al. (1997) were the first to propose an average
coefficient of 0.07 to link water table and subsidence rate. His
proposed range is 0.04-0.09, however, his measurement for the low
bound correspond to a plantation more than 30 years old. We
therefore conserve the symmetry around 0.07. Couwenberg et al.
Wösten et al.
(2010) find a coefficient of 0.09 for the first 50 cm but suggest the
Subsidence / cm yr-1 Wösten et (1997);
r 0.05 0.09 Uniform correlation could be not applying beyond this depth. Hooijer et al.,
Drainage depth cm-1 al., (1997) Couwenberg
(2012) examine the relation for an acacia plantation and find a
et al. (2010)
slope of 0.0498 with however an intercept value of 1.5 cm yr-1. For
0.7m drainage, this regression is consistent with the linear relation
from Wösten et al. (1997). They note that they could not find a
clear relation on the palm plantation with more homogenous
subsidence rates.
Jauhiainen
Peat bulk density profiles are reported in Hooijer et al. (2012),
et al. Jauhiainen et
Couwenberg and Hooijer (2013), decreases significantly along the
(2012); al. (2012) ;
peat profile, that vary between 0.06 to values up to 0.15 g cm-3 for
Couwenberg Hooijer et al.
the top 10 cm. Couwenberg et al. (2010) used a density of 0.068 g
Peat bulk density BD g cm-3 0.06 0.12 Uniform et al. (2010) ;
cm-3 for lower peat layers and Jauhiainen et al. (2012) values in the
(2010) ; Couwenberg
range 0.06-0.12 g cm-3. Couwenberg and Hooijer (2013) observe
Couwenberg & Hooijer
values around 0.12 g cm-3 for the upper 0.5m peat layer and around
& Hooijer (2013)
0.08 g cm-3 for lower layer.
(2013)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 158

 Couwenberg Couwenberg et al. (2010) report range in the literature of 35-100%
Jauhiainen et
et al. but applies in his calculation a range of 40-60%. Page et al (2011)
al. (2012) ;
Oxidation rate Ox % 40 92 Uniform (2010) ; performs various analysis using 40% and 60% oxidation rate.
Hooijer et al.
Page et al. Jauhiainien et al. (2012) find higher oxidation rate of 80% and
(2012)
(2011) Hooijer et al. (2012) report a measure oxidation rate of 92%.
Jauhiainen et al. (2012), Hooijer et al. (2012), Couwenberg &
Couwenberg Hooijer (2013), Couwenberg et al. (2010) all use a carbon fraction
et al. of 55%. Page et al. (2011) use 60%, whereas Agus et al. (2013)
Page et al.
Carbon fraction Fc % 50 60 Uniform (2010) ; note that variation of carbon fraction over the peat profile must be
(2011)
Agus et al. better taken into account. Couwenberg et al. (2010) report some
(2013) possible slightly lower carbon fraction on peat with average in some
samples at 50%.

RESULTS
Unit Range (95%) Distribution
Parameter Notation Mean Comment
Min Max 25% 50% 75%
The range of subsidence obtained covers well values reported by
Subsidence rate
S cm yr-1 3.4 7.0 5.1 4.3 5.0 5.8 the subsidence literature (see e.g. discussion in Hooijer et al.
(=r ∙ WT)
(2012))
Emission per cm
tCO2 yr- Emission per cm drainage range encompasses here values from
drainage 1 -1
e ha 0.39 1.52 0.84 0.62 0.8 1.02 Agus et al. (2013): 0.72, Hooijer et al. (2010): 0.91 and Jauhiainen
(=100 ∙ r ∙ BD ∙ Ox ∙
cm-1 et al. (2012): 0.71 with an intercept.
Fc ∙ 44/12)
Emission factor tCO2 yr- Emission factor obtained match well the range from filtered
EF 1
27 113 61 44 57 74
(= e ∙ WT) ha-1 literature as shown in Figure 31.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 159

Figure 30: Distribution of central estimates of studies and range of uncertainty. Values are reported according to

Table 19 & Table 20 statistics. For papers analyzing different sites, the different findings were reported

separately. For subsidence, the range of uncertainty corresponds to sensitivity analysis on subsidence rate or peat

bulk density. For Flux chamber studies, the standard deviation is reported, except for Marwanto & Agus (2013)

where only the min and max were available. In each group, results are ordered by year of publication. Acacia

values are displayed separately

Figure 31: Distribution used for the distribution of emission factors based on simplified subsidence assumptions

and comparison with literature values. The upper part of the graph shows the distribution assumed. The lower

part features the articles used in Figure 30, after a filtering process. Numbers in the lower part refer to the study

names in Figure 30

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 160

 II.4 Represent expansion of oil palm plantations into peat land

Motivation for Improvement

Palm oil production is a significant source of GHG emissions when new plantations are developed
on peat land. Until recently, spatially explicit data on recent development of plantations in
Southeast Asia was scarce. As a consequence, future expansion patterns are difficult to predict,
and the number of local drivers and extent of policy-driven uncertainty make this dynamics difficult
to model. Currently, GLOBIOM does not represent in Indonesia and Malaysia internal transportation
costs, which means plantation expansion is only allocated on a crop suitability basis. Therefore, it
was decided in the context of this project to ground assumptions on location of palm plantation
expansion on the basis of current literature findings. The model will then use different possible
allocation within a plausible range as an input in the Monte-Carlo sensitivity runs. Expansion into
peat land will then be associated emission factors derived from Section II.3.

Methodological approach
We look at the literature findings on two different aspects: first, the estimation of current peat land
occupied by palm oil plantation, to get insight into the average share of plantations that expanded
into peat land in the past, and the trend of expansion; second, the projection patterns assumed for
future expansion of plantations into peat land, also a topic of exploration of some papers.
Historical expansion: We reviewed five studies which assess recent development of palm
plantation on peat based on a remote sensing analysis (see Table 23).112 Gunarso et al. (2013)
report that in 2010, palm plantations grown on peat accounted for 1.7 million ha in Indonesia and
ca. 721,000 ha in Malaysia, which represents 22% and 18% of the total plantation area,
respectively. Miettinen et al. (2012) found 1.3 million ha in Indonesia and ca. 780,000 ha in
Malaysia. However, a considerable area of Peninsular Malaysia and East Kalimantan was not
included in their analysis due to persistent cloud cover on satellite images, which might partly
explain the lower estimations as compared to Gunarso et al. (2013). The third study covering both
countries found notably lower numbers with ca. 508,000 ha of converted peat land in Indonesia
and ca. 371,000 ha in Malaysia (Koh, Miettinen, Liew, & Ghazoul, 2011). This difference is likely
due to the coarser scale satellite imagery applied in their study113, which did not allow for
identifying i) immature palm plantations (<80% canopy cover) and ii) small patches of plantations
(<200 ha). Palm expansion after 2002 could therefore not be considered, whereas the annual
expansion rate ranged between 8-10% in Indonesia and 3-6% in Malaysia on that period (Gunarso
et al., 2013).

Dynamics of expansion: we analyse in details the expansion patterns provided by Gunarso et al.
(2013).114

112
Some older assessments also exist that have based their estimation on analysis of palm concession maps and not on remote sensing
analysis of current planted areas. This usually led to higher estimate of peat land occupation (for instance, 25% in Indonesia in Hooijer et
al., 2006) because the full concession area can be sometimes little developed.112 This however suggests that future expansion could still
drive larger share of expansion into peat land, if all currently attributed concession areas were developed.
113
For most papers, current distribution of plantations were analysed using high to medium resolution satellite imagery (5-30m
resolution). Visual interpretation of satellite images followed by manual delineation of oil palm stands was the most common approach to
identify plantation areas (Carlson et al., 2012; Gunarso et al., 2013; Miettinen et al., 2012), sometimes combined with object-oriented
digital classification (Omar et al., 2010). Koh et al. (2011) use lower resolution imagery (250m resolution) and applied a different
classification algorithm.
114
We rely here on data on planted areas from Gurnaso et al. (2013). It is noteworthy that these statistics based on remote sensing
differ from some governmental statistics for planted areas. Areas reported for Malaysia are higher in 2000 (3,467 Mha versus 3,056 Mha

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 161

These show a strong increase in the share of expansion in plantations occurring on peat land in
Malaysia over the years 2000, in particular on the period 2005-2010 where this rate reached 46%.
This increase in share of Malaysian expansion on peat is also confirmed by the Malaysian Palm Oil
Board (Omar et al., 2010) that recorded on the period 2003-2009 a share of 34% of plantation
going on peat.115 Miettinen et al. (2012) found occupation rate even stronger on a short period
with 52% (2007-2010), observing very large implantation on peat in Sarawak.116 In Indonesia,
marginal expansion also goes increasingly to peat, with a rate of expansion of 25.4% in 2005-
2010, versus 22.2% in 2000-2005. This is mainly driven by a strongly increasing trend in Sumatra
(from 28% to 51% in five years) and in Kalimantan (from 5% to 15%). These statistics on the
trend of peat disappearance are in line with observation from Miettinen et al. (2012).

Future plantation expansion on peat: in order to project possible rate of expansion into peat,
we reviewed four additional studies that project the likely future expansion of oil palm plantations
in different provinces. Some works rely on an extrapolation of observed trend on the basis of a
detailed spatial analysis (Miettinen et al. (2012)). Some other studies prefer a spatially explicit
modelling based on suitability criteria (EPA 2012, Harris et al., 2013). A last approach looks at
marginal occupation patterns by assuming areas currently under lease will be developed in the
future (Carlson et al., 2012). Policy intervention in the form of a peat moratorium are sometimes
considered, as in Harris et al. (2013). However, in the case of this latter scenario, macro-regional
land use patterns are not strongly affected.117 Table 24 provides an overview of the ratio of total
future plantation area that is expected to occur on peat according to studies. For Indonesia, the
estimations for 2020 range from 13% for the marginal projection of EPA (2012) to 28% for the
study of Miettinen et al. (2012). For Malaysia, the lower bound is represented by an estimated 7%
for 2020 and 2030 in the BAU scenario by Harris et al. (2013).118
Authors usually disagree on the direction of the trend in marginal expansion patterns. EPA (2012)
and Harris (2013) keep an assumption of constant rate of expansion of peat, but use a 20 years
average and do not take into account the higher levels observed on the decade 2000. Miettinen et
al. (2012) and Carlson et al. (2013) assume increasing trend on the basis of recently observed
estimations. Among sources of uncertainty, an important factor is, on the one hand, the
localization of future production across provinces with very different patterns, and on the other
hand, on the effect of changing policies in each province.

for Malaysian department of statistics) and 2010 (5,230 Mha versus 4,202 Mha). For Indonesia, reported statistics are lower for 2000
(3,678 Mha versus 4,158 Mha for Indonesian Ministry of Agriculture) and closer in 2010 (7,724 Mha versus 7,700 Mha).
115
Using satellite imageries, Omar et al. (2010) report that the average share of plantation on peat increased from 8.2% in 2003 and
13.3% in 2009 in Malaysia, and that 37.4% of plantations in Sarawak were on peat in 2009. This confirms the strong increase on the
recent period even if the rate of occupation is slightly lower than in Gunarso et al. (2013) that report 46% for Sarawak in 2010 and
17.8% for total Malaysia.
116
Miettinen et al. (2012) do not provide statistics on total planted, only on area planted on peat and we use here statistics from Gunarso
et al. (2013). In the case of Sarawak, we find that rates in Miettinen et al. (2012) exceeds 100% of marginal expansion on peatland for
2007-2010, which shows some disagreement between the two studies on expansion patterns of plantation in that region.
117
The moratorium scenario in Harris et al. (2013) affects marginally land use change projections but relies on the assumption that peat
conversion will be completely stopped after 2020, both in Indonesia and in Malaysia, while plantations will go on expanding. In our
approach, we use for our peatland conversion scenario on less extreme scenarios, relying on historical observations on different periods.
118
After examinations of spreadsheets from Harris et al. (2013), it was observed that the low rate for Malaysia was due to a cell error and
that 7% was used instead of 14% for future expansion.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 162

Recalculating marginal expansion rate based on three policy developments: In order to
properly disentangle these effects, we apply the scenarios of projections of plantations based on
Harris et al. (2013) and Miettinen et al. (2012) across provinces, and assume various development
in the trends of marginal expansion in each province. For Harris et al. (2013), we use the business-
as-usual scenario for Malaysia only, due to some too irrealistic trends observed for projections in
Indonesia.119 For the approach from Miettinen et al. (2012), we consider two possible linear trends
in plantation expansion, that we calculate for the 2000-2010 and the 2005-2010 periods.
Last, we also look at what the results would be if the expansion pattern observed over the past
years (2012-2013) would continue for the next decade.120 Projections were considered at the level
of the three regions per country, as in Figure 32. The second effect we isolate is the marginal rate
of expansion in each region to be applied. We consider three different development in each
province:
1 A no regulation scenario (“Trend 10 years”), where the increasing trend on peatland
encroachment observed over the past 10 years go on increasing;
2 a stabilization scenario (“Current stable”), where the expansion into peatland remains at the
level observed on 2005-2010, without further increase;
3 a policy shift scenario (“Return to hist.”), where the trend of expansion into peatland decrease
to come back to historical average on the period 1990-2005.

We did not consider scenarios of complete enforced ban of expansion into peatland due the
continuation of expansion pressure observed in Indonesia121 and the high level of opposition to
such regulation in the most exposed States, in particular in Sarawak.122
The results of our sensitivity analysis on expansion share are presented in Table 25.
As a consequence, we choose to reflect the full range of estimated values in our Monte-Carlo
analysis for average expansion and assume the share of plantation going into peat land on the
period 2010-2030 to be:
 For Indonesia: average of 32% (range 11%-57%)
 For Malaysia: average of 34% (range 14%-52%)

119
Projections from Harris et al. (2013) were leading to lower expansion into peatland due to a slowing down of production in
Sumatra Island. However, the recent trends show that such projections were not realistic, as expansion of palm oil has reached
6.6 Mha in 2013 according Indonesian official statistics. Some serious limitations of the Harris study have also been pointed
during the comment period to the study offered by US EPA (see ICCT, 2012 -
http://www.theicct.org/sites/default/files/ICCT_EPA-palm-NODA-comments_Apr2012.pdf).
120
We base our analysis of recent statistics on data from the Indonesian official statistics (www.bps.go.id) and the Malaysian Palm Oil
Board statistics (http://bepi.mpob.gov.my). According to these statistics, expansion of plantation in Indonesia would have occurred for
60% in Sumatra and for 38% in Kalimantan (in 2013, no data found for 2012). For Malaysia, expansion was 61% in Sarawak and 20% in
Malaysian Peninsula and 19% in Sabah. Overall, two third of expansion took place in Indonesia.
121
USDA reported for the year 2013 10 Mha of oil palm plantation in Indonesia, which challenges optimistic scenarios where production
would have declined in most dynamics regions such as Sumatra, an assumption found in Harris et al. (2013) scenarios.
122
According to the Malaysian Palm Oil Board statistics, plantation expansion in Sarawak would have been 8% in 2013, and total planted
area would represent 1.16 Mha in December 2013. Expansion is most likely to continue as the State of Sarawak has announced an
objective of 3 million ha. International pressure has been put on Sarawak producers to reduce their expansion into peatland, in particular
with the threat of Wilmar, an international oil trader representing half of the Sarawak production purchase to ban palm oil sourced from
plantations on peat. The federation of producers (SOPPOA) opposed this measure and still claims 1.2 million ha more peatland with the
backing of the government of Sarawak preoccupied by the situation of smallholders. Malaysian producers support a Malaysian standard
on palm oil but are critical of standards proposed by the Roundtable of Sustainable Palm Oil, supposed to defend protectionist views of
NGOs and to deny possibility of peat agriculture.
Sources: accessed June 2014.
http://www.theborneopost.com/2014/02/15/standing-firm-against-palm-oil-boycott-threat/
http://www.theborneopost.com/2014/01/17/soppoa-wilmars-declaration-detrimental-to-local-industry/
http://www.thestar.com.my/Business/Business-News/2014/02/18/Planters-Its-unfair/
http://www.newsarawaktribune.com/news/22041/SOPPOA-supports-govt-policy-on-oil-palm-devt-in-Sarawak/
http://mypalmoil.blogspot.co.at/2014/04/sarawak-oil-palm-planters-back-mspo.html
Possibility of other outlets than Western world
http://www.thestar.com.my/Business/Business-News/2014/03/10/Sarawak-plans-to-sell-CPO-in-Middle-East/

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 163

These ranges of values both show mean values and uncertainty bounds of comparable magnitude.
The uncertainty range covers the historical rates observed on the period 2005-2010 (25.4% for
Indonesia and 46% for Malaysia, due to the strong surge in the Sarawak state123).

Implications for model results

Alongside emission factors for drained peat, these estimations of the expansion patterns into peat
will be used to calculate in the model a plausible range of total emissions attributable to oil palm
production in Indonesia and Malaysia. No peat land emissions will be considered for other regions
than Southeast Asia due their more marginal contribution to overall wetlands emissions associated
to palm oil.

123
If trends observed on the period 2005-2010 were to continue in Sarawak, palm plantation could convert the total initial peat area in
that State, i.e. 1.3 to 1.4 Mha (according to Gurnaso et al. (2013) and Mittienen et al. (2012), respectively), by the end of the decade
2020. Factoring in this consideration in the calculation leads to a lower rate for the subsequent period (2020-2030), which explains that
the average rate over 2010-2030 hardly exceeds 50% in our estimation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 164

Table 23: Estimates of total oil palm planting area in 2010 (in ha), oil palm planting area on peat (in ha) and % of total planting area by region for Indonesia and Malaysia

according to four studies with varying coverage

Miettinen et al.
Source Gunarso et al. (2013) Koh et al. (2011) Carlson et al. (2012)
(2012)

Total Planting Share Total Planting Total Planting Share Share

Share Planting area
Region planting area on on planting area on planting area on peat on on
on peat on peat (ha)
area (ha) peat (ha) peat area (ha) peat (ha) area (ha) (ha) peat peat*

Sumatra 4,743,308 1,395,733 29% 3,871,839 464,554 12% n.d. n.d. 1,047,000.00 22%

Kalimantan 2,896,952 307,515 11% 1,100,105 43,184 4% 3,164,005 402,166 13% 314,000.00 11%

Papua 83,622 1,727 2% n.d. n.d. n.d. n.d. n.d.

7,723,88 4,971,94
Total Indonesia 1,704,975 22% 507,738 10% n.d. n.d. 1,361,000 18%
2 4

Omar et al. (2010) (data from Miettinen et al.

Gunarso et al. (2013) Koh et al. (2011)
2008/09) (2012)

Peninsular
1,510,809 215,984 14% 2,005,833 236,820 12% 2,503,682 207,458 8% 238,000 16%
Malaysia

Sarawak 1,033,260 475,946 46% 357,915 103,841 29% 1,167,172 437,174 37% 494,000 48%

Sabah 1,510,809 29,028 2% 918,739 30,166 3% 1,340,317 21,405 2% 50,000 3%

4,054,87 3,282,48 5,011,17

Total Malaysia 720,958 18% 370,827 11% 666,038 13% 782,000 19%
8 7 1

* Calculated from Gurnaso et al. (2013) data on total oil palm plantation areas. Source: authors’ compilation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 165

Table 24: Overview of studies that project oil palm expansion on peat in the future

Percent of plantations
Percent of plantations
on peat
Study Affiliation / on peat (historical) Underlying
Methodology Study area In 2020 (2030) Comment
(year) funder assumptions
Indonesia Malaysia Indonesia Malaysia

Historical average
22% 7%
Indonesia, 22% 18% (constant rate of
Roundtable for (22%) (7%) plantations on peat)
Harris et Malaysia,
Sustainable GEOMOD
al. (2013) Papua New
Palm Oil
Guinea 19% 13% Peat Moratorium (no
22% 18% further expansion)
(17%) (12%)

Miettinen Univ. Singapore

Extrapolation Linear projection
et al. / International Indonesia,
from spatial 18% 19% 28% 42% based on 2007-2010
(2012a, Malaysia
Council on Clean trends
2012b) period
Transportation

Indonesia,
U.S. GEOMOD 15% 10% Historical projection Sensitive to the
Malaysia
EPA Environmental ratio of mature –
22% 13%
(2012) Protection Indonesia, Projected incremental immature palms
GEOMOD 13% 9%
Agency Malaysia expansion Projected to 2022

Development of all oil

Carlson et Univ. Yale &
Static model Kalimantan 13% n.d. 17% n.d. palm leases issued
al. (2012) Stanford
until 2012 by 2020

Source: authors’ compilation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 166

Table 25: Projected expansion on peatland according to literature land use scenarios and local patterns of expansion

on peat

Indonesia 2010-2030 Malaysia 2010-2030

Local expansion pattern on peat Local expansion pattern on peat

Regional
Current Return Current
land use Trend 10 yrs Trend 10 yrs Return to Hist
stable to Hist stable
scenario

Proj. Harris et al.

31% 17% 6% 45% 44% 24%
(2013)

Proj. Linear 2000-

51% 31% 14% 36% 32% 14%
2010

Proj. Linear 2005-

44% 25% 11% 41% 38% 18%
2010

Proj. Linear 2012-

57% 38% 19% 52%b 49% 20%
2013

Min 11% a 14%

Mean 32% a 34%

Max 57% a 52%

a
We did not keep results based on Harris et al. (2013) projections for our summary statistics of Indonesia because projected
numbers were at odd with current developments.
b
This scenario leads to complete use of peatland in Sarawak by 2030, which decrease the expansion rate into peatland in that r
egion at the end of the period. See note 124.
Source: authors calculation.

90%
Share of new plantations into peat

80%
70%
60%
50% Historical avg.
40%
30%
1970-1990
20%
1990-2000
10%
2000-2005
0%
2005-2010

Figure 32: Percentage expansion of plantation into peat land in different regions of Indonesia and Malaysia over

time. Source: Gurnaso et al. (2013)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 167

II.5 Expand the inclusion of soil organic carbon (SOC) worldwide

Motivation for improvements

Soil organic carbon (SOC) is a prominent C stock and its representation is important for a
comprehensive accounting of GHG emissions from agriculture and LUC. In previous versions of
GLOBIOM, a refined SOC accounting design had been developed for the European Union. As ILUC
from biofuels occurs in- and outside of the EU, there was a need to expand the representation of SOC
accounting in the model to other regions of the world.

Methodological approach
We complemented the initial dataset of soil organic carbon in GLOBIOM with data from the
Harmonized World Soil Database v1.2 (HWSD, see FAO et al., 2012). This database is a spatially
explicit layer of soil information in the different regions of the world, such as organic carbon, water
storage capacity, soil depth etc. The information on SOC is here used as input in GLOBIOM at the grid
level. This database therefore complements the EU datasets already in the model (Lugato et al., 2013
for cropland, Jones et al., 2005 for other land use types). A summary of average SOC content by
large region and land use type is displayed in Table 26.
In order to track in the model changes in SOC content in the different regions, we then applied a Tier
1 approach124 based on GHG accounting IPCC guidelines (IPCC, 2006). The formula applied is as
follows (Equation 2.25):

SOC = ∑ (c,s,i) ( SOCREF * FLU * FMG * FI * A) , where:

c, s, i are respectively the climate zones, soil types and management systems in the
region
SOCREF is the carbon stock of reference, calculated using the initial 2000 data and the
initial management information
FLU is the land use factor informing on type of use among crop cover, flooded areas
for rice, perennial crops, or set aside land;
FMG is the management factor informing on tillage practice
FI is the input factor informing on level of fertilizer input and use of manure
A is the land use area for climate zone c, soil type s and management type i.

Default values of FLU, FMG, FI are sourced from Table 5.5 and Table 6.2 of IPCC guidelines. SOCREF is
determined in our setting on the basis of the HWSD and EU specific datasets, correcting for the land
use, management and input factors when relevant. Therefore, we do not use here the averaged IPCC
default value, in order to fully benefit from the spatially explicit information. Differences between
IPCC reference SOC values and HWSD values are documented in Carré et al. (2010).

LUC and management information (crop type, fertilizer level) are calculated endogenously in the
model. Tillage practice is another important component determining level of SOC. Because tillage
practice is not explicitly modelled at the global level in GLOBIOM, we assume for most regions a full
tillage practice for cropland, except for the EU and some large countries where more precise
management information could be retrieved (USA, Brazil, Argentina). The tillage assumptions are
summarized in Table 27. One additional management changed is considered for the agricultural
residues where the effect of removing residues on C stock is accounted for (see Section II.1).

124
Tier 1 approach corresponds to the default methodology proposed by IPCC guidelines, when more local information is not available.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 168

In order to keep the full consistency between SOC in Europe and Rest of the World, and prevent
asymmetrical calculations, the same IPCC Tier 1 formula is here applied for EU and for the Rest of the
World.

Implications for model results

Additional SOC at world level in the accounting will allow a more comprehensive coverage of indirect
emissions related to expansion of agricultural land. The approach remains here however simplified,
based on a Tier 1 approach but follows recommendation from JRC on accounting of soil organic
carbon (Carré et al., 2010). A limitation of the approach is the impossibility to consider C stock
change associated to restoration of degraded land, due to the too limited data on C stock and
locations of degraded areas at world level.

Table 26: Reference level for soil organic carbon in t C/ha by GLOBIOM region. Reference levels correspond to C

stocks in an undisturbed grassland area

SOCREF
Australia / New Zealand 20
Argentina 27
Brazil 34
Canada 58
China 33
Congo Basin 35
Former USSR 48
India 27
Indonesia 48
Japan 56
Malaysia 34
Mexico 29
Mid-East North Africa 17
Pacific Islands 39
Rest of Central America 40
Rest of Eastern Europe 34
Rest of Western Europe 58
Rest of South America 35
Rest of South Asia 22
Rest of Southeast Asia 32
South Africa 20
South Korea 33
Eastern Africa 26
Southern Africa 24
Western Africa 22
Turkey 26
Ukraine 47
USA 35
EU28 59

Note: Reference SOC level is calculated based on HWSD, except for the EU where a more detailed map is used (Jones et al.,

2005). In that latter case, it has been calculated as the average SOC stocks on forests, grassland and other natural vegetation.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 169

Table 27: Average tillage assumption for GLOBIOM regions in the base year
Share
Region Full tillage Share reduced tillage Share no tillage

European Union 76% 18% 6%

Canada 40% 30% 30%
USA 75% -- 25%
Brazil 65% -- 35%
Argentina 30% -- 70%
Paraguay 28% -- 72%
Australia 75% -- 25%
Rest of the world 100% -- --

Source: PICCMAT project data for the European Union http://climatechangeintelligence.baastel.be/piccmat/ , Canadian Agricultural

Census for Canada, Derpsch et al. (2010) for other regions.

II.6 Include forest regrowth and reversion time on unmanaged land

Motivation for improvement

GLOBIOM usually assumed that, when cropland or grassland is abandoned, it can either be turned
into managed forest or forest plantation, if economically profitable, or abandoned into other natural
vegetation. Carbon stock in living biomass for other natural vegetation was taken from the Ruesch
and Gibbs (2008) database. However, even if not actively managed as forest or plantation,
abandoned land can lead to progressive forest regrowth in some regions, with carbon stock higher
than typically assumed for natural grassland, for instance; it was therefore decided to implement a
more detailed representation. The question of the timing of this carbon sequestration is also
important, because the model solves by 10 year time steps and the evaluation period of the policy is
limited to 20 years.

Methodological approach
IPCC (2006) guidelines distinguish two types of carbon reversion associated to forest regrowth,
natural and artificial regeneration.125 Considering that managed forests are already represented in
GLOBIOM, we only look here at the carbon accumulation in land converted into unmanaged forest.
The typical type of conversion of land to Forest is assumed according to IPCC to have an average
transition period of 20 years, after which the land can be considered classified as Forest again. IPCC
default table provides a single value per type of forest for different regions (see overview in Table
28). However, GLOBIOM contains geographically information on the carbon stocks in natural forest
for each Production Unit in the model. We therefore use the geographical heterogeneity of stock
distribution to vary geographically the default value for carbon growth. For each production unit, we
assume that the annual growth rate equals 2% of the total carbon stock in living biomass of a mature
forest.

125
“Land is converted to Forest Land by afforestation and reforestation, either by natural or artificial regeneration (including plantations).
The anthropogenic conversion includes promotion of natural re-growth (e.g., by improving the water balance of soil by drainage),
establishment of plantations on non-forest lands or previously unmanaged Forest land, lands of settlements and industrial sites,
abandonment of croplands, pastures or other managed lands, which re-grow to forest. […] Land conversion may result in an initial loss of
carbon due to changes in biomass, dead organic matter, and soil carbon. But natural regeneration or plantation practices lead to carbon
accumulation and that is related to changes in the area of plantations and their biomass stocks.” [IPCC (2006) Guidelines, Vol 4., Chapter 4,
p. 4.30]

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 170

However, forest regrowth does not systematically take place when land is abandoned, because of
various environmental factors. Indeed, the abandoned land can be too infertile or degraded, as
indicated by IPCC guidelines.126 A land classified as “Other land” (i.e. not cropland, grassland, forest
land, wetland or settlement) is under IPCC accounting rule considered by default not accumulating
carbon (see IPCC, 2006, Vol.4, Chap. 9). To distinguish what share of land can be considered
returning to forest or only to other natural vegetation (with constant carbon stock), we apply the
same method as suggested by the U.S. Environmental Protection Agency (EPA, 2010). Within each
Production Unit, we assume that, forest regrowth takes place at the same share on abandoned land
as is the share of forest already observed on fertile land (forest, other natural vegetation, grassland).

Implications for model results

Carbon balance in the model is changed with this improvement, with better representation of carbon
sequestration when land is abandoned. Former and new carbon stocks associated to abandoned land
are summarised in Table 29. In the former accounting approach, other natural land carbon stock was
allocated to abandoned land using data from Ruesch and Gibbs but without any dynamic
consideration. Therefore, after 10 years, the carbon stock was already maximum, a pattern that is
changed with the new approach, which decreases the carbon stock typically reached after 10 years.
Additionally, the new approach considers possibility to reach higher level of stock directly through
forest regrowth (e.g. in Europe).

Table 28: Carbon accumulation in living biomass for natural forest regrowth in different regions (first 20 years)

Average
Average
above- Ratio
above and
ground below-
below
biomass ground to
Region Ecological zone ground C
growth above
accumulation
(tonnes ground
a
(tonnes C /
d.m. / ha / biomass
ha / yr)
yr)
Tropical and
subtropical zones
Africa Tropical rainforest 10 0.37 6.4
Africa Tropical moist deciduous forest 5 0.20 – 0.24 7.5
Africa Tropical and subtropical dry forest 2.4 0.28 – 0.56 1.6
Tropical shrubland and subtropical
Africa 0.2 – 0.7 0.32 – 0.4 0.3
steppe
Tropical and subtropical montain
Africa 2.0 – 5.0 0.27 2.1
systems
Asia (continental) Tropical rain forest 7.0 0.37 4.5
Tropical moist deciduous and
Asia (continental) 9.0 0.20 – 0.24 13.5
subtropical humid forest
Asia (continental) Tropical and subtropical dry forest 6.0 0.28 – 0.56 4.0

126
“Some abandoned lands may be too infertile, saline, or eroded for forest re-growth to occur. In this case, either the land remains in its
current state or it may further degrade and lose organic matter. Those lands that remain constant with respect to carbon flux can be
ignored. However, in some countries, the degradation of abandoned lands may be a significant problem and could be an important source of
CO2. Where lands continue to degrade, both above-ground biomass and soil carbon may decline rapidly, e.g., due to erosion. The carbon in
eroded soil could be re-deposited in rivers, lakes or other lands downstream. For countries with significant areas of such lands, this issue
should be considered in a more refined calculation.” [IPCC (2006) Guidelines, Vol 4., Chapter 4, p. 4.30]

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 171

 Average
Average
above- Ratio
above and
ground below-
below
biomass ground to
Region Ecological zone ground C
growth above
accumulation
(tonnes ground
a
(tonnes C /
d.m. / ha / biomass
ha / yr)
yr)
Tropical shrubland and subtropical
Asia (continental) 5.0 0.32 – 0.4 3.2
steppe
Tropical and subtropical mountain
Asia (continental) 1.0 – 5.0 0.27 2.1
systems
Asia (insular) Tropical rain forest 13 0.37 8.4
Tropical moist deciduous and
Asia (insular) 11 0.20 – 0.24 16.5
subtropical humid forest
Asia (insular) Tropical and subtropical dry forest 7.0 0.28 – 0.56 4.7
Tropical shrubland and subtropical
Asia (insular) 2.0 0.32 – 0.4 1.3
steppe
Tropical and subtropical mountain
Asia (insular) 3.0 – 12 0.27 4.5
systems
North America Tropical rain forest 0.9 – 18 0.37 6.1
South America Tropical rain forest 11 0.37 7.1
North and South Tropical moist deciduous and
7.0 0.20 – 0.24 10.5
America subtropical humid forest
North and South
Tropical and subtropical dry forest 4.0 0.28 – 0.56 2.7
America
North and South Tropical shrubland and subtropical
4.0 0.32 – 0.4 2.6
America steppe
North and South Tropical and subtropical mountain
1.8 – 5.0 0.27 2.0
America systems
Temperate zones
Europe Temperate oceanic forest 2.3 0.40 – 0.46 1.5
Europe Temperate continental forest 4.0 0.40 – 0.46 2.7
Europe Temperate mountain systems 3.0 0.40 – 0.46 2.0
North America Temperate oceanic forest 15 0.40 – 0.46 10.1
North America Temperate continental forest 4.9 0.40 – 0.46 3.3
North America Temperate mountain systems 3.0 0.40 – 0.46 2.0
South America Temperate oceanic forest 2.4 – 8.9 0.40 – 0.46 3.8
Asia Temperate continental forest 4.9 0.40 – 0.46 3.3
Asia Temperate mountain systems 3.0 0.40 – 0.46 2.0
Boreal zones
Asia, Europe, North
Boreal coniferous forest 0.1 – 2.1 0.39 0.7
America
Asia, Europe, North
Boreal tundra woodland 0.4 0.39 0.3
America
Asia, Europe, North Boreal mountain systems 1.0 – 1.1 0.39 0.7

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 172

 Average
Average
above- Ratio
above and
ground below-
below
biomass ground to
Region Ecological zone ground C
growth above
accumulation
(tonnes ground
a
(tonnes C /
d.m. / ha / biomass
ha / yr)
yr)
America
a
For the below to above ground ratio, the coefficient selected correspond to biomass density below or equal to a growth period of

20 years. Source: IPCC (2006), Vol. 4, Chap. 4, Table 4.4 and Table 4.9. Last column calculated using carbon fraction value of

0.47.

Table 29: Average carbon stock from sequestration in abandoned land in old and new GLOBIOM approach (tonnes

C/ha). Weighting within each region is done by agricultural area, as this is the areas of particular interest for

abandonment
Regrowth mix
Forest annual 10 20 Full Forest full Natural vegetation
Region
regrowth rate years years regrowth regrowth full regrowth
Latin America 1.8 14.0 22.7 38.4 88.1 27.4
South Asia 0.9 7.5 14.1 31.7 42.9 27.3
North America 1.0 7.6 11.9 24.4 50.3 14.5
Europe 1.5 9.9 17.2 37.9 73.2 7.1
Eastern Asia 0.5 5.1 9.2 20.4 26.1 18.6
Southeast Asia 1.5 12.7 22.8 45.9 73.2 29.7
Former Soviet
Union 0.9 4.8 7.3 14.6 41.2 4.7
Sub-Saharan
Africa 1.3 7.2 13.6 34.9 37.0 31.5
Oceania 0.6 4.8 8.9 24.8 25.7 24.3
Middle East
North Africa 0.5 4.0 7.3 12.8 20.9 11.8

II.7 Refine co-product substitution

Motivation for improvements

Co-products are an essential component of the lifecycle analysis of biofuel production. Biodiesel from
rapeseed, soybeans, and sunflower leads to significant amounts of protein meals being delivered on
the market. These products are used in animal diet as protein supplement and the biofuel sector
strongly interacts with the livestock sector through this channel. Similarly, dried distillers’ grains with
solubles (DDGS) and sugar beet fibers, generated by cereals and sugar beet processing respectively,
are also used for feed preparation and displace consumption of other agricultural products. To
understand the final balance associated to these changes, it is important to understand which
products are being substituted to which extent through increased supply of co-products from the
biofuel processing chain.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 173

The present improvement focuses on the representation of the feed nutrients in terms of energy and
proteins, and complements efforts on the modeling of oilseeds markets from Section II.9 on
substitution mechanisms and II.12 on oilseeds transformation chains.

Methodological approach
GLOBIOM already incorporates a precise description of animal diets for each livestock system and
species, which can be used to best represent the substitution patterns for biofuel co-products. We
calculate for each species modified diet specifications incorporating more co-products, subject to
some maximum incorporation constraints.

Diet specifications are calculated taking into account digestibility patterns and metabolisable energy
and proteins for the different animal types. Feed requirements are calibrated on the current data
used in GLOBIOM, and derived from the RUMINANT model (Herrero et al., 2013). For each feed item,
we calculate the exact nutrient content using feed tables from U.S. National Research Council (NRC,
1982), as presented in Table 30.127 These tables contain all major crop types traditionally used for
feeding, including protein meals and DDGS. In the case of DDGS, however, as technology evolved a
lot over time in terms of protein extraction efficiency, we used more recent data on DDGS
characteristics. Because composition of co-products can vary across places and refineries, sensitivity
analysis will also explore slight variation around the values presented in Table 30.

This information on feed composition is used to specify substitution patterns for each animal type, by
ensuring that both the energy and protein balance are preserved. This diet substitution pattern is
applied only to the livestock systems based on grain and protein meals consumption (of type
intensive, mixed-intensive or mixed-extensive). Substitution with co-products for grass-based
production systems is not considered. To represent substitution, we allow mobility in the feed intake
of the animals while satisfying two inequalities directly coded in the model: 1) the crude protein
intake should be higher or equal to initial intake; 2) the metabolisable energy intake should be higher
or equal to initial intake. These equations are applied to protein meals and to the main feed grain
type used in the region (usually corn or wheat), to adjust on energy content. For cattle, where we
have more detailed information, we directly use the average of Net energy for growth and for
maintenance, whereas for dairy cows, we refer to the Net Energy for lactation to capture the specific
dietary needs in the respective livestock sectors.128

Because each species and each feedstuff have different characteristics, the replacement results used
differ depending on the livestock sector and the biofuel pathway. The advantage of this approach is
to be able to directly trace the substitution efficiency on the basis of nutrient content of co-products,
instead of relying on substitution coefficients from the literature. The large number of animals
coexisting in different systems and rations mixes guarantee smooth substitution profile leading to
continuous transition patterns in the feed substitution.

127
Although this source might appear old, changes in feedstuff nutrient composition remained limited, as illustrated by a comparison with
more recent tables (for instance for beef, NRC, 2000).
128
Different beef and dairy cattle have different feed requirements. Maintenance energy intake is required for all animals to ensure the
appropriate level of feed for normal metabolism, at equilibrium, without production of any other output. For beef cattle, it needs to be
supplemented by energy for growing with different ratios depending on the stage of development of the animal. For dairy cows, milk
production requires an additional regular intake of energy that leads to the lactation energy requirement (that includes maintenance
energy).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 174

Table 31 shows the substitution patterns obtained when applying the composition found in Table 30.
We calculate in this table a simple bilateral substitution between one biofuel by-product and two feed
products. Sign indicate if the co-product replaces a feedstuff (positive) or requires an additional
provision of cereals to preserve the energy balance (negative).

For example, one unit of rapeseed meal for beef triggers an additional consumption of 0.085 unit of
corn while substituting 0.832 unit of soya meal. Because several protein meals can substitute with
each other, some more complex substitution can also appear. For instance, wheat DDGS can
substitute with soybean meal - and with some cereals to satisfy the complete energy balance - but
soybean meal can also in turn substitute with some other oilseed meals.

Table 31 is therefore only illustrative of the simplest substitution patterns with a pair of feed
products. Additionally, in the case of the US, we directly used another source (Hoffman and Baker,
2011) that specifies the rate of substitution observed in that region, with a greater substitution
towards corn than protein meals. For the US, we therefore assume that 1 tonne corn DDGS substitute
with 1.1 tonne corn for ruminants and with 0.8 tonne corn and 0.2 tonne soybean meal for swine.

In complement to substitution possibilities, for each animal type, incorporation of DDGS is limited due
to the nutrient characteristics of co-products, some of them not directly accounted for in the model
substitution patterns. In particular, DDGS too high incorporation rates can lead to an oversupply of
proteins and phosphorus, leading to waste disposal issues that affect manure management (Hoffman
& Baker, 2011). For DDGS, we therefore capped the incorporation levels at some selected values on
the basis of a literature review by Hoffman and Baker (2011). These incorporation constraints are
provided in Table 32 and were chosen at the mid-range of low and high values in the literature,
except when higher incorporation rates were already observed in U.S. statistics (beef cattle).

Implications for model results

The new representation of co-product substitution in the model will allow to specify the substitution
of animal feeding with different diet possibilities specific to each species, on the direct basis of
nutrient composition and their properties per type of animal. For instance, wheat DDGS will
substitute more cereals in the ruminant sector than with the non-ruminant, and sugar beet pulp will
be little used by the poultry sector due to poor digestibility. We also observe that, as long as other
nutrient constraints (e.g. amino-acids) are not taken into account,129 rapeseed and sunflower meal
can appear as appealing substitutes with other protein sources due to their high protein level, but
may require cereals complement to preserve the energy balance, which was not captured before.
This new design will allow for a more accurate accounting of the substitution possibilities of co-
products in the feed and to better measure the LUC effects implications associated their
incorporation.

129
For sake of time, it was unfortunately not possible in the framework of this exercise to develop more sophisticated substitution rules than
the one developed here, although considering additional component requirements would probably affect further the substitution possibilities.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 175

Table 30: List of metabolisable energy and protein content associated to the different feed crops and supplements in

the model. All values below are expressed for dry matter feed. ME = metabolisable energy, NEm = net energy for

maintenance, NEg = net energy for growth, NEl = net energy forlactation, MEn = metabilisable energy, nitrogen

corrected (for poultry)

Feed stuff Ruminan Ruminan Ruminan Dairy Chicken Swine Crude Crude
t ME t t cattle MEn ME protei Fiber
(Mcal/k NEm(Mc NEg(Mca NEl(Mcal (kcal/k (kcal/ n (%) (%)
g) al/kg) l/kg) /kg) g) kg)
Barley grain 3.29 2 1.35 1.94 2843 3299 13.5 5.7
Dry bean 3.29 2 1.35 1.94 2593 3.772 25.3 5
Corn grain 3.42 2.09 1.42 2.01 3818 3724 10.9 2.9
Corn silage 2.62 1.55 0.94 1.57 NA 2981 8.3 25.1
Oats grain 2.98 1.79 1.17 1.77 2862 3012 13.3 12.1
Pea 3.42 2.09 1.42 2.01 2385 3416 25.3 6.9
Potato 3.16 1.91 1.27 1.87 NA 3516 9.5 2.4
Rapeseed meal solv
2.62 1.55 0.94 1.57 1924 2935 40.6 13.2
extd
Rye, grain 3.29 2 1.35 1.94 3001 3327 13.8 2.5
Soybean seeds 3.60 2.22 1.52 2.11 3674 3905 42.8 0.1
Soybean meal solv
3.29 2 1.35 1.94 2485 3155 49.9 7
extd, 44% protein
Soybean oil 8.23 5.25 4.02 4.66 8667 7283 1.4 NA
Sugar beet pulp,
with molasses, 2.93 1.76 1.14 1.74 719 3139 10.1 16.5
dehydrated
Sunflower meal, wo
hulls, meal solv 2.45 1.44 0.82 1.47 2242 2851 49.8 12.2
extd
Triticale grain 3.29 2 1.35 1.94 3521 3396 17.6 4.4
Wheat grain 3.47 2.12 1.45 2.04 3401 3660 16 2.9
Wheat durum grain 3.34 2.03 1.37 1.96 3652 3492 15.9 2.5
Source : National Research Council, 1982, 2000
Corn distillers
grains with
3.90 2.38 1.69 2.28 2531 3790 31.2 8.6
solubles,
dehydrated
Wheat distillers
grains with
3.75 2.29 1.62 2.14 2406 3472 36.6 7.6
soluble,
dehydrated
Sources:
Wheat DDGS : Noblet, Cozannet & Skiba (2012) for pigs and poultry; extrapolated from Kalscheur et al. (2012) for ruminant.
Corn DDGS: Kalscheur et al. (2012) for ruminant; extrapolated from Anderson et al. (2012) for pigs and Noblet, Cozannet & Skiba (2012)
for poultry.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 176

Table 31: Substitution pattern for each animal species for one unit of co-product consumed by the livestock sector.

Positive values correspond to a replacement of feed, negative value to a joint addition of another feedstuff to

preserve the energy balance. For example, one kg of rapeseed meal for beef triggers an additional consumption of

0.08590 kg of corn while substituting 0.8332 kg of soya meal

Sugar beet Rapeseed Sunflower

Corn DDGS Wheat DDGS
pulp meal meal
Feed item
SUBSTITUTE FOR CORN & SOYBEAN MEALS
Beef

Corn 0.711 0.523 0.800 -0.085 -0.390

Soya meal* 0.470 0.619 0.028 0.832 1.083

Dairy
Corn 0.673 0.452 0.849 -0.005 -0.294
Soya meal* 0.478 0.635 0.017 0.815 1.062

Swine
Corn 0.559 0.382 0.824 0.121 -0.098
Soya meal* 0.494 0.650 0.022 0.787 1.019

Poultry
Corn 0.298 0.178 0.066 -0.030 -0.073
Soya meal* 0.560 0.695 0.188 0.820 1.014

SUBSTITUTE FOR WHEAT & SOYBEAN MEALS

Beef

Wheat 0.791 0.582 0.890 -0.094 -0.434

Soya meal* 0.371 0.547 -0.083 0.844 1.137

Dairy
Wheat 0.753 0.506 0.950 -0.006 -0.329
Soya meal* 0.384 0.571 -0.102 0.816 1.103

Swine
Wheat 0.686 0.437 0.944 0.139 -0.112
Soya meal* 0.405 0.593 -0.100 0.769 1.034

Poultry
Wheat 0.375 0.224 0.083 -0.038 -0.091
Soya meal* 0.505 0.662 0.176 0.826 1.027

Note: Soybean meals are not represented here as they are largely used already as feed in rows. Their substitution values can be

found by reading the table from row to column and inverting the cereal contribution. For instance, 0.832 unit of soybean meal

substitute for beef with 1 unit of rapeseed meal and 0.085 unit of corn (corn is now replaced as the negative sign needs to be

inversed).

Table 32: Maximum incorporation constraint for DDGS as percent of daily dry matter intake

Animal type Observed Maximum incorporation in literature Value in

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 177

 incorporation GLOBIOM
rate
US Widwest
Low High
(2007)
a
Beef 22% 10% 30% 30% b

c
Dairy 8% 10% 30% 20%
d
Swine 10% 10% 30%-50% 20%
Poultry NA 10% 15% 12.5%
a
Statistics reported here are based on calculation for cows. Beef cattle on feed high bound up to 40%.
b
High bound taken to take into account observed rate.
c
Incorporation statistics reported here for dairy cows (not replacement heifers).
d
Low bound for market swine, high bound for breeding swine.

Source: Hoffman and Baker (2011).

II.8 Represent multi-cropping

Motivation for improvement

In several regions of the world, the possibility of harvesting more than one crop per year in a same
field has been used to increase output per hectare. Most famous examples are the multi-harvest of
rice in Southeast Asia or the soybean-maize double cropping practice in Latin America. GLOBIOM was
not taking into account so far this possibility and annually harvested areas of cropland were
calculated on the basis of harvested areas of each crop, without any specific correction. Multi-
cropping (or inter-cropping) possibilities were therefore not considered. Additionally, when cropland
area was found larger than harvested areas, the “unused” cropland was considered kept constant, to
reflect the presence of other not referenced crops or various conservation uses. No change in
cropland harvest frequency was then represented. Therefore, it was decided to better represent the
trend of multi-cropping in the baseline of GLOBIOM by introducing some cropping intensity change
and the potential of this development to free some agricultural land.

Useful definitions [largely based on Ray and Foley (2013)]

Harvested area: Area of crop that has been harvested through one year, possibly several time in case of successive cropping
seasons in the same year.
Annually harvested cropland: Area of cropland which is used for cultivation (possibly several times a year).
Total standing cropland: total area of land declared as cropland, including fallow land.
Cropland harvest frequency (CHF) or cropping intensity: defined as Harvested area divided by Total standing cropland.
Double-/Multi-cropping: practice of harvesting two/several crops successively in a same year on the same cropland area.
Inter-cropping: practice of planting several crops simultaneously in the same field, with alternate rows of crop of the different
species.

Methodological approach
First, FAO statistics were used to calculate cropland harvest frequency (CHF) of the different regions.
CHF calculation is not sufficient to identify all places where muti-cropping could be observed due to
disparity of cropping practices within a region. However, it provides a sufficient criteria to locate
some of them.
Indeed, if this ratio is greater than one, some areas of land have necessarily been used to grow
several crops in the same year. If this ratio is below one, a share of cropland has necessarily not

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 178

been used for production, but this does not exclude multi-cropping practices in some other locations
in the region.

Nine countries were found with CHF > 1 for the calibration year (2000), which reveals the presence
of multi-cropping in these regions. The list of regions and corresponding countries can be found in
Table 33. They are consistent with assessment of multi-cropping location in the literature (Langeveld
et al., 2013; Siebert et al., 2010). For these regions, annual crop yield assumptions were adjusted in
GLOBIOM to better reflect the current average output per hectare of harvested cropland and per
year. As a consequence, cultivated areas were decreased in these regions, and land areas
erroneously allocated to the cultivation of the crop were reclassified as “other natural vegetation”.
For regions with CHF >1, we also implemented a trend in the baseline for cropland harvest
frequency. For this we used the trend on the period 2000-2011, following a methodology similar to
Ray & Foley (2013), who have calculated trends in ratio of harvested land over cropland area over
time.130

In the case of China and India, we refined our estimates by a closer look to the literature specific to
these regions. For China, we relied a remote sensing historical analysis of change in multi-cropping
patterns (Zuo et al., 2013). According to this study, CHF in 2005 is found to be 1.53 and multi-
cropping efficiency is assessed to be 87.6% with significant barriers to multi-cropping improvements
(max CHF would be 1.75). We assume that the maximum would be reached in 2030. For the case of
India, we use another remove sensing analysis (Biradar and Xiao, 2011). These authors observe than
Indian CHF was 1.267 in 1993, and 1.371 in 2005. We use this trend to refine our projections for
India.

For regions with CHF<1, it is not possible to derive frequency of harvest for the different crops,
without studying some specific national datasets, a process too time consuming for this project. As an
exception to this general rule, we had a closer look on the case of Brazil, well known for his
increasingly use of multi-cropping practices (on corn and soybeans). According to Spera et al.
(2014), use of double cropping in the State of Matto Grosso grew from 500,000 ha in 2001 to 2.8
Mha in 2011. Therefore we also apply the trend on cropland harvest frequency in Brazil (+0.9% per
year on frequency), although the cropland harvest frequency is lower than 1 in this region.

Implications for model result

The first effect of multi-cropping representation will be, for the countries with CHF > 1, a reevaluation
of annually harvested cropland area for the base year 2000, after adjusting the yield values.
Additionally, the trend on yield will be modified for these regions as well as for Brazil. So far, the
exogenous yield trend was only representing the effect of technical change, but it will now also
incorporate an additional component for change in management related to multi-cropping. Yields are
likely to grow faster for the regions concerned, reducing the impact of additional production in the
baseline on LUC.

130
It should be noted that the trend in historical data may also be associated to change in fallow land (decreasing) and not only to multi-
cropping. However, as unused cropland is kept constant in the model (see improvement 27 for more explanations), this approach is
consistent to replicate the trend in cropland harvest frequency.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 179

Table 33: Cropland harvest frequency and associated indicators according to FAOSTAT in 1999-2001

Harvested area – Maximum cropland

Annual growth
Region with cropland Cropland harvest harvest frequency
(1000 ha, only rate
multi-cropping frequency (2000) (Ray and Foley,
>1Mha reported) (2000-2011)
2013)
China* 29,089 1.53 0.4% 1.75
Nigeria 8,537 1.26 -1.7% 2.00
India** 6,514 1.32 0.7 % 1.63
Bangladesh 5,544 1.63 1.1% 1.99
VietNam 3,865 1.47 -0.5% 1.95
Philippines 2,779 1.28 0.2% 2.00
Myanmar 2,551 1.24 1.6% 1.80
Nepal 2,052 1.84 0.8% 1.06
Egypt 1,271 1.38 0.5% 1.01
Others (<1 Mha) 1,347 1.02 -- --
TOTAL 63,549 1.13 -- --
Brazil 0.78 0.9% 1.71
World 0.82

Source: authors’ own calculation using FAOSTAT data, except for last column from Ray and Foley (2013). Note that we report here

growth rates for the index, whereas Ray and Foley report rate of annual change. * For China, we use data from Zuo et al, 2013.

** For India, data from Biradar and Xiao, 2011.

II.9 Represent imperfect substitution between vegetable oils

Motivation for improvements

Vegetable oils were represented in the standard version of GLOBIOM with distinct demand functions,
and the level consumed was only determined by an exogenous shifter for the income effect and an
own-price demand elasticity for the price effect. No distinction was made between demand for food
and demand for industrial use, to the exception of biofuel use. However, vegetable markets are to
some extent connected, as illustrated by the strong correlation between the different oil prices. It
was agreed that a better representation of these linkages should be introduced into GLOBIOM, by
introducing some substitution possibilities between vegetable oil on the supply side, while keeping in
mind the restrictions to such substitution related to the different properties of these oils, the specific
needs of industries, as well as the preferences of consumers.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 180

Methodological approach
The question of patterns of change in the oilseed market is complex and we investigated different
sets of statistics to better understand the mechanisms at play, provided by the industry, by FAOSTAT
and by the USDA.131 Stylized facts were examined to address a certain number of questions in the
debate. We observed the following points:
 Food consumption per capita of vegetable oil has been relatively stable in Europe for rapeseed
over the past decade (-9% according to USDA). But sunflower oil consumption as food has
notably increased between 2002 and 2012 (USDA: +38%), as well as palm oil (+63% between
2000 and 2012). Soybean oil has decreased in the same time by 30% according to USDA. In the
EU, the use of soybeans and rapeseed has decreased to the benefit of sunflower and palm oil;
 Most of substitution in the EU on vegetable oils has been observed through imports and on the
industrial uses market. By contrast however, the substitution patterns in the US were larger for
final consumption, and soybean oil used as food was significantly substituted by palm oil (Figure
34);
 Decrease in EU food consumption of rapeseed has remained limited compared to total increase
in supply (see Figure 7 and Figure 34). The main sources of additional supply have been
increased rapeseed production, and additional rapeseed imports (see Figure 36);
 Palm oil imports to the EU have significantly expanded over the period 2000-2012 (see Figure
36). Parts of these imports have been driven by a direct use by the industrial sector, in particular
biofuels. But one third of these imports have also been absorbed by the food sector. The food
sector has absorbed a similar quantity of sunflower oil, half of it being imported. These products
compensated in the food sector for rapeseed and soybean oil transferred to the industrial uses.

For instance, we analysed whether the food consumption varied for the different oils types in the EU,
or whether difference of price between rapeseed and palm oil could explain some changes in trade
patterns. On the basis of this analysis, we concluded that some substitution of vegetable oil was
observed in food demand but was overall relatively limited compared to the industrial demand.
Therefore, a relatively low elasticity of substitution should be used in the case of the EU. On the
model side, in order to implement this limited substitution effect, we created an aggregated
vegetable oil food item, in which the fluctuation of the different oil shares is relatively constrained.
For this purpose, the objective function of GLOBIOM was modified to include some non-linear costs
associated to the change in composition of the vegetable oil aggregate. In the version of IFPRI-
MIRAGE used in Al-Riffai, Dimaranan & Laborde (2010), an elasticity of substitution of 2 was used in
the different regions for substitution of vegetable oils and a trade Armington elasticity of 10.132 When
prices increase, both rapeseed imports and other oil demand react to compensate the shock.

An analysis of historical development in the EU for oilseed markets shows that rapeseed oil and
rapeseed imports increased five times more than food demand for rapeseed oil decreased, between
2000 and 2012 (Figure 36). In MIRAGE, the contribution of rapeseed demand change through
substitution was found contributing more than two times more to the new demand for rapeseed oil.133

131
Methodology applied by USDA to split consumption across uses was found more consistent with the data provided by the industry than
the one from FAOSTAT. The latter allocate a large part of consumption to other uses, whereas such use is usually better allocated in USDA
and FEDIOL databook. Unfortunately, the USDA data do not provide the decomposition of demand between the different industrial uses.
Therefore, the substitutability of vegetable oil within the non-biofuel industrial uses is not discussed here.
132
Data retrieved on ec.europa.eu/energy/renewables/studies/doc/land_use_change/iluc_report_annex_1.xls
133
Analysis of MIRAGE results from Al Riffai et al., 2010 suggests that for a shock of 324,000 tonnes of rapeseed oil to the biofuel, the EU
food market provides 65,000 tonnes, ie 20%.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 181

Uncertainty on the right substitution level is therefore key for the magnitude of responses on the
vegetable oil market. As starting point, we will consider in our analysis an elasticity value of 0.8 for
the EU, smaller than MIRAGE. The value of this elasticity will be part of the sensitivity analysis.
However, for other regions like the US, observed recent changes in consumption of soybeans suggest
that higher substitution is possible within the food sector. We will therefore keep a value of 2 in the
USA. These values will be varied through the Monte-Carlo, in order to capture effect of having lower
or higher substitution effects.

Implications for model results

Vegetable oil markets were already connected in GLOBIOM through demand for biofuel use. The
present improvement in the model will introduce some possibility of substitution on the food market
side, but with a more limited possibilities, reflecting the stickiness observed in the past time series, in
particular in the case of the EU. Trade should therefore remain for the EU the most important driver
of propagation of demand shock. In the rest of the world, markets will also be connected through the
industrial demand and through the food market, in particular for some regions like the US showing
larger substitution patterns.

Table 34: Analysis of vegetable oil final consumption in the EU in the biofuel scenario from Al-Riffai et al., 2010

(using appendix results tables)

Food Food
Ratio food
Feedstock use for biofuels consumption consumption Food
change /
(1000 t) change (%) level (1000 consumption
feedstock
(Al-Riffai et al., 2010, Table (Al-Riffai et t) change
demand for
S6b) al., 2010, (USDA, (1000 t)
biofuel
Table S7) 2010)
Baseline Scenario
Difference
2020 2020
Palm oil 824 1008 185 1.08% 2750 30 0.161
Rapeseed
4997 5320 324 -2.39% 2733 -65 -0.202
oil
Soybean oil 2978 3441 463 -0.41% 1290 -5 -0.011
Sunflower
430 511 81 0.30% 3191 10 0.121
oil

shows that as the rapeseed consumption increased by 6 million tons in the EU in the 2000s, rapeseed decreased by 500,000 tons, ie the
food sector did not contribute more than 8%, ie 2.5 times less. See Table 34 for calculation details.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 182

 18
16
kg per cap per year

14
12
10
8 Oil, Soybean
6 Oil, Sunflowerseed
4 Oil, Rapeseed
2
0
1999/2000
2000/2001
2001/2002
2002/2003
2003/2004
2004/2005
2005/2006
2006/2007
2007/2008
2008/2009
2009/2010
2010/2011
2011/2012
2012/2013
2013/2014
Figure 33: Consumption per capita of vegetable oil by EU consumer according to USDA PSD statistics

0,8

0,6
Change in kg per cap per year

0,4

0,2 Rapeseed oil

Soybean oil
0
Sunflower oil
1991-2000

2001-2010

1991-2000

2001-2010

1991-2000

2001-2010

-0,2 Palm oil

-0,4

-0,6 EU28 North America World

-0,8

Figure 34: Change in consumption per capita of vegetal oil in EU28, North America and World on two periods 1991-

2000 and 2001-2010. Rate of change are obtained by regression of consumption per capita on each period using

USDA PSD data on food use

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 183

 0
1.000
2.000
3.000
4.000

0
5.000
0

10.000
15.000
20.000
25.000
30.000
35.000
40.000
45.000
1.000
2.000
3.000
4.000
5.000
6.000
7.000
8.000
9.000
10.000
1989/1990 1989/1990 1999/2000
2000/2001
1991/1992 1991/1992
2001/2002
1993/1994 1993/1994
2002/2003
1995/1996 1995/1996 2003/2004
1997/1998 1997/1998 2004/2005
1999/2000 1999/2000 2005/2006
Oil, Palm

2006/2007

Food Use
Food Use

Food Use
2001/2002 2001/2002
Oil, Soybean
Oil, Rapeseed

2007/2008
2003/2004 2003/2004
2008/2009
2005/2006 2005/2006
Oil, Sunflowerseed

2009/2010
2007/2008 2007/2008 2010/2011

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU

2009/2010 2009/2010 2011/2012

2014 according to USDA PSD database. (1000 tonnes)

2011/2012 2011/2012 2012/2013
2013/2014 2013/2014 2013/2014

184
EU28

World
North America

0
500
1.000
1.500
2.000
2.500
0
2.000
4.000
6.000
8.000

0
10.000
12.000
14.000
16.000
18.000
20.000

2.000
4.000
6.000
8.000
1989/1990 1999/2000
1989/1990
1991/1992 2000/2001
1991/1992
1993/1994 2001/2002
1993/1994
1995/1996 2002/2003
1995/1996 2003/2004
1997/1998 1997/1998
2004/2005
1999/2000 1999/2000
2005/2006
Oil, Palm

2001/2002 2001/2002 2006/2007

Oil, Soybean

Industrial use
Industrial use

Industrial use
2003/2004
Oil, Rapeseed

2003/2004 2007/2008
2005/2006 2005/2006 2008/2009
Oil, Sunflowerseed

2007/2008 2009/2010
2007/2008
2010/2011
2009/2010 2009/2010
2011/2012

Figure 35: Vegetable oil consumption in the food and the industrial sectors (including biofuels) between 1990 and
2011/2012 2011/2012
2012/2013
2013/2014 2013/2014 2013/2014
 7.000

6.000

5.000

4.000 Rapeseed

3.000 Soybean
Sunflower
2.000
Palm oil
1.000

0
Industrial Food Crushed Crushed Imported
-1.000 demand demand EU oilseeds vegetable
oilseeds imports oil

Figure 36: Changes in EU market between 2000 and 2012 for four vegetable oil types (1000 tonnes, source: USDA)

II.10 Separate representation of Argentina, Indonesia, Malaysia and

Ukraine

Motivation for improvements

Argentina, Indonesia and Malaysia are important player on the international biofuel market today:
Argentina is a major producer of soybeans and Indonesia and Malaysia concentrate a large majority
of the palm oil production. Ukraine could play an important role in the future as a supplier of
agricultural products to the EU market, in particular cereals but also rapeseed and sunflower. In order
to better represent how these countries can interact with the international markets, it was
recommended for the purpose of this project to single them out in order to more precisely trace their
trade and how their production level can influence land use patterns, in response to policy scenarios
on biofuels.

Methodological approach
Most model parameters for the parameterization of the supply side in GLOBIOM (land use, crop area
and yield, animal distribution and systems, etc.) are provided at the Simulation Unit level and
sourced from biophysical models and downscaling of some national datasets. For these data, input
were reprocessed and made compatible with the new regional levels. The demand side had to be
disaggregated, using data on quantities and prices from FAOSTAT. Argentina was separated out from
previous “Rest of South America” region; Indonesia and Malaysia – previously Rest of South East Asia
and Ukraine – previously Former USSR we also singled out. New price elasticities were sourced from
USDA. Bilateral trade flows were recalibrated for all the new regions based on COMTRADE and tariffs
from MAcMap, following the methodology used so far in GLOBIOM.
Although most data on the supply side were already available through the different input datasets,
some adjustments had to be performed to represent adequately some production patterns of the new
regions. Indeed, 2000 data from SPAM (Spatial Production Allocation Model), used for the calibration
of the initial crop areas, was not available for rape and sunflower in some regions, in particular
Ukraine. Some special treatment had to be applied to allocate these crops spatially.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 185

Initial crop areas were distributed across Simulation Units und management system using the SPAM
information on potential “pre-crops” in the rotations of sunflower and rapeseed (typically barley,
corn, or wheat).

Implications for model results

The new regional mapping of GLOBIOM now gives access to more precise characterization of trade
and uses of products in the new regions. Demand quantities, trade flows, prices can now be reported
for these regions separately. On the supply side, LUC patterns can be more precisely connected to
trade, as market resolution has been increased in the new areas of interest and localization of
production changes is now more precisely assessed.

Table 35: List of regions newly represented in GLOBIOM for the ILUC study (new countries in bold)

GLOBIOM region Definition

ANZ Australia, New Zealand
Argentina Argentina
Brazil Brazil
Canada Canada
China China
Cameroon, Central African Republic, Congo Republic, Democratic Republic of Congo,
Congo Basin:
Equatorial Guinea, Gabon
Baltic: Estonia, Latvia, Lithuania
EU28, each East: Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia,
country is treated Central: Austria, Belgium, Germany, France, Luxembourg, Netherlands
as separate region North: Denmark, Finland, Ireland, Sweden, United Kingdom
South: Cyprus, Greece, Italy, Malta, Portugal, Spain
Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russian
Former USSR
Federation, Tajikistan, Turkmenistan, Uzbekistan
India India
Indonesia Indonesia
Japan Japan
Malaysia Malaysia
Mexico Mexico
Middle East and Algeria, Bahrain, Egypt, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Morocco,
North Africa Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, Yemen
Pacific Islands Fiji Islands, Kiribati, Papua New Guinea, Samoa, Solomon Islands, Tonga, Vanuatu
Bahamas, Barbados, Belize, Bermuda, Costa Rica, Cuba, Dominica, Dominican Republic,
El Salvador, Grenada, Guatemala, Haiti, Honduras, Jamaica, Nicaragua, Netherland
RCAM
Antilles, Panama, St Lucia, St Vincent, Trinidad and Tobago

RCEU Albania, Bosnia and Herzegovina, Macedonia, Serbia-Montenegro

ROWE Gibraltar, Iceland, Norway, Switzerland
Bolivia, Chile, Colombia, Ecuador, Guyana, Paraguay, Peru, Suriname, Uruguay,
RSAM Venezuela

RSAS Afghanistan, Bangladesh, Bhutan, Maldives, Nepal, Pakistan, Sri Lanka

RSEA OPA Brunei Daressalam, Singapore, Myanmar, Philippines, Thailand
RSEA PAC Cambodia, Korea DPR, Laos, Mongolia, Viet Nam
South Africa South Africa
South Korea South Korea
Burundi, Ethiopia, Kenya, Rwanda, Tanzania, Uganda
Eastern Africa

Angola, Botswana, Comoros, Lesotho, Madagascar, Malawi, Mauritius, Mozambique,

Southern Africa Namibia, Reunion, Swaziland, Zambia, Zimbabwe

Western Africa Benin, Burkina Faso, Cape Verde, Chad, Coted Ivoire, Djibouti, Eritrea, Gambia, Ghana,

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 186

 GLOBIOM region Definition
Guinea, Guinea Bissau, Liberia, Mali, Mauritania, Niger, Nigeria, Senegal, SierraLeone,
Somalia, Sudan, Togo
Turkey Turkey
Ukraine Ukraine
United States of
United States of America
America

II.11 Represent unused agricultural land in Europe

Motivation for improvements

Agriculture represented 44% of land use in the European Union in 2000. In GLOBIOM, this land is
represented through various land categories: “Cropland” corresponding to the share of this land used
to produce the crops represented in GLOBIOM; “Grassland” corresponding to areas used for ruminant
grazing. A part of this land is also occupied by other agricultural activities that are not represented
for the moment in GLOBIOM (e.g. vegetables, vineyards, orchards). These are identified as “other
agricultural land” and this land is kept fixed in the model. Beside these managed land, some
unmanaged land are also input to the model: “natural forest”, “wetlands”, and “other natural land”,
that contain all remaining types of fertile areas.
Unused agricultural land falls in three categories:
 set-aside land is represented in the EU directly in the crop rotation as a crop management option
subject to profit maximization;
 other fallow land declared as cropland is part of the “Other agricultural land category”;
 abandoned land no longer declared as cropland is accounted outside of the agricultural land
under the category “other natural land”.

The extent of “other natural land” in GLOBIOM is usually much greater than the “other agricultural
land” category. Therefore, even if “other agricultural land” is fixed, the potential agricultural land
expansion is large. However, depending on the location, “forest” land can also be used in the
expansion. In the standard version of GLOBIOM, both “forest” and “other natural land” are used when
agricultural land expand, with proportions determined by the calibration parameters and based on
observations of past land use changes.
It has been argued that the potential contribution of unused land in Europe has been underestimated
in past assessments of LUC dynamics and should be better represented in GLOBIOM. For that reason,
it was decided to perform some scenarios where the access to unused land would be facilitated in the
EU as well as in Ukraine, a large potential supplier of agricultural products to Europe which already
today provides significant quantities of biofuel feedstock to the EU.

Methodological approach
This improvement is performed as a new scenario (scenario C) with a change of parameterisation of
the model. For this scenario, we model possibilities of expansion through the third unused agricultural
land category detailed above. Access cost to “other natural land” is therefore reduced for this land
use type in all countries of the European Union as well as for Ukraine.
For all EU regions, this land use category is large. In the case of Ukraine, input data were also
improved to better account for recent assessment on abandoned land based from Alcantara et al.
(2013) who find that 9.2 Mha of farmland is currently abandoned in Ukraine.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 187

This land area was therefore classified under “Other natural vegetation” instead of “Other agricultural
land”, homogeneously across the country.

Implications for model results

Scenario C has been especially designed to represent the effect of improved access to unused
agricultural land. The effect of the change in access cost in scenario C will be an increased share of
agricultural expansion in the EU and Ukraine, with greater use of the “other natural vegetation” land
use category.

Table 36: Land use in the EU and in Ukraine in the GLOBIOM nomenclature in 2000 (1000 ha)

Other
Other agricultural
Country Cropland Pasture Forest natural Wetlands
land
vegetation
Austria 1,319 1,705 97 3,828 494 16
Belgium 809 675 104 653 191 10
Bulgaria 2,570 1,511 192 3,507 2,303 11
Croatia 852 1,609 488 2,129 301
Cyprus 71 46 195 148 410
CzechRep 2,846 631 137 2,569 1,170 8
Denmark 2,264 107 80 526 906 62
Estonia 610 48 7 2,170 1,076 188
Finland 1,712 460 28 21,953 4,893 867
France 17,989 7,637 736 13,787 9,783 181
Germany 11,505 4,360 851 10,751 4,784 178
Greece 2,173 1,836 187 1,188 6,054 39
Hungary 3,229 1,025 366 1,661 2,093 63
Ireland 264 2,778 126 461 2,039 1,002
Italy 7,091 3,328 641 7,452 6,553 58
Latvia 940 231 73 3,148 1,661 149
Lithuania 1,490 408 405 1,924 1,862 55
Luxembourg 38 46 17 89 48
Malta 5 0 14 6
Netherlands 851 932 149 345 710 31
Poland 12,067 3,774 1,072 8,879 3,752 132
Portugal 1,263 1,102 104 3,825 1,491 25
Romania 6,815 4,836 354 6,698 2,591 247
Slovakia 1,355 494 78 1,980 647 5
Slovenia 154 276 11 1,188 284 3
Spain 11,185 6,495 603 12,580 13,514 93
Sweden 2,734 206 120 24,077 8,911 3,514
UK 5,539 7,950 275 2,603 5,579 438
Ukraine 14,419 17,392 7,415 8,893 10,361 259

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 188

II.12 Refine biofuel feedstock processing coefficients

Motivation for improvements

The past assessment of ILUC has raised some concerns about the conversion coefficients to be used
at different stages of the processing of agricultural materials into biofuels. GLOBIOM offers an explicit
representation of conversion technologies and it was decided to document the current conversion
assumptions to give opportunity to agriculture and industry stakeholder to comment on the
assumptions for i) oilseeds crushing supply chains (improvement 34), ii) bioethanol and biodiesel
transformation chains (improvement 35).

Methodological approach
A data document was compiled containing most important assumptions for GLOBIOM supply chains.
This document will be made public once all input from stakeholders will have been reviewed and
initial GLOBIOM assumptions improved when relevant. A list of input and comments on assumptions
received during the consultation period is provided in Table 37. Corrections performed are also
reported in that table when they were considered relevant. In some cases, reported issues led to
direct adjustments in the initial GLOBIOM data (e.g. on biofuel supply chains), or only in adjustment
of coefficients along the baseline (e.g. for crushing rates that vary over time).
Comments received were in particular on the following topics:
 Crushing rate values in Europe and rest of the world (FEDIOL)
 Conversion efficiency for corn and maize (Epure)
 Conversion efficiency for sugar beet and sugar cane (CGB)
 Final use of vegetable oils (FEDIOL)
 Production level of sugar beet (CGB)
 Conversion efficiency of sugar beet (CGB)
 Conversion efficiency of sugar cane (CGB)

Implications for model results

Adjustments of supply chains parameters will allow a more precise description of land use
requirements associated to the different feedstocks, and improve the assessment of indirect LUC
effects.

II.13 References

ADEME (2002) ‘Comite Nationale des Coproduits, fiche no. 1: Paille de cereals’, ADEME and IDELE134

Agus, F., Handayani, E., van Noordwijk, M., Idris, K., & Sabiham, S. (2010). Root respiration
interferes with peat CO2 emission measurement. In 19th World Congress Congress of Soil Science,
Soil Solutions for a Changing World (pp. 50–53). Brisbane, Australia.

134

http://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CCYQFjAAahUKEwiwiJOg54fGAhVEgXIKHRlSBuQ&url=http
%3A%2F%2Fidele.fr%2F%3FeID%3Dcmis_download%26oID%3Dworkspace%3A%2F%2FSpacesStore%2F7df6433c-00e5-49a8-a6c5-
b7e41edd9883&ei=JZV5VfDhA8SCygOZpJmgDg&usg=AFQjCNH01V-Y9kqhD_tFleWjN5-qGvwhOA&bvm=bv.95277229,d.bGQ&cad=rja

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 189

Agus, F., Henson, I., Sahardjo, B. H., Harris, N., van Noordwijk, M., & Killeen, T. (2013). Review of
emission factors for assessment of CO2 emission from land use change to oil palm in Southeast Asia
(pp. 7–28). Kuala Lumpur.

Alcantara, C., Kuemmerle, T., Baumann, M., Bragina, E. V, Griffiths, P., Hostert, P., … Radeloff, V. C.
(2013). Mapping the extent of abandoned farmland in Central and Eastern Europe using MODIS time
series satellite data. Environmental Research Letters, 8(3), 35035. Retrieved from
http://stacks.iop.org/1748-9326/8/i=3/a=035035

Al-Riffai, P., Dimaranan, B., & Laborde, D. (2010). Global Trade and Environmental Impact Study of
the EU Biofuels Mandate. Final Report for the Directorate General for Trade of the European
Commission.

Anderson, P. V., Kerr, B. J., Weber, T. E., Ziemer, C. J., & Shurson, G. C. (2012). Determination and
prediction of digestible and metabolizable energy from chemical analysis of corn coproducts fed to
finishing pigs. Journal of Animal Science, 90, 1242–1254.

Asikainen, A., Liiri, H., Peltola, S., Karjalainen, T., & Laitila, J. (2008). Forest energy potential in
Europe (EU 27). Metla, Finnish Forest Research Institute. Retrieved from
http://www.metla.fi/julkaisut/workingpapers/2008/mwp069-en.htm

Biograce. (2014). Biograce version 4c (April 2014).

BIOMASS FUTURES. (2012). Atlas of EU biomass potentials. Deliverable 3.3: Spatially detailed and
quantified overview of EU biomass potential taking into account the main criteria determining
biomass availability from different sources. Retrieved from
http://www.biomassfutures.eu/public_docs/final_deliverables/WP3/D3.3%20%20Atlas%20of%20tech
nical%20and%20economic%20biomass%20potential.pdf

Biradar, C. M. & Xiao, X. (2011), 'Quantifying the area and spatial distribution of double- and triple-
cropping croplands in India with multi-temporal MODIS imagery in 2005', International Journal of
Remote Sensing 32(2), 367-386.

Blanco-Canqui, H., & Lal, R. (2007). Soil and crop response to harvesting corn residues for biofuel
production. Geoderma, 141, 355–362. doi:10.1016/j.geoderma.2007.06.012

Carlson, K. M., Curran, L. M., Asner, G. P., Pittman, A. M., Trigg, S. N., & Marion Adeney, J. (2012).
Carbon emissions from forest conversion by Kalimantan oil palm plantations. Nature Climate Change,
3(3), 283–287. doi:10.1038/nclimate1702

Carré., F., Hiederer, R., Blujdea, V., & Koeble, R. (2010). Background Guide for the Calculation of
Land Carbon Stocks in the Biofuels Sustainability Scheme Drawing on the 2006 IPCC Guidelines for
National Greenhouse Gas Inventories. Luxembourg: Office for Official Publications of the European
Communities.

Comeau, L., Hergoualc’h, K., Smith, J., & Verchot, L. (2013). Conversion of intact peat swamp forest
to oil palm plantation: effects on soil CO 2 fluxes in Jambi (No. 110). Bogor, Indonesia.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 190

Couwenberg, J. (2009). Emission factors for managed peat soils: an analysis of IPCC default values.
Wageningen.

Couwenberg, J., Dommain, R., & Joosten, H. (2010). Greenhouse gas fluxes from tropical peatlands
in south-east Asia. Global Change Biology, 16(6), 1715–1732. doi:10.1111/j.1365-
2486.2009.02016.x

Couwenberg, J., & Hooijer, A. (2013). Towards robust subsidence-based soil carbon emission factors
for peat soils in south-east Asia, with special reference to oil palm plantations. Mires & Peat, 12, 1–
13.

Dariah, A., Marwanto, S., & Agus, F. (2013). Root- and peat-based CO2 emissions from oil palm
plantations. Mitigation and Adaptation Strategies for Global Change, (12).

Derpsch, R., Friedrich, T., Kassam, A., & Hongwen, L. (2010). Current status of adoption of no-till
farming in the world and some of its main benefits. International Journal of Agriculture and Biological
Engineering, 3(1), 1–26. doi:10.3965/j.issn.1934-6344.2010.01.0-0

Dimitriou, I. (2013). Techno-economic assessment and uncertainty analysis of thermochemical

processes for second generation biofuels. Aston University.

Ecofys (2013), Low ILUC potential of wastes and residues for biofuels. Straw, forestry residues, UCO,
corn cobs. Retrieved from http://www.ecofys.com/files/files/ecofys-2013-low-iluc-potential-of-
wastes-and-residues.pdf.

Edwards, R., Mahieu, V., Griesemann, J. C., Larive, J. F., & Rickeard, D. J. (2004). Well-to-wheels
Analysis of Future Automotive Fuels and Powertrains in the {E}uropean Context. SAE Transactions,
113(4), 1072–1084.

EPA. (2010). Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis. Retrieved from
http://www.epa.gov/otaq/fuels/renewablefuels/regulations.htm

EPA. (2012). Notice of Data Availability Concerning Fuels Produced from Palm Oil Under the RFS
Program. Federal Register, 77(18), 4300–4318.

FAO. (2010). Global Forest Resources Assessment. Retrieved from

http://www.fao.org/forestry/fra/fra2010/en/

FAO. (2014). FAOSTAT Gateway.

FAO, IIASA, ISRIC, ISS-CAS, & JRC. (2012). Harmonized World Soil Database (version 1.2).
Retrieved from http://webarchive.iiasa.ac.at/Research/LUC/External-World-soil-database/HTML/

Gunarso, P., Hartoyo, M. E., Agus, F., & Killeen, T. (2013). Oil palm and land use change in
Indonesia, Malaysia and Papua New Guinea (pp. 29–64). Singapore.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 191

Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D. Thau, S. V.
Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, and J. R. G.
Townshend. 2013. “Hansen/UMD/Google/USGS/NASA Tree Cover Loss and Gain Area.” University of
Maryland, Google, USGS, and NASA. Accessed through Global Forest Watch on 05 Feb 2015.
www.globalforestwatch.org.

Harris, N., Killeen, T., Brown, K., Netzer, M., & Gunarso, P. (2013). Pojections of oil palm expansion
in Indonesia, Malaysia and Papua New Guinea from 2010 to 2050. rspo.org (pp. 89–112). Jakarta.
Hergoualc’h, K., & Verchot, L. V. (2011). Stocks and fluxes of carbon associated with land use change
in Southeast Asian tropical peatlands: A review. Global Biogeochemical Cycles, 25(2), n/a–n/a.
doi:10.1029/2009GB003718

Hergoualc’h, K., & Verchot, L. V. (2013). Greenhouse gas emission factors for land use and land-use
change in Southeast Asian peatlands. Mitigation and Adaptation Strategies for Global Change.
doi:10.1007/s11027-013-9511-x

Herrero, M., Havlík, P., Valin, H., Notenbaert, A., Rufino, M. C., Thornton, P. K., … Obersteiner, M.
(2013). Biomass use, production, feed efficiencies, and greenhouse gas emissions from global
livestock systems. Proceedings of the National Academy of Sciences of the United States of America,
110(52), 20888–93. Retrieved from
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3876224&tool=pmcentrez&rendertype=a
bstract

HGCA (2014) HGCA/AHDB/ADAS, ‘Straw incorporation review’, AHDB Research Review No. 81, July
2014, by Fiona Nicholson, Daniel Kindred, Anne Bhogal, Susie Roques, Jonny Kerley, Susan Twining,
Tom Brassington, Peter Gladders, Helen Balshaw, Sarah Cook and Steve Ellis

Hirano, T., Segah, H., Harada, T., Limin, S., June, T., Hirata, R., & Osaki, M. (2007). Carbon dioxide
balance of a tropical peat swamp forest in Kalimantan, Indonesia. Global Change Biology, 13(2),
412–425. doi:10.1111/j.1365-2486.2006.01301.x

Hirano, T., Segah, H., Kusin, K., Limin, S., Takahashi, H., & Osaki, M. (2012). Effects of disturbances
on the carbon balance of tropical peat swamp forests. Global Change Biology, 18(11), 3410–3422.
doi:10.1111/j.1365-2486.2012.02793.x

Hoffman, L. A., & Baker, A. (2011). Estimating the Substitution of Distillers’ Grains for Corn and
Soybean Meal in the U.S. Feed Complex.

Hooijer, A., Page, S., Canadell, J. G., Silvius, M., Kwadijk, J., Wösten, H., & Jauhiainen, J. (2010).
Current and future CO2 emissions from drained peatlands in Southeast Asia. Biogeosciences, 7(5),
1505–1514. doi:10.5194/bg-7-1505-2010

Hooijer, A., Page, S., Jauhiainen, J., Lee, W. ., Lu, X. X., Idris, A., & Anshari, G. (2012). Subsidence
and carbon loss in drained tropical peatlands. Biogeosciences, 9(3), 1053–1071. doi:10.5194/bg-9-
1053-2012

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 192

Hooijer, A., Silvius, M., Wösten, H., & Page, S. E. (2006). PEAT-CO2. Assessment of CO2 emissions
from drained peatlands in SE Asia (No. Q3943). Delft, Netherlands.

Husnain, H., Wigena, I. G. P., Dariah, A., Marwanto, S., Setyanto, P., & Agus, F. (2014). CO2
emissions from tropical drained peat in Sumatra, Indonesia. Mitigation and Adaptation Strategies for
Global Change. doi:10.1007/s11027-014-9550-y

IEA. (2011). Task 37 Energy from Biogas.

IPCC. (2006). IPCC guidelines for national greenhouse gas inventories. (S. Eggleston, L. Buendia, K.
Miwa, T. Ngara, & K. Tanabe, Eds.). Institute for Global Environmental Strategies, Hayama, Japan.
IPCC. (2013). Drained Inland Organic Soils. In M. Drösler, A. Freibauer, G. Pan, & L. V. Verchot
(Eds.), 2013 Supplement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories:
Wetlands (2013th ed., pp. 1–79). Geneva: IPCC, Switzerland.

IRENA. (2013). Road transport: the cost of renewable solutions - Preliminary findings.

Jauhiainen, J., Hooijer, A., & Page, S. E. (2012). Carbon dioxide emissions from an Acacia plantation
on peatland in Sumatra, Indonesia. Biogeosciences, 9(2), 617–630. doi:10.5194/bg-9-617-2012

Johnson, J. M.-F., & Barbour, N. W. (2010). Crop yield and greenhouse gas responses to stover
harvest on glacial till Mollisol. In 19th World Congress of Soil Science, Soil Solutions for a Changing
World. Brisbane, Australia.

Jones, R. J. A., Hiederer, R., Rusco, E., & Montanarella, L. (2005). Estimating organic carbon in the
soils of Europe for policy support. European Journal of Soil Science, 56(5), 655–671.
doi:10.1111/j.1365-2389.2005.00728.x

Kalscheur, K. F., Garcia, A. D., Schingoethe, D. J., Royon, F. D., & Hippen, A. R. (2012). Feeding
biofuel co-products to dairy cattle. In H. P.S.Makkar (Ed.), Biofuel co-products as livestock feed -
Opportunities and challenges. Rome: FAO.

Khasanah, N., van Noordwijk, M., Ekadinata, A., Dewi, S., Rahayu, S., Ningsih, H., … Octaviani, R.
(2012). The carbon footprint of Indonesian palm oil production (No. 25) (p. 10). Bogor, Indonesia.

Killeen, T. J., Harris, N. L., Brown, K., Netzer, M., & Gunarso, P. (2013). Projections of Oil Palm
Expansion in Indonesia, Malaysia and Papua New Guinea from 2010 to 2050. rspo.org (pp. 89–112).

Koh, L. P., Miettinen, J., Liew, S. C., & Ghazoul, J. (2011). Remotely sensed evidence of tropical
peatland conversion to oil palm. Proceedings of the National Academy of Sciences of the United
States of America, 108(12), 5127–32. doi:10.1073/pnas.1018776108

Langeveld, J. W. A., Dixon, J., van Keulen, H., & Quist-Wessel, P. M. F. (2013). Analysing the effect
of biofuel expansion on land use in major producing countries: evidence of increased multiple
cropping.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 193

Limpert, E., Stahel, W. A., & Abbt, M. (2001). Log-normal Distributions across the Sciences: Keys
and Clues. BioScience, 51(5), 341–352.

Lugato, E., Panagos, P., Bampa, F., Jones, A., & Montanarella, L. (2013). A new baseline of organic
carbon stock in European agricultural soils using a modelling approach. Global Change Biology, n/a–
n/a. doi:10.1111/gcb.12292

Marwanto, S., & Agus, F. (2013). Is CO2 flux from oil palm plantations on peatland controlled by soil
moisture and/or soil and air temperatures? Mitigation and Adaptation Strategies for Global Change,
(12). doi:10.1007/s11027-013-9518-3
Melling, L., Goh, K. J., Chaddy, A., & Hatano, R. (2014). Soil CO2 Fluxes from Different Ages of Oil
Palm in Tropical Peatland of Sarawak, Malaysia. In A. E. Hartemink & K. McSweeney (Eds.), Soil
Carbon (1st ed., pp. 447 – 455). Cham: Springer International Publishing. doi:10.1007/978-3-319-
04084-4

Melling, L., Hatano, R., & Goh, K. (2005). Soil CO2 flux from three ecosystems in tropical peatland of
Sarawak, Malaysia. Tellus B, 57B, 1–11.

Melling, L., & Henson, I. (2011). Greenhouse gas exchange of tropical Peatlands–a review. Journal of
Oil Palm Research, 23(August), 1087–1095.

Miettinen, J., Hooijer, A., Shi, C., Tollenaar, D., Vernimmen, R., Liew, S. C., … Page, S. E. (2012).
Extent of industrial plantations on Southeast Asian peatlands in 2010 with analysis of historical
expansion and future projections. GCB Bioenergy, 4(6), 908–918. doi:10.1111/j.1757-
1707.2012.01172.x

Mutert, E., Fairhurst, T., & Uexkull, H. von. (1999). Agronomic management of oil palms on deep
peat. Better Crops International, 13(1), 22–27.
National Research Council. (2000). Nutrient Requirement of Beef Cattle: Seventh Revised Edition:
Update 2000. Washington, D.C.

Noblet, J., Cozannet, P., & Skiba, F. (2012). Nutritional value and utilization of wheat dried distillers
grain with solubles in pigs and poultry. In H. P.S.Makkar (Ed.), Biofuel co-products as livestock feed -
Opportunities and challenges. Rome: FAO.

Omar, W., Abd Aziz, N., Tarmizi, A., Harun, M. H., & Kushairi, A. (2010). Mapping of Oil Palm
Cultivation on Peatland in Malaysia (No. 529) (pp. 3–6). Kuala Lumpur, Malaysia.
Page, S. E., Morrison, R., Malins, C., Hooijer, A., Rieley, J. O., & Jauhiainen, J. (2011). Review of peat
surface greenhouse gas emissions from oil palm plantations in Southeast Asia (No. 15). Washington
D.C.

Ray, D. K., & Foley, J. A. (2013). Increasing global crop harvest frequency: recent trends and future
directions. Environmental Research Letters, 8(4), 44041. Retrieved from http://stacks.iop.org/1748-
9326/8/i=4/a=044041

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 194

Ruesch, A. S., & Gibbs, H. K. (2008). New IPCC Tier-1 Global Biomass Carbon Map For the Year
2000. Oak Ridge, Tennessee. Retrieved from
http://cdiac.ornl.gov/epubs/ndp/global_carbon/carbon_documentation.html

Scarlat, N., Martinov, M., & Dallemand, J.-F. (2010). Assessment of the availability of agricultural
crop residues in the European Union: potential and limitations for bioenergy use. Waste Management,
30, 1889–1897.

Schrier-Uijl, A., & Anshari, G. (2013). Methods for determining greenhouse gas emissions and carbon
stocks from oil palm plantations and their sourroundings in tropical peatlands. rspo.org (pp. 169–
191). Kuala Lumpur.

Schrier-Uijl, A., Silvius, M., & Parish, F. (2013). Environmental and social impacts of oil palm
cultivation on tropical peat–a scientific review (pp. 131–168). Wageningen.

Siebert, S., Portmann, F. T., & Döll, P. (2010). Global Patterns of Cropland Use Intensity. Remote
Sensing, 2(7), 1625–1643. doi:10.3390/rs2071625

Spera, S. A., Cohn, A. S., VanWey, L. K., Mustard, J. F., Rudorff, B. F., Risso, J., & Adami, M. (2014).
Recent cropping frequency, expansion, and abandonment in Mato Grosso, Brazil had selective land
characteristics. Environmental Research Letters, 9(6), 064010. doi:10.1088/1748-9326/9/6/064010

Uryu, Y., Mott, C., Foead, N., Budiman, A., Setiabudi, & Takakai, F. (2008). Deforestation, forest
degradation, biodiversity loss and CO2 emissions in Riau, Sumatra, Indonesia. One Indonesian
Province’s Forest and Peat Soil …. Jakarta, Indonesia.

Verwer, C., Meer, P. Van Der, & Nabuurs, G. (2008). Review of carbon flux estimates and other
greenhouse gas emissions from oil palm cultivation on Tropical peatlands-Identifying the gaps in
Knowledge. Wageningen.

Wilhelm, W. W., Doran, J. W., & Power, J. F. (1986). Corn and Soybean Yield Response to Crop
Residue Management Under No-Tillage Production Systems1. Agronomy Journal, 78(1), 184.
doi:10.2134/agronj1986.00021962007800010036x

Wösten, H., Ismail, A. B., & van Wijk, A. L. M. (1997). Peat subsidence and its practical implications:
a case study in Malaysia. Geoderma, 78(1-2), 25–36.

Zuo, L.-J.; Wang, X.; Liu, F. & Yi, L. (2013), 'Spatial Exploration of Multiple Cropping Efficiency in
China Based on Time Series Remote Sensing Data and Econometric Model ', Journal of Integrative
Agriculture 12(5), 903 - 913.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 195

Table 37: List of comments received on supply chains specifications in GLOBIOM and actions taken

Correction
Source Comment Comment
necessary
We compared EU crushing rates with our numbers.
Values are very close to our estimates but with
FEDIOL
Crushing rates for the EU slightly higher moisture content (2-3%). These Yes
(Jan 14)
estimates were updated in contribution from FEDIOL
from May 2014.
We compared these data with our FAO and EUROSTAT
FEDIOL Split end-use for the EU in 2011- sources and we found some consistent shares of uses
(Jan 14) 2012 Yes
overall. The data will be compared again after the
baseline has been produced.
EU crushing coefficients were found very close for
rapeseed and sunflower, and identical for soybeans to
FEDIOL
Crushing rates for major GLOBIOM assumptions. We adjusted our coefficient to
(May Yes
producing countries
2014) FEDIOL values for the year 2010 and following. For
other countries than EU, our estimates were also very
close but adjusted as well.
Due to its too loose connection to LUC dynamics, it
has been decided not to change the supply chain
EBB (April representation in the model to introduce glycerin.
Production of glycerin No
2014)
Although glycerin can impact the life cycle assessment
of biofuels, only LUC emissions are the focus of the
study.
Estimates provided are 6 to 13% above JRC values.
Epure
Corn ethanol conversion However, US EPA usually uses a value of around 417
(June Yes
coefficient in the EU: 400-427 l/t
2014) l/ton. This value will be used as a reference unless
more specific data on the EU are provided by JRC.
The value proposed by Epure are 1-3% above
Epure Wheat ethanol conversion
Biograce values. Except if authoritative reference is
(June coefficient in the EU: 0.29-0.295 No
2014) t/t. provided, we assumed Biograce default was
acceptable.
Biograce value for corn was updated with latest
Epure Wheat DDGS and corn DDGS
version of Wells to Tank analysis from JRC (2014
(June output should be comparable: Yes
2014) 0.29-0.32 t/t crop at 10% mc version 4a). Yields are now for DDGS of 0.31 tons
DDGS 0% mc for wheat and corn at 10% mc.
The numbers provided is 0.294 with 0% mc, which is
Epure
Wheat DDGS conversion ratio is equivalent to 0.326 with 10% mc. The value from
(June No
too low. Biograce for wheat DDGS seems therefore in line with
2014)
Epure input.
Epure Sugar processing co-products We did not change the supply chains to add products
(June vinasse and carbonate lime are that do not interact with LUC dynamics, because the No
2014) not represented. project only looks at LUC emissions.
Epure Sugar content assumed for sugar We analysed in details statistics received from CGB.
(June beet is too low: average of 17.6% We concluded that in order to best reflect the Yes
2014) / should be assumed for past 5 heterogeneity of sugar content across production of

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 196

 Correction
Source Comment Comment
necessary
CGB years. member states, the most consistent approach in
(March GLOBIOM was to recalculate all yield values for beet
2014) at 16% sugar content. We will for this use CIBE
information on yield in ton sugar / ha per Member
State and divide by 0.16 to obtain beet yield and
production.
Epure As explained above, our production statistics will be
Sugar production volume are not
(June updated taking harmonizing yield at 16% sugar Yes
correct at EU MS level
2014) content.
CGB This is currently the way it is done in GLOBIOM: we
Average yield values should be
(March use a 3 year average on the period 1999-2001 for the No
used rather than point estimates
2014) base year yield level.
We take yield improvement into account in our
Yield improvement should be baseline; our yield improvement assumptions will be
CGB
taken into account, sugar beet calibrated on CAPRI model projections used by DG
(March Yes
had a strong yield improvement Agriculture. If longer time series on sugar yield per ha
2014)
over past years. are provided, we can also introduce a trend on sugar
content in beet.
Sugar cane area not harvest
CGB After check, area reported by FAOSTAT as “harvested”
should be accounted to reflect
(March for Brazil correspond to the total area under sugar No
correct apparent sugar cane yield
2014) cane.
(different from field yield)
We will adjust our production statistics to reflect the
Sugar beet sugar content for actual average sugar content of each Member state
ethanol production should take and correct production for 16% sugar equivalent.
CGB
into account the fact that EU Therefore, the conversion of ethanol will be the same
(March No
ethanol producers have higher for all EU beet based on this 16% and no further
2014)
sugar content in beet than adjustment will be needed. The yield will be equal to
average EU. the actual 18.2% once ethanol production will be
adequately allocated across member states.
We checked the value used in our tables based on
Biograce/JRC of 1.77 GJ / ton sugar cane. This
Sugar cane conversion rate corresponds after conversion to 83.6 liter ethanol /
CGB
should be checked to reflect ton sugar cane. This is slightly lower than the 86.3
(March No
dehydrated conversion efficiency liter reported by CGB but considering the past
2014)
instead of hydrated one. average sugar content over 5 years was found to be
138 kg / ton sugar cane, this seems consistent (CGB
assumed for their calculation 142 kg/ton sugar cane).
As explained above, we will indeed adjust our
Sugar beet conversion factor from
CGB production and yield values in GLOBIOM to reflect
JRC should be applied to beet at
(March production of beet at 16% sugar content and not at Yes
16% sugar content, not to actual
2014) actual content. The JRC conversion factor will
yield.
therefore remain relevant.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 197

Annex III Technical background of modelling

III.1 Modelling supply of forestry residues

In order to estimate the land use emissions effect of using forestry residues, we proceeded along two
steps: i) determining the supply curves of residues in the different EU forest; ii) assessing the carbon
impact of increasing residues removals.

III.1.1 Calculation of sustainable potential

The “logging residues” potential considered here include:

 Losses from harvesting of roundwood at the forest site (i.e. rooten wood, piece of wood
unsuitable for roundwood), excluding bark (bark is assumed to be harvested and delivered to the
industry with roundwood);
 Branches, including all branches attached to the tree stem;
 Tops, as the stemwood section dimensionally unsuitable for production of roundwood (i.e. the
roundwood top diameter threshold was Country adjusted in the G4M forest model, e.g. 7-10 cm
of diameter).

The “theoretical potential volume” of logging residues is obtained from the G4M forest model
developed at IIASA, for the year 2020. The volume of logging residues from branches is calculated
with biomass expansion factors applied on the tree stem volume in the G4M model. The volume of
residues in tops is calculated as a difference between total stem and roundwood volume. The wood
harvesting losses are sourced from country-level data.
The potential volume is calculated on the basis of spatially explicit information on a 5 × 5 km forest
grid and divided between clear cuttings, thinnings and thinnings of young forests.

A “sustainable harvestable potential” is then calculated by applying the following restrictions:

A “technical recovery rate” is applied for the forest operations carried out in each cell of the forest
grid. This rate is estimated to be 70% after mechanized cutting (Nurmi 2007; Wihersaari 2005) and
60% after motor manual operations. The recovery rate reflects the percentage of branches left on the
forest stand due to the technical difficultly to be collected (e.g. sparse small branches). The
mechanization degree in EU countries is defined according to Asikainen et al. (2008).

A “techno-ecological restriction” is then also applied in order to exclude sensitive sites. The restriction
excluded from calculations stands difficult to be harvested for technical reasons (i.e. steep slope,
scarce accessibility, poor ground bearing capacity) and sensible for ecological reasons (i.e. soils poor
in nutrients, steep slopes sensible to erosion). According in Asikainen et al. (2008), in the EU after
applying these restrictions, 75% of clear cuts and 45% of thinnings are available for harvesting
logging residues. In the thinning of young forests, an average of previous values was assumed
(60%), in order to reflect developing technologies for whole tree harvesting in such sites.

Once applying the two restrictions above, the sustainable harvestable potential is estimated to 39%
of the initial theoretical potential (186 Mm3). After removal of the current uses, and exclusion of

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 198

residues whose marketable price would be higher than current wood chip prices, we finally find a
total remaining potential for the EU of 14.4 Mm3 available for biofuels.

III.1.2 Supply cost calculations

The supply costs are calculated as sum of costs for piling the biomass, extraction, chipping and road
transportation to industries. The calculations are performed on a spatially explicit grid of 5 ×5 km.
Road transportation distances from each forest cell to industrial hubs located in the major cities in
each country are with geographical information systems (GIS) analysis. According to the
mechanization degree in the different EU countries (Asikainen 2008), the piling of biomass is
assumed to be carried out by a forest harvester (i.e. mechanized) or by a forest operator equipped
with chainsaw (i.e. motor-manual). The efficiencies for forest operations in each cell are calculated
according to literature models (Brunberg 2007; Di Fulvio & Bergström 2013; Stampfer et al. 2003;
Nurmi 2007; Ghaffariyan et al. 2013) and different parameters are used in case of thinning or clear
cuts. The efficiency in road transportation of wood chips is modeled according to Johansson & Liss
(2006), trucking load capacities are adjusted according to maximum allowable payloads (European
Commission 2014) and transportation distances are determined from the GIS calculations for each
forest cell. The unitary costs for each operation are obtained in some countries of reference (i.e.
Sweden for harvester, forwarder, chipper and truck and trailer; Austria for operator with chainsaw)
and adapted to each of the EU countries by using specific econometrics relations for fixed costs (i.e.
Risk Adjusted Discount Rates (c.f. Benitez et al. 2007)), for labor cost (i.e. Purchase Power Parity
Index (World Bank, 2014)) and for operational costs (i.e. fuel prices according to GIZ (2013)).

The sum of harvesting and transportation cost for each forest cell determines the supply cost
associated to the amount of logging residues extractable. The costs are aggregated in each country in
cost supply curves, showing the cumulative amounts of biomass deliverable when increasing
progressively the supply cost.

III.1.3 Soil carbon losses

The modeling of the soil carbon losses due to a sustainable harvesting of logging residues are
estimated according to Repo et al. (2014) for the EU. These authors estimate that the sustainable
share of extractable residues range from 2 to 44% of potential, depending on the country. The
carbon losses estimates are dependent on the decomposition time of soil litter which is function of
temperature and precipitation in each country.

Losses in Repo et al (2014) study also include the impact of removing stumps (i.e. in total branches,
tops and stumps are considered). According to Strömgren et al. (2013), stumps account for the
largest share of residues impacting soil carbon and other residues account for 42% of total carbon
losses. Therefore, only 42% of the values from Repo et al. (2014) are considered in our analysis, as
we consider the impact from removing branches and tops only. Associated carbon losses are
presented in Table 1.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 199

Table 38: Development of the average litter and soil carbon loss on forest land resulting from sustainable removals

of forest harvest residues (i.e. branches and tops) in the EU countries

Soil Carbon Loss (t C ha-1)

5 years 20 years 50 years 80 years

Germany 0.6 1.5 2.7 2.9
United Kingdom 0.5 1.3 2.1 2.5
Czech Republic 0.5 1.1 2.1 1.9
Denmark 0.4 1.1 1.6 1.8
Luxembourg 0.4 1.0 1.5 1.8
Sweden 0.3 0.8 1.3 1.6
Italy 0.3 0.7 1.3 1.5
Finland 0.2 0.7 1.2 1.4
Latvia 0.3 0.8 1.2 1.4
Hungary 0.3 0.7 1.2 1.3
Netherlands 0.3 0.6 1.0 1.2
Ireland 0.2 0.7 1.0 1.2
Belgium 0.3 0.6 1.0 1.1
Poland 0.3 0.6 1.1 1.1
France 0.2 0.6 1.0 1.0
Austria 0.2 0.5 0.9 0.9
Slovakia 0.2 0.5 0.8 0.8
Estonia 0.2 0.5 0.7 0.8
Bulgaria 0.1 0.4 0.7 0.8
Romania 0.1 0.3 0.5 0.5
Slovenia 0.1 0.3 0.4 0.4
Portugal 0.1 0.2 0.4 0.4

Spain 0.1 0.2 0.4 0.4

Greece 0.1 0.1 0.3 0.4

Lithuania 0.0 0.1 0.1 0.2

Source: Repo et al. (2014)

III.1.4 References

Asikainen, A., Liiri, H., Peltola, S., Karjalainen, T., Laitila, J. 2008. Forest energy potential in Europe
(EU 27). Working Papers of the Finnish Forest Research Institute 69. 33 p. ISBN 978-951-40-2080-3.
Available at: http://www.metla.fi/julkaisut/workingpapers/2008/mwp069.htm

Nurmi J. 2007. Recovery of logging residues for energy from spruce (Pices abies) dominated stands.
Biomass and Bioenergy, 31 (2007) 375–380.

Wihersaari, M., 2005. Greenhouse gas emissions from final harvest fuel chip production in Finland.
Biomass and Bioenergy, 28 (5): 435-443

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 200

Brunberg T. 2007. Underlag för produktionsnormer för extra stora engreppskördare i slutavverkning.
(Basic data for productivity norms for extra large single-grip harvesters in final felling). Redogörelse
från Skogforsk nr 2, Skogforsk The Forest Research Institute of Sweden, Uppsala.

Di Fulvio, F., Bergström, D., 2013. Analyses of a single-machine system for harvesting pulpwood
and/or energy-wood in early thinnings. International Journal of Forest Engineering, 24 (1): 2-15.

Ghaffariyan M., R, Spinelli, R., Brown, M., 2013. A model to predict productivity of different chipping
operations. Southern Forests: a Journal of Forest Science. 75 (3), 129-136.

Stampfer, K.; Limbeck-Lilienau, B.; Kanzian, Ch.; Viertler, K. 2003: Baumverfahren im Seilgelände
Verfahrensbeispiele. – Wien: Eigenverlag des FPP Kooperationsabkommens Forst-Platte-Papier, 27 S.

Johansson, J., Liss J., E., 2006. Utvärdering av nytt ekipage för vidaretransport av bränsleflis.
Högskolan Dalarna, Institutionen för matematik, naturvetenskap och teknink. Arbetsdokument nr 3.
Grapenberg. 25 pp.

European Commission 2014. EU transport in figures. Statistical Pocketbook 2014. Publications Office
of the European Union, 2014. http://ec.europa.eu/transport/facts-fundings/statistics/index_en.htm

Benitez P.,C. McCallun I. Obersteiner, M., Yamagata, Y. 2007, Regional Externalities, Heijman, Wim
(Ed.) Springer 2007, XIV, 342 p

The World Bank http://data.worldbank.org/ accessed on October 2014

Deutche Gesellschaft fur Internationale Zusammenarbeit (GIZ) GmbH. Fuel Prices 2012.
www.giz.de/transport

Schlamadinger B, Spitzer J, Kohlmaier GH, Lüdeke M. 1995. Carbon balance of bioenergy from
logging
residues. Biomass & Bioenergy 8(4):221–234.

Strömgren M., Egnell G., Olsson B. A. 2013. Carbon stocks in four forest stands in Sweden 25 years
after harvesting of slash and stumps. Forest Ecology and Management, 290: 59-66.

Repo A., Böttcher H., Kindermann G., Liski J. 2014. Sustainability of forest bioenergy in Europe: land-
use-related carbon dioxide emissions of forest harvest residues. GCB Bioenergy,
doi:10.1111/gcbb.12179.

III.2 Amortisation of emissions over 50 instead of 20 years

One might assume that with the choice to distribute LUC emissions over 20 years, there will be no
further LUC emissions after that period. This section assesses how the results would change if a
longer reference period is adopted.

The timing of GHG emissions over time for the different GHG emission and sink sources are
represented in Figure 37, for a stylized 1% shock (land use footprint with the back dashed line). In
line with the policy scenario B chosen for this study, land use change occurs on the period 2010
(2008 in reality, but simplified for illustration purposes) to 2020 as new areas are allocated to
biofuels as the EU approaches its 10% target for renewable energy in transport.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 201

This leads to natural vegetation conversion emissions during this period and in parallel, land
cover change for cropping leads to some sequestration in the agricultural biomass, that are
accounted on the period of the deployment.135 These emissions are then divided by the reference
period considered for land use emission accounting, 20 years in the case of this study.
Natural vegetation reversion (or foregone sequestration) on abandoned land leads to some carbon
sequestration that does not any longer occurs if this land is used for agriculture. This constitutes an
additional source of emissions, the time span of which goes much beyond the reference period. In the
current accounting, foregone sequestration is accounted on the period 2010-2030. Accounting for a
longer time-frame would increase the amount of cumulated emissions, and beyond 50 years of forest
regrowth, we consider here that the forest is reverted, i.e.cumulative emissions from natural
vegetation reversion would then equal those of clearing of a forest in the same location.136
Soil organic carbon is usually released for long periods after cropland management is changed, e.g.
in case of tillage of new land. However, IPCC Tier 1 approach simplifies the representation of SOC
emissions and considers all emissions occur in the first 20 years after management change. This is
also the approach taken here, therefore we account in our setting for all SOC emissions for mineral
soils. Changing the reference period of 20 years would proportionally lower these emission values,
because, for this element, the cumulated emissions remain unchanged.
Peatland oxidation (carbon emissions from organic soils) follows a different dynamics than that
from mineral soil, because continuous drainage maintains the level of carbon emissions, which
continue until the total oxidation of the peat. As moderate peat depth is considered 1-2 meters,
whereas deep peat can be more that 4 meters deep. At a subsidence rate of 5 cm/year in first
decades, usually declining over time due to peat compaction, time for total exhaust of carbon stock
can reach easily 50 to 100 years.137

We compare in Table 39 what the LUC emission factors would be in the case of the EU2020 mix
scenario if the reference period was not any longer 20 years but was instead 50 years. Under such
assumption, emission from natural vegetation conversion, agricultural biomass and soil organic
carbon decrease proportionally (by a factor 20/50). However, emissions from peatland are
unchanged and emission from foregone sequestration are even slightly higher because the full forest
regrowth is now considered. As a consequence, the decrease in the total LUC value is not as large as
it would have occurred by just rescaling all emission factors by 20/50. For the EU2020 mix scenario,
the decrease in LUC value is only 19%. In other words, the cumulative LUC emissions double if the
reference period is extended from 20 to 50 years, due to long living emissions sources, such as
peatland oxidation and natural vegetation reversion.

Contributions of the different sources vary however significantly across feedstocks. Therefore, the
change in LUC value can be more notable in the case of some particular feedstocks, as illustrated in
Figure 2. The largest differences are observed in the case of soybean oil (-49%), because a large
contribution to emissions is coming from natural vegetation conversion, to the difference of palm oil
where peatland is the largest contributor and remains much less affected. Other notable changes are

135
For palm plantation, the C regrowth could go slightly beyond 2020 but still would fall within the first 20 years.
136
The reference period of 2010-2030 was chosen here to align with the reference period for natural vegetation conversion and reversion.
Taking a later 20 year reference period than 2010-2030 would potentially increase further the cumulated reversion value because all biofuel
deployment would already have taken place. It should however be kept in mind that the pace of C sequestration for reversion is highly
uncertain in the short to medium run. Long term estimates for foregone sequestration– beyond 50 years – are more accurate in case of
forest regrowth, because most of the regrowth can be assumed to have taken place.
137
Note that because our modelling of peatland drainage is not spatially explicit, it is not possible here to know the exact timing of peatland
emissions. In any case, no large-scale map of peatland depth in Indonesia and Malaysia could be found at the time of this study.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 202

for perennials that now show positive emissions (2 gCO2/MJ instead of -12 gCO2/MJ) and short
rotation coppices also strongly decrease their benefits (-5 gCO2/MJ instead of -29 gCO2/MJ). This is
because reversion of forest in the long term is a better sink than these feedstocks. For such long time
horizon, these feedstocks remain beneficial only in case where they are grown in areas where natural
forest would not regrow.

Emission sources and time

Emission per year

2010 2015 2020 2025 2030 2035 2040 2045 2050 2055 2060 2065 2070 2075 2080

Natural vegetation conversion Peatland oxidation Agricultural biomass

Natural vegetation reversion SOC Land use biofuels

Figure 37. GHG emission flow overtime in this study, in difference to the baseline, corresponding to a marginal shock

of 1%. Magnitude of emissions represented here are not corresponding to a specific feedstock. The dashed line

represent the land occupied by the biofuel feedstock, expanding on 2010-2020 and then stable after the shock.

Table 39. Change in accounting of each emission source for the EU 2020 mix scenario with 20 years and 50 years

reference period.
gCO2/MJ Reference period 20 years Reference period 50 years
Natural vegetation conversion 56 22
Natural vegetation reversion 7 9
Agricultural biomass -37 -15
Soil organic carbon 15 6
Peatland oxidation 57 57
Natural vegetation conversion 56 22
Total 97 79

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 203

 250

200

150
gCO2/MJ

100

50

-50

50 years 20 years

Figure 38. LUC value for different feedstocks with reference period of 50 years compared to 20 years

Finally, in this context, it is also interesting to briefly reflect on an alternative policy scenario which
was not assessed in this study. Both, scenario B and B1 assume no further change in biofuel use from
2020, i.e. continued LUC emissions during the reference period (and - as explained above – with LUC
emission even beyond the reference period in reality). The European Commission has adopted the
view that “food-based biofuels should not receive public support after 2020”138: Assuming such a
change in EU policy towards a phase-out of (conventional) biofuels , LUC emissions would be reduced
accordingly, because the new cropland initially converted due to the biofuel policy would be taken
over by expanding agricultural production for non-biofuels uses, the net cropland take of 64Mha
which is eight times higher than the net cropland conversion of 8Mha due to biofuels during 2008-
2020. Consequently, as long as overall net land take continues at a strong pace, the EU has, in
theory, the option to revert to a “zero LUC” situation139 by phasing out (conventional) biofuels. It
would mean that the LUC emissions transitorily attributable to the RED stimulus for biofuels are
gradually transferred to the overall expanding agricultural system where they are not accounted for.

III.3

138
COM(2014)15
139
Even when reverting to a no biofuels policy a net carbon debt may still remain, not least because the EU biofuels policy anticipated the
overall land use change as compared to the baseline.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 204

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 205
Annex IV Data used in the GLOBIOM model

IV.1 Parameters

Table 39: Energy content of various biofuel types, as Lower Heating Value (LHV) at 0% mc
a)
Bioethanol 26.8 MJLHV/kg
a)
Biodiesel (FAME) 37.2 MJLHV/kg
a)
Biodiesel (HVO) 44 MJLHV/kg
a)
Biodiesel (Fischer Tropsch) 44 MJLHV/kg
b)
Butanol 33 MJLHV/kg
b)
Methanol 20 MJLHV/kg
b)
Bio DME 28 MJLHV/kg
b)
Methane (upgraded biogas) 50 MJLHV/kg (or about 33 MJLHV/m3)
a) (Biograce 2014).
b) Renewable Energy Directive 2009/28/EC.

IV.2 Land cover data

Land use data are important for GLOBIOM because it constitutes the backbone of the bottom-up
modelling structure (see Annex I). In particular it indicates where crops are grown (informing on their
potential yield), where other uses of land compete with it (livestock on grassland, forestry) and
where some land is available.

Land cover at the global level is based on the Global Land Cover 2000 dataset (GLC2000) but more
detailed land cover maps exist for the EU. The European Environment Agency in particular
disseminate the CORINE land cover maps, that provide information on base year 2000 land cover for
Europe at a 1x1 km resolution. We build on this information to represent the land cover in Europe at
a detailed level. GLOBIOM cropland areas mainly include CORINE class 210 arable land and
heterogeneous areas (class 240) and is adjusted in Europe to match the harvested area in GLOBIOM
(including fallow). Forest areas in GLOBIOM consist of total forests (class 310) harmonized with forest
areas from the G4M model. For grassland, pastures (class 230) is used. However, these areas are
then adjusted in relation to grazing quantities to represent only productive grassland. This allows to
represent the possibility of expansion of livestock within the current grassland areas, and the
possibility to convert unused grassland to other uses. The heterogeneous areas cover (class 240) is
then used as a buffer for this adjusment. Other cropland which represents crop not covered currently
by the model is calculated using EUROSTAT data. The remaining CORINE land cover classes artificial
areas (class 100), permanent crops (orchards, vineyard, etc., class 220), open space (i.e. natural
land with sparse or no vegetation, class 330), wetlands (class 400) and water bodies (class 500) are
kept constant over time.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 206

Table 40: Land cover data used as input for GLOBIOM in 2000 (1,000 ha) [Source: Corine land cover 2000]

Other arable
Grassland 2

vegetation

Wetlands
Cropland

natural
1

Forest

Other
Total

land
Austria 7,515 1,284 1,600 3,828 629 159 16
Belgium 2,456 813 867 653 0 112 10
Bulgaria 10,363 3,293 2,276 3,507 877 398 11
Croatia 5,196 669 22 2,129 301 2,074 0
Cyprus 861 63 0 148 456 195 0
Czech Republic 7,409 2,495 1,955 2,568 163 220 8
Denmark 3,945 1,920 1,349 526 0 88 62
Estonia 4,100 613 807 2,169 307 15 188
Finland 29,912 2,050 180 21,953 4,838 24 867
France 51,514 18,027 14,251 13,783 3,404 1,868 181
Germany 32,687 11,312 9,543 10,754 0 898 178
Greece 12,383 2,232 2,310 1,185 5,773 845 39
Hungary 8,618 4,197 1,866 1,661 215 616 63
Ireland 6,671 673 4,056 461 435 43 1,002
Italy 27,210 7,704 5,246 7,457 3,518 3,227 58
Latvia 6,204 1,137 1,631 3,148 68 72 149
Lithuania 6,153 1,479 2,230 1,924 145 320 55
Luxembourg 241 37 23 89 6 85 0
Malta 23 3 5 0 0 14 0
Netherlands 3,026 851 1,582 345 37 180 31
Poland 29,769 12,389 6,502 8,878 347 1,521 132
Portugal 8,492 1,732 2,088 3,825 0 821 25
Romania 22,135 8,777 2,575 6,698 2,991 847 247
Slovakia 4,600 1,374 984 1,979 144 114 5
Slovenia 1,933 148 499 1,189 56 39 3
Spain 47,527 12,538 7,759 12,583 9,248 5,307 93
Sweden 39,561 2,577 9,298 24,077 0 96 3,514
UK 22,395 5,539 8,337 2,602 5,123 356 438
1
This does not include here artificial areas and mountains, deserts, lakes and other not relevant areas. Total is
therefore lower than country official area.
2
Before adjustment in GLOBIOM to distinguish productive grassland from not grazed areas.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 207

IV.3 Carbon stocks

The table below contains the average carbon stock values in tonnes per hectare per region.

Table 42: Average carbon stock values per carbon pool per region (t/ha) [source: GLOBIOM model based on

estimates from Forest Resource Assessment 2010 (FAO) for forestry and Ruesch and Gibbs (2008) dataset for living

biomass carbon stock in other natural vegetation and in grassland.]

Forest Forest Other Natural Land Grassland

Above-and-Below- Dead Organic Above-and-Below- Above-and-Below-
Ground-Biomass Matter Ground-Biomass Ground-Biomass
Latin America 128 14 26 7
South Asia 64 9 29 3
North America 66 19 10 3
EU28 88 22 9 3
Eastern Asia 43 11 14 2
Southeast Asia 120 13 29 5
Russia and 59 21 4 4
neighbouring
countries formerly
part of the USSR
Sub-saharan Africa 111 12 35 4
Oceania 69 15 17 3
Middle East and 71 15 13 2
North Africa

IV.4 Crop yields

For the EU, EPIC simulations are performed for different crop rotations and tillage systems
(conventional, reduced, and minimum tillage) with statistically computed fertilizer rates and irrigation
management. Crop rotations have been derived from crop shares calculated from EUROSTAT
statistics on crop areas in NUTS2 regions explicitly taking into account data on relative crop shares
and agronomic constraints such as maximum frequency in a rotation. Average NUTS2 EPIC yields are
harmonized with EUROSTAT/CAPRI data to match production and area data in the base year. The
yield values are based on a 1998-2002 average. For 2010 we implement yield changes according to
historic data while we apply an exogenous yield trend 2010 onwards which has been estimated based
on 1998-2012 data (see Section 2.2.3 for yield trends in the baseline). Endogenous yield response is
operated in the model through system shifts, as explained in Annex I, Section I.2.4).

In the case of sugar beet, the yield value indicated below are raw data not adjusted by sugar content.
These data have then been rescaled in the model to represent yield at 16% sugar content.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 208

Table 41: Crop yield (fresh matter tonne/hectare)

Sunflower seeds

Durum wheat
Sugar beet

Soft wheat
Rapeseed

Soybean
Potato
Barley

Peas
Corn

Oats

Rye
Dry matter content (%) 89 85 90 20 91 90 24 94 85 85 85 85

Austria 4.5 11 2.7 30.3 2.7 2.3 64.7 2.6 5.1 3.9 4 4

Belgium 6.5 11 3.9 44.7 3.2 65.7 7.6 5.3 5

Bulgaria 3.1 2.8 0.8 10.2 1.2 0.9 14.6 1.1 2.9 3.7 1.7 1.9

Croatia 3.2 5.2 0.7 13.2 2.1 2 33.5 1.8

Cyprus 1.8 0.9 20.3

Czech Republic 3.8 6.6 2.3 20.5 2.7 1.4 46.5 2.1 4.7 4 4

Denmark 5.5 3.3 37.8 2.8 55.8 7.7 5.4 5.2

Estonia 1.9 1.6 14.3 1.4 2.1 1.7 1.9

Finland 3.4 2 24.2 1.4 33.9 0.5 3.4 2.6 3.1

France 6.1 8.9 4.6 39.2 3 2.6 70.9 2.3 7.4 3.5 1

Germany 6.2 9.2 3.4 39 3.5 58.1 2.4 7.7 4.2 3.1 3.7

Greece 2.9 9.2 1.8 24.2 2.6 66.7 1.5 1.8 2.6 5.2 5.3

Hungary 3.2 5.6 1.9 20.9 1.7 2 47.6 1.7 4 3.8 1.6 2.1

Ireland 7.1 2.3 31.7 3.5 41.5 9.4 3.2 2.5

Italy 3.7 9.7 2.1 24.7 1.1 3.7 46.9 2.1 3.8 2.8 4.3 4.7

Latvia 1.9 1.6 13.4 1.6 31.4 2.7 5 5.8

Lithuania 2.5 2.3 1.4 13 1.3 30.4 3.4 2.1 2.1

Luxembourg 6.5 11 3.9 44.7 3.2 7.6 2.4 2.2

Malta 4.6 1.5 19 7.4

Netherlands 6.1 9.3 3.8 45.7 2.7 60.1 5.1 8.2 2.4 2.9

Poland 3.3 6.4 1.6 17.7 2.3 2.9 37 1.1 3.9 1.5 2

Portugal 1.4 6.2 0.5 14.3 58.5 0.5 0.9 1.8 1.7 2.5

Romania 2.7 2.8 1.4 13.8 1.2 1.3 19.4 1.1 2.9 1.6 2.2

Slovakia 2.7 4.3 1.6 14.5 1.9 1.2 37.3 1.6 3.6 4.1 1.6 2.5

Slovenia 3.5 6.3 2.4 20 2.3 2.1 42.6 2.4 4.4 2.6 3.1

Spain 2.8 9.3 0.7 26.2 1.4 2.3 64 0.9 2.7 2.6 1.8 1.5

Sweden 4.1 7.7 2.7 32.7 2.2 46.6 6.1 3.7 5.4

UK 5.6 3.7 39 3.5 46.1 3.1 7.5 5.8 5.8

At the global level, yields are also estimated through EPIC estimation, and harmonised with FAO
statistics by country. Yield values from FAO are used from the period 1998-2002.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 209

Reference values are reported below. For sugar cane in Brazil and palm oil in Malaysia and Indonesia,
FAO harvesting yield values are corrected to correspond to average yield by planted area instead of
harvested area.

Table 42: Crop yields in selected world regions and countries (fresh matter tonne/ha)

Sunflower seed
Oil palm fruit

Sugar cane
Rapeseed

Sorghum
Potatoes

Soybean
Cassava

Wheat
Barley

Corn

Rice
Dry matter content (%) 89 21 85 53 20 91 85 90 89 25 94 85
Australia 1.9 6.0 34.3 1.2 8.9 1.9 2.8 85.7 1.0 1.8
Argentina 2.4 10.0 5.7 26.9 1.4 5.4 2.5 4.8 63.9 1.8 2.3
Brazil 2.0 13.3 3.0 10.0 17.4 1.6 3.0 2.5 1.7 69.3 1.6 1.7
Canada 2.8 7.2 27.3 1.4 2.4 1.6 2.2
China 3.0 16.1 4.9 14.2 14.3 1.5 6.3 1.8 3.6 66.5 1.6 3.8
Congo Basin 0.6 7.8 1.1 8.2 4.7 0.9 0.6 1.2 18.7 1.3
Former_USSR 1.6 2.5 10.7 0.7 2.7 1.0 1.2 0.9 1.6
India 2.0 26.0 1.8 17.8 0.9 2.9 1.0 0.8 69.3 0.6 2.7
Indonesia 12.6 2.8 17.5 14.8 4.3 1.2 55.9
Japan 3.5 2.4 31.4 1.5 6.5 1.8 65.0 3.7
Malaysia 10.1 2.5 18.1 3.1 0.3 75.6
Mexico 2.2 13.4 2.5 14.8 22.9 1.2 4.5 1.6 3.2 73.9 0.8 4.7
Middle East North Africa 0.9 5.4 20.2 1.7 6.3 1.6 2.1 105.0 0.9 1.9
Central America 0.9 6.0 1.5 17.2 21.3 3.3 2.5 1.2 46.7 1.6
Rest of South America 1.3 12.0 2.4 15.0 12.1 2.3 4.6 2.3 2.4 74.6 1.1 2.4
Rest of South Asia 1.0 8.5 1.8 12.3 0.8 3.2 0.8 0.6 45.9 1.4 2.3
South Korea 3.8 4.0 24.8 1.3 6.6 1.4 1.4 3.2
Eastern Africa 1.1 10.1 1.3 13.3 7.6 0.8 1.6 1.1 1.2 86.1 0.7 1.3
Southern Africa 3.7 7.0 1.1 12.1 10.2 1.9 1.9 0.7 68.8 0.6 4.2
Western Africa 0.8 9.7 1.4 3.3 5.2 1.6 0.8 0.8 57.2 0.7 1.8
Turkey 2.2 4.2 26.1 2.3 5.7 2.8 1.5 2.1
Ukraine 2.1 3.0 10.4 0.9 3.4 1.1 1.0 1.1 2.6
USA 3.2 2.5 8.4 40.4 1.5 6.9 2.6 3.9 78.2 1.5 2.7

IV.5 Bioenergy transformation pathways

For most biofuel pathways, the total feedstock to fuel conversion is described as one step. For some
pathways, the conversion is described in two steps via an intermediate product (e.g. vegetable oil).
Conversion coefficients are applied worldwide, except where indicated otherwise. These coefficients
are kept constant over time.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 210

IV.5.1 Production of ethanol

Table 43: Conversion of corn to ethanol

Product Region Unit Input Output

Corn tonne (15% mc) -1
a)
Ethanol USA GJ 8.68
tonne (0% mc) 0.324
b)
EU & ROW GJ 8.72
tonne (0% mc) 0.325
a)
Corn DDGS USA tonne (0% mc) 0.304
GJ 5.42
c)
EU & ROW tonne (0% mc) 0.295
GJ 5.26
a) 2.76 gallon (= 10.5 litre) ethanol and 17 lbs (=7.7 kg) of dried distillers grains per bushel corn (=25.4012 kg at 15.5% mc) (EPA,
2010), with LHV corn at 18.5 MJ/kg at 0% mc, LHV ethanol at 26.81 MJ/kg at 0% mc and LHV DDGS at 16.0 MJ/kg at 10% mc.
b) Edwards et al. (2004). Revision V4 (2014). Pathway “Production of Ethanol from Corn (Community produced) (steam from natural gas
CHP)”. Overall yield is 0.6032 MJ ethanol/MJ corn, with LHV corn at 17 MJ/kg at 0% mc and LHV ethanol at 26.81 MJ/kg at 0% mc.
c) Ibid, Yield of DDGS is 1.392 tonne DDGS/tonne ethanol, with DDGS at 10% mc. LHV DDGS is 16.0 MJ/kg at 10% mc. Biograce does not
make distinction between the energy content of corn DDGS and wheat DDGS. Note that Globiom will not use LHV for DDGS but more
metabolizable energy by animal.

Table 44: Conversion of wheat to ethanol

Product Region Unit Input Output

Wheat tonne (15% mc) -1
a)
Ethanol Global GJ 7.68
tonne (0% mc) 0.286
b)
Wheat DDGS Global tonne (0% mc) 0.294
GJ 5.22
a) Biograce (2014). Pathway “Production of Ethanol from Wheat (steam from natural gas CHP)”. Overall yield is 0.5313 MJ
ethanol/MJ wheat, with LHV wheat at 17.0 MJ/kg at 0% mc and LHV ethanol at 26.81 MJ/kg at 0% mc
b) Ibid. Yield of DDGS is 1.14 tonne DDGS/tonne ethanol, with DDGS at 10% mc. LHV DDGS at 16.0 MJ/kg at 10% mc.

Since the starch content of rye is approximately the same as for wheat, the same conversion
efficiencies and costs are assumed. Wheat and rye are processed in the same ethanol facility, with
feedstock mix depending on availability and cost.

Table 45: Conversion of rye to ethanol

Product Region Unit Input Output

Rye tonne (15% mc) -1
Ethanol Global GJ 7.68
tonne (0% mc) 0.286
Rye DDGS Global tonne (0% mc) 0.294
GJ 5.22

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 211

Table 46: Conversion of sugar beet to ethanol

Product Region Unit Input Output

tonne (76% mc, 16%
Sugar beet -1
sugar content)
a)
Ethanol Global GJ 2.13
tonne 0.079
b)
Sugar fibre Global tonne (0% mc) 0.055
GJ 0.857
a) Biograce (2014). Pathway “Production of Ethanol from Sugar beet (steam from NG boiler). Overall yield is 0.5436 MJ ethanol/MJ
sugar beet, with LHV sugar beet at 16.3 MJ/kg at 0% mc and LHV ethanol at 26.81 MJ/kg at 0% mc.
b) Ibid. Yield of co-product is 0.219 MJ sugar beet pulp/MJ sugar beet, with LHV sugar beet pulp at 15.6 MJ/kg at 0% mc.

Table 47: Conversion of sugar cane to ethanol

Product Region Unit Input Output

Sugar cane tonne (75% mc) -1
a)
Ethanol Global GJ 1.77
tonne (0% mc) 0.066
b)
Bagasse Global kWh N/A
a) Biograce (2014). Pathway “Production of Ethanol from Sugarcane”. Overall yield is 0.3607 MJ ethanol/MJ sugar cane, with LHV
sugar cane at 19.6 MJ/kg at 0% mc and LHV ethanol at 26.81 MJ/kg at 0% mc.
b) Electricity cogeneration is not explicitly represented in GLOBIOM for sugar cane processing but accounted for through the absence
of energy cost for production.

IV.5.2 Oilseed crushing

The crushing ratios currently used in the model are derived from data provided by national statistic
offices and accessible through Eurostat or FAOSTAT. Crushing rates and crushing efficiency are then
reproduced in the model as they appear and kept constant over time. Within EU, national statistics
display some variations that may not necessarily correspond to differences in technologies used but
most likely in heterogeneity in crop processed. For that reason, we only use one average EU crushing
rate.

The crushing ratios for oil used here should be interpreted as fresh seed to crude oil crushing ratios.
Conversion to biodiesel later requires a vegetable oil refining stage that is accounted for separately.
Cake and oil do not sum to 100% due to seed moisture extraction and in some cases additional
losses.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 212

 a)
Table 48: Crushing ratio oilseeds (1999-2001 average) . Weight fractions
Rapeseed Soybean Sunflower

Cake Oil Total Cake Oil Total SunC SunO Total

Dry matter content (%) 89 100 (91) 89 100 (90) 92 100 (94)

(Total refers to seed)

b)
EU28 56 42.5 98.5 80 18 98 54 44.5 98.5
Other regions
Brazil 79 20 99
Canada 56 42 98 78 18 96
China 62 36 98 82 18 100 50 35 85
Former USSR 54 43 97
India 60 35 95 80 18 98
Japan 57 42 99 77 19 96
Mexico 80 15 95
Middle East & North Africa 80 17 97
Rest of South America 80 19 99 57 41 98
Rest of South Asia 62 33 95
South-East Asia 80 18 98
South Korea 76 18 94
Turkey 45 38 83
USA 79 19 98
a) FAOSTAT, UN Food and Agriculture Organization and FEDIOL. We report here ratios for crushed quantities higher than 1 million
tonnes.
b) EUROSTAT/CAPRI database and FEDIOL.

IV.5.3 Vegetable oil refining

The use of crude vegetable oil as a feedstock for biodiesel involves a refining stage, leading to some
losses. We currently apply 4% by mass loss for all regions and vegetable oil types.

Table 49: Vegetable oil refining

Product Unit Input Output

Crude vegetable oil tonne -1
a)
Refined vegetable oil tonne 0.960
a) Edwards et al. (2004).

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 213

IV.5.4 Production of FAME biodiesel via (trans)esterification

Table 50: Conversion of vegetable oil to FAME biodiesel

Product Region Unit Input Output

Refined oil tonne (0% mc) -1
a)
FAME Global GJ 36.6
tonne (0% mc) 0.983
a) Biograce (2014). Pathway “Production of FAME from Rapeseed (steam from natural gas boiler)”. Yield is 0.9936 MJ FAME/MJ
refined oil, with LHV FAME at 37.2 MJ/kg at 0% mc. The LHV refined oil is not given by Biograce, we assume it is similar to that of soybean
and palm oil at 36.8 MJ/kg at 0% mc. Furthermore, refined glycerol is coproduced at 105.6 kg / tonne FAME. The co-production of glycerol is
accounted for in the processing costs. The Globiom model does not take into account trickle down effects of glycerol, even though it can be
used to produce biofuels, such as biomethanol (production of fuels on basis of residues is separately accounted for).

IV.5.5 Production of HVO biodiesel via hydrotreatment

Table 51: Conversion of vegetable oil to HVO biodiesel

Product Region Unit Input Output

Vegetable oil tonne (0% mc) 1
a)
HVO Global GJ 34.8
tonne (0% mc) 0.791
a) Biograce (2014). Pathway “Production of HVO from Rapeseed (steam from natural gas boiler)”. Yield is 0.967 MJ HVO/MJ oil (not
refined), with LHV HVO at 44.0 MJ/kg at 0% mc. Assume that LHV vegetable oil is 36.0 MJ/kg at 0% mc. Biograce does not specify co-
product, although other sources mention gasoline and propane as side products.

IV.5.6 Production of biogas

Table 52: Conversion of maize silage to biogas

Product Region Unit Input Output

Maize silage tonne (0% mc) 1
a)
Biogas Global GJ 9.9
tonne (0% mc) 0.198
a) Typical yield from (IEA, 2011) slide 5, biogas from whole crop maize is 178 – 400 m3 methane per tonne dry matter (mainly
depending on the feedstock quality and retention time), so we use 300 m3 as average value, with lower heating value methane at 33
MJ/m3. However, about 25% of energy produced is used to drive the complete process (digester and upgrading). Density methane is 0.66
kg/m3.

IV.5.7 Production of cellulosic ethanol via hydrolysis-fermentation

This pathway is included as a container of future technologies producing alcohols from lignocellulosic
biomass.

Table 53: Conversion of wood to cellulose ethanol

Product Region Unit Input Output

Wood tonne (0% mc) -1
a)
Ethanol GJ 6.99
tonne (0% mc) 0.348
a) IRENA (2013) Table 4.2: The average yield is 440 liters of ethanol per tonne (0% mc) wood. However the same report explains yield
beyond 330 liters of ethanol per tonne are not economically profitable, therefore we assume here a yield of 330 liters per tonne. Assume
LHV ethanol at 26.81 MJ/kg at 0% mc as in other tables above, and a density of 0.79 kg/litre.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 214

IV.5.8 Production of diesel via gasification and Fischer-Tropsch synthesis

This pathway should be seen as a container of future technologies producing diesel-like fuels from
lignocellulosic biomass.

Table 54: Conversion of wood to Fischer-Tropsch diesel

Product Region Unit Input Output

Wood tonne (0% mc) -1
a)
FT Diesel GJ 9.37
tonne (0% mc) 0.213
a) Dimitriou (2013) compares FT diesel production via entrained flow gasification and circulating fluidised bed gasification and finds
comparable outcomes. We have used the parameters for the CFB pathway here: At an input of 120 tonne/hr wet biomass (30% mc and LHV
13,056 MJ/kg), thus 84 tonne at 0% mc, the output is 17.93 tonne/hr FT diesel at 43.92 MJ/kg.

IV.6 Co-product replacement coefficients

Insertion of co-product in animal feed is part of selected improvements to GLOBIOM. See Annex II.7
for full data.The table below reports substitution coefficients used for co-product substitution.

In the case of the US, as the substitution of co-products have been over recent year performed more
on energy than protein basis, we apply the coefficients reported by USDA (Hoffman & Baker, 2011).

Table 55: Substitution pattern applied in the EU for each animal species for one unit of co-product consumed by the

livestock sector. Positive values correspond to a replacement of feed, negative value to a joint addition of another

feedstuff to preserve the energy balance

Sugar beet Rapeseed
Feed item Corn DDGS Wheat DDGS Sunflower meal
pulp meal
SUBSTITUTE FOR CORN & SOYBEAN MEALS
Beef
Corn 0.711 0.523 0.800 -0.085 -0.390
Soya meal* 0.470 0.619 0.028 0.832 1.083

Dairy
Corn 0.673 0.452 0.849 -0.005 -0.294
Soya meal* 0.478 0.635 0.017 0.815 1.062

Swine
Corn 0.559 0.382 0.824 0.121 -0.098
Soya meal* 0.494 0.650 0.022 0.787 1.019

Poultry
Corn 0.298 0.178 0.066 -0.030 -0.073
Soya meal* 0.560 0.695 0.188 0.820 1.014

SUBSTITUTE FOR WHEAT & SOYBEAN MEALS

Beef
Wheat 0.791 0.582 0.890 -0.094 -0.434
Soya meal* 0.371 0.547 -0.083 0.844 1.137

Dairy
Wheat 0.753 0.506 0.950 -0.006 -0.329

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 215

 Sugar beet Rapeseed
Feed item Corn DDGS Wheat DDGS Sunflower meal
pulp meal
Soya meal* 0.384 0.571 -0.102 0.816 1.103

Swine
Wheat 0.686 0.437 0.944 0.139 -0.112
Soya meal* 0.405 0.593 -0.100 0.769 1.034

Poultry
Wheat 0.375 0.224 0.083 -0.038 -0.091
Soya meal* 0.505 0.662 0.176 0.826 1.027

Note: Soybean meals are not represented here as they are largely used already as feed in rows. Their
substitution values can be found by reading the table from row to column and inverting the cereal
contribution. For instance, 0.833 unit of soybean meal substitute for beef with 1 unit of rapeseed
meal and 0.09 unit of corn (corn is now replaced as the negative sign needs to be inversed).

IV.7 Biofuel feedstock demand

Tables below provide statistics for the model base year on different uses of biofuel feedstocks. This
information is useful to understand with what uses the incorporation of feedstocks conflict with on the
markets. In GLOBIOM, food and feed are represented separately and other uses are assumed to
respond similarly to food in case of price changes. These uses are distinguished between food, feed
and other uses (which in principle include biofuels, although the biofuels production in 1999-2001
was relatively limited). These data represent EU consumption, therefore it includes imports to the EU
market, but not exports from the EU. Food and Other uses can include some industrial processing,
but crushing is accounted as specific category in the case of oilseeds, because corresponding supply
chains are explicitly represented in GLOBIOM.

IV.7.1 Demand for ethanol feedstocks (1,000 t FM, average 1999-2001)

Barley Corn Oats Rye Sugar beet Wheat

Austria Food 3 133 10 111 1,498 578
Austria Feed 652 1,357 155 89 78 384
Austria Other uses 242 669 13 18 1,234 89
Belgium Food 50 107 1 14 2,737 1,353
Belgium Feed 538 842 80 6 413 1,403
Belgium Other uses 293 125 1 8 2,282 474
Bulgaria Food 50 339 2 10 1,207 1,570
Bulgaria Feed 485 972 67 22 50 1,060
Bulgaria Other uses 219 110 10 4 351 333
Croatia Food 1 343 4 9 811 461
Croatia Feed 168 1,287 65 0 14 139
Croatia Other uses 70 115 4 1 390 76
Cyprus Food 3 23 0 0 0 71
Cyprus Feed 320 181 0 0 0 37
Cyprus Other uses 102 19 0 0 0 6
CzechRep Food 112 81 24 204 2,475 1,142
CzechRep Feed 1,287 320 112 45 0 2,288
CzechRep Other uses 458 16 18 17 792 397
Denmark Food 10 16 32 84 1,019 501

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 216

 Barley Corn Oats Rye Sugar beet Wheat
Denmark Feed 2,736 53 181 103 592 3,370
Denmark Other uses 645 5 15 29 452 269
Estonia Food 9 2 4 52 0 70
Estonia Feed 246 10 81 16 0 142
Estonia Other uses 61 1 16 13 0 19
Finland Food 8 7 47 100 1,072 348
Finland Feed 1,173 51 819 1 134 144
Finland Other uses 455 27 88 14 336 123
France Food 160 200 36 35 11,907 8,129
France Feed 3,370 5,830 662 105 2,067 11,331
France Other uses 559 935 18 4 7,172 1,890
Germany Food 239 765 191 984 15,122 6,783
Germany Feed 8,626 3,105 1,084 1,812 2,640 9,374
Germany Other uses 2,859 641 107 508 2,004 1,645
Greece Food 12 16 11 9 2,120 2,165
Greece Feed 449 2,460 92 16 223 49
Greece Other uses 50 34 8 4 420 177
Hungary Food 17 249 3 13 2,021 1,510
Hungary Feed 750 3,972 133 72 83 1,065
Hungary Other uses 238 605 14 9 979 303
Ireland Food 30 65 10 3 659 380
Ireland Feed 918 153 101 0 395 775
Ireland Other uses 340 26 11 0 1,296 62
Italy Food 14 610 7 4 8,015 9,221
Italy Feed 1,778 9,474 353 20 1,676 1,041
Italy Other uses 320 85 30 1 2,157 658
Latvia Food 14 0 11 42 340 170
Latvia Feed 222 21 74 47 22 160
Latvia Other uses 65 1 14 27 78 47
Lithuania Food 33 12 7 126 625 413
Lithuania Feed 773 30 78 157 30 481
Lithuania Other uses 178 4 19 79 54 152
Luxembourg Food 50 107 1 14 2,737 1,353
Luxembourg Feed 538 842 80 6 413 1,403
Luxembourg Other uses 293 125 1 8 2,282 474
Malta Food 1 4 0 0 0 60
Malta Feed 48 70 1 0 0 5
Malta Other uses 3 1 0 0 0 1
Netherlands Food 11 71 28 75 3,258 1,003
Netherlands Feed 585 1,169 36 74 2,954 1,920
Netherlands Other uses 394 411 1 3 2,136 989
Poland Food 261 62 47 1,265 8,355 4,500
Poland Feed 2,847 1,214 4,635 2,995 417 4,404
Poland Other uses 895 76 802 1,066 1,261 1,322
Portugal Food 37 105 25 49 2,211 1,130
Portugal Feed 173 1,681 85 1 407 554
Portugal Other uses 154 298 8 9 504 91
Romania Food 2 870 9 18 3,474 3,795
Romania Feed 391 6,776 334 9 166 910
Romania Other uses 679 767 53 4 493 911
Slovakia Food 6 14 5 58 942 600
Slovakia Feed 318 390 27 10 51 701
Slovakia Other uses 179 158 6 14 197 160
Slovenia Food 15 52 1 10 545 224
Slovenia Feed 90 475 8 5 29 70

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 217

 Barley Corn Oats Rye Sugar beet Wheat
Slovenia Other uses 43 47 1 1 139 16
Spain Food 457 147 28 29 7,194 4,080
Spain Feed 8,010 5,888 695 264 1,949 4,375
Spain Other uses 1,313 1,041 108 20 2,458 514
Sweden Food 156 20 57 120 1,953 673
Sweden Feed 1,149 6 694 31 617 837
Sweden Other uses 271 51 104 13 221 353
UK Food 114 1,153 221 50 11,516 5,976
UK Feed 3,484 313 222 8 3,179 6,623
UK Other uses 1,888 94 27 1 2,353 1,064
Rest of World Food 19,774 117,799 N/A* N/A* N/A* 376,142
Rest of World Feed 61,249 371,358 N/A* N/A* N/A* 72,674
Rest of World Other uses 4,165 544,170 N/A* N/A* N/A* 25,749

Source: Consolidated EUROSTAT/CAPRI database for EU countries; FAOSTAT for Rest of the World.
*N/A refers to sectors that are only represented in the EU for this version of GLOBIOM.

IV.7.2 Demand for biodiesel feedstocks (1,000 t, average 1999-2001)

Rape- Sun- Soy- Rape- Sunflo Soy-

Rape- Sun- Soy-
seed flower bean seed wer bean
seed flower bean
oil oil oil cake cake cake
Austria Food 0 2 7 9 26 14 0 0 0
Austria Feed 2 13 10 4 0 1 83 54 492
Austria Processing 175 105 10 0 0 0 0 0 0
Austria Other uses 9 1 4 32 27 15 1 1 9
Belgium Food 0 0 0 33 39 61 0 0 0
Belgium Feed 0 0 1 43 3 15 167 88 1,328
Belgium Processing 565 126 1,272 0 0 0 0 0 0
Belgium Other uses 0 0 0 59 30 121 0 0 0
Bulgaria Food 0 27 1 5 137 8 0 0 0
Bulgaria Feed 0 0 2 0 2 3 9 174 62
Bulgaria Processing 11 368 7 0 0 0 0 0 0
Bulgaria Other uses 0 22 1 2 12 9 0 0 0
Croatia Food 0 1 0 8 22 7 0 0 0
Croatia Feed 0 3 87 0 0 0 9 26 95
Croatia Processing 14 46 37 0 0 0 0 0 0
Croatia Other uses 1 3 13 1 1 0 0 0 1
Cyprus Food 0 0 0 1 5 4 0 0 0
Cyprus Feed 0 0 1 0 0 0 0 10 96
Cyprus Processing 0 0 4 0 0 0 0 0 0
Cyprus Other uses 0 0 0 2 0 1 0 0 1
CzechRep Food 0 0 2 92 20 33 0 0 2
CzechRep Feed 10 0 1 12 1 3 189 9 467
CzechRep Processing 583 33 16 0 0 0 0 0 0
CzechRep Other uses 10 1 0 114 2 6 0 0 5
Denmark Food 0 5 0 102 2 50 0 0 0
Denmark Feed 42 6 21 0 0 0 421 276 1,549
Denmark Processing 304 70 70 0 0 0 0 0 0
Denmark Other uses 18 0 0 0 0 0 0 0 0
Estonia Food 0 0 0 5 1 5 0 0 0
Estonia Feed 1 0 0 0 0 0 16 9 22
Estonia Processing 41 1 0 0 0 0 0 0 0
Estonia Other uses 1 0 0 11 0 0 0 1 1
Finland Food 0 0 1 12 3 13 0 0 1
Finland Feed 107 7 15 1 0 1 82 2 193

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 218

 Rape- Sun- Soy- Rape- Sunflo Soy-
Rape- Sun- Soy-
seed flower bean seed wer bean
seed flower bean
oil oil oil cake cake cake
Finland Processing 67 5 130 0 0 0 0 0 0
Finland Other uses 2 0 1 12 0 3 0 0 2
France Food 0 0 2 144 355 23 0 0 0
France Feed 281 129 399 62 33 47 1,001 896 4,668
France Processing 1,312 1,272 523 0 0 0 0 0 0
France Other uses 85 50 29 322 72 18 0 0 0
Germany Food 0 22 45 540 197 244 0 0 0
Germany Feed 42 44 0 15 1 39 1,566 257 3,960
Germany Processing 4,331 295 3,974 0 0 0 0 0 0
Germany Other uses 114 2 0 439 24 24 2 0 11
Greece Food 0 6 1 0 57 40 0 0 0
Greece Feed 0 0 0 0 0 0 0 39 424
Greece Processing 0 68 333 0 0 0 0 0 0
Greece Other uses 0 0 0 0 0 0 0 0 1
Hungary Food 0 9 5 1 87 3 0 0 0
Hungary Feed 1 17 27 1 0 1 18 297 743
Hungary Processing 90 362 41 0 0 0 0 0 0
Hungary Other uses 5 21 5 31 6 22 0 4 7
Ireland Food 0 0 0 16 11 21 0 0 0
Ireland Feed 0 0 4 0 0 0 122 146 336
Ireland Processing 10 1 32 0 0 0 0 0 0
Ireland Other uses 0 0 0 0 0 0 0 0 0
Italy Food 0 0 0 62 214 236 0 0 0
Italy Feed 8 24 174 3 6 20 97 560 3,592
Italy Processing 46 559 1,567 0 0 0 0 0 0
Italy Other uses 2 9 53 16 33 7 0 0 0
Latvia Food 0 0 0 8 3 8 0 0 2
Latvia Feed 0 0 2 1 0 2 6 6 19
Latvia Processing 8 3 2 0 0 0 0 0 0
Latvia Other uses 1 0 0 8 0 2 0 0 1
Lithuania Food 0 0 0 12 4 12 0 0 0
Lithuania Feed 4 0 0 6 1 5 2 12 63
Lithuania Processing 5 3 0 0 0 0 0 0 0
Lithuania Other uses 7 0 0 6 0 1 0 0 0
Luxembourg Food 0 0 0 33 39 61 0 0 0
Luxembourg Feed 0 0 1 43 3 15 167 88 1,328
Luxembourg Processing 565 126 1,272 0 0 0 0 0 0
Luxembourg Other uses 0 0 0 59 30 121 0 0 0
Malta Food 0 0 0 2 1 2 0 0 0
Malta Feed 1 1 1 0 0 0 0 2 25
Malta Processing 0 0 3 0 0 0 0 0 0
Netherlands Food 0 3 3 55 29 82 0 0 5
Netherlands Feed 7 2 171 6 10 53 579 596 2187
Netherlands Processing 152 572 4,092 0 0 0 0 0 0
Netherlands Other uses 7 0 18 101 32 68 0 6 510
Poland Food 0 17 1 169 15 17 0 0 0
Poland Feed 0 2 0 9 4 31 315 66 1,137
Poland Processing 804 1 11 0 0 0 0 0 0
Poland Other uses 70 1 0 158 15 71 0 0 1
Portugal Food 0 0 0 0 128 13 0 0 0
Portugal Feed 0 0 128 0 0 11 1 184 952
Portugal Processing 1 276 687 0 0 0 0 0 0
Portugal Other uses 0 5 8 0 17 16 0 2 12
Romania Food 0 19 0 2 266 3 0 0 0

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 219

 Rape- Sun- Soy- Rape- Sunflo Soy-
Rape- Sun- Soy-
seed flower bean seed wer bean
seed flower bean
oil oil oil cake cake cake
Romania Feed 2 3 16 1 0 5 0 353 122
Romania Processing 17 731 102 0 0 0 0 0 0
Romania Other uses 3 34 21 4 37 8 0 0 1
Slovakia Food 0 1 0 31 14 3 0 0 1
Slovakia Feed 0 1 1 4 0 0 46 8 182
Slovakia Processing 145 42 2 0 0 0 0 0 0
Slovakia Other uses 3 2 1 36 1 1 0 0 2
Slovenia Food 0 0 1 7 9 11 0 0 1
Slovenia Feed 0 0 0 3 0 1 0 20 120
Slovenia Processing 33 0 0 0 0 0 0 0 0
Slovenia Other uses 0 0 0 8 1 3 0 0 3
Spain Food 0 39 1 18 450 263 0 0 220
Spain Feed 4 118 338 0 0 0 84 667 4,175
Spain Processing 40 1,230 2,684 0 0 0 0 0 0
Spain Other uses 1 26 1 4 36 18 0 9 0
Sweden Food 0 0 1 66 8 14 0 0 0
Sweden Feed 41 0 0 7 0 3 201 20 331
Sweden Processing 236 13 8 0 0 0 0 0 0
Sweden Other uses 6 0 0 65 1 4 0 0 0
UK Food 0 22 3 692 137 145 0 0 0
UK Feed 188 1 127 0 0 0 824 458 2,034
UK Processing 1,389 9 823 0 0 0 0 0 0
UK Other uses 28 1 0 0 0 0 0 0 0
Rest of the 17,38
Food 619 381 9,290 6,264 5,461 0 0 0
World 8
Rest of the 82,54
Feed 1,990 1,702 9,835 0 0 0 13,778 5,314
World 4
Rest of the 125,35
Processing 22,716 15,099 0 0 0 0 0 0
World 6
Rest of the
Other uses 1,171 416 5,279 2,825 704 5,535 0 0 0
World
Source: Consolidated EUROSTAT/CAPRI database for EU countries; FAOSTAT for Rest of the World.

IV.8 Demand elasticitities

In GLOBIOM, demand for food react to prices and the response magnitude is determined by the
values of the demand elasticities. For instance, an elasticity of -0.1 means that for a 10% increase in
price, the quantity of consumption will change by 10% x -0.1 = -1%.
Crop product elasticities are based on data provided by USDA that estimated demand elasticities per
categories of product (e.g. cereals, sugar, vegetable oil) at the consumer level. These elasticities are
applied in GLOBIOM to the demand in each crop providing the product. Because USDA data do not go
at the level of detail of each product separately, we assume that all sub-product within one category
have the same values140 (e.g. barley has identical values to corn), which can mask some potentially
more heterogeneous response in some particular cases. For two countries of particular importance for
the future of food demand, China and India, we relied on nationally estimated data to obtain more
precise estimates and describe better the differences between some sub-products.

140
For some regional aggregates, values can still differ because weights for the aggregation (consumption in each country) can vary from
one product to another.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 220

Table 56: Demand elasticities

Barley, Corn, Palm oil, Rapeseed

Whear, Oats, Rye , oil, Sunfower oil, Potato, Peas Sugar
Rice Soybean oil

Austria -0.05 -0.05 -0.21 -0.27

Belgium -0.05 -0.08 -0.23 -0.28

Bulgaria -0.20 -0.23 -0.35 -0.53

Croatia -0.16 -0.19 -0.32 -0.49

Cyprus -0.05 -0.18 -0.22 -0.33

CzechRep -0.10 -0.14 -0.29 -0.41

Denmark -0.07 -0.11 -0.25 -0.21

Estonia -0.12 -0.17 -0.31 -0.49

Finland -0.05 -0.10 -0.25 -0.34

France -0.05 -0.07 -0.23 -0.28

Germany -0.05 -0.08 -0.23 -0.26

Greece -0.19 -0.22 -0.31 -0.39

Hungary -0.14 -0.18 -0.31 -0.46

Ireland -0.05 -0.09 -0.24 -0.37

Italy -0.05 -0.06 -0.24 -0.27

Latvia -0.16 -0.19 -0.33 -0.54

Lithuania -0.09 -0.14 -0.31 -0.52

Luxembourg -0.05 -0.05 -0.16 -0.11

Malta -0.05 -0.08 -0.26 -0.33

Netherlands -0.05 -0.08 -0.23 -0.30

Poland -0.15 -0.19 -0.32 -0.50

Portugal -0.05 -0.08 -0.27 -0.38

Romania -0.21 -0.24 -0.36 -0.52

Slovakia -0.14 -0.18 -0.31 -0.48

Slovenia -0.09 -0.13 -0.28 -0.42

Spain -0.05 -0.05 -0.22 -0.38

Sweden -0.08 -0.12 -0.25 -0.31

UK -0.05 -0.05 -0.21 -0.28

Source: (Muhammad, et al. 2011)

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 221

 Sugar cane

Sunflower

Vegetable
Soybean
Potato

Wheat
Barley
Rest of the world

Corn

Rice

oil
Australia -0.05 -0.05 -0.24 -0.05 -0.08 -0.27 -0.09 -0.05 -0.08
Argentina -0.18 -0.18 -0.34 -0.18 -0.22 -0.44 -0.22 -0.18 -0.22

Brazil -0.27 -0.27 -0.38 -0.27 -0.29 -0.53 -0.29 -0.27 -0.29

Canada -0.06 -0.06 -0.23 -0.06 -0.1 -0.24 -0.1 -0.06 -0.1
a
China -0.39 -0.48 -0.45 -0.35 -0.4 -0.63 -0.3 -0.4
CongoBasin -0.39 -0.45 -0.49 -0.44 -0.46 -0.61 -0.43 -0.45

Former_USSR -0.21 -0.27 -0.35 -0.21 -0.22 -0.55 -0.23 -0.22 -0.23
b
India -0.39 -0.19 -0.46 -0.25 -0.4 -0.63 -0.4 -0.34 -0.4

Indonesia -0.32 -0.32 -0.42 -0.32 -0.34 -0.59 -0.32 -0.34

Japan -0.06 -0.06 -0.24 -0.06 -0.1 -0.25 -0.1 -0.06 -0.1

Malaysia -0.28 -0.28 -0.39 -0.28 -0.3 -0.55 -0.28 -0.3

Mexico -0.14 -0.14 -0.32 -0.14 -0.18 -0.51 -0.18 -0.14 -0.18

MidEastNorthAfr -0.29 -0.27 -0.38 -0.27 -0.31 -0.54 -0.29 -0.28 -0.29
Central America -0.33 -0.33 -0.44 -0.33 -0.35 -0.54 -0.35 -0.33 -0.32

Rest of Latin America -0.26 -0.27 -0.39 -0.28 -0.31 -0.56 -0.28 -0.27 -0.35
Rest of South Asia -0.37 -0.37 -0.45 -0.38 -0.37 -0.63 -0.38 -0.36 -0.26
Southeast Asia -0.35 -0.35 -0.43 -0.35 -0.36 -0.60 -0.33 -0.35 -0.06

SouthAfrReg -0.27 -0.27 -0.38 -0.27 -0.29 -0.66 -0.29 -0.27 -0.3
SouthKorea -0.19 -0.19 -0.31 -0.19 -0.21 -0.39 -0.21 -0.19 -0.38

Eastern Africa -0.45 -0.43 -0.48 -0.42 -0.44 -0.67 -0.43 -0.44 -0.32

Southern Africa -0.41 -0.46 -0.5 -0.42 -0.46 -0.64 -0.47 -0.42 -0.4
Western Africa -0.41 -0.41 -0.46 -0.41 -0.41 -0.66 -0.41 -0.39 -0.29
Turkey -0.25 -0.25 -0.37 -0.25 -0.27 -0.52 -0.27 -0.25 -0.21
Source: (Muhammad, et al. 2011) except for :
a
Zhuang R., Abbott P., 2005. Price Elasticities of Key Agricultural Commodities in China. Paper presented at the AAEA Annual

Meeting.
b
Kumar, P., Kumar, A., Parappurathu, S., Raju, S.S. 2011. Estimation of Demand Elasticity for Food Commodities in India.

Agricultural Economics Research Review 24.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 222

Annex V Sensitivity and uncertainty analyses

V.1 Most important uncertainties in LUC modelling

A sensitivity analysis on the model was carried out through a set of Monte Carlo runs. This means
that the calculation is carried out repeatedly with randomly varied parameters. In this case, about
250 runs have been performed for each of the feedstock specific scenarios A, A1, A2 and the NREAP
scenario. For this analysis, 11 parameters were varied along the specifications reported in Table 55
below.

The first set of elasticities to be varied is related to the model behavioral responses. Elasticities were
varied for demand response, trade response, management response (irrigation), vegetable oil
substitution and impact of the biofuel policy on the feedstock yield. These elasticities determine how
much LUC occur and in what regions.
A second set of elasticities concerns biophysical parameters. Co-product protein content is the first
important one, as it determines the degree of substitution of co-products with other oilseed meals.
Additional testing was applied on the impact of removing yield residues on yield and soil organic
carbon. Last, the emission factors for peat land as well as the share of (palm oil) plantation
expanding into peat land were varied for Indonesia and Malaysia.
In the Monte Carlo analysis, the chosen parameters are randomly varied, but still this involves a pre-
defined distribution shape141. Some parameters are varied between -50% and +100%. For
parameters that are known with more accuracy, the range and shape of variation is pre-set in line
with this project’s Data document and Improvement Document. Parameters can be changed by a
same amount when the uncertainty is not region or product specific (correlated parameters – see last
column of the table). When the uncertainty is specific to each region or product – the variation in
parameters is different for each region and/or product.

Two important settings for the Monte-Carlo analysis are also reported in Table 55 below. The first one
is related to the parameter distribution shape. Because elasticities are parameters calculated as log
response (percentage change of quantity compared to percentage change of price), we varied them
along a loguniform distribution (a distribution where log of the parameter is uniformly distributed).
We then consider that the response can be, for instance for trade elasticity (-50% to 100%), twice
stronger or twice smaller. When parameters are known with relatively more accuracy (e.g. demand
elasticities), the range of values considered was narrower. Because the central value of the
distribution is not necessarily more plausible than another point in the distribution, we also preferred
a loguniform distribution to a lognormal one. Biophysical parameters were associated different
distribution shapes, either uniform when no better information was known on the distribution, or for
peatland related factors, along the distributions determined in the improvement document.

141
Most values are varied along a loguniform distribution, because the central value is not necessarily more plausible than other points in
the distribution. Biophysical parameters were varied along different distribution shapes, either uniform when no better information was
known, or along the distribution determined in the Improvement document.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 223

Table 57: Parameter variation used for the Monte Carlo analysis

Value range Correlation of

Distribution
Parameter parameter between
Minimum Maximum assumption
products/regions
Behavioral
parameters
Correlated across regions
Demand elasticity - 33% +50% loguniform
and products
Not correlated across
Trade elasticity -50% +100% loguniform
regions and products
Water supply Not correlated across
-50% +100% loguniform
elasticity regions
Vegetable oil Not correlated across
-50% +100% loguniform
substitution elasticity regions
Land expansion Not correlated across
-50% +100% loguniform
elasticity regions and land use types
Yield response Elasticity model Same assumption for all
Elasticity model uniform
feedstock + 0.2 regions
Biophysical
parameters
Co-product protein Correlated across regions
-10% +10% uniform
content and products
Soil carbon impact Same assumption for all
-10% 0% uniform
straw EU regions
Same assumption for all
EU regions
Yield impact straw -4% 0% uniform
Correlated with SOC
impact
-1
Peat land emissions 113 tCO2 ha Same assumptions for
27 tCO2 ha-1 yr-1 lognormal
factor yr-1 Indonesia and Malaysia
Palm expansion into Same assumptions for
12% 54% lognormal
peat land Indonesia and Malaysia

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 224

How to read results in this section?
Each graph of this section presents the detailed results of the sensitivity analysis performed for the
feedstocks scenarios. The total distribution of results (right-hand side of the figure in grey) is
decomposed across the different sources of GHG emissions accounted for in the study. These
categories are the same as the ones used in the results section for the calculation of cumulated
emissions and they follow the same color codes.

For each emission source, a box-and-whisker plot is used to represent the distribution of results. The
colored box indicates the 25-75% range of central values, with the thick bar marking the median
value. The limit of the whiskers indicate the 5%-95% distribution limits. Single points represent the
outliers outside if this range.

The histogram on the right-hand side of the figures replicates the distribution shown in the results
section. The length of the bars is proportional to the number of runs which lead to a value in the y-
axis. The distribution corresponds to the values of the last box-and-whisker plot.

Number of runs for the sensitivity analysis

Limit of 5-95%
uncertainty
range

Median value
25-75% uncertainty
range

Outliers not in the 5-

95% uncertainty range

Natural Soil Peat land Sum of

Reverted Biomass
vegetation organic drainage contribution
vegetation in
biomass
agricul- carbon
ture

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 225

V.2 Detailed results per scenario

V.2.1 Maize

Uncertainty on emissions associated to maize come mostly from natural vegetation emissions due to
the combination of reduction of deforestation in Latin America due to substitution of soybean meal
and DDGS, and to increase of palm oil in Southeast Asia to replace soybean oil.

Negative ILUC is observed for a certain number of cases, when a decrease of cropland expansion in
Latin America, is not counterbalanced by emissions in Southeast Asia (low substitution of vegetable
oil). However, these cases concern only values in the first quartile of the distribution.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 226

V.2.2 Wheat

Uncertainty on emissions associated to wheat come mostly from natural vegetation emissions due to
the combination of reduction of deforestation in Latin America due to substitution of soybean meal
and rapeseed meal, and to increase of palm oil in Southeast Asia to replace soybean oil. Effects is
slightly higher in the case of wheat than for maize because of the slightly higher protein content
assumed for wheat DDGS in comparison to maize DDGS.

Negative LUC is observed for a significant number of cases, corresponding to decrease of cropland
expansion in Latin America. This effect is however often counterbalanced by emissions in Southeast
Asia (substitution of vegetable oil). As a consequence, about one third of the distribution correspond
to negative values.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 227

V.2.3 Barley

In the case of barley, similar effects are observed as for wheat and maize. Due to the slightly lower
yield of barley compared to wheat, emissions are overall slightly higher and the number of negative
emission case is reduced.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 228

V.2.4 Sugar beet

For sugar beet, the main source of uncertainty comes from natural vegetation emissions but these
remain relatively limited compared to those of other feedstocks. As a consequence, the distribution is
more skewed with a range of values between 0 and 50 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 229

V.2.5 Sugar cane

Concerning sugar cane, uncertainty is considerable as soon as behavioral parameters are varied.
Indeed, depending on the response of land use, natural vegetation emissions can reach high values.
Most sugar cane is located in the South of Brazil in the region of Sao Paulo where agricultural land is
well developed and far from the Amazon and Cerrado. However, some other sectors, in particular
cattle, are present both in the South and on the agricultural expansion frontier, which can generate
some leakage. The possibility of land displacement from the South to the agricultural expansion
frontier in Brazil leads to high upper-tail of emissions up to 200 gCO2-eq. At the same time, if no
natural vegetation emission occur, the sequestration effect of sugar cane plantations through
agricultural biomass and soil organic carbon can lead to some negative emissions.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 230

V.2.6 Rapeseed

Uncertainty in emissions for rapeseed come mainly from two sources: i) conversion of natural
vegetation, directly dependent on the degree of leakage to palm oil on the vegetable market, and on
the substitution effect between rapeseed meal and soybean meal; ii) degree of peatland emissions
due to palm oil. When combining these two sources of uncertainty, the range of results fall with a
symmetrical distribution with most values between 0 and 100 gCO2.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 231

V.2.7 Sunflower

Like rapeseed, sunflower uncertainty is particularly strong for natural vegetation emissions and
peatland emissions. Due a lower vegetable oil yield, the distribution is shifted up compared to
rapeseed emissions, with a larger leakage to palm oil per unit of energy.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 232

V.2.8 Soybean

Soybean land use emissions are particularly uncertain on the natural vegetation conversion side. This
is due to the various responses that land use can have in Latin America (e.g. from an expansion on
the deforestation frontier to a large response through multi-cropping), but also to the low yield in
vegetable oil of soybeans. Large uncertainty through palm oil leakage and peatland emissions also
participate to a high dispersion of the results. Overall, the range of emissions lead to the central part
of the distribution (second and third quartile) in the 50-150 gCO2-eq range, with a high upper-tail of
emissions up to almost 400 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 233

V.2.9 Palm oil

Palm oil emission uncertainty is for a large part associated to peatland emissions, whereas natural
vegetation emission uncertainty is found of relatively lower magnitude. This can be explained by the
relatively high yield of palm plantations and the fact that not all plantation expansion necessarily lead
to deforestation. The final distribution of palm oil emissions therefore directly compares to the one
found for peatland emissions in Indonesia and Malaysia. The range is relatively large, from values
close to zero to around 500 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 234

V.2.10 Perennials

Uncertainty in the case of perennials appears mainly associated to natural vegetation emissions;
however this source only has a secondary impact on the final distribution of the results, more
impacted by the effect of agricultural biomass and soil organic carbon. Overall results are found to
follow distribution with mostly negative values, with minimum around -30gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 235

V.2.11 Short rotation plantations

Uncertainty for short rotation plantation is associated mainly to natural vegetation emissions, driven
by different patterns of LUC. These however play a secondary role compared to sequestration of
carbon in the plantation biomass. Overall, the distribution of results remain negative, with a limited
range of results from -40 to -10 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 236

V.2.12 Cereal straw

Yield impact also leads to some reallocation of land use with uneven implications for natural
vegetation emissions. The largest part of emissions is in the range 0-20 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 237

V.2.13 Starch group

For starchy crops, the largest uncertainty comes from land conversion emissions. Impact of other
source participates to the overall level of emission levels, but with lower dispersion. Overall, the
second and third quartile of emissions are located in the range 0-30 gCO2-eq, whereas some negative
values can appear for the first quartile of values.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 238

V.2.14 Sugar group

In the case of sugar crops, the results are significantly shaped by the little dispersion of sugar beet
emissions, with most of the distribution in the 15-35 gCO2-eq range. However, due to uncertainty on
land use conversion emissions from sugar cane, a upper-tail is found with some possible values close
to 100 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 239

V.2.15 Vegetable oil group

Vegetable oil emissions are strongly influenced by the high values found for palm oil and soybean oil.
In particular, large uncertainties appear that are related to land use conversion emissions, but also
with significant magnitude to peatland emissions. The overall dispersion of the vegetable oil group is
therefore relatively high, with most of the distribution higher than 50 gCO2-eq and some values
higher than 200 gCO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 240

V.2.16 EU 2020 biofuel mix scenario

For the EU 2020 biofuel mix scenario, sources of uncertainty come mainly from natural vegetation
biomass and peatland emissions, in particular due to the contribution of vegetable oil for biodiesel.
The final distribution of effects is quite large, with values ranging from 20 to 150g CO2-eq.

Ecofys-IIASA-E4tech - The LUC impact of biofuels consumed in the EU 241

ECOFYS Netherlands B.V. | Kanaalweg 15G | 3526 KL Utrecht| T +31 (0)30 662-3300 | F +31 (0)30 662-3301 | E [email protected] | I www.ecofys.com
ECOFYS Netherlands B.V.

Kanaalweg 15G
3526 KL Utrecht

T: +31 (0) 30 662-3300

F: +31 (0) 30 662-3301

E: [email protected]
I: www.ecofys.com