Journal of Hydrology 267 (2002) 80–93

www.elsevier.com/locate/jhydrol

Land-use impacts on storm-runoff generation: scenarios of land-use

change and simulation of hydrological response in a meso-scale
catchment in SW-Germany
Daniel Niehoff a, Uta Fritscha, Axel Bronstertb,*
a
Potsdam Institute for Climate Impact Research, 14412 Potsdam, Germany
b
University of Potsdam, Institute for Geoecology, 14415 Potsdam, Germany

Abstract
The impact of land-use change impacts on storm-runoff generation is presented based on a simulation study composed of
three parts: (1) generation of spatially explicit land-use scenarios; (2) spatially distributed and process based hydrological
modelling of runoff generation; (3) application of this procedure and demonstration of results for a predominantly agricultural
meso-scale catchment in a loessy-soil landscape in SW Germany.
Land-use scenarios are a prerequisite for assessing the influence of potential changes of land-use and/or land-cover on runoff
generation. The land-use change modelling kit (LUCK) provides a method for the spatial transformation of overall trends of
land-use into spatially distributed scenarios of land-use patterns, taking into account their topology in a true position mode. The
assignment of land-use categories to each grid cell is realised in a spatially explicit manner, dependent on an evaluation of the
site characteristics as well as its neighbourhood relationships.
Based on these land-use scenarios, the influence of altered land-use characteristics on flooding is simulated using a modified
version of the physically based hydrological model WaSiM-ETH. In order to extend the model’s capabilities with respect to
adequate representation of land-use related runoff generation mechanisms, some additional mechanisms have been introduced:
(1) A macropore module accounting for fast infiltration processes; (2) A siltation module decreasing hydraulic conductivity of
the soil surface as a function of precipitation intensity and vegetation coverage. (3) Sub-grid variability considering the
impervious and sealed portion of a grid cell. The location and lateral interaction of landscape elements within a catchment is
captured by spatially explicit modelling on the basis of gridded information provided by the scenarios.
The whole simulation procedure is applied to a meso-scale catchment in SW Germany. The results show that the influence of
land-use conditions on storm-runoff generation depends greatly on the rainfall event characteristics and on the related spatial
scale, i.e. the influence is only relevant for convective storm events with high precipitation intensities in contrast to long-lasting
advective storm events with low precipitation intensities. However, convective events—and thus land-use conditions—are of
very minor relevance for the formation of floods in large river basins because this type of rainfall event is usually restricted to
small-scale occurrence. q 2002 Elsevier Science B.V. All rights reserved.
Keywords: Land-use change; Scenario; Storm-runoff; Runoff generation; Distributed hydrological model; Flooding

1. Motivation

Both the landscape and the river systems in large

* Corresponding author. Fax: þ 49-331-977-2092. parts of Europe and elsewhere have undergone major
E-mail address: [email protected] (Axel Bronstert). changes in the past, and there is no doubt that these
0022-1694/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 1 6 9 4 ( 0 2 ) 0 0 1 4 2 - 7
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 81

environmental changes have altered the flooding differ markedly. Convective events have a small
conditions in this region. But due to the complexity spatial extension (not more then a few tens of km2), a
of the processes involved, the magnitude of their limited duration (usually not more then a few hours)
impact on storm-runoff generation and subsequent but high rainfall intensities. In contrast, advective
flood discharge into the river system is still highly rainfall events are originated by frontal atmospheric
uncertain. systems, which results in large regions being affected
The work presented in this paper focuses on three by the rain over a comparatively long period of time,
main questions, always accounting for the boundary but yielding much smaller rainfall intensities.
conditions of land-use dynamics for which statements Additionally, convective rainfall usually appears as
are made and for the spatial and temporal scale of the a thunderstorm with dry antecedent catchment
hydrological and meteorological processes involved: conditions, while advective rainfall mainly occurs
during late autumn, winter, and early spring with
1. Which kind of land-use changes have been usually wet antecedent catchment conditions, some-
observed in the past, which kind of land-use times connected with snow melt. These very different
changes can be expected in the future and what will conditions concerning rainfall intensity and antece-
be their probable spatial distribution in the land- dent catchment wetness lead to significant differences
scape? in runoff generation processes and the relevance of
2. Which runoff generation mechanisms (e.g. infiltra- land-cover and soil-surface conditions, i.e. only if
tion excess overland flow, saturation excess, rainfall intensities are high (e.g. convective rainfall)
subsurface stormflow, quick groundwater outflow) and the surface has a rather small conductivity (e.g.
are likely to be affected by land-use and land-cover dry soil and/or crusted or sealed surface) runoff
changes and how can they be represented by a generation may be considerably controlled by the
hydrological model? land-cover or soil-surface conditions.
3. How does the influence of land-use and land-cover The hydrological response of the above-mentioned
changes on storm-runoff generation depend on catchments to convective and advective rainstorms is
rainfall event characteristics and temporal scale being simulated for present land-use as well as for
and what is the related significance of catchment various possible future land-use conditions, involving
characteristics and spatial scale? both process-oriented hydrological modelling and the
delineation of land-use scenarios. This article contains
The investigation does not address the influences exemplary results for only one of the three catch-
of river training conditions and retention along the ments. The results of the whole study have been
river courses on flood-wave propagation. documented by Bronstert et al. (2001).
Prior to this study, no satisfactory method was
available for the generation of spatially distributed
2. Study set-up land-use scenarios. Therefore this task is performed
using a new approach, the land-use change modelling
This article is based on a comprehensive study kit (LUCK), which is described in Section 3.2.
investigating the impacts of land-surface conditions In order to quantify the influences of land-use on
on storm-runoff generation at the lower meso-scale. storm-runoff generation, the deterministic and dis-
For this purpose, three different catchments (100 – tributed hydrological model WaSiM-ETH (Schulla,
500 km2) within the Rhine basin have been selected 1997) has been chosen, which originally focused on
which represent different characteristic land-use climate change impacts on the catchment water
patterns with either predominantly urban, agricultural balance. In order to improve the representation of
or forestal structure. land-use characteristics with respect to runoff gener-
The study includes the investigation of floods ation, the soil module has been extended, now
originated by both convective and advective rainfall explicitly allowing for macropore infiltration, siltation
events. The space-time scales and the associated and crusting of the soil surface, connection of sealed
rainfall intensities of these different rainfall types surfaces to the sewer system and decentralised water
82 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

retention in the landscape. The underlying mech- sented here deals with the hydrological cycle,
anisms of these extensions are outlined in Section focusing on the impacts on storm-runoff generation.
4. A more detailed description can be found in Designing scenarios is a widespread technique in
Katzenmaier et al. (2001). The combination of a business, planning and policy consulting, because it
land-use scenario and well adapted modelling of offers an opportunity for assessment of the present and
storm-runoff mechanisms is applied for a predo- possible future situation and the mitigation of
minantly agricultural meso-scale catchment in a mismanagement. Due to its future-oriented and
loessy-soil landscape in SW Germany. The results are flexible character, this method is also used for
presented in Section 5. environmental studies. The scenario is to be under-
stood as a projection rather than a prediction. The
automated generation of land-use/land-cover scen-
3. Land-use and land-cover scenarios arios and modelling of landscape development is a
rather new branch in environmental studies (Wenkel,
1999). As well as the geophysical aspect, the
3.1. Background anthropogenic aspects also need to be considered.
Designing such scenarios consists of two tasks: the
Land-use refers to human activities that are determination of land-use trends and the spatial
directly related to land, making use of its resources transformation of these trends into spatially distrib-
and interfering in the ecological processes that uted land-use patterns. Impact studies of environmen-
determine the functioning of land-cover (Veldkamp tal changes often neglect neighbourhood relationships
and Fresco, 1996). It is controlled by the potential and the position of land-use alterations, by altering
value of the land for agricultural, forest, urban, or only the proportion of different land-use types. This
nature protection use and is governed by multilevel study utilises trends in land-use change derived from
economic and socio-cultural interactions. In Europe external analysis and presents a new technique for the
land-use is categorised into the fields: urban areas, spatially explicit allocation of land-use change in the
agriculture, and forestry. Land-cover refers to the landscape.
surface appearance of the landscape, which is mainly
affected by its use, its cultivation and the seasonal 3.2. Land-use change modelling kit (LUCK)
phenology. Changes of land-use are caused by
modified biophysical or human demands that arise The developed approach for scenario generation
from changed natural, economic or political con- (LUCK) works in a grid-based discretization-mode of
ditions (O’Callaghan, 1996). The consequences are the catchment (Fritsch et al., 2001) representing the
either modification or conversion: modification spatial distribution of land-use types in the landscape.
implies a change of condition within a type, caused The spatially averaged, large-scale trends of land-use
by different cultivation techniques or management development have to be provided by external analysis.
strategies; conversions include a transition from one The potential conversion of the land-use types of each
land-use type to another. grid is based on an evaluation of the characteristics of
Land-use/land-cover patterns are highly dynamic each grid as well as on its neighbourhood relation-
and rarely in a stable equilibrium. The strongest ships. Typical land-use patterns and the influence of
seasonal variation of land-cover occurs predomi- so-called axes of infrastructural development, e.g.
nantly for arable land, but is an important factor for roads or railways, are included in this analysis. A
other land-use types, too. Land-use changes may have combination of these criteria forms the potential of
regional or even global effects as a result of the each grid to become subject to changes. Since land-
accumulation of many small-scale, local changes. use changes happen successively, the procedure tries
Alterations in land-use exert an influence on the to imitate these dynamics by an iterative procedure.
ecosystem as a whole, because they affect the water Land-use changes for the main land-use categories
cycle, biodiversity, radiation budgets and many other are considered within LUCK by three different
processes (Riebsame et al., 1994). The study pre- modules for urbanisation, agricultural, and forestal
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 83

land-use. As the conversion of one land-use type to use type ‘settlement’ in the immediate vicinity of
another follows strictly economic-hierarchical prin- each grid is determined.
ciples, urbanisation is ranked highest, followed by
agriculture. This ranking enables the generation of Land-use conversion develops gradually, which is
spatially distributed land-use change scenarios driven simulated by LUCK using an iterative procedure. In
by multiple overall trends. In its present version, order to prescribe the final overall proportion of
LUCK is adjusted to CORINE land-use data and a settlement area, this procedure requires as an external
general soil map of the scale 1: 200 000. Details of the input a so-called ‘scenario target’, which is the
whole concept are provided by Fritsch (2002). spatially averaged trend of land-use development.
Then, the spatial allocation is determined by the
3.2.1. The urbanisation module synthesis of the two criteria of evaluation and
The development of urban areas is given highest neighbourhood relationships, as explained above.
priority, based on the general trends of settlement area The technique for the increase of urban areas is
in Europe in the 20th century. The settlement areas summarised in Fig. 1.
have steadily increased, in particular after the second The iteration procedure is composed of the
world war, and the trend is still persistent. In following steps: firstly, the requirements for grids
Germany, planning regulations operate for settlement that might change from their present land-use into
expansion, but still communal self-administration settlement are set to a maximum. Only those grid cells
undermines the intention of balanced growth (Kleyer, will be converted that are both most suitable and have
1996). Rural and urban settlements have experienced a maximum number of neighbouring cells already
polycentric growth, which leads to a ring-like used as settlement areas. For grid cells with identical
structure of an older centre and newer districts around conditions, the conversion decision is made randomly.
the kernel. Deviations from a perfect circle are caused During the iteration procedure, the number of already
either by limitations of the ecosystem or by infra- converted cells is constantly compared to the number
structure properties which stimulate the growth of set by the scenario target. The two criteria (suitability
built-up areas close to the axes of development, such and neighbourhood) are successively given a less
as main roads or railways. The scenario technique for stringent user-determined threshold which labels grid
urbanisation tries to imitate this evolution by the help cells as unsuitable. The neighbourhood analysis is
of the following criteria: performed after each iteration loop to update the
relationships.
1. Evaluation of site characteristics: each grid cell is
evaluated with regard to its suitability to be 3.2.2. Validation of the urbanisation module
converted from its actual land-use to settlement. The validation of the urbanisation technique has
Only steep slopes (structural constraints) and areas been performed by means of historical land-use data
of nature conservation (legal constraints) are of the Lein catchment (described in Section 5.1). The
excluded from potentially being converted to land-use that is identifiable on a topographical map of
urban land. Axes of infrastructural development the area dating from 1844 is used as the starting
have a positive influence on economic develop- condition for the simulation of settlement increase
ment and promote the growth of settlement areas. from historical conditions up to the present rate of
2. Neighbourhood analysis: neighbourhood relation- settlement. In view of the simplicity of the scenario
ships play a major role for land-use pattern technique, the correlation of more than 50% between
development. Due to planning regulations and the modelled and the present conditions indicates an
economic efficiency, new settlement areas need to adequate result for a simulation of settlement devel-
be connected to existing settlements. A so-called opment with LUCK. Moreover, one can see that the
focal analysis is used to ensure that settlement general pattern of the settlement areas is represented
growth only takes place adjacent to built-up areas. well. Fig. 2 shows the comparison of the historical
By analysing the adjacent neighbours of each grid, situation, the simulated present conditions and the
the amount of existing neighbours with the land- present land-use.
 84 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

Fig. 1. Scheme of the scenario module for urban development.

3.2.3. The module for the set-aside of arable land from settlement land-use. In terms of seasonal
During the last decades the profitability of phenology and agro-management, arable land has
agricultural production has decreased continuously. very dynamic features, which are almost independent
This trend has been engendered by subsidies for crop of neighbourhood relationships to other agricultural
production on one side and closure premiums for fields. Neighbourhood relationships might be relevant
marginal revenue sites on the other side. The EU-wide only in the case of an expanding adjacent settlement
agreement Agenda 2000 prescribes ceasing cultiva- area. However, in order to avoid redundancy with the
tion on 10% of current arable land. The agricultural urbanisation module, the influence of urbanisation is
development module of LUCK rates the possible not considered within this module.
selection for set-aside areas with the aid of an The underlying idea of this tool is the evaluation of
estimation of yield potential. the yield potential of each grid, which is a combined
The category of agricultural land-use differs valuation of soil fertility and its suitability for
completely in its development and characteristics mechanised farming. The procedure has been adjusted

Fig. 2. Comparison of historical data, simulated present conditions and present conditions according to the CORINE data basis (the land-use
class ‘complex plot structures’ refers to sub-scale plots like small gardens).
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 85

to the data basis and derives from the standardised mainly due to the occurrence of macropores (Section
approach of Marks et al. (1992). Spatial information 4.1), reducing effects due to a possible siltation of the
on topography, soil texture and soil type allow an soil (Section 4.2). Details and more comprehensive
estimation of classified values of field capacity, application examples are given by Niehoff (2002).
hydraulic conductivity, porosity, nutrient availability,
root depth, slopes, and altitude. These attributes are 4.1. Macropore flow
the basis for the derivation of the so-called fertility
potential, water potential and topographic potential. Following the proposal of Beven and Germann
Combined with the potential for mechanised farming (1982), it is possible to classify the porosity of natural
(derived from soil depth and slope data), these four soil into micropores and macropores. Although the
potentials represent the features of the arable land that fraction of macropore volume is small compared with
are relevant for the crop yields. Climate is considered the total pore volume, macropores are capable of
to be homogeneous at the meso-scale, so it is not conducting a significant water flux because of their
included in this analysis. The calculated overall yield high transport capacity. For specific conditions, e.g.
potential is standardised to values ranging from one soil surfaces with little matrix hydraulic conductivity
for the lowest and 100 for the highest evaluation and/or very high rainfall intensity, a large part of
value. This procedure leads to a standardised map of infiltration may pass through the macropores into
yield potentials. Finally, the selection of grids to be deeper soil layers, bypassing the matrix close to the
set-aside is based on the evaluation results. If grid soil surface. Approximating these features leads to the
cells are classified with the same yield potential, the assumption of two infiltration components, i.e. the rate
procedure performs a random selection out of those into the soil matrix and the rate into the macropore
cells with the same values. system. Following the model approach by Bronstert
(1999), the net precipitation rate is completely
absorbed by the soil matrix as long as the matrix
4. Extensions of the hydrological model infiltration capacity is not exceeded, and the infiltra-
tion into the macropores is neglected as long as the
WaSiM-ETH has been chosen here because it matrix can absorb the total rainfall intensity. The
includes most of the processes relevant for runoff infiltration component into the macropore system is
generation, it considers the spatial distribution of switched on when the net precipitation intensity
catchment characteristics, and is based on spatial and exceeds the matrix infiltration rate. Thus, infiltration
temporal dynamics of climate variables. However, excess overland flow is generated if the precipitation
some extensions were developed in order to improve intensity exceeds the sum of the two infiltration
the representation of the influence of land-cover and components.
the unsaturated zone on infiltration processes. Infiltra- Based on the approach of Bronstert (1999), a
tion is a crucial process in modelling runoff generation macropore module has been added into WaSiM-ETH,
of a catchment. Particularly under extreme precipi- by means of treating macropores as a single linear
tation conditions it contributes strongly to the water storage which interacts with the soil matrix. Infiltra-
being transferred to the river system, either directly by tion into the macropores occurs as infiltration-excess
generating infiltration excess overland flow or or as saturation-excess of the soil matrix. Exfiltration
indirectly by influencing the extent of saturated out of the macropores into the soil matrix is limited to
surface areas, thus causing saturation excess overland unsaturated conditions and is controlled by the
flow. Various effects of the actual soil surface and saturation deficit of the matrix and the storage
land-cover conditions have to be accounted for to coefficient of the single linear storage. Macropore
cover the whole range of possible infiltration capacity storage capacity is given as the product of macro-
of the catchment. Therefore, the extensions to porosity and the average depth of the macropore layer.
WASIM-ETH developed here focus on soil and As information on the interaction between macro-
land-cover characteristics, which either increase or pores and the soil matrix is sparse, the macropore
reduce the infiltration capacity. Increasing effects are storage coefficient is subject to calibration.
86 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

The influence of macropores on storm-runoff the macropores. Hydraulic conductivity of the soil
generation depends enormously on rainfall character- surface is reduced by simply multiplying it with a
istics and antecedent soil moisture conditions. This is factor Csilt that takes values between zero (imperme-
illustrated by Fig. 3, which compares a flood resulting able soil surface) and one (no siltation). Csilt depends
from a convective storm event to one induced by an on the precipitation intensity, the actual canopy cover
advective storm event. The convective event is and a given maximum reduction factor Cmax derivable
characterised by high rainfall intensities and low from experimental studies. Once a low-permeable
antecedent soil moisture, whereas the advective event layer has developed at the soil surface, regeneration
with much lower intensities is accompanied by high from the aggregate breakdown takes place over a
antecedent soil moisture. The simulations refer to the longer period of time (up to several months) or
Lein catchment, which is described in Section 5.1. abruptly, when sowing or harvesting is done.
The two flood events are in the same order of When applying the siltation module to the Lein
magnitude, both having a return period of about three catchment (see Section 5.1) with parameter values as
years. As Fig. 3a demonstrates, generation of infiltra- they are measured at the plot scale, an unrealistically
tion-excess overland flow and resultant direct runoff is high amount of infiltration-excess is simulated and
overestimated by far when infiltration is calculated catchment runoff is drastically overrated. Fig. 4
solely as micropore infiltration controlled by soil contains the results obtained from a sensitivity
hydraulic conductivity. This is also true for the analysis conducted with data from the Lein catchment
advective event (Fig. 3b), but because of lower for the factor Cmax. The sensitivity analysis was
precipitation intensities and higher soil moisture the performed separately for five convective and six
consequences are less obvious. In this context it is advective storm events which induced floods with
important to notice that for advective events like the return periods of between two and eight years. The
one in February 1990, which contributed to a huge right-hand side of Fig. 4 shows mean values for the
flood in the Rhine basin, infiltration-excess overland impact of Cmax on runoff volume and runoff
flow is of very minor importance anyway. maximum value and suggests a moderate increase of
catchment runoff for convective storm events with
4.2. Siltation effects high precipitation intensities and no significant impact
for advective storms. The diagram on the left, on the
Due to their characteristic particle size distribution, other hand, demonstrates that only one convective
loess soils are generally susceptible to aggregate event met the siltation criteria of high precipitation
breakdown and siltation during high intensity rainfall, intensities associated with a low canopy covering that
resulting in surface sealing and a drastic decrease in occurred because cereals had been harvested shortly
hydraulic conductivity at the soil surface as well as a before. For this event (September 1987), sensitivity of
decline in macropore connectivity (e.g. Römkens Cmax is so extraordinarily high, that even for a
et al., 1995). In the past, the impact of siltation on moderate two-fold maximal reduction of hydraulic
infiltration and the production of infiltration-excess conductivity ðCmax ¼ 0:5Þ peak runoff is nearly
overland flow had been studied extensively with the doubled. In contrast to this exaggerated simulation,
help of sprinkler experiments at the plot scale. The the observed runoff coefficient of this flood event was
reduction of hydraulic conductivity obtained from only 6%, which is in the same range as for the other
such experiments is in the range of one order of convective events, where no siltation was simulated.
magnitude (e.g. Roth et al., 1995). Much less is known This means that siltation effects at this meso-scale
about the magnitude of macropore disconnection due catchment are much less relevant than indicated by
to siltation and the impact of siltation on runoff the model results. We assume that one reason for this
generation at the catchment scale. is that the sub-grid spatial variability of runoff
The siltation module, which has been developed as generation processes and possible re-infiltration
an extension for WaSiM-ETH, takes into account the processes are not known and included within the
decrease in hydraulic conductivity at the soil surface present model. These findings support the observation
as well as the reduced inflow of infiltration-excess into of Roth et al. (1995) that a large percentage of the
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 87

Fig. 3. Impact of macropores on runoff generation; illustrated by a simulation of two flood events in the Lein catchment with a return period of
approximately three years as response to (a) a convective and (b) an advective rainstorm event.

infiltration-excess generated locally re-infiltrates in averaging neglects the non-linearity of infiltration

areas which are not affected by siltation. Conse- processes, and inevitably leads to an overestimation of
quently, the empirical evidence obtained for siltation the influence of built-up areas on storm-runoff
at the plot scale cannot simply be adopted in order to generation and underrates the compensating effect of
describe runoff generation at the catchment scale. permeable (‘green’) areas within settlements.
In order to take into account this pronounced form
4.3. Sealed surfaces and urban sewer systems of heterogeneity within urban grid cells (often
referred to as subgrid variability), each grid cell is
Urban areas, on the one hand, consist of imperme- divided into a sealed and an unsealed part according to
able surfaces which allow no or only very little the degree of sealing of the cell’s actual land-use type
infiltration and are often connected to a sewer system. (see Fig. 5). The sealed part of a grid cell is assumed
One the other hand, they also contain greens, parks, to be connected to a sewer system. Another advantage
green strips or gardens, where better infiltration and of this procedure is that it allows for a distinction
soil storage conditions can be found. Lumping of soil between densely settled areas and loose settlement.
and land-use parameters in settlement areas by simple But the approach is not limited to settlement areas:

Fig. 4. Sensitivity analysis for the hydraulic conductivity reduction factor Cmax used by the siltation module within the extended WaSiM-ETH.
88 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

Fig. 5. Model concept for sealed surfaces within a grid cell and their connection to the sewer system.

linear infrastructure like roads within grid cells with and therefore most relevant driving force for land-use
agricultural or forestal land-use can also be con- conversion. Due to the vicinity to a prospering
sidered in this way. In most other grid-based models industrial region the study area has experienced a
(e.g. TOPMODEL (Beven et al., 1984); SHE (Abbott steady growth of built-up areas during recent decades,
et al., 1986)), such linear landscape elements are caused by settlement expansion.
disregarded, because they cannot be represented in As an example, the result of an increase of
conventional grid discretization with cell sizes of settlement area in the Lein catchment is shown in
100 £ 100 m and more. Fig. 6. As an scenario target an increase of 50% for the
settlement from 7.4% up to 11.1% of the catchment
area is prescribed, which is close to the German
5. Simulation results average of 11.8%. As an infrastructural development
axis, the course of the lower river has been used,
5.1. The study area which is very similar to the regional road network.
The simulation shows a result (Fig. 6, right), where
Situated in south-western Germany, the Lein the development of urban areas is more pronounced in
catchment is part of the Kraichgau region and is the towns located along the development axis and
characterised by a gently sloping terrain that reaches much less in the others.
from an altitude of 160 –350 m above sea level. The
catchment drains an area of 115 km2. Geologically the 5.2.2. Hydrological impacts
area is structured by marl and gypsum, which are The hydrological model works on the same basis of
partly covered by a loess stratum of up to 20 m gridded spatial data as the scenario technique. There-
thickness. The soils are predominantly represented by fore, the land-use maps shown in Fig. 6 for both
luvisols and cambisols, which provide good cultiva- present and scenario conditions can be used directly to
tion conditions for crop land. This is reflected by a evaluate the impact of the urbanisation scenario on
60% agricultural land-use. Nevertheless this area has storm-runoff generation.
experienced a decided increase of settlement due to its Fig. 7 is a comparison of two flood events in the
vicinity to a prospering industrial region. Lein catchment which have already been described in
Section 4.1. It contains simulation results for present
5.2. Urbanisation land-use conditions as well as for the urbanisation
scenario shown in Fig. 6. The comparison demon-
5.2.1. Land-use scenario strates that the increase in flood volume and peak
Since the urbanisation tool has been validated on runoff due to urbanisation is much more distinct for
historical data, the method is applied to the Lein the convective storm event than for the advective one,
catchment. Although this area is predominantly under although the precipitation volume and the peak flow
agricultural land-use, urbanisation has become more are of the same order of magnitude for both events and
and more important, because it is the most profitable represent a return period of approximately 2– 3 years
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 89

Fig. 6. Spatial distributed scenarios of urbanisation of the Lein catchment with an increase of 50% of the settlement and industrial areas.

in both cases. The markedly slighter effect on the or resulting high soil moisture), the smaller the
advective event is the result of (1) higher antecedent contribution of infiltration excess overland flow to
soil moisture which levels differences in soil charac- total runoff, and thus the less important surface and
teristics as well as (2) lower precipitation intensities land-cover conditions are.
which prevent the occurance of infiltration excess
overland flow and an overflow of the sewer system. 5.3. Altered agricultural management
This argumentation is also supported by a com-
parison of various advective events with different 5.3.1. Scenario
return periods, as it is shown in Table 1. The The accomplishment of the European Union resol-
comparison reveals a strong negative correlation ution (Agenda 2000) requires that 10% of agricultural
between the impact of urbanisation on runoff and production land be set-aside until the year 2006. Since
the baseflow contribution to the flood event. This the Lein catchment is provided with fertile soils,
shows that the wetter the antecedent catchment agricultural production will continue to be profitable
conditions (indicating high groundwater levels and/ and this resolution will be the external driving force for

Fig. 7. Hydrological impact of urbanisation in the Lein catchment; simulation of two flood events as a response to (a) a convective storm event
and (b) an advective storm event under present and scenario conditions.
 90 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

Table 1
Increase in runoff volume and maximum due to a 50% growth of settlement and industrial areas in the Lein catchment; the events are sorted by
the urbanisation impact on runoff volume

Year, month Increase in runoff compared Simulated baseflow contribution to volume (%) Duration (h) Return period approx. (a)
to present conditions

Maximum (%) Volume (%)

1990, February 3.4 3.7 19 150 2

1993, December 5.9 2.7 17 250 8
1997, February 3.9 2.7 19 150 7
1982, December 1.7 1.5 27 225 3
1983, May 0.6 0.9 39 300 4
1988, March 0.0 0.0 52 650 3
Mean 2.6 1.8 29 290 4.5

the development of the arable land in this region. Under Fig. 8) is used directly as an input to the hydrological
the premise that only the most marginal revenue sites are model. The land-cover classification ‘set-aside area’
chosen for set-aside and presuming that tenures allow is then considered to be almost bare (open) land,
flexible land shifting, Fig. 8 shows a scenario for a future because no crops are cultivated in this area. This is
agriculture land-use. parameterised by a small interception loss, and a
Based on the integrated yield estimation, the sparse vegetation cover which in case of high rainfall
agricultural module of LUCK has selected mainly intensities may trigger siltation and disconnect
those areas potentially to be set aside that are either macropores from the soil surface (Section 4.2). It
too wet, too steep or have a shallow soil thickness has to be mentioned that these fields, if set-aside
which interferes with the agricultural use. Further- permanently, may develop a new, more natural
more, the fertile loess-covered soils are excluded from vegetation cover, which may lead to higher values
potential set-aside measures. of interception storage, and better infiltration con-
ditions (protected soil and efficient macropores). This
5.3.2. Hydrological impacts possible re-development of natural vegetation is not
Again, the altered map generated by LUCK (see considered in the parameterisation applied here.

Fig. 8. Agricultural scenario for the Lein catchment with 10% of set-aside areas. Left: present situation; right: scenario.
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 91

The simulated impact of the set-aside scenario on (1) Land-cover characteristics and their influence on
storm-runoff generation is marginal for both events the infiltration conditions of the soil surface.
illustrated in Fig. 9. The slight increase in runoff (2) Linkage of land-cover, soil-macroporosity, and
visible in Fig. 9a is due to an assumed less-dense unsaturated zone dynamics controlling infiltra-
canopy cover on the set-aside fields at the end of June tion excess.
(vegetation period). Nearly no change in runoff is (3) Temporal and spatial dynamics of rainfall events
calculated for the advective event shown in Fig. 9b. and runoff generation processes.
This is also realistic, because soil storage conditions (4) Initial conditions of the flood event, particu-
are not substantially affected by this measure, whereas larly regarding antecedent moisture conditions
possible modifications of the infiltration conditions in the catchment.
are not relevant during rainstorms with low precipi-
tation intensities. A combination of spatially distributed land-use
scenarios and a process-oriented hydrological model
seems to be an adequate tool to investigate potential
6. Conclusions land-use change impacts on flood generation. WaSiM-
ETH has been chosen for this study because it
6.1. Land-use scenarios considers the spatial distribution of catchment charac-
teristics and spatial and temporal dynamics of climate
LUCK has proven to be a useful tool for land- variables. The extensions presented in Section 4 were
use scenario generation. It provides a spatially developed to better represent land-cover and soil-
distributed specification of land-use changes in surface conditions and the unsaturated zone in order to
meso-scale areas, based on basic spatial data of improve modelling of infiltration and runoff gener-
land-use, soil and topography. Combined scenarios ation processes. Thereby, on the one hand the
of different land-use categories are possible additional parameters and process descriptions have
following the described hierarchy. The resulting led to an increase in model uncertainty. On the other
scenarios can be used as input for studies of hand, the influences of land-cover characteristics on
various environmental questions, provided that storm-runoff generation are now represented in a
land-use is adequately parameterised and the comprehensible way that allows us to track complex
spatial distribution is considered. interactions and coherence, which otherwise would
The most crucial component of the described not be obvious. Some general conclusions can be
method is the evaluation of the land representing the drawn from what has been learnt during model
suitability for the different land-use categories. The development and application:
weighting of the different attributes is a subjective
matter, which needs to be properly validated. The (1) The influence of land-use on storm-runoff
evaluation performed in this study is predominantly generation is stronger for convective storm
based on biophysical facts. However, since land-use is events with high precipitation intensities than
a societal issue, socio-economic aspects also influence for long advective storm events with low
decisions on land-use change. Thus, land-use changes precipitation intensities, because only storm
often happen almost independently of the natural events originated by high rainfall intensities are
conditions, i.e. the presented scenarios have only at least partially controlled by the conditions of
limited prognostic value. the land-cover and/or the soil-surface.
(2) Convective storm events, however, are of very
6.2. Hydrological modelling of land-use changes minor relevance for the formation of floods in the
large river basins of Central Europe because the
Modelling the influence of land-use and land-use extent of convective rainstorms is usually
changes on storm-runoff generation requires an restricted to local occurrence.
adequate consideration of the following factors and (3) Precipitation volume as well as antecedent soil
aspects: moisture conditions and groundwater levels are
 92 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93

Fig. 9. Hydrological impact of a set-aside scenario in the Lein catchment; simulation of two flood events in the Lein catchment as a response to
(a) a convective storm event and (b) an advective storm event under present and scenario conditions.

of major importance for the degree to which Activities (IRMA), and by the German Federal
land-use can influence storm-runoff generation. Environmental Agency (UBA).
The magnitude of a flood peak or the return
period of a flood event respectively are less
meaningful indicators in this respect. References

However, there are limitations to such modelling Abbott, M.B., Bathurst, J.C., Cunge, A., O’Connell, P.E.,
studies, since land-use and hydrological models are Rassmussen, J., 1986. An introduction to the SHE, 2: structure
accompanied by a high degree of uncertainty. This of a physically-based, distributed modelling system. J. Hydrol.
87, 61–77.
uncertainty is due to insufficient data availability or
Beven, K., Germann, P.F., 1982. Macropores and water flow in
quality and related space-time heterogeneity (data soils. Wat. Resour. Res. 18, 1311–1325.
uncertainty), insufficient knowledge on the physics Beven, K., Kirkby, M., Schoffield, N., Tagg, A., 1984. Testing a
and the stochastic features of the processes involved, physically-based flood forecasting model (TOPMODEL) for
in particular during extreme precipitation periods three UK catchments. J. Hydrol. 69, 119– 143.
Bronstert, A., 1999. Capabilities and limitations of detailed
(process uncertainty), and simplifications inherent in
hillslope hydrological modelling. Hydrological Processes 13,
the model structure (model uncertainty). For less 21 –48.
complex hydrological models (small number of Bronstert, A., Fritsch, U., Katzenmaier, D., 2001. Quantifizierung
parameters), a rigorous uncertainty estimation can des Einflusses der Landnutzung und -bedeckung auf den
be carried out, such as described by Freer et al. (1996). Hochwasserabfluss in Flussgebieten unter Berücksichtigung
der aktuellen Bodenbearbeitung, des Zustandes der Vegeta-
However, due to the large number of parameters and
tionsdecke und möglicher Klimaänderungen. Report for the
long computing times involved, this procedure is not German Federal Environmental Agency, Berlin, 216 p., (URL:
easily transferable to detailed process-oriented hydro- http://www.umweltbundesamt.de/wasser).
logical models like, for instance, the one applied for Freer, J., Beven, K.J., Ambroise, B., 1996. Bayasian estimation of
the Lein catchment (Sections 4 and 5). uncertainty in runoff prediction and the value of data: an
application of the GLUE approach. Wat. Resour. Res. 32,
2161–2173.
Fritsch, U., Katzenmaier, D., Bronstert, A., 2001. Land-use
Acknowledgements scenarios for flood risk assessment studies. In: Mander, Ü.,
Printsmann, A., Palang, H. (Eds.), Development of European
Landscapes, vol. 2. International Association for Landscape
This work has been funded by the European Union Ecology (IALE), Tartu, pp. 324–330.
as a contribution to the INTERREG Rhine-Maas Fritsch, 2002. Entwicklung von Landnutzungsszenarien für land-
 D. Niehoff et al. / Journal of Hydrology 267 (2002) 80–93 93

schaftsökologische Fragestellungen. Brandenburgische Umwelt Riebsame, W.E., Meyer, W.E., Turner, B.L., 1994. Modeling land
Berichte, Heft 12, Universität Potsdam, 150 p. use and cover as part of global environmental change. Clim.
Katzenmaier, D., Fritsch, U., Bronstert, A., 2001. Quantifizierung Change 28, 45 –64.
des Einflusses von Landnutzung und dezentraler Versickerung Römkens, M.J.M., Luk, S.H., Poesen, J.W.A., Mermut, A.R., 1995.
auf die Hochwasserentstehung. In: Heiden, S., Erb, R., Sieker, Rain infiltration into loess soils from different geographic
F. (Eds.), Hochwasserschutz heute—Nachhaltiges Wasserman- regions. Catena 25, 21–32.
agement, Erich Schmidt Verlag, Berlin. Roth, C.H., Helming, K., Fohrer, N., 1995. Surface sealing and
Kleyer, M., 1996. In: Konold, W., (Ed.), Urbanisierungsprozesse in runoff generation on soils derived from loess and pleistocene
der Kulturlandschaft, Naturlandschaft – Kulturlandschaft, sediments. Z. Pflanzenernährung Bodenkunde 158, 43 –53.
Landsberg, pp. 229 –242. Schulla, J., 1997. Hydrologische Modellierung von Flussgebieten
Marks, R.M., Müller, M., Leser, H., Klink, H.J. (Eds.), 1992. zur Abschätzung der Folgen von Klimaänderungen. Zürcher
Anleitung zur Bewertung des Leistungsvermögens des Land- Geographische Schriften, Heft 69 Zürich, Switzerland.
schaftshaushalts, Selbstverlag des Zentralausschusses der Veldkamp, A., Fresco, L.O., 1996. CLUE: a conceptual model to
deutschen Landeskunde, Trier. study the conversion of land use and its effects. Ecol. Modell.
Niehoff, D., 2002. Modellierung des Einflusses der Landnutzung 85, 253–270.
auf die Hochwasserentstehung in der Mesoskala. Brandenbur- Wenkel, K.-O., 1999. Dynamische Landschaftsmodelle für die
gische Umwelt Berichte, Heft 11, Universität Potsdam, 144 p. Angewandte Landschaftsökologie. In: Schneider-Sliwa, R.,
O’Callaghan, J.O., 1996. Land use: the interaction of economics, Schaub, D., Gerold, G. (Eds.), Angewandte Landschaftsökolo-
ecology and hydrology. London. gie, Springer, Berlin, pp. 107–133.