REPORTS

5. Supporting Online Material is available on Science 11. B. D. Harfe et al., Genes Dev. 12, 2623 (1998). 24. R. D. Burdine, C. S. Branda, M. J. Stern, Development
Online. 12. P. M. Loria, J. Hodgkin, O. Hobert, J. Neurosci. 24, 125, 1083 (1998).
6. All transcriptional reporters were made by polymer- 2191 (2004). 25. R. J. Johnston, O. Hobert, Nature 426, 845 (2003).
ase chain reaction (PCR) fusion, from the predicted 13. We used two separate arrays because of possible 26. From pPD95.67, gift of A. Fire (Stanford University).
start to the next predicted upstream gene, and extra- promoter interference that we have experienced 27. N. Chen, I. Greenwald, Dev. Cell 6, 183 (2004).
chromosomal arrays were marked with pha-1(þ). when using the same promoter to drive two different 28. D. D. Shaye, I. Greenwald, Development 132, 5081,
Expression of mir-61 transcription was studied by gene products in the same array. For the assessment (2005).
using the arIs107 integrant, which displayed the same of many targets, the assay could be improved by 29. We are grateful to X. Zhou for performing microinjections;
pattern as extrachromosomal arrays. integrating the transgene expressing the miRNA. and to O. Hobert, R. Mann, and D. Shaye for discussion
7. mir-61 is also expressed in cells of the somatic gonad 14. I. S. Greenwald, P. W. Sternberg, H. R. Horvitz, Cell and insightful comments on the manuscript. This work
in which LIN-12 is active, and this expression is also 34, 435 (1983). was supported by NIH grant CA095389 (to I.G.). I.G. is an
lost when the LBSs are mutated (Fig. 1, B and C). In 15. V. L. Tybulewicz, Curr. Opin. Immunol. 17, 267 (2005). investigator with the Howard Hughes Medical Institute.
contrast, expression in other tissues where we have 16. K. R. Norman et al., Cell 123, 119 (2005). Supporting Online Material
no evidence that lin-12 activity is functionally rel- 17. M. Sundaram, I. Greenwald, Genetics 135, 765 (1993). www.sciencemag.org/cgi/content/full/1119481/DC1
evant, such as intestinal cells, is unaffected. These 18. E. J. Hubbard, G. Wu, J. Kitajewski, I. Greenwald, Genes Materials and Methods
observations suggest that the loss of expression in Dev. 11, 3182 (1997). SOM Text
P5.p and P7.p reflects lack of response to lin-12 and 19. C. Wen, I. Greenwald, J. Cell Biol. 145, 1165 (1999). Figs. S1 to S5
not loss of a general enhancer. 20. C. W. Cowan et al., Neuron 46, 205 (2005). References and Notes
8. V. Ambros, Nature 431, 350 (2004). 21. P. W. Reddien, S. Cameron, H. R. Horvitz, Nature 412,
9. C. Z. Chen, L. Li, H. F. Lodish, D. P. Bartel, Science 198 (2001). 29 August 2005; accepted 13 October 2005
303, 83 (2004). 22. A. S. Yoo, unpublished observations. Published online 20 October 2005;
10. S. Chang, R. J. Johnston Jr., C. Frokjaer-Jensen, S. Lockery, 23. M. Rebeiz, N. L. Reeves, J. W. Posakony, Proc. Natl. 10.1126/science.1119481
O. Hobert, Nature 430, 785 (2004). Acad. Sci. U.S.A. 99, 9888 (2002). Include this information when citing this paper.

2080, relative to baseline conditions in 1990

Ecosystem Service Supply (5). Socioeconomic trends were developed from
the global Intergovernmental Panel on Climate
and Vulnerability to Global Change Special Report on Emission Scenarios
(IPCC SRES) storylines B1, B2, A1FI, and A2
Change in Europe for EU15þ (4, 6, 7) (table S1). With this com-
mon starting point, socioeconomic changes
Dagmar Schröter,1,2* Wolfgang Cramer,1 Rik Leemans,3 relate directly to climatic changes through green-
I. Colin Prentice,4 Miguel B. Araújo,5,6 Nigel W. Arnell,7 house gas concentrations and to land-use changes
through climatic and socioeconomic drivers, such
Alberte Bondeau,1 Harald Bugmann,8 Timothy R. Carter,9 as demand for food. Four general circulation
Carlos A. Gracia,10 Anne C. de la Vega-Leinert,1 Markus Erhard,11 models (GCMs)—the Hadley Centre Coupled
Frank Ewert,3 Margaret Glendining,12 Joanna I. House,4 Model Version 3 (HadCM3), the National Center
Susanna Kankaanpää,9 Richard J. T. Klein,1 Sandra Lavorel,13,14 for Atmospheric Research–Parallel Climate
Marcus Lindner,15 Marc J. Metzger,3 Jeannette Meyer,15 Model (NCAR-PCM), the Second Generation
Timothy D. Mitchell,16 Isabelle Reginster,17 Mark Rounsevell,17 1
Potsdam Institute for Climate Impact Research, 14473
Santi Sabaté,10 Stephen Sitch,1 Ben Smith,18 Jo Smith,19 Potsdam, Germany. 2Center for International Develop-
Pete Smith,19 Martin T. Sykes,18 Kirsten Thonicke,4 ment, Harvard University, Cambridge, MA 02138, USA.
3
Wilfried Thuiller,20 Gill Tuck,12 Sönke Zaehle,1 Bärbel Zierl8 Department of Environmental Sciences, Wageningen
University, 6700 AA Wageningen, Netherlands. 4De-
partment of Earth Sciences, University of Bristol, BS8
Global change will alter the supply of ecosystem services that are vital for 1RJ Bristol, UK. 5School of Geography and Environ-
human well-being. To investigate ecosystem service supply during the 21st ment, University of Oxford, OX1 3TB Oxford, UK.
6
century, we used a range of ecosystem models and scenarios of climate and Museo Nacional de Ciencas Naturales, 28006 Madrid,
land-use change to conduct a Europe-wide assessment. Large changes in climate Spain. 7Tyndall Centre for Climate Change Research,
School of Geography, University of Southampton,
and land use typically resulted in large changes in ecosystem service supply. Southampton SO17 1BJ, UK. 8Department of Environ-
Some of these trends may be positive (for example, increases in forest area and mental Sciences, Eidgenössische Technische Hochschule,
productivity) or offer opportunities (for example, ‘‘surplus land’’ for agricultural 8092 Zürich, Switzerland. 9Finnish Environment Institute,
extensification and bioenergy production). However, many changes increase 00251 Helsinki, Finland. 10Center for Ecological Research
and Forestry Applications, University of Barcelona, 08193
vulnerability as a result of a decreasing supply of ecosystem services (for ex- Barcelona, Spain. 11Institute for Meteorology and Climate
ample, declining soil fertility, declining water availability, increasing risk of Research, Forschungszentrum Karlsruhe, 82467 Garmisch-
forest fires), especially in the Mediterranean and mountain regions. Partenkirchen, Germany. 12Agriculture and the Envi-
ronment Division, Rothamsted Research, AL5 2JQ
Harpenden, UK. 13Laboratoire d’Ecologie Alpine, CNRS,
To sustain a future in which the Earth_s life- models. A dialogue with stakeholders from Université Joseph Fourier, 38041 Grenoble, France.
support systems are maintained and human relevant sectors was conducted throughout the 14
Centre d’Ecologie Fonctionnelle et Evolutive, CNRS,
needs are met, human activities must first be study (4). Montpellier, France. 15European Forest Institute, 80100
recognized as an integral component of eco- Our assessment was based on multiple Joensuu, Finland. 16 Tyndall Centre for Climate Change
Research, University of East Anglia, NR4 7TJ Norwich,
systems (1, 2). Scenarios of global change raise scenarios for major global change drivers UK. 17Département de Géographie, Université Catholique
concern about alterations in ecosystem services (socioeconomic factors, atmospheric green- de Louvain, 1348 Louvain-la-Neuve, Belgium. 18Depart-
such as food production and water supply, but house gas concentrations, climate factors, and ment of Physical Geography and Ecosystems Analysis,
the potential trajectories of change, especially land use). The scenarios were quantified for Lund University, 22362 Lund, Sweden. 19School of Bio-
logical Sciences, University of Aberdeen, AB24 3UU
at the regional scale, are poorly characterized Europe (15 pre-2004 European Union mem- Aberdeen, UK. 20Kirstenbosch Research Center, South
(3). We investigated the changing supply of bers, plus Norway and Switzerland, henceforth African National Biodiversity Institute, 7735 Cape Town,
ecosystem services in a spatially explicit vul- referred to as EU15þ) during the 21st century South Africa.
nerability assessment of Europe, using multiple at 10¶-by-10¶ latitude/longitude grid resolution, *To whom correspondence should be addressed.
global change scenarios and a set of ecosystem and for periods ending in 2020, 2050, and E-mail: [email protected]

www.sciencemag.org SCIENCE VOL 310 25 NOVEMBER 2005 1333

 REPORTS
Coupled Global Climate Model (CGCM2), and Table 1. Summary of the basic socioeconomic, atmospheric, and climatic drivers based on model outputs
the Commonwealth Scientific and Industrial forced by SRES scenarios. Population and atmospheric CO2 concentration estimates are for the year 2080.
Research Organisation–Climate Model Version For the climatic indicators, 30-year averages 2051 to 2080 compared with 1961 to 1990 are shown. In this
study, we focused on the HadCM3 climate model and the A2 storyline. The EU15þ population in 1990 was
2 (CSIRO2)—were used to simulate climatic 376 million people. The GCMs were forced with these concentrations plus CO2 equivalents accounting for
changes (4). Out of 16 combinations of story- the other greenhouse gases. The atmospheric CO2 concentration in 1990 was 354 parts per million (ppm)
lines and GCMs, we selected seven scenarios by volume. Precipitation changes (%) on the Iberian Peninsula are given by season: JJA, summer (June, July,
for interpretation: B1, B2, A1FI, A2 calcu- August); DJF, winter (December, January, February).
lated with HadCM3 (variation across story-
lines, Bsocioeconomic options[), and A2 Climate model
Scenarios by 2080
calculated additionally with three other GCMs
HadCM3 NCAR-PCM CGCM2 CSIRO2
(variation across climate models, Bclimatic
uncertainty[) (Table 1) (4). Storyline B1
Temperature-change scenarios in Europe Population (106) 376 376 376 376
vary regionally but show a clear trend toward CO2 concentration (ppm) 518 518 518 518
D Temperature (-C) 3.1 – – –
warming. The average projected temperature D Precipitation (%)
increase in Europe ranged from 2.1- to 4.4-C Europe 4.8 – – –
(across storylines) and from 2.7- to 3.4-C for Iberian Peninsula JJA –17 – – –
the A2 storyline (across GCMs) (Table 1), Iberian Peninsula DJF 7 – – –
with the strongest warming consistently in the Storyline B2
high latitudes (fig. S1). Seasonal and regional Population (106) 346 346 346 346
CO2 concentration (ppm) 567 567 567 567
variation of changes in precipitation was D Temperature (-C) 2.1 – – –
considerable. Generally, all scenarios con- D Precipitation (%)
curred on decreasing precipitation in the south Europe 2.7 – – –
of Europe, particularly in summer (Table 1), Iberian Peninsula JJA –14 – – –
and increasing precipitation over much of Iberian Peninsula DJF 7 – – –
northern Europe (fig. S2). Storyline A1FI
Population (106) 376 376 376 376
Land-use scenarios (4) showed little varia- CO2 concentration (ppm) 779 779 779 779
tion based solely upon different GCMs, in- D Temperature (-C) 4.4 – – –
dicating that socioeconomic assumptions had a D Precipitation (%)
greater effect on land use than did climatic Europe –0.5 – – –
drivers. The general trends were of reductions Iberian Peninsula JJA –27 – – –
in agricultural areas for food production, partly Iberian Peninsula DJF 2 – – –
Storyline A2
compensated for by increases in bioenergy pro- Population (106) 419 419 419 419
duction and forests, as well as small increases in CO2 concentration (ppm) 709 709 709 709
urban and protected areas (Table 2). In the A D Temperature (-C) 2.8 2.7 3.4 2.7
(economic) scenarios, the decline in agri- D Precipitation (%)
cultural land was especially pronounced Europe 0.5 2.3 0.0 –0.6
(Fig. 1), mainly owing to assumptions about Iberian Peninsula JJA –22 –18 –26 –19
Iberian Peninsula DJF 10 0 1 –3
the role of technological development (8).
The land that becomes Bsurplus[ to the re-
quirement of food production would allow trast, the available choice of bioenergy crops Mediterranean population would be living in
balancing the production of other ecosystem decreased in southern Europe owing to in- watersheds with increased water stress (14% in
services against food production, for example creased drought, unless production systems are B2). In this region, water scarcity would likely
through extensification (9) or bioenergy pro- adapted (Table 2). be aggravated by higher extractions per capita
duction (10). Changes in the provision of water affect for irrigation and tourism (15).
We next examined the changing supply of humans directly and indirectly through effects Case studies for the Rhine, RhHne, and
a number of ecosystem services owing to on other ecosystem services. At the global Danube basins, as well as for small Alpine
global change in Europe. The selected services scale, increases in population and consumption catchments, indicated climate-induced changes
reflect the availability of modeling tools ade- alone will reduce water availability (3, 12, 13). in the timing of runoff (4). These result from
quate for pan-European assessment and the We quantified the implications of population impacts of rising temperatures on snow-cover
aim for a broad range of terrestrial services and climate change on water availability in dynamics, which enhanced winter runoff, re-
covering the four service categories identified EU15þ using a macroscale hydrological model duced summer runoff (Table 2), and shifted
by the Millennium Ecosystem Assessment (1). (4). In 1995, approximately 193 million people monthly peak flows by up to two months
The European Commission proposed a out of a total EU15þ population of 383 mil- earlier than at present (16). This reduced water
target of doubling the contribution of renew- lion lived under water stress (water availability supply at peak demand times and increased the
able energy sources to the EU_s total primary G1700 m3 capita–1 yearj1) (14). In the absence risk of winter floods. Changes in snow-cover
energy needs to 12% by 2010 (11). Biomass of climate change, these numbers decreased by dynamics directly affect biodiversity at high
energy will add to this goal. We assessed the 2080 where population decreased (scenario elevations. Moreover, navigation and hydro-
potential distribution of 26 bioenergy crops B2, Table 1). In contrast, population and cli- power potential would be altered.
under changing climatic conditions (4). The po- mate change increased in the numbers of people In addition to its importance for water sup-
tential distribution of bioenergy crops increased living in water-stressed watersheds and ex- ply and biodiversity conservation, snow cover
in northern Europe as a result of increasing acerbated water deficiency for many already is of course indispensable for winter tourism.
temperatures (Table 2). These potential gains stressed areas (Table 2), particularly in south- The Alpine case studies indicated a rise in the
are optimistic, given that restricting soil ern Europe (Fig. 2). Under the A1FI, A2, and elevation of reliable snow cover from about
conditions are not taken into account. In con- B1 scenarios, between 20 and 38% of the 1300 m today to 1500 to 1750 m at the end of

1334 25 NOVEMBER 2005 VOL 310 SCIENCE www.sciencemag.org

 REPORTS
Table 2. Summary of land-use drivers and global change impacts for Europe, time period 2080 compared with baseline (1990), unless otherwise noted (4).

Storyline B1 B2 A1FI A2 A2 A2 A2
GCM HadCM3 HadCM3 HadCM3 HadCM3 NCAR-PCM CGCM2 CSIRO2

Land-use model outputs forced by climate, CO2, and interpretations of SRES storylines
Land-use change (%)*
Cropland (for food production) –7.0 –6.4 –10.7 –10.4 –10.6 –10.7 –10.6
Grassland (for livestock) –1.1 –6.7 –8.7 –10.0 –10.1 –10.2 –10.0
Forest 3.5 5.6 0.8 0.7 1.0 1.0 1.2
Urban 0.05 0.06 0.09 0.08 0.07 0.07 0.07
Bioenergy production 3.4 7.4 8.7 8.7 9.1 8.6 8.6
Protected 6.1 6.1 6.1 6.1 6.1 6.1 6.1
Surplus 1.1 0.0 9.8 10.9 10.5 11.2 10.8
Impacts as estimated by ecosystem models
D Potential distribution of bioenergy crops (%)y
Overall 3 4 1 3 6 7 5
Latitude 35 to 45 –7 –6 –13 –8 –1 –3 –2
Latitude 45 to 55 –1 0 –6 –2 4 8 –6
Latitude 55 to 65 12 13 12 13 11 14 15
Latitude 65 to 71 18 22 32 23 19 16 34
Additional people living under water stress (106)z 44.3 25.8 44.3 15.7 7.5 11.7 5.8
People living under increased water stress (106)` 31.0 38.2 45.7 35.6 18.4 69.6 25.4
D Alpine summer runoff (%)¬ –24 -23 –46 –34 –12 –27 –20
D Elevation of reliable snow cover (m)¬ 230 180 450 310 200 230 390
Species loss per grid cell (minimum to maximum %)P –7 to –58 –8 to –53 –8 to –59 –8 to –55
D Area burnt, Iberian Peninsula (%) 112 57 80 55 –1 37 8
D Wood increment (%) –10.0 9.7 3.8 4.4 2.9 2.9 6.2
Cumulative carbon balance (Pg C)L 2.2 2.4 1.8 3.0 4.9 4.1 3.7
Average carbon flux (% of emissions)** 2.5 2.7 2.1 3.5 5.5 4.7 4.2
D Soil organic carbon (Pg C)yy
Total –0.1 –0.9 –4.1 –4.4 –4.3 –4.5 –4.8
Cropland –4.3 –4.3 –5.9 –5.6 –5.4 –5.5 –5.8
Grassland 1.5 –1.2 –2.2 –2.7 –2.7 –2.7 –2.8
Forest 2.8 3.6 1.0 1.1 1.3 1.3 0.7
*Baseline areas (% of EU15þ): Cropland, 23.0%; grassland, 17.2%; forest, 31.0%; urban, 1.5%; other (shrubland, barren land, wetland, inland waters, sea, permanent ice, and snow), 27.3%.
For all scenarios, it is assumed that 20% of the area of Europe will become designated as ‘‘protected’’ by 2080. This was based on a judgment made from past and current increases in
protected-areas coverage in Europe, the latter being due to member-state responses to the need for implementation of the NATURA 2000 network. Although this target was the same for
all scenarios, it was assumed that it would be reached for different reasons: The economic scenarios require areas for recreation for a richer population, whereas the environmental scenarios
require areas designated for conservation purposes (tables S1 and S2). ‘‘Surplus’’ is land that is left over when the demand for all land-use types is satisfied. .Change in potential
distribution of 26 bioenergy crops (% land area) due to climate change. The estimates do not take soil conditions into account. -Additional people (millions) living in stressed
watersheds due to climate change (compared with the hypothetical case of no climate change). Water-related resource problems are likely when water availability falls below the threshold
of 1700 m3 capita–1 year –1 (14). `People (millions) living in already water-stressed watersheds (less than 1700 m3 capita–1 year –1), where climate change further reduces water
availability by more than 10%. ¬Average of five Alpine case studies. PYear 2050 compared with the baseline (1990). The range of minimum (full instantaneous dispersal) to
maximum (zero dispersal) loss is shown. Plants, mammals, reptiles, amphibians, and breeding birds were considered. This indicator records only losses from a specific grid cell and does not
take potential gains into account. The indicator does not make a statement about potential losses of the species from Europe or about extinction. LCumulative land-atmosphere carbon
flux between 1990 and 2080. Positive values denote fluxes to land. **Average yearly land-atmosphere flux (1990 to 2080) relative to EU15þ CO2 emissions in 1990. ..Change in
cumulative soil organic carbon content in mineral soil down to a depth of 30 cm.

the 21st century (Table 2) (16). A 300-m rise grate instantaneously to newly suitable hab- cork oak (Quercus suber), holm oak (Q. ilex),
of the snow line would reduce the proportion itats, the relative potential gain of plant species aleppo pine (Pinus halepensis), and maritime
of Swiss ski areas with sufficient snow from in Mediterranean regions was relatively high pine (P. pinaster). These changes would have
currently about 85 to 63% (17). because of habitat expansion (fig. S3B). How- implications for the sense of place and cultural
Biodiversity is essential to ecosystem pro- ever, unhindered expansion is unlikely because identity of the inhabitants, traditional forms of
cesses in ways that are not yet fully understood of the concurrent impacts of other drivers such land use, and the tourism sector.
(18), and it is considered worth protecting in as land use, nitrogen deposition, and biotic We assessed the potential impacts of
its own right (3). We used a statistical model- exchange, especially in the Mediterranean re- management and global change on the overall
ing framework to project the distribution of gion (20). Flexible management of nature wood production from European forest using an
more than 2000 plant and animal species across reserve areas may conserve species. However, inventory-based model (4). In line with other
Europe (4). These simulations do not incorpo- stakeholders pointed out great difficulties in industrialized areas, but opposed to global
rate effects of land-use change, because at the changing existing reserve boundaries under cur- trends (3, 24), the total European forest area
resolution of this study these were confounded rent policies and land-ownership restrictions. was projected to increase (Table 2). Climate
with climate effects (19). We therefore present To obtain more detailed results on tree change resulted in increased forest growth
conservative estimates that neglect effects of species in the Mediterranean region, we used a (Table 2), especially in northern Europe. The
habitat loss or landscape fragmentation (20). process-based tree-growth model (4). The sim- impact of increased summer drought in south-
Projections of species loss per grid cell showed ulations corroborated negative effects on veg- ern Europe was partly mitigated by higher
changes under all scenarios (Table 2). Moun- etation, especially over the long term, owing to precipitation in spring and increased water-use
tains and Mediterranean species were dispro- increased drought. Furthermore, the area burnt efficiency in response to rising atmospheric
portionately sensitive to climate change (fig. by forest fires increased in this region under all CO2 concentrations. Increasing forest area in-
S3A) (4, 21), in agreement with recent ob- but one scenario (4) (Table 2). The distribution creased annual wood increment because of a
servations (22) and projections (23). Under the of a number of typical tree species is likely to high proportion of young stands. When low
unrealistic assumption that all species can mi- decrease in the Mediterranean region, such as wood demand led to less intensive manage-

www.sciencemag.org SCIENCE VOL 310 25 NOVEMBER 2005 1335

 REPORTS
ment (B scenarios), forests grew old and less The total amount of carbon stored in ter- biosphere currently acts as a net carbon sink
productive, and increment decreased by 10.0% restrial biosphere is an important factor in (26) (Table 2). Land-use change affected this
in the B1 scenario (in B2, afforestation climate regulation (25). The net carbon land- sink positively through decreases in agricul-
counteracted this effect; Table 2). In general, atmosphere flux is determined by net primary tural land and increased afforestation. Further-
management had a greater influence on wood production and carbon losses due to soil more, CO2 fertilization enhanced net primary
production in Europe than climate or land-use heterotrophic respiration, fire, harvesting, production. However, soil carbon losses due
change. As corroborated by stakeholders, for- and land-use change. The aggregate land- to warming balanced these effects by 2050
est management is influenced more strongly atmosphere flux over Europe was estimated and led to carbon releases by the end of this
by actions outside the forest sector, such as using a dynamic global vegetation model (4). century. The temperature effect on soil car-
trade and policies, than from within. Our results confirm that Europe_s terrestrial bon losses is confirmed by recent experi-
mental and modeling studies (27–29) and by
separate calculations using a soil carbon
model (4). Although afforestation led to a
net increase in soil organic carbon in forest
soils despite the losses due to warming, the
total amount of carbon in European soils
decreased (Table 2).
Stakeholders were primarily interested in
the efficacy of land-use changes as a tool for
mitigation. We found that the choice of land
use is relevant concerning the average yearly
carbon uptake and the emission reduction
target of the European Union. However, car-
bon uptake remains small compared to fossil
fuel emissions even under the land-use change
scenario with maximum increase in forest area
(Table 2).
Stakeholders from the agricultural sector
were interested in soil organic matter content
as a key factor in the carbon cycle and as an
indicator of soil fertility. However, their great-
est concern was the total amount of land avail-
able for farming. This may reflect that current
agricultural subsidies disconnect farmers_ suc-
cess from actual ecosystem service supply,
such as soil fertility and crop production. In
some regions it is therefore questionable
whether land that is Bsurplus[ to food demands
Fig. 1. Change in cropland area (for food production) by 2080 compared with the baseline would readily be open for other uses.
(percentage of EU15þ area) for the four storylines [A1FI (A1), A2, B1, and B2] with climate The trends in European change drivers dif-
calculated by HadCM3. fer from global trends (3, 24) in several ways:

Fig. 2. Stress status of water

basins by 2080 considering cli-
mate change and population
growth, compared with the
hypothetical case of no climate
change. Stressed water basins
have less than 1700 m3 capitaj1
yearj1 (14). Only water basins
that affect the EU15þ are
shown. No significant change
means that changes in average
annual runoff for these stressed
watersheds are less than 10%.
Panels marked A1, A2, B1, B2
show the four storylines (A1 is
A1FI) based on HadCM3 climate
(2051 to 2080) and respective population sizes. Panels marked HadCM3, CSIRO2, CGCM2, and PCM show the four GCMs (2051 to 2080; PCM is NCAR-PCM) and A2
population size.

1336 25 NOVEMBER 2005 VOL 310 SCIENCE www.sciencemag.org

 REPORTS
Population increases moderately if at all, the energy—White paper for a community strategy 28. W. Knorr, I. C. Prentice, J. I. House, E. A. Holland,
and action plan’’ European Communities Tech. Nature 433, 298 (2005).
extent of urbanization is relatively small, forest Report No. 599 (1997). 29. E.-D. Schulze, A. Freibauer, Nature 437, 205 (2005).
area increases, and demand for agricultural 12. United Nations Environment Programme, ‘‘GEO-3: 30. World Tourism Organization, First International Con-
land decreases. This allows changes in land Global environmental outlook report 3’’ (UNEP, ference on Climate Change and Tourism, Djerba,
management that could decrease vulnerability. United Nations Environment Programme, Earthscan, Tunisia, 9 to 11 April 2003.
London, 2002). 31. This work was funded by the EU-project ATEAM
Problematic trends in the EU15þ are mostly 13. N. W. Arnell, Glob. Environ. Change 14, 31 (2004). (Advanced Terrestrial Ecosystem Assessment and
climate related. 14. M. Falkenmark, J. Lundquist, C. Widstrand, Nat. Modelling, EVK2-2000-00075). We gratefully ac-
The range of potential impacts in Europe Resour. Forum 13, 258 (1989). knowledge the support of our EU scientific officer
15. In addition to increasing the number of people D. Peter, and our external observers P. Leadley, D.
covers socioeconomic options (storylines) and served with water in a region, tourists’ water Ojima, and R. Muetzelfeldt. We thank M. Welp for
variation among GCMs. For most ecosystem consumption has been shown to be far in excess of advice on stakeholder dialogue and our stakeholders
services the A1FI scenario produced the biggest that of local residents (30). for their voluntary collaboration (a full list is
16. B. Zierl, H. Bugmann, Water Resour. Res. 41, WO2028 available in the supporting online material). Part of
negative impacts, and the B scenarios seemed (2005). this work was carried out while D.S. was hosted by
preferable. However, a division into either 17. H. Elsässer, P. Messerli, Mt. Res. Dev. 21, 335 (2001). the Science, Environment, and Development Group
Beconomic[ (A scenarios) or Bequitable and 18. D. U. Hooper et al., Ecol. Monogr. 75, 3 (2005). at the Center for International Development, Har-
19. W. Thuiller, Glob. Change Biol. 10, 2020 (2004). vard University.
environmental[ (B scenarios) does not reflect all 20. O. E. Sala et al., Science 287, 1770 (2000).
societal choices, given that sustainability does 21. W. Thuiller, S. Lavorel, M. B. Araújo, M. T. Sykes, I. C. Supporting Online Material
not forbid economic prosperity (3). The four Prentice, Proc. Natl. Acad. Sci. U.S.A. 102, 8245 (2005). www.sciencemag.org/cgi/contents/full/1115233/DC1
22. G.-R. Walther et al., Nature 416, 389 (2002). Materials and Methods
storylines help explore but do not contain our 23. M. Gottfried, H. Pauli, K. Reiter, G. Grabherr, Diversity Figs. S1 to S3
optimal future pathway. Distrib. 5, 241 (1995). Tables S1 and S2
Among all European regions, the Mediter- 24. J. A. Foley et al., Science 309, 570 (2005). References and Notes
ranean appeared most vulnerable to global 25. A. D. McGuire et al., Global Biogeochem. Cycles 15,
183 (2001). 24 May 2005; accepted 18 October 2005
change. Multiple potential impacts were pro- 26. G. J. Nabuurs et al., Glob. Change Biol. 9, 152 (2003). Published online 27 October 2005;
jected, related primarily to increased temper- 27. C. Fang, P. Smith, J. B. Moncrieff, J. U. Smith, Nature 10.1126/science.1115233
atures and reduced precipitation. The impacts 433, 57 (2005). Include this information when citing this paper.
included water shortages, increased risk of
forest fires, northward shifts in the distribution
of typical tree species, and losses of agricultural
potential. Mountain regions also seemed vul- Representation of Action-Specific
nerable because of a rise in the elevation of
snow cover and altered river runoff regimes. Reward Values in the Striatum
The sustained active participation of stake-
holders indicated that global change is an issue Kazuyuki Samejima,1*. Yasumasa Ueda,2 Kenji Doya,1,3
of concern to them, albeit among many other Minoru Kimura2*
concerns. The development of adaptation
strategies, such as for reduced water use and The estimation of the reward an action will yield is critical in decision-making. To
long-term soil preservation, can build on our elucidate the role of the basal ganglia in this process, we recorded striatal neurons
study but requires further understanding of the of monkeys who chose between left and right handle turns, based on the
interplay between stakeholders and their envi- estimated reward probabilities of the actions. During a delay period before the
ronment in the context of local, national, and choices, the activity of more than one-third of striatal projection neurons was
EU-wide constraints and regulations. selective to the values of one of the two actions. Fewer neurons were tuned to
relative values or action choice. These results suggest representation of action
References and Notes values in the striatum, which can guide action selection in the basal ganglia circuit.
1. Millennium Ecosystem Assessment, Ecosystems and
Human Well-Being: A Framework for Assessment Animals and humans flexibly choose actions in the action values by the errors of estimated
(Island Press, Washington, DC, 2003).
2. M. Palmer et al., Science 304, 1251 (2004). pursuit of their specific goals in the en- action values. Reinforcement learning models
3. W. V. Reid et al., Millennium Ecosystem Assessment vironment on a trial-and-error basis (1, 2). The- of the basal ganglia have been put forward (4–6).
Synthesis Report (Island Press, Washington, 2005). ories of reinforcement learning (3) describe The midbrain dopamine neurons encode errors
4. Materials and methods are available as supporting
material on Science Online. reward-based decision-making and adaptive of reward expectation (7–9) and motivation (9),
5. The periods represent 30-year averages: baseline choice of actions by the following three steps: and they regulate the plasticity of the cortico-
1990 (mean over 1961 to 1990), 2020 (mean over (i) The organism estimates the action value, striatal synapses (10, 11). Neuronal discharge
1991 to 2020), 2050 (mean over 2021 to 2050), and
2080 (mean over 2051 to 2080). defined as how much reward value (probability rates in the cerebral cortex (12–15) and stri-
6. The economic A scenarios represent a world focused times volume) an action will yield. (ii) It atum (16–18) are modulated by rewards that
on material consumption, whereas the environmental selects an action by comparing the action are estimated by sensory cues and behavioral
B scenarios focus on sustainability, equity, and
environment. The second dimension distinguishes
values of multiple alternatives. (iii) It updates responses. These observations are consistent
globalization (dimension 1) from regionalization (di- 1
with action selection through the reinforcement
mension 2). Department of Computational Neurobiology, ATR learning rule (3) and with the notion of
7. N. Nakicenovic, R. Swart, Eds., Intergovernmental Computational Neuroscience Laboratories, 619-0288
Panel on Climate Change Special Report on Emissions Kyoto, Japan. 2Department of Physiology, Kyoto stimulus-response learning (19, 20). However,
Scenarios (Cambridge Univ. Press, Cambridge, 2000). Prefectural University of Medicine, 602-8566 Kyoto, two critical questions remain unanswered: Do
8. F. Ewert, M. D. A. Rounsevell, I. Reginster, M. J. Japan. 3Initial Research Project, Okinawa Institute of the striatal neurons acquire action values in
Metzger, R. Leemans, Agric. Ecosyst. Environ. 107, Science and Technology, 904-2234 Okinawa, Japan. their activity through learning? How is the
101 (2005).
9. We refer to extensification as the transition of a land- *To whom correspondence should be addressed. striatal neuron activity involved in reward-
use type with high intensity of use to a lower intensity E-mail: [email protected] (K.S.); mkimura@ based action selection? Here we show by using
(e.g., improved grassland to seminatural cover). koto.kpu-m.ac.jp (M.K.)
.Present address: Brain Science Research Center,
a reward-based, free-choice paradigm that the
10. M. D. A. Rounsevell, F. Ewert, I. Reginster, R. Leemans,
T. R. Carter, Agric. Ecosyst. Environ. 107, 117 (2005). Tamagawa University Research Institute, 194-8610 striatal neurons learn to encode the action
11. ‘‘Energy for the future: Renewable sources of Tokyo, Japan. values through trial-and-error learning and

www.sciencemag.org SCIENCE VOL 310 25 NOVEMBER 2005 1337