R EPORTS

Forecasting Agriculturally Driven Global Environmental Change David Tilman,1* Joseph Fargione,1 Brian Wolff,1 Carla D’Antonio,2 Andrew Dobson,3 Robert Howarth,4 David Schindler,5 William H. Schlesinger,6 Daniel Simberloff,7 Deborah Swackhamer8

During the next 50 years, which is likely to be the final period of rapid agri- cultural expansion, demand for food by a wealthier and 50% larger global will be a major driver of global environmental change. Should past dependences of the global environmental impacts of agriculture on human population and consumption continue, 109 hectares of natural would be converted to agriculture by 2050. This would be accompanied by 2.4- to 2.7-fold increases in nitrogen- and phosphorus-driven eutrophication of terrestrial, freshwater, and near-shore marine ecosystems, and comparable increases in pesticide use. This eutrophication and destruction would cause unprecedented simplification, loss of ecosystem services, and species extinctions. Significant scientific advances and regulatory, technolog- ical, and policy changes are needed to control the environmental impacts of agricultural expansion.

During the first 35 years of the Green Revolu- ecosystems by the use and release of limiting tion, global grain production doubled, greatly resources that influence ecosystem functioning reducing food shortages, but at high environ- (nitrogen, phosphorus, and water), release of mental cost (1–5). In addition to its effects on pesticides, and conversion of natural ecosys- greenhouse gases (1, 6, 7), agriculture affects tems to agriculture. These sources of global change may rival change in environ- 1Department of , Evolution and Behavior, Uni- mental and societal impacts (2, 8). Population versity of Minnesota, 1987 Upper Buford Circle, St. size and per capita consumption are assumed to Paul, MN 55108, USA. 2Department of Integrative be the two greatest drivers of global environ- Biology, University of California, Berkeley, CA 94720, mental change. Humans currently appropriate USA. 3Department of Ecology & Evolutionary Biology, Princeton University, Princeton, NJ 08540, USA. 4The more than a third of the production of terrestrial Oceans Program, Environmental Defense, and the ecosystems and about half of usable freshwa- Ecosystems Center, Marine Biological Lab, Woods ters, have doubled terrestrial nitrogen supply Hole, MA 02543, USA. 5University of Alberta, Z-811 and phosphorus liberation, have manufactured Biological Sciences Building, Edmonton, Alberta, T6G and released globally significant quantities of 2E9, Canada. 6The Phytotron, Duke University, Durham, NC 27708, USA. 7Department of Ecology & pesticides, and have initiated a major extinction Evolutionary Biology, University of Tennessee, Knox- event (2–4, 8–10). Global population, which ville, TN 37996, USA. 8Environmental and Occupa- increased 3.7-fold during the 20th century, to 6 tional Health, University of Minnesota, 420 Delaware billion people (11), is forecast to increase to 7.5 Street SE, Minneapolis, MN 55455, USA. billion by the year 2020 and to about 9 billion *To whom correspondence should be addressed. E- by 2050 (12). Constant-dollar global per capita mail: [email protected] www.sciencemag.org SCIENCE VOL 292 13 APRIL 2001 281 R EPORTS gross domestic product (GDP) increased 4.6- jectories of the past 35 or more years. Because ognizing that substantial changes in future pop- fold in the 20th century (13) and is projected to these trajectories include in them the impacts of ulation and economic growth, agricultural poli- be 1.3 times current levels by 2020 and 2.4 past technological developments, changes in cies, climate, and other factors would affect our times current levels by 2050 (14, 15). How choices, and environmental regula- results. Detailed regional forecasts and forecasts might projected increases in population and tions, our forecasts implicitly assume similar based on mechanistic models that couple re- wealth influence the global environment? The technological, regulatory, and behavioral gional economies, agriculture, and the environ- prospects of are widely recog- changes in the future. Shifts in these could ment are also needed and would complement nized (16). Here, we explore the nonclimatic cause major deviations from our forecasts. our simpler global approach. global environmental impacts of agricultural We use univariate and multiple regressions Four forecasts were made for each variable: expansion during the coming 20 to 50 years. to forecast future global trends for each of seven by a linear fit to its temporal trend (Fig. 1), We use past global trends and their dependence environmental variables related to agriculture extrapolated to the years 2020 and 2050; by the on global population and GDP to empirically (Table 1). Because of the exponential nature of fitted dependence of each variable on population forecast the potential global environmental im- past global population and economic growth, we size (13, 17, 18), combined with the global pacts of agriculture. Like economic forecasting, had anticipated exponential temporal trends for projected (12) for 2020 and ecological forecasting is notoriously difficult these variables. Surprisingly, each was a linear, 2050; by the linear dependence of each variable and imprecise. Our forecasts are not predic- and almost equally strong, function of time, on GDP (13, 17), combined with global GDP tions, but rather are estimates of environmental population, and GDP (Fig. 1 and Table 1). We projections (14, 15) for 2020 and 2050; and by impacts should agriculture continue on the tra- thus use linear fits in our forecasts, while rec- multiple regression fitting each variable to year,

Fig. 1. (A) Trends in annual rates of application of ni- trogenous fertilizer (N) ex- pressed as mass of N, and of phosphate fertilizer (P) ex- pressed as mass of P2O5, for all nations of the world ex- cept the former USSR (18, 19), and trends in global to- tal area of irrigated crop land (H2O) (18). (B) Trends in global total area of land in pasture or crops (18). (C) Trend in global pesticide production rates, measured as millions of metric tons per year (30). (D) Trend in ex- penditures on pesticide im- ports (18) summed across all nations of the world, trans- formed to constant 1996 U.S. dollars. All trends are as dependent on global popula- tion and GDP as on time (Table 1).

Table 1. Univariate and multivariate forecasts for years 2020 and 2050, based on **P Ͻ 0.0001; *P Ͻ 0.01; NS, P Ͼ 0.05. The value in 2000 is based on temporal trends observed in the past 35 to 40 years and their dependence on population extrapolation from the latest available data, generally 1998. Mean projections are and GDP. Parentheses show R2 values for each regression. Levels of significance: means of the three univariate and the one multivariate projection.

Fertilizer (106 MT) Pesticide Irrigated land Crop land Pasture land Imported (106 ha) Produced (109 ha) (109 ha) NP (109 1996 (106 MT) U.S.$)

Value in 2000 87.0 34.3 280 3.75 11.8 1.54 3.47 Mean projections Forecast 2020 135 47.6 367 6.55 18.5 1.66 3.67 Forecast 2050 236 83.7 529 10.1 32.2 1.89 4.01 Individual projections for 2050 Univariate By year 186 62.0 465 7.33 25.8 1.79 3.90 (0.986**) (0.927**) (0.998**) (0.946*) (0.957**) (0.976**) (0.977**) By population 166 56.2 417 8.02 22.2 1.73 3.79 (0.980**) (0.910**) (0.996**) (0.990*) (0.951**) (0.974**) (0.979**) By GDP 343 98.3 761 18.1 48.8 2.20 4.59 (0.964**) (0.904**) (0.992**) (0.995*) (0.955**) (0.973**) (0.977**) Multivariate 249 118 473 7.06 32.0 1.83 3.75 (0.989**) (0.979**) (0.998**) (0.994NS) (0.960**) (0.977**) (0.982**)

282 13 APRIL 2001 VOL 292 SCIENCE www.sciencemag.org R EPORTS population, and per capita GDP, combined with 26). Eutrophication is the biggest pollution could lead to the loss of about a third of remain- projected values for these in 2020 and 2050. We problem in most coastal waters (23), and, with ing tropical and temperate forests, savannas, and present all four forecasts for 2050 to illustrate overfishing and aquaculture (27), is a major grasslands and of the services, including carbon similarity and variability, and mean forecasts for threat to marine . Agricultural nutri- storage (33), provided by these ecosystems. Ad- 2020 and 2050 (Table 1). The averages for 2020 ent pollution has led to increased blooms of ditional natural habitat would be lost worldwide allow a mid-course evaluation of the 50-year toxic algae in many coastal systems and to the to urban and suburban development, to road- forecasts. large hypoxic (“dead”) zone in the Gulf of Mex- ways, and to the rotation of low-quality lands The doubling of global food production dur- ico (24, 28). In total, projected increases in N through agriculture. Species extinction is an ir- ing the past 35 years was accompanied by large and P fertilization and irrigation would cause reversible impact of habitat destruction. Interac- increases in global nitrogen (N) and phosphorus significant losses of biodiversity, as well as tions between climate change, species invasions, (P) fertilization and irrigation [Fig. 1A and (5)]. marked changes in the composition and func- and could cause further If past trends in N and P fertilization (18, 19) tioning of both terrestrial and aquatic ecosys- diversity losses, because many species may be and irrigation (18) and their dependence on tems (2, 3, 8, 23, 25, 26, 28). unable to migrate through fragmented population and GDP continue, our mean fore- Although society benefits from pesticides, to reach regions with suitable and soils cast is for global N fertilization to be 1.6-fold some cause environmental degradation or affect (34). times present amounts by 2020 and 2.7 times human health (29, 30). Some pesticides, de- Just as demand for energy is the major cause present values by 2050 (Table 1). By 2050, N pending on persistence and volatility, disperse of increasing atmospheric greenhouse gases, de- fertilization alone would annually add 236 ϫ globally (29), bioaccumulate in food chains (31) mand for agricultural products may be the major 106 MT of N to terrestrial ecosystems (20), and have impacts on human health and the driver of future nonclimatic global change. Our compared with 140 ϫ 106 MT from all natural health of other species far from points of release forecasts have high variance, but even the low- sources (2). Individual forecasts for N fertiliza- and many years after release. If past patterns est projections are cause for concern. The pro- tion in 2050 range from a 1.9-fold increase continue, global pesticide production (30), jected 50% increase in global population and based on its dependence on population to a which has increased for 40 years (Fig. 1C), demand for diets richer in meat by a wealthier 3.9-fold increase based on GDP. P fertilization would be 1.7 times that at present by 2020 and world are projected to double global food de- is forecast to be 1.4 times current amounts in 2.7 times the present amount by 2050 (Table 1). mand by 2050 (22), creating an environmental 2020 and 2.4 times current amounts in 2050. P Projections for 2050 range from 1.9- to 4.8-fold challenge that may rival, and significantly inter- estimates for 2050 range from 1.6-fold to 3.4- increases. World trade in pesticides (18), anoth- act with, climatic change. The actual impacts of fold increases (20). Irrigated area (18), a mea- er estimate of trends in pesticide use, would be agricultural expansion will depend on how large sure of agricultural demand for water, is forecast 1.6 times present levels by 2020 and 2.7 times the expansion actually is and on how it is to be 1.3 times the current area in 2020 and 1.9 present levels by 2050 (Fig. 1D and Table 1). achieved. Our projections of global environ- times as great in 2050. Should trends continue, by 2050, humans and mental impacts assume a continuation of past Humans annually already release as much N other organisms in natural and managed ecosys- practices, i.e., mainly of agricultural intensifica- and P to terrestrial ecosystems as all natural tems would be exposed to markedly elevated tion by means of fertilization, irrigation, pesti- sources (2, 3). The large projected increases in levels of pesticides. cide application, and crop breeding. We implic- N, P, and irrigation water [Table 1 and (20)] Land use and habitat conversion are, in es- itly assume that the increasing yields of the would have significant environmental impacts. sence, a zero-sum game: land converted to ag- Green Revolution can continue unabated for 50 Irrigation increases salt and nutrient loading to riculture to meet global food demand comes more years. If this does not occur, perhaps be- downstream aquatic ecosystems, can cause from forests, grasslands, and other natural hab- cause of water shortages, evolution of resistant salinization of soils, and has impacts on streams itats. Increases in agricultural land, a major pests and pathogens, emergence of new pests and rivers because of damming and removal of quantified cause of global habitat destruction, and pathogens, or diminishing returns from fer- water (21). In many areas, there is insufficient are a conservative estimate of losses of native tilization and selection for higher-yielding vari- water for projected demands (21, 22). N and P ecosystems. Global trends for pastureland [Fig. eties (1, 18, 35, 36), the projected food demand leakage from agricultural systems causes major 1B and (18)] suggest a net increase of 2.0 ϫ 108 would be met only if the agricultural land base environmental problems (2, 3, 8, 23). About half hectares of pasture by 2020 and of 5.4 ϫ 108 increased more than we have projected, i.e., by of fertilizer N and P is captured in harvested hectares by 2050 (Table 1). If past trends (Fig. an extensification of agriculture. Alternatively, crops (23–25) and, after consumption, enters 1B) continue, global cropland (18) would in- food demand could be lowered if the trend human and livestock waste streams. About 70% crease by a net of 1.2 ϫ 108 hectares by 2020 toward diets richer in meat were reversed or if of harvested crops are fed to livestock in devel- and of 3.5 ϫ 108 hectares by 2050 (Table 1). global population stabilized at a lower than pro- oped countries (23), but few livestock wastes are The combined total represents an average global jected level. treated for N and P removal. Thus, much N and agricultural land base in 2050 that would be The Green Revolution greatly reduced world P from fertilizer and animal wastes enters sur- 18% larger than at present. These are net global hunger. Comparable advances in agricultural face and groundwater (3, 25), and N also is changes. Because analyses like those of Table 1, production are needed during the coming 50 volatilized to the atmosphere as ammonia and but for developed countries, project a net with- years to assure a sufficient, secure, and equitable deposited regionally (23–25). drawal of 1.4 ϫ 108 ha of land from agriculture global food supply (1), but these advances must The major environmental consequence of P by 2050, the net loss of natural ecosystems to follow new trajectories if the problems we have addition is eutrophication of surface waters, par- cropland and pasture in developing countries by identified are to be minimized. An environmen- ticularly freshwater lakes and streams (3). For 2050 would be 109 ha, about half of all poten- tally sustainable revolution (1), a greener revo- N, consequences include eutrophication of estu- tially suitable remaining land (22, 32). lution, is needed. It must be based on the total aries and coastal seas, loss of biodiversity and The conversion of 109 hectares of land to costs and benefits of agriculture, including agri- changes in species compositions in terrestrial agriculture would represent the worldwide loss culture-dependent gains and losses in values of and aquatic ecosystems, groundwater pollution of natural ecosystems larger than the United such ecosystem goods and services as potable with nitrate and nitrite, increases in the green- States. Because of regional availabilities of suit- water, biodiversity, carbon storage, pest control, house gas N2O, increases in NOx and resulting able land, this expansion of agricultural land is pollination, fisheries, and recreation (37, 38). tropospheric smog and ozone, and acidification expected to occur predominately in Latin Amer- Existing knowledge, if widely used, could of soils and sensitive freshwaters (2, 8, 23, 25, ica and sub-Saharan central Africa (1, 22). It significantly reduce the environmental impacts

www.sciencemag.org SCIENCE VOL 292 13 APRIL 2001 283 R EPORTS

of agriculture and increase . Inte- 35). Methods are needed to efficiently close the growth of 2.7% yearϪ1 from 1995 to 2050. We used a grated pest management, application of site- from soil to crop to livestock and more conservative 2.5% based on analysis of the 40- year trend (13) in the annual rate of growth of constant- and time-appropriate amounts of agricultural back to agricultural soil, and to prevent the dollar global per capita GDP. chemicals and water, use of cover crops on occurrence and the spread to humans of live- 16. J. T. Houghton et al., Climate Change 1995: The fallow lands and buffer strips between cultivat- stock pathogens. Ways to better control crop Science of Climate Change (Cambridge Univ. Press, Cambridge, 1996). ed fields and drainage areas, and appropriate pathogens and pests are needed, such as by 17. World Development Indicators, CD-ROM (World deployment of more productive crops can in- greater use of natural enemies, crop diversity Bank, Washington, DC, 2000) crease yields while reducing water, fertilizer, (40), and biotechnology, if deployed so as to 18. Food and Agriculture Organization of the United Nations, FAOSTAT homepage [on-line] (2001). Avail- and pesticide use and movement to nonagricul- reduce evolution of pest resistance. Methods to able at: http://apps.fao.org/. Cropland refers to the tural habitats (6, 7, 21, 23, 35, 39–41). Treat- forecast quantitatively the impact on ecosystem FAO land-use category of “arable and permanent ment of animal wastes is necessary, especially functioning of loss of habitat, loss of biodiver- crops.” Pasture is FAO’s “permanent pasture.” in developed countries, where more than a third sity, changes in species composition, and in- 19. Fertilizer projections are for the entire world except the former USSR, whose economic problems added of fertilizer N passes through livestock (23). creased nutrient inputs need development. Be- variance to the relationship. Currently, animal wastes receive little or no cause most agricultural expansion will occur in 20. If the former USSR is included, the 2050 global fertilizer projection for N is 270 ϫ 106 MT yearϪ1 treatment and are a major source of surface developing countries, the discovery and adop- Ϫ and for P is 110 ϫ 106 MT year 1. If N fixation by water pollution and terrestrial N deposition (23, tion of appropriate practices likely would re- combustion of fossil fuels and by crops doubled by 28). Preservation and restoration of wetlands quire aid from developed countries, including 2050, these plus N fertilization (including the USSR) Ϫ and riparian zones can remove N by denitrifi- International Monetary Fund and World Bank would add 390 ϫ 106 MT year 1, 2.8 times more N than all natural processes combined. cation before it reaches watercourses and can loans, or debt forgiveness. Moreover, regional 21. S. L. Postel, Pillar of Sand: Can the Irrigation Miracle trap P in soils. differences in food demand and in the potential Last? (Norton, New York, 1999). Comprehensive land-use planning could of extensification versus intensification to meet 22. N. Alexandratos, Proc. Natl. Acad. Sci. U.S.A. 96, 5908 (1999). mitigate some effects of agricultural expansion. these needs (21, 22, 32, 35, 44) means that, 23. National Research Council, Clean Coastal Waters: Under- Some agricultural impacts could be ameliorated although the problems are global, solutions must standing and Reducing the Effects of Nutrient Pollution if the 1.4 ϫ 108 hectares projected for removal be local, regional, and global. (National Academy Press, Washington, DC, 2000). 24. A. F. Bouwman et al., Global Biogeochem. Cycles 11, from agriculture in developed nations were re- If global population stabilizes at 8.5 to 10 561 (1997). stored to provide ecosystem services (37), such billion people, the next 50 years may be the final 25. R. W. Howarth et al., Biogeochemistry 35, 75 (1996). as carbon storage, preservation of biodiversity, episode of rapid global agricultural expansion. 26. E. A. Holland, F. Dentener, B. Braswell, J. Sulzman, and production of potable water. Alternatively, During this period, agriculture has the potential Biogeochemistry 46, 7 (1999). 27. R. L. Naylor et al., Nature 405, 1017 (2000). if kept in agriculture, this land could save a to have massive, irreversible environmental im- 28. J. A. Downing et al., Gulf of Mexico Hypoxia: Land and comparable area of natural ecosystems in devel- pacts. The minimization of these impacts, while Sea Interactions (Task Force Report No. 134, Council oping nations from destruction if food so pro- providing sufficient and equitably distributed for Agricultural Science and Technology, Ames, IA, 1999). duced could meet demands of developing na- food, will be a great challenge. Although there 29. Assessment Report: Arctic Pollution Issues (Arctic tions. The capability of the remaining natural are likely to be mechanisms and policies that Monitoring and Assessment Program, Oslo, Norway, lands to supply ecosystem services and to pre- can reduce, or perhaps reverse, many of the 1998). 30. World Health Organization, Public Health Impacts of serve biodiversity could be increased by plan- trends that we have identified, these solu- Pesticides Used in Agriculture (WHO & UN Environ- ning the pattern and location of agricultural tions will not be achieved unless far more re- ment Program, Geneva, 1990). development so as to save biodiversity hot sources are dedicated to their discovery and 31. K. A. Kidd, D. W. Schindler, D. C. G. Muir, W. L. Lockhart, R. H. Hesslein, Science 269, 240 (1995). spots; to minimize fragmentation; to maximize implementation. 32. N. Alexandratos, World Agriculture: Toward 2010 the range of ecosystem types preserved; and to (Wiley, Chichester, UK, 1995). preserve wetlands and riparian zones that pro- References and Notes 33. W. H. Schlesinger, Biogeochemistry: An Analysis of Glob- tect surface waters from inputs of nutrients, 1. G. Conway, The Doubly Green Revolution (Penguin al Change (Academic Press, San Diego, CA, 1997). Books, London, 1997). 34. O. E. Sala et al., Science 287, 1770 (2000). pesticides, eroded soil and pathogens. Such ac- 2. P. M. Vitousek, H. A. Mooney, J. Lubchenco, J. M. 35. K. G. Cassman, Proc. Natl. Acad. Sci. U.S.A. 96, 5952 tions would continue a global trend of setting Melillo, Science 277, 494 (1997). (1999). land aside as nature reserves and national parks 3. S. R. Carpenter et al., Ecol. Appl. 8, 559 (1998). 36. V. W. Ruttan, Proc. Natl. Acad. Sci. U.S.A. 96, 5960 (1999). 4. D. W. Schindler, Ambio 28, 350 (1999). 37. G. C. Daily, Nature’s Services: Societal Dependence on (42). Cumulatively worldwide, an area roughly 5. D. Tilman, Proc. Natl. Acad. Sci. U.S.A. 96, 5995 (1999). Natural Ecosystems (Island Press, Washington, DC, the size of the Indian subcontinent is designated 6. P. A. Matson, W. J. Parton, A. G. Power, M. J. Swift, 1997). for conservation of biodiversity. Many pre- Science 277, 504 (1997). 38. G. C. Daily et al., Science 289, 395 (2000). 7. P. A. Matson, W. McDowell, A. Townsend, P. Vitousek, 39. J. Walsh, Preserving the Options: Food Production and serves, though, are inadequately protected, and Biogeochemistry 46, 67 (1999). Sustainability (CGIAR, Washington, DC, 1991). some may be sustainably protected only if in- 8. National Research Council, Global Change Ecosys- 40. Y. Zhu et al., Nature 406, 718 (2000). corporated into local economies (43). tems Research (National Academy Press, Washington, 41. P. L. Woomer, M. J. Swift, Eds., The Biological Man- DC, 2000). agement of Tropical Soil Fertility (Wiley, Chichester, Even the best available technologies, fully 9. S. L. Postel, G. C. Daily, P. R. Ehrlich, Science 271, 785 UK, 1994). deployed, cannot prevent many of the forecasted (1996). 42. M. J. B. Green, J. Paine, State of the World’s Protected problems. Major international programs are 10. S. L. Pimm, G. J. Russell, J. L. Gittleman, T. M. Brooks, Areas at the End of the Twentieth Century, IUCN (The needed to develop new technologies and poli- Science 269, 347 (1995). Conservation Union) Protected Areas Symposium, Al- 11. J. E. Cohen, How Many People Can the Earth Support? bany, Western Australia, 23 to 29 November 1997 cies for ecologically sustainable agriculture. Re- (Norton, New York, 1995). (World Council on Protected Areas, Gland, Switzerland, gion-appropriate education, incentives, and le- 12. World Population Prospects: The 1998 Revision, vol. I, 1998). Available at www.wcmc.org.uk/protected_areas/ gal restrictions will be required to encourage Comprehensive Tables (UN Department of Economic albany.pdf and Social Affairs, Population Division, New York, 43. D. Janzen, Proc. Natl. Acad. Sci. U.S.A. 96, 5987 (1999). adoption. The research needs are diverse. We 1999). 44. J. Groot, J. R. Penning, F. W. T. De Vries, P. W. J. must seek, by breeding and biotechnology, 13. A. Maddison, Monitoring the World Economy 1820– Uithol, Nutr. Cycl. Agroecosys. 50, 181 (1998). gains in the fundamental efficiency of crop N, P, 1992 (Development Center of the Organization for 45. We thank the National Center for Ecological Analysis Economic Cooperation and Development, Paris, 1995). and Synthesis for support; N. Larson and L. Johnson for and water use (21, 35, 36). Advances in preci- GDP was extended to 1998 using annual percent in- assistance; R. Cook for advice on statistical methods; and sion agriculture that decrease N and P inputs are creases in (17). V. Ruttan, S. Hobbie, S. Polasky, J. Reichman, G. Daily, D. needed, as are methods that manage soil organic 14. National Research Council, Our Common Journey: A Wilcove, J. Lubchenco, D. Janzen, J. Clark, J. Cohen, S. Transition Toward Sustainability (National Academy Carpenter, P. Vitousek, and E. Sandlin for comments. matter and microbial communities to reduce Press, Washington, DC, 1999). nutrient leaching and to optimize soil fertility (6, 15. Reference (14) forecasts constant-dollar global GDP 22 January 2001; accepted 8 March 2001

284 13 APRIL 2001 VOL 292 SCIENCE www.sciencemag.org