Proc. Natl. Acad. Sci.

USA
Vol. 96, pp. 5995–6000, May 1999
Colloquium Paper

This paper was presented at the National Academy of Sciences colloquium ‘‘Plants and Population: Is There Time?’’
held December 5–6, 1998, at the Arnold and Mabel Beckman Center in Irvine, CA.

Global environmental impacts of agricultural expansion: The need

for sustainable and efficient practices
DAVID TILMAN
Department of Ecology, Evolution, and Behavior, University of Minnesota, 1987 Upper Buford Circle, St. Paul, MN 55108

ABSTRACT The recent intensification of agriculture, and easy answers are uncovered. Rather, a new long-term, multi-
the prospects of future intensification, will have major detri- disciplinary research program is needed to develop agricul-
mental impacts on the nonagricultural terrestrial and aquatic tural methods that can feed a growing world and still preserve
ecosystems of the world. The doubling of agricultural food the vital services provided to humanity by the world’s natural
production during the past 35 years was associated with a ecosystems.
6.87-fold increase in nitrogen fertilization, a 3.48-fold increase Current agricultural practices involve deliberately maintain-
in phosphorus fertilization, a 1.68-fold increase in the amount ing ecosystems in a highly simplified, disturbed, and nutrient-
of irrigated cropland, and a 1.1-fold increase in land in rich state. To maximize crop yields, crop plant varieties are
cultivation. Based on a simple linear extension of past trends, carefully selected to match local growing conditions. Limiting
the anticipated next doubling of global food production would factors, especially water, mineral nitrogen, and mineral phos-
be associated with approximately 3-fold increases in nitrogen phate, are supplied in excess, and pests are actively controlled.
and phosphorus fertilization rates, a doubling of the irrigated These three features of modern agriculture—control of crops
land area, and an 18% increase in cropland. These projected and their genetics, of soil fertility via chemical fertilization and
changes would have dramatic impacts on the diversity, com- irrigation, and of pests (weeds, insects, and pathogens) via
position, and functioning of the remaining natural ecosystems chemical pesticides—are the hallmarks of the green revolution.
of the world, and on their ability to provide society with a They have caused four once-rare plants (barley, maize, rice,
variety of essential ecosystem services. The largest impacts and wheat) to become the dominant plants on earth as humans
would be on freshwater and marine ecosystems, which would became the dominant animal. Indeed, these four annual
be greatly eutrophied by high rates of nitrogen and phospho- grasses now occupy, respectively, 67 million hectares, 140
rus release from agricultural fields. Aquatic nutrient eu- million hectares, 151 million hectares, and 230 million hect-
trophication can lead to loss of biodiversity, outbreaks of ares, each, worldwide, which is 39.8% of global cropland. For
nuisance species, shifts in the structure of food chains, and comparison, the total forested area of the United States,
impairment of fisheries. Because of aerial redistribution of including Alaska, is 298 million hectares. Entire regions of the
various forms of nitrogen, agricultural intensification also world now are dominated by virtual monocultures of a given
would eutrophy many natural terrestrial ecosystems and crop. These monocultures have replaced natural ecosystems
contribute to atmospheric accumulation of greenhouse gases. that once contained hundreds to even thousands of plant
These detrimental environmental impacts of agriculture can species, thousands of insect species, and many species of
be minimized only if there is much more efficient use and vertebrates. Thus, agriculture has caused a significant simpli-
recycling of nitrogen and phosphorus in agroecosystems. fication and homogenization of the world’s ecosystems.
It is as difficult to predict the future of agriculture now as it
would have been to anticipate, in 1950, the successes and
The agricultural achievements of the past 35 years have been
impacts of the green revolution. However, some insights may
impressive. Grain production, mainly from wheat, rice, and
be provided by an analysis of the broad trends that occurred
maize, has increased at a rate greater than human population.
during the recent doubling of global food production. These
This has decreased the number of malnourished people even
trends may give some insight into the global environmental
as the earth’s human population doubled to 5.8 billion. Al-
impacts that the anticipated second doubling of agricultural
though the estimates vary widely, world population is projected
productivity may have. Next, I consider insights that ecology
to increase about 75% before leveling off at about 10 billion.
may offer into the sustainability and stability of agricultural
In combination with increasing demand for meat in developing
ecosystems. Finally, I pose the major environmental challenges
countries and the use of grains as livestock feed, this increased
that face humanity as global human population and demand
population density should cause world demand for grain
for food continues to increase.
production to more than double. This raises several important
questions. If it is possible for world food production to double, The Ecology of Doubling Crop Production
again, within the next four or five decades, what impacts would
this doubling have on the functioning of the nonagricultural The Food and Agriculture Organization (FAO) database (1)
ecosystems of the world, and on the services they provide to provides a wealth of information on agricultural activities for
humanity? What routes might be used to decrease such individual nations, regions, and the world from 1961 to the
impacts? I explore these questions first by asking what the present. Using the FAO data, let’s look at the pattern of world
global ecological impacts of ‘‘more of the same’’ agriculture food production during this period and the factors that allowed
might be, and then by considering practices that might de- it to almost double. The majority of the food crops grown on
crease such impacts. In particular, insights are sought in the the arable lands of the earth are cereals (barley, maize, rice,
parallels between natural and agricultural ecosystems, but no and wheat), coarse grains, and root crops. For convenience, I

PNAS is available online at www.pnas.org. Abbreviation: FAO, Food and Agriculture Organization.

5995
5996 Colloquium Paper: Tilman Proc. Natl. Acad. Sci. USA 96 (1999)

will call the sum of these world food production. In 1996,

cereals comprised 57% of this total, coarse grains 25%, and
root crops 18%. By using this measure, world food production,
as estimated from the FAO database (1), almost doubled
(increased 1.97-fold) from 1961 to 1996 (Fig. 1). Comparable
patterns, and comparable ecological implications, occur if just
cereal production was considered, or if production for just
Europe and the United States, for which better data are
available, was considered.
Many factors contributed to the recent doubling of world
food production. The development of higher-yielding strains of
crops and better agricultural practices were important, as were
increased use of herbicides for weed control and insecticides
and fungicides for pest control. In addition, there were marked
increases in the amounts of nitrogen and phosphorus fertilizers
applied each year worldwide, in the proportion of arable land
that was irrigated, and in the total amount of land that was
cultivated annually worldwide (Fig. 2). It was the combined
effects of all of these factors, and more, that allowed world
food production to double in 35 years.
The FAO data (1) show that this recent doubling of world
food production was accompanied by 6.87- and 3.48-fold
increases in the global annual rate of nitrogen and phosphorus
fertilization, respectively, by a doubling in the amount of land
that was irrigated, and by a 10% increase in the amount of land
in cultivation (Fig. 2). What might be required to allow food
production to double again? A simple, naive and optimistic
scenario might assume that, during the next four decades, all
of the relationships of Fig. 2 would remain linear and gains in
crop genetics, weed and pest control, and cultivation practices
would continue at their previous pace. The assumption of
linearity can be used to predict the rates of nitrogen and
phosphorus fertilization and irrigation, and the increase in
amount of cultivated land needed to double food production.
Even this scenario, though, would require, based on the linear
regression of Fig. 2 A, that the global rate of application of
nitrogen fertilizer increase from about 75 3 106 metric tons per
year to 235 3 106 metric tons per year. Nitrogen fixed by
legume crops also would need to more than triple. Comparable
calculations, based on the regression of Fig. 2B, predict that
the global annual rate of application of phosphorus fertilizer
would have to increase from about 37 3 106 metric tons per
year to 94 3 106 metric tons per year for food production to
double. Similarly, the worldwide proportion of arable lands
that are irrigated would have to increase from the current 17%
to about 32% (based on extrapolation of Fig. 2C), and the total
amount of land in cultivation would have to increase from
about 1.47 3 109 hectares to 1.73 3 109 hectares (extrapolation
of Fig. 2D). These changes represent a worldwide tripling of

FIG. 2. The relationship between annual global food production

(cereals 1 coarse grains 1 root crops) and agricultural inputs, based
on FAO data (1). (A) Global annual nitrogen fertilization rate. (B)
Global annual phosphorus fertilization rate. (C) Proportion of arable
lands that are irrigated. (D) Total land surface in agricultural crop
production.

the annual rates of N and P fertilization, a doubling of the

FIG. 1. Based on FAO data (1), world food production, measured
extent of irrigation, and an 18% increase in the amount of land
as the sum of cereals, coarse grains and root crops, almost doubled farmed.
from 1961 to 1996. A linear regression, and 95% and 99% confidence Such linear projections of yields may be overly optimistic for
intervals for the regression, are shown. a variety of reasons. First, the yield of a crop is a saturating
 Colloquium Paper: Tilman Proc. Natl. Acad. Sci. USA 96 (1999) 5997

function of the rate of supply of its limiting resource. Adding between the amounts of nitrogen in the major rivers of the
fertilizer to already well-fertilized areas, such as productive world and the magnitude of agricultural nitrogen inputs to
croplands in the developed nations that produce the majority their watersheds (6).
of the world’s cereals, will have little impact on yield. Signif- The long-term ecological impacts of increased rates of
icant regional gains, though, can be achieved in many devel- agricultural nitrogen and phosphorus input will depend on the
oping countries by appropriately fertilizing croplands not levels to which these nutrients accumulate in various nonag-
currently receiving fertilizer. Similar saturating yield curves ricultural ecosystems. These levels are uncertain because of the
occur for phosphorus and irrigation. In total, such saturating complexities of the global biogeochemistry of nitrogen and
curves imply that it may be difficult to increase yields at a pace phosphorus. These nutrients accumulate in a variety of forms
similar to that of the past four decades. Second, the easiest and in many different sinks (arable soil organic matter, ground-
greatest gains from crop breeding programs may have already water, freshwater and marine ecosystems and their sediments,
occurred (2). Annual gains in yields from breeding programs
nonagricultural ecosystems, atmospheric nitrous oxide) after
are decreasing, and the research costs associated with these
agricultural application, but the eventual sizes of these pools
gains are escalating (2). This is not surprising. Given a fixed
gene pool, the responses to a given selective regime are most will depend on biologically and physically driven rates of
rapid initially and increasingly slower through time. Such yield transfer in and out of these pools (5). For agricultural nitrogen,
gains represent genetic movement on the morphological and one critical step will be the rate and location of denitrification,
physiology tradeoff surface on which plant species have dif- especially complete denitrification to N2. The transport of
ferentiated and evolved. The closer a given variety is to the phosphorus to nonagricultural ecosystems especially will de-
optimum point on this tradeoff surface, the less will be the gain pend on erosion and surface flow. As emphasized by Socolow
from further selection. Once most of the original genetic (7), a scientific effort comparable to that on the global carbon
variance preserved within crop landraces and remaining wild cycle will be needed to understand the impacts on global
relatives has been exploited, future breeding-based yield gains biogeochemistry of elevated rates of agricultural nitrogen and
are likely to be small or difficult to obtain. Marked yield gains phosphorus application.
from crop breeding then would require that plants overcome Nitrogen and phosphorus are the two most important
major morphological and physiological constraints that no limiting nutrients of terrestrial, freshwater, and marine eco-
organisms have overcome during hundreds of millions of years systems (3, 8–11). The impacts of elevated levels of a major
of evolution. Organisms that greatly overcame such barriers, limiting nutrient are well documented. Nutrient addition
perhaps through gene transfers, would be supercompetitive causes dominance by a few, often formerly rare plant and
species that could potentially invade into and change the animal species, and the loss of species diversity (e.g. refs. 3, 9,
structure of nonagricultural ecosystems (3). 12–15). Both effects are approximately proportional to the
These concerns lead me to wonder if global food production cumulative magnitude of nutrient addition. High rates of
can be doubled by a continuation of past practices. The other nitrogen deposition caused by intensive, nitrogen-rich agricul-
route for a major increase in food production is a marked ture in the Netherlands were a major cause of the conversion
increase in arable land, which the FAO data suggest has played
of species-rich native heathlands into monoculture grasslands
a modest role in the past 30 years. Because the best land
and then forest (16). At high rates of nutrient addition,
already has been cultivated, the amount of land dedicated to
nuisance plant species often dominate both terrestrial and
agriculture may have to increase disproportionately to the gain
in global food production. aquatic ecosystems. For instance, high rates of nitrogen addi-
tion cause prairie grasslands to become virtual monocultures
Ecological Impacts of Doubling Global Food Production of an otherwise extremely rare nonnative agricultural weed
(17). Bluegreen algal species, some toxic, often dominate
If these simple extrapolations of past practices are any indi- lakes, rivers and streams that receive high rates of P and N
cation, doubling global food production will triple the annual loading. Similarly, blooms of toxic red algae and of pathogenic
rates of nitrogen and phosphorus release to the globe. Current taxa such as Pfisteria occur in nutrient-polluted marine habi-
rates of agricultural nitrogen production, via both production tats. Anoxic conditions associated with high rates of phospho-
of fertilizer and cultivation of legume crops, already approx- rus and nitrogen loading cause fish die-offs in both freshwater
imately equal the natural (preindustrial) rate of addition of and marine ecosystems (18).
biologically active nitrogen to the globe (4). Point-source Unless its efficiency is increased, a doubling of irrigation
releases of phosphorus are tightly regulated in developed would pose additional environmental problems. Humans al-
nations because phosphorus is a major limiting nutrient in ready impact a large portion of the terrestrial hydrologic cycle
aquatic ecosystems and increases in its supply rate harm water (19). Additional irrigation would divert more water from
quality and aquatic foodweb structure. A tripling of global aquatic ecosystems and impact groundwaters and surface
phosphorus supply rates is likely to adversely impact many waters via additional leaching of agrochemicals.
aquatic ecosystems, especially those that have significant in- A conservative estimate, based on the assumption that
puts of eroded agricultural soils or phosphorus-rich wastes future yield gains can match those of the past 35 years, is that
from livestock and poultry. Nitrogen is much more motile in
doubling global food production will require 18% more arable
soil than phosphorus because soil bacteria can convert ammo-
land. Even this 18% increase would require the loss of 268
nia to nitrate and nitrite, which are readily leached from soil
million hectares of nonagricultural ecosystems worldwide,
(5). Denitrification by bacteria also can convert nitrate into
nitrous oxide, a potent greenhouse gas. In addition, ammonia, comparable in size to cultivating all of the currently forested
which is both directly applied as fertilizer and created via land of the United States. A doubling of food production may
bacterial degradation of animal waste and other organic require a much greater increase in land dedicated to agricul-
compounds, is highly volatile. It is transported via air and ture. The resulting ecosystem destruction would vastly increase
deposited on other ecosystems with precipitation. These nu- the proportion of the world’s species threatened with extinc-
merous modes of transport mean that agricultural nitrogen, tion. It also would cause a massive release of CO2 from land
less than half of which stays in a field or is harvested with a clearing and tilling (5). Because high-diversity ecosystems
crop, impacts both terrestrial and aquatic ecosystems as a generally occur on infertile soils (3, 9, 20), the conversion of
eutrophier, and impacts global climate because of is role as a less-fertile ecosystems to agriculture would disproportionately
greenhouse gas. Indeed, there is a direct and quantitative link impact world biodiversity.
5998 Colloquium Paper: Tilman Proc. Natl. Acad. Sci. USA 96 (1999)

Agriculture and the Loss of Ecosystem Services livestock factories or to heavily fertilized fields, even though
these are now major sources of nutrient loading to many
A doubling of global food production would have major aquatic ecosystems (18). The development of more nutrient-
impacts on the ability of nonagricultural ecosystems to provide efficient crops also could have major environmental benefits.
services (21) vital to humanity. Existing nonagricultural eco- If crops could be bred to consume a larger proportion of soil
systems provide, at no cost, pure, drinkable water. In contrast, nitrate and ammonium, this would decrease the amount of
the groundwater associated with intensive agricultural ecosys- unconsumed soil nitrate and ammonium that would be lost via
tems often contains sufficiently high concentrations of nitrite leaching and volatilization. This would decrease impacts on
and nitrates or of pesticides and their residues as to be unfit for off-site ecosystems. Breeding programs that increased crop
human consumption. Expensive treatment is required to make yields would decrease some of the future impacts of agriculture
it potable. The biodiversity of nonagroecosystems provides by decreasing the amount of additional land that would have
many services to agriculture. For instance, the genetic diversity to be brought into agricultural production.
of both wild relatives of crop plants and unrelated organisms The ecosystems of the world now are dominated by humans
is used to increase yields and to reduce impacts of agricultural (25). The implications of human domination, including im-
pests and pathogens. However, the maintenance of the wild pacts from expanding agricultural activities, must be better
biodiversity needed for future development of crops and understood and incorporated into policy. This will require an
medicines occurs mainly in nonagricultural ecosystems, the on-going, iterative process in which science and policy regu-
very ecosystems threatened by agricultural expansion and lating agricultural practices advance hand-in-hand, much as is
nutrient release. Agriculture depends on soil fertility, fertility being done for the climate issue by the Intergovernmental
created by the ecosystems destroyed when lands are converted Panel on Climate Change. This will require predictive, mech-
to agriculture. Especially on sandy soils, the best way to regain anistic models of the impacts of agriculture on nonagricultural
soil fertility lost because of tilling is to allow re-establishment ecosystems.
of the native ecosystems. Many agricultural crops depend on
the pollination services provided by insects, birds, or mammals Ecological Insights into Agricultural Impacts and
that live in nearby nonagricultural ecosystems (18). Similarly, Sustainability
agricultural crops benefit from biocontrol agents, such as
parasitic and predatory insects, birds, and bats, that live in What might be done to decrease the environmental impacts of
neighboring nonagricultural ecosystems and that decrease agriculture while maintaining or improving its productivity,
outbreaks of agricultural pests. Nonagricultural ecosystems, stability, or sustainability? This major challenge will have no
such as forests on slopes and wetlands, help meter the release single, easy solution. Partial answers will come from increases
of water into streams and rivers, and thus help in flood control. in the precision and efficiency of nutrient and pesticide use,
If properly managed, natural ecosystems also can produce a from advances in crop genetics including advances from bio-
sustainable supply of goods used by society, including timber technology, and from a variety of engineering solutions. Some
and fiber, fish, and game. additional insights may come from a consideration of the
This brief overview of ecosystem services (21) demonstrates principles that govern the functioning of all ecosystems, in-
that society, and agriculture, depend on many services pro- cluding agroecosystems. Ecosystem functioning is known to
vided by nonagricultural ecosystems. Although it is difficult to depend on the traits of the species ecosystem’s contain (their
establish economic values for such services (22), it is clear that, composition), the number of species they contain (their species
when possible, technological substitutes for lost ecosystem diversity), and the physical conditions they experience, espe-
services can be extremely expensive. This highlights the need cially disturbance regimes. A consideration of the principles
for public policy to consider the short-term and long-term governing the impacts of composition, diversity, and distur-
costs of actions that decrease the ability of nonagricultural bance on ecosystems may suggest ways to decrease impacts of
ecosystems to provide vital ecosystem services to society. agriculture or to make it more productive, stable, or sustain-
able. It is critical to realize that these principles apply within
More of the Same Will Not Work a given ecosystem type. They describe differences in function-
ing of otherwise identical ecosystems that share the same
The global agricultural enterprise is passing a threshold. It has species pool and differ only in which and how many species
gone from being a minor source of off-site environmental they contain. These principles were not derived from, and do
degradation 35 years ago to becoming the major source of not apply to, comparisons among different ecosystem types,
nitrogen and phosphorus loading to terrestrial, freshwater, and such as cattail swamps versus prairies, or mangrove versus
marine ecosystems. If this loading increases as projected here, upland forest, or tropical versus temperate forests.
agriculture will adversely transform most of the remaining A fundamental principle of epidemiology and ecology is that
natural, nonagricultural ecosystems of the world. Because the the severity and extent of a disease or pest outbreak depends
global environmental impact of agriculture on natural ecosys- on the density of the host population. At low host population
tems and the services they provide may be as serious a problem densities, there is a low chance of contagious spread. However,
as global climate change, the impacts of agriculture merit more at high host densities, a disease or pest can spread epidemically
study. throughout the population. An unavoidable effect of high
A ‘‘more of the same’’ approach to the doubling of agricul- diversity is that most species have lower densities than in low
tural production will have significant environmental costs, diversity communities. For instance, on average a species is
costs that could be lowered by processes that increase the about one-fourth as abundant in a four-species community as
efficiency of fertilizer use, such as precision agriculture (23) in monoculture. This simple effect caused a variety of plant leaf
and by incentives for their use. Methods that increase the fungal diseases to have lower rates of occurrence at higher
nutrient efficiency of the overall agricultural production pro- plant diversity in a field experiment.
cess also are needed. For instance, wastes from large-scale Agriculture has transformed once-rare plants into some of
animal operations are rich in N and P. Unless properly recycled the most abundant species on earth. Maize, which once
into arable fields, or subjected to tertiary sewage treatment to occurred in scattered multispecies mixtures on nutrient-poor
remove nitrogen and phosphorus, such wastes can be a major or disturbed soils, now covers 140 million hectares of the earth.
source of N and P loading to nonagricultural ecosystems (24). Potential pathogens and pests that never had encountered
However, the regulations that apply to municipal sewage and maize now do so frequently. Pests and pathogens that formerly
factory effluents often have not been applied to large-scale could not have maintained populations on maize now encoun-
 Colloquium Paper: Tilman Proc. Natl. Acad. Sci. USA 96 (1999) 5999

ter hosts growing at much greater local and regional densities managed to optimize yield in the face of constraints on nutrient
and with higher tissue nutritional levels. Just as humans have release to the environment.
accumulated diseases as densities increased during the past Recent theory has predicted (35) and recent field experi-
2,000 years (26), so, too, will major crops continue to accu- ments have shown (36, 37) that the rates of loss of limiting
mulate diseases and pests. Southern corn blight is one such nutrients from terrestrial ecosystems are lower at higher plant
disease. A strain of western corn rootworm that is newly diversity, and are equally impacted by species composition.
adapted to living on both corn and soybeans is an emerging Cultivation has major effects on soil fertility. Within the first
pest. Wheat head rust is another disease. The latter virtually 50 years of tilling, 40–70% of the original store of soil organic
eliminated wheat as a major rotation crop from Indiana and matter (carbon and nitrogen) is lost (42). For porous sandy
Illinois in the 1920s and now is doing so in western Minnesota soils, which start with relatively low organic matter and
and eastern North and South Dakota. Plant diseases and pests nitrogen, the loss of fertility during farming can be so great that
can have devastating impacts. The American chestnut, once a the soils cannot be sustainably farmed. Recovery of soil C and
dominant tree of eastern U.S. forests, and the American elm N should be more rapid if abandoned fields are planted with
both were virtually eliminated after pathogens, to which no a high-diversity mixture of appropriate plant species. On the
known resistance occurs, invaded North America. Similarly, sandy soils of my research site in central Minnesota, native
novel pests or pathogens or strains of pathogens could either warm-season prairie grasses and legumes, combined, signifi-
greatly reduce the area in which wheat, rice, or maize can be cantly increase the rate of recovery of soil fertility after
grown or, perhaps, eliminate these as viable crops. A major agriculture (43). Programs designed to restore soil fertility,
protection against these possibilities is diversity—the diversity such as land set-aside programs, may be more successful if such
of crops deployed in a region, the diversity of substitute crops, lands are planted to high-diversity mixtures of appropriate
and the diversity of genetic resistances within crops. species.
All else being equal, the stability of the total rate of plant Finally, in higher diversity ecosystems, there is more
production in an ecosystem depends on both the species complete use of limiting resources (36, 37, 41). The resulting
diversity of the plant community and its species composition lower concentrations of unconsumed soil nutrients decreases
(e.g. refs. 15 and 27–29). The stability of primary productivity the number of other species that invade an ecosystem (32).
is greater for ecosystems containing greater plant diversity (15, Weeds are a major pest of agriculture. In North America,
28). This results from three underlying processes. First, the most weedy species are non-native annuals introduced from
same statistical averaging process that causes more diverse Europe or Asia. The ability of newly introduced weeds to
portfolios of stocks to be more stable than less diverse port- spread across a landscape will depend on the spatial pattern
folios applies to ecosystems (30–32). Second, interspecific of agricultural and native high-diversity ecosystems. Land-
competition causes negative covariances in the abundances of scapes with an appropriate balance of agricultural and
species, and such compensatory effects can act to more greatly natural ecosystems may be more resistant to invasion by new
stabilize more diverse ecosystems (15, 32). Third, the increase weedy species.
in ecosystem productivity that occurs as diversity increases,
termed overyielding, also tends to stabilize primary produc- Conclusions
tivity at higher diversity (32). The greater stability of more
diverse ecosystems means that diversity has an insurance value A hallmark of modern agriculture is its use of monocultures
by minimizing year-to-year variance in yields. Greater stability grown on fertilized soils. Ecological principles suggest that
of agricultural yields might be attained by growing, as a single such monocultures will be relatively unstable, will have high
crop, a mixture of appropriately chosen genotypes of a given leaching loss of nutrients, will be susceptible to invasion by
species, such as a mixture of high-yielding hybrid varieties. weedy species, and will have high incidences of diseases and
The plant species diversity of an ecosystem, and its plant pests—all of which do occur. Although ecological principles
species composition, influence its primary productivity (33– may predict these problems, they do not seem to offer any easy
38). Total primary productivity increases about 35–70% as solutions to them. Agriculture, and society, seem to be facing
plant species diversity increases from one to about 20 species. tough tradeoffs. Agricultural ecosystems have become incred-
ibly good at producing food, but these increased yields have
Such effects have a series of alternative theoretical explana-
environmental costs that cannot be ignored, especially if the
tions (27, 32, 39 – 41). The two major classes of explanations
rates of nitrogen and phosphorus fertilization triple and the
are the sampling effect and niche differentiation. The sam-
amount of land irrigated doubles. The tradition in agriculture
pling effect implies that the increase in productivity associ-
has been to maximize production and minimize the cost of
ated with greater plant diversity is caused by the higher
food with little regard to impacts on the environment and the
probability that a more productive species or variety will be
services it provides to society. As the world enters an era in
present in a more diverse plot. The niche differentiation
which global food production is likely to double, it is critical
effect is based on complementary use of different limiting
that agricultural practices be modified to minimize environ-
resources by different species. One strain or species may grow
mental impacts even though many such practices are likely to
best during the cooler portion of the growing season, and
increase the costs of production.
another during the warmer portion. Or one may better exploit
soil nutrients in deeper soils and another at shallower depths. 1. Food and Agriculture Organization (1997) FAOSTAT (Food and
Such differing abilities to use limiting resources cause produc- Agriculture Organization of the United Nations, Rome).
tivity to increase with diversity (41). 2. Ruttan, V. W. (1999) Proc. Natl. Acad. Sci. USA 96, 5960–5967.
Under conditions typical of high-intensity agriculture (fer- 3. Tilman, D. (1982) Resource Competition and Community Struc-
tilized, irrigated fields in which light limits the growth of all ture: Monographs in Population Biology (Princeton Univ. Press,
plants), the sampling effect theory should apply, with maximal Princeton).
yields provided by the appropriate monoculture. All major 4. Vitousek, P. M., Aber, J. D., Howarth, R. W., Likens, G. E.,
grain crops (corn, wheat, rice, barley, etc.), soybeans, sugar Matson, P. A., Schindler, D. W., Schlesinger, W. H. & Tilman, D.
(1997) Ecol. Appl. 7, 737–750.
cane, and most other crops are grown in monoculture. How- 5. Schlesinger, W. H. (1991) Biogeochemistry: An Analysis of Global
ever hay, some crops harvested for fodder, and grasslands Change (Academic, San Diego).
maintained for grazing often are grown under conditions in 6. Howarth, R. W., Billen, G., Swaney, D., Townsend, A., Jaworski,
which niche differences could allow benefits from diversity. N., Lajtha, K., Downing, J. A., Elmgren, R., Caraco, N., Jordan,
Crop diversity also may be of benefit when arable lands are T., et al. (1996) Biogeochemistry 35, 181–226.
6000 Colloquium Paper: Tilman Proc. Natl. Acad. Sci. USA 96 (1999)

7. Socolow, R. H. (1999) Proc. Natl. Acad. Sci. USA 96, 6001–6008. 27. McNaughton, S. J. (1993) in Biodiversity and Ecosystem Function,
8. Vitousek, P. M. (1984) Ecology 65, 285–298. eds., Schulze, E.-D. & Mooney, H. A. (Springer, Berlin), pp.
9. Tilman, D. (1980) Am. Nat. 116, 362–393. 361–383.
10. Tilman, D., Kilham, S. S. & Kilham, P. (1982) Annu. Rev. Ecol. 28. Tilman, D. & Downing, J. A. (1994) Nature (London) 367,
System. 13, 349–372. 363–365.
11. Smith, V. H., Tilman, G. D. & Nekola, J. C. (1999) Environ. 29. Naeem, S. & Li, S. (1997) Nature (London) 390, 507–509.
Pollution, in press. 30. Doak, D. F., Bigger, D., Harding, E. K., Marvier, M. A.,
12. Lawes, J. & Gilbert, J. (1880) Philos. Trans. R. Soc. 171, 289–416. O’Malley, R. E. & Thomson, D. (1998) Am. Nat. 151, 264–276.
13. Thurston, J. (1969) in Ecological Aspects of the Mineral Nutrition 31. Tilman, D., Lehman, C. L. & Bristow, C. E. (1998) Am. Nat. 151,
of Plants, ed. Rorison, I. (Blackwell Scientific, Oxford), pp. 3–10. 277–282.
14. Patrick, R. (1963) Ann. N.Y. Acad. Sci. 108, 359–365. 32. Tilman, D. (1999) Ecology, in press.
15. Tilman, D. (1996) Ecology 77, 350–363. 33. Naeem, S., Thompson, L. J., Lawler, S. P., Lawton, J. H. &
16. Aerts, R. & Berendse, F. (1988) Vegetatio 76, 63–69. Woodfin, R. M. (1994) Nature (London) 368, 734–737.
17. Tilman, D. (1987) Ecol. Monogr. 57, 189–214. 34. Naeem, S., Thompson, L. J., Lawler, S. P., Lawton, J. H. &
18. Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Woodfin, R. M. (1995) Philos. Trans. R. Soc. London Ser. B 347,
249–262.
Sharpley, A. N. & Smith, V. H. (1998) Ecol. Appl. 8, 559–568.
35. Naeem, S., Håkenson, K., Lawton, J. H., Crawley, M. J. &
19. Postel, S. L., Daily, G. C. & Ehrlich, P. R. (1996) Science 271,
Thompson, L. J. (1996) Oikos 76, 259–264.
785–788. 36. Tilman, D., Wedin, D. & Knops, J. (1996) Nature (London) 379,
20. Huston, M. A. (1979) Am. Nat. 113, 81–101. 718–720.
21. Daily, G. C. (1997) Nature’s Services: Societal Dependence on 37. Tilman, D., Knops, J., Wedin, D., Reich, P., Ritchie, M. &
Natural Ecosystems (Island, Washington, DC). Siemann, E. (1997) Science 277, 1300–1302.
22. Costanza, R., d’Arge, R., de Groot, R., Farber, S., Grasso, M., 38. Symstad, A. J., Tilman, D., Willson, J. & Knops, J. M. H. (1998)
Hannon, B., Limburg, K., Naeem, S., O’Neill, R. V., Paruelo, J., Oikos 81, 389–397.
et al. (1997) Nature (London) 387, 253–260. 39. Swift, M. J. & Anderson, J. M. (1993) in Biodiversity and
23. Matson, P. A., Parton, W. J., Power, A. G. & Swift, M. J. (1997) Ecosystem Function, eds. Schulze, E.-D. & Mooney, H. A.
Science 277, 504–509. (Springer, Berlin), pp. 15–41.
24. Drinkwater, L. E., Wagoner, P. & Sarrantonio, M. (1998) Nature 40. Huston, M. A. (1997) Oecologia 110, 449–460.
(London) 396, 262–265. 41. Tilman, D., Lehman, C. L. & Thomson, K. T. (1997) Proc. Natl.
25. Vitousek, P. M., Mooney, H. A., Lubchenco, J. & Melillo, J. M. Acad. Sci. USA 94, 1857–1861.
(1997) Science 277, 494–499. 42. Parton, W. J. & Rasmussen, P. E. (1994) Soil Sci. Soc. Am. J. 58,
26. Crosby, A. W. (1986) Ecological Imperialism: The Biological Expan- 530–536.
sion of Europe 900–1900 (Cambridge Univ. Press, Cambridge). 43. Knops, J. M. H. & Tilman, D. (1999) Ecology, in press.