Introduction
Ever since Malthus, the sufficiency of the global food supply to feed the human population has been challenged. One side of the current debate claims that green-revolution methods—involving high-yielding plant and animal varieties, mechanized tillage, synthetic fertilizers and biocides, and now transgenic crops—are essential in order to produce adequate food for the growing human populationReference Borlaug1–4. Green-revolution agriculture has been a stunning technological achievement. Even with the doubling of the human population in the past 40 years, more than enough food has been produced to meet the caloric requirements for all of the world's people, if food were distributed more equitablyReference Smil5. Yet Malthusian doubts remain about the future. Indeed, given the projection of 9 to 10 billion people by 20506 and the global trends of increased meat consumption and decreasing grain harvests per capita4, advocates argue that a more intensified version of green-revolution agriculture represents our only hope of feeding the world. Another side of the debate notes that these methods of food production have incurred substantial direct and indirect costs and may represent a Faustian bargain. The environmental price of green-revolution agriculture includes increased soil erosion, surface and groundwater contamination, release of greenhouse gases, increased pest resistance, and loss of biodiversity7–Reference Relyea14. Advocates on this side argue that more sustainable methods of food production are essential over the long termReference Tilman, Cassman, Matson, Naylor and Polasky15–17.
If the latter view is correct, then we seem to be pursuing a short-term solution that jeopardizes long-term environmental sustainability. A central issue is the assertion that alternative forms of agriculture, such as organic methods, are incapable of producing as much food as intensive conventional methods doReference Borlaug1, Reference Trewavas3, Reference Smil5. A corollary is that organic agriculture requires more land to produce food than conventional agriculture does, thus offsetting any environmental benefits of organic productionReference Green, Cornell, Scharlemann and Balmford18. Additionally, critics have argued that there is insufficient organically acceptable fertilizer to produce enough organic food without substantially increasing the land area devoted to agricultureReference Trewavas3.
Here, we evaluate the potential contribution of organic agriculture to the global food supply. Specifically, we investigate the principal objections against organic agriculture making a significant contribution—low yields and insufficient quantities of organic nitrogen fertilizers. The term ‘organic’ here refers to farming practices that may be called agroecological, sustainable, or ecological; utilize natural (non-synthetic) nutrient-cycling processes; exclude or rarely use synthetic pesticides; and sustain or regenerate soil quality. These practices may include cover crops, manures, compost, crop rotation, intercropping, and biological pest control. We are not referring to any particular certification criteria and include non-certified organic examples in our data.
Methods
We compiled data from the published literature about the current global food supply, comparative yields between organic and non-organic production methods, and biological nitrogen fixation by leguminous crops. These data were the basis for estimating the global food supply that could be grown by organic methods and the amount of nitrogen that could become available through increased use of cover crops as green manures.
Estimation of the global food supply
Estimation of the global food supply grown by organic methods involved compiling data about current global food production, deriving ratios of the yields obtained from organic versus non-organic production methods, and applying these yield ratios to current global production values.
Global food production
Summary data from the Food and Agricultural Organization (FAO) for 200119 document the current global food supply—grown primarily by conventional methods in most of the developed world and primarily by low-intensive methods in most of the developing world. The FAO provides estimates of the current food supply in 20 general food categories19 which we modified for our study. We combined three pairs of categories (into sugars and sweeteners, vegetable oils and oilcrops, meat and offals). We omitted from consideration three categories (spices, stimulants, and ‘miscellaneous’), because they contribute few calories and little nutritional value to the daily diet and lack comparative data for organic versus non-organic production. In addition, we reported data for seafood and ‘other aquatic products’ but did not estimate yield ratios for these categories, since most of these foods are currently harvested from the wild. Alcoholic beverages were reported since they contribute significantly to the average daily caloric intake, but no assessment of organic yields was made. The data presented for yield ratios pertain to ten categories covering the major plant and animal components of human diets.
Food-production data of the FAO include both commercial and domestic production and exclude losses during harvest. Pre-harvest crop losses are not included in the estimates; these losses may be substantialReference Oerke, Dehne, Schönbeck and Weber20 but are not necessarily more serious for organic production, since a host of methods is available for managing pestsReference Flint21, Reference Morales22. For each country or region, the FAO data for the food supply available for human consumption take into account food production, exports, imports, and stocks, as well as losses of production to become livestock feed, seed, or waste19. ‘Waste’ refers to post-harvest loss during storage, transport, and processing. We compiled this information for the world, for developed countries, and for developing countries, following the FAO classification of countries as developed or developing.
Deriving yield ratios
We estimated the global organic food supply by multiplying the amount of food in the current (2001) food supply by a ratio comparing average organic:non-organic yields. Comparisons of organic to non-organic production are available for many plant foods and a few animal foods. For each of 293 comparisons of organic or semi-organic production to locally prevalent methods under field conditions, the yield ratio is the ratio of organic:non-organic production. A ratio of 0.96, for example, signifies that the organic yield is 96% that of the conventional yield for the same crop. The comparisons include 160 cases with conventional methods and 133 cases with low-intensive methods. Most examples are from the peer-reviewed, published literature; a minority come from conference proceedings, technical reports, or the Web site of an agricultural research station. Like Stanhill's 1990 survey of organic and conventional productionReference Stanhill23, our data include numerous comparisons from paired farms and controlled experiments at research stations. The studies range in observation length from a single growing season to over 20 years. Despite the observation that yields following conversion from conventional to organic production initially decline and then may increase with timeReference Leigh and Johnson24, Reference Pimentel, Hepperly, Hanson, Douds and Seidel25 (but see ref. 23), we included studies regardless of duration. All of Stanhill's examples (which are included here) were from the developed world, whereas our dataset also includes diverse examples from the developing world. No attempt was made to bias the results in favor of organic yields; many examples from developed and developing countries exhibit low comparative yields. We avoided generalizations based on country-wide or regional average yields by organic or conventional methods. Some examples are based on yields before and after conversion to organic methods on the same farm.
We grouped examples into ten general food categories and determined the average yield ratio for all cases in each food category. For the complete dataset and sources, see Appendix 1. Table 1 presents the average yield ratios of these food categories for all studies combined (the world), studies in developed countries, and studies in developing countries. If no data were available (e.g., tree nuts) for estimating global organic production, then we used the average yield ratio for all plant foods, or all animal foods where relevant. For individual studies in which several yield ratios were reported for a single crop (e.g., 0.80–2.00) grown under the same treatment, we took the average as the value for the study. When different treatments were described, we listed a value for each treatment. Averaging the yield ratios across each general food category reduced the effects of unusually high or low yield ratios from individual studies. As these studies come from many regions in developed and developing countries, the average yield ratios are based on a broad range of soils and climates. The average yield ratio is not intended as a predictor of the yield difference for a specific crop or region but as a general indicator of the potential yield performance of organic relative to other methods of production.
Table 1. Average yield ratio (organic:non-organic) and standard error (S.E.) for ten individual food categories recognized by the FAO19 and three summary categories. Average yield ratio based on data from 91 studies (see Appendix 1 for data and sources). (A) All countries. (B) Developed countries. (C) Developing countries.
Studies in the global south usually demonstrate increases in yields following conversion to organic methods (Table 1C), but these studies are not comparable with those in the developed world. At present, agriculture in developing countries is generally less intensive than in the developed world. Organic production is often compared with local, resource-poor methods of subsistence farming, which may exhibit low yields because of limited access by farmers to natural resources, purchased inputs, or extension services. While adoption of green-revolution methods has typically increased yields, so has intensification by organic methodsReference Giller, McDonagh, Cadisch, Syers and Rimmer26. Such methods more often result in non-certified than in certified organic production, since most food produced is for local consumption where certification is not at issueReference Scialabba and Hattam27. Data from these studies are relevant for our inquiry, which seeks quantitative comparisons between organic production and prior methods, whether by conventional or subsistence practices, since both prevailing methods contribute to global food production.
Estimating the global food supply
Using the average yield ratio for each food category, we estimated the amount of food that could be grown organically by multiplying the amount of food currently produced times the average yield ratio (Tables 2 and 3). Following the FAO methodology19, this estimate was then proportionally reduced for imports, exports, and losses (e.g., Table 2, column D) to give the estimated organic food supply after losses (e.g., Table 2, column G), which is the food supply available for human consumption. We assumed that all food currently produced is grown by non-organic methods, as the global area of certified organic agriculture is only 0.3%Reference Willer and Yussefi28.
Table 2. Actual (2001) food supply and estimates for Model 1. Data for world food supply from FAO Statistical Database19.
1 Average yield ratio for all plant foods (developed countries) was used, since no comparative yield data were available for this food category.
2 Average yield ratio for all animal foods (developed countries) was used, since no comparative data were available for this food category.
Mg=megagram=metric ton.
Table 3. Actual (2001) food supply and estimates for Model 2. Data for world food supply from FAO Statistical Database19; data for yield ratios from Table 1.
1 Ratio is greater than 1.0 because of imports. All values in column (C) include imports but these are typically a small proportion of food production; for tree nuts, however, about one-third of the supply for the developed world is imported from the developing world.
2 Average yield ratio for all plant foods (developed countries) was used, since no comparative yield data were available for this food category.
3 Average yield ratio for all animal foods (developed countries) was used, since no comparative yield data were available for this food category.
4 Average yield ratio for all plant foods (developing countries) was used, since no comparative yield data were available for this food category.
5 Average yield ratio (developing countries) for all plant and animal foods was used, since no comparative yield data were available for this food category; the average for all foods was a more conservative estimate than the average for animal foods alone.
We constructed two models of global food production grown by organic methods. Model 1 applied the organic:non-organic (conventional) yield ratios derived from studies in developed countries to the entire agricultural land base (Table 2). This model effectively assumes that, if converted to organic production, the low-intensity agriculture present in much of the developing world would have the same or a slight reduction in yields that has been reported for the developed world, where green-revolution methods now dominate. Model 2 applied the yield ratios derived from studies in the developed world to food production in the developed world, and the yield ratios derived from studies in the developing world to food production in the developing world (Table 3). The sum of these separate estimates provides the global estimate.
In Model 1, the standard error of the estimate was calculated for an affine transformation (i.e., rescaled to world food production)Reference Stark29. In Model 2, the estimated global organic food production was the sum of two regional calculations—the yield ratios from the developed world times the current food production in the developed world and the yield ratios from the developing world times the current food production in the developing world. The standard error of the global estimate was determined for the sum of two independent random variablesReference Stark29.
For Model 2, we did not adjust for the amount of imported food in each food category. These amounts ranged from 4.9 to 75.8% (imported as a proportion of total food supply before losses) for the developed-world food supply and from 0.7 to 22.7% for the developing-world food supply19. Adjusting for imports in Model 2 would elevate slightly to greatly the estimates of the organic food supply in developed countries (Table 3, column F, because a proportion of the actual food supply would be multiplied by the higher average yield ratios for developing countries) and would diminish slightly the estimates of the organic food supply in the developing world (Table 3, column K, because a proportion of the actual food supply would be multiplied by the lower average yield ratios for the developed world). The overall results would be qualitatively similar.
Additional model assumptions
Both models were based on the pattern of food production and the amount of land devoted to crops and pasture in 2001. The models estimate the kinds and relative amounts of food that are currently produced and consumed, including the same pattern of total and per-capita consumption of meat, sugars, and alcoholic beverages. Additional assumptions include (1) the same proportion of foods grown for animal feed (e.g., 36% of global grain production), (2) the same proportion of food wasted (e.g., 10% of starchy roots), and (3) the same nutritional value of food (e.g., for protein and fat content in each food category), even though changes in some of these practices would benefit human or environmental health. Finally, we made no assumptions about food distribution and availability, even though changes in accessibility are necessary to achieve global food security. These assumptions establish the boundary conditions for the models but are not intended as an assessment of the sustainability of the current global food system.
Calories per capita
The calories per capita resulting from Models 1 and 2 were estimated by multiplying the average yield ratios (organic:non-organic) in each food category by the FAO estimate of per-capita calories currently available in that food category19.
Nitrogen availability with cover crops
The main limiting macronutrient for agricultural production is biologically available nitrogen (N) in most areas, with phosphorus limiting in certain tropical regionsReference Brady and Weil30. For phosphorus and potassium, the raw materials for fertility in organic and conventional systems come largely from mineral sources31 and are not analyzed here.
Nitrogen amendments in organic farming derive from crop residues, animal manures, compost, and biologically fixed N from leguminous plantsReference Lampkin32. A common practice in temperate regions is to grow a leguminous cover crop during the winter fallow period, between food crops, or as a relay crop during the growing season. Such crops are called green manures when they are not harvested but plowed back into the soil for the benefit of the subsequent crop. In tropical regions, leguminous cover crops can be grown between plantings of other crops and may fix substantial amounts of N in just 46–60 daysReference Boddey, de Moraes, Alves and Urquiaga33. To estimate the amount of N that is potentially available for organic production, we considered only what could be derived from leguminous green manures grown between normal cropping periods. Nitrogen already derived from animal manure, compost, grain legume crops, or other methods was excluded from the calculations, as we assumed no change in their use. The global estimate of N availability was determined from the rates of N availability or N-fertilizer equivalency reported in 77 studies—33 for temperate regions and 44 for tropical regions, including three studies from arid regions and 18 studies of paddy rice. N availability values in kg ha−1 were obtained from studies as either ‘fertilizer-replacement value,’ determined as the amount of N fertilizer needed to achieve equivalent yields to those obtained using N from cover crops, or calculated as 66% of N fixed by a cover crop becoming available for plant uptake during the growing season following the cover cropReference Hoyt and Hargrove34. The full dataset and sources are listed in Appendix 2. We estimated the total amount of N available for plant uptake by multiplying the area currently in crop production (but not already in leguminous forage production—large-scale plantings of perennial legume systems) by the average amount (kg ha−1) of N available to the subsequent crop from leguminous crops during winter fallow or between crops (Table 4, Appendix 2).
Table 4. Estimated nitrogen available for plant uptake from biological nitrogen fixation with leguminous cover crops, for the world and the US. For A, and F, data are from FAO Statistical Data Base19 and USDA National Agriculture Statistics35; for B, data for the world are from Gallaway et al., 1995Reference Galloway, Schlesinger, Levy, Michaels and Schnoor36, and for the US from USDA-ERSReference Vesterby and Krupa37 and the USDA National Agriculture Statistics35; for D, data are from sources listed in Appendix 2. Estimates are based on land area not currently in leguminous forage production.
Results and Discussion
Estimates of food and caloric production under organic agriculture
Figure 1 compares the estimates from Models 1 and 2 to the current food supply. According to Model 1, the estimated organic food supply is similar in magnitude to the current food supply for most food categories (grains, sweeteners, tree nuts, oil crops and vegetable oils, fruits, meat, animal fats, milk, and eggs). This similarity occurs because the average yield ratios for these categories range from 0.93 to 1.06 (Figure 1, Tables 1B and 2). For other food categories (starchy roots, legumes, and vegetables), the average yield ratios range from 0.82 to 0.89, resulting in somewhat lower production levels. The average yield ratio for all 160 examples from developed countries is 0.92, close to Stanhill's average relative yield of 0.91Reference Stanhill23. According to Model 2, the estimated organic food supply exceeds the current food supply in all food categories, with most estimates over 50% greater than the amount of food currently produced (Figure 1). The higher estimates in Model 2 result from the high average yield ratios of organic versus current methods of production in the developing world (Tables 1C and 3). The average yield ratio for the 133 examples from the developing world is 1.80. We consider Model 2 more realistic because it uses average yield ratios specific to each region of the world.
Figure 1. Estimates of the global food supply from two models of organic production compared with the actual food supply in 2001. Standard errors are given for food categories with multiple studies of yield ratios (see Table 1 and Appendix 1).
These two models likely bracket the best estimate of global organic food production. Model 1 may underestimate the potential yield ratios of organic to conventional production, since many agricultural soils in developed countries have been degraded by years of tillage, synthetic fertilizers, and pesticide residues. Conversion to organic methods on such soils typically results in an initial decrease in yields, relative to conventional methods, followed by an increase in yields as soil quality is restored7, Reference Pimentel, Hepperly, Hanson, Douds and Seidel25. Model 2 may overestimate the yield ratios for the developing world to the extent that green-revolution methods are practiced.
Both models suggest that organic methods could sustain the current human population, in terms of daily caloric intake (Table 5). The current world food supply after losses19 provides 2786 kcal person−1 day−1. The average caloric requirement for a healthy adult38 is between 2200 and 2500 kcal day−1. Model 1 yielded 2641 kcal person−1 day−1, which is above the recommended value, even if slightly less than the current availability of calories. Model 2 yielded 4381 kcal person−1 day−1, which is 57% greater than current availability. This estimate suggests that organic production has the potential to support a substantially larger human population than currently exists. Significantly, both models have high yields of grains, which constitute the major caloric component of the human diet. Under Model 1, the grain yield is 93% that of current production. Under Model 2, the grain yield is 145% that of current production (Table 5).
Table 5. Caloric values for the actual food supply (2001, data from FAO19) and for the organic food supply estimated in Models 1 and 2 (Tables 2 and 3). For alcoholic beverages, seafood, and other aquatic products, no change in caloric intake was assumed.
The most unexpected aspect of this study is the consistently high yield ratios from the developing world (Table A1, Appendix 1). These high yields are obtained when farmers incorporate intensive agroecological techniques, such as crop rotation, cover cropping, agroforestry, addition of organic fertilizers, or more efficient water managementReference Pretty, Morison and Hine16, Reference Uphoff39. In some instances, organic-intensive methods resulted in higher yields than conventional methods for the same crop in the same setting (e.g., the system of rice intensification (SRI) in ten developing countriesReference Uphoff39). Critics have argued that some of these examples exceed the intrinsic yield limits set by crop genetics and the environmental contextReference Sheehy, Peng, Dobermann, Mitchell, Ferrer, Yang, Zou, Zhong and Huang40. (Such controversy surrounds the ‘SRI’ and our data include studies from both sides of this controversy.) Yet alternative agricultural methods may elicit a different pathway of gene expression than conventional methods doReference Kumar, Mills, Anderson and Mattoo41. Thus, yield limits for conventionally grown crops may not predict the yield limits under alternative methods.
Crop rotation and yield-time adjustment
Organic grain production frequently uses a different rotation system than conventional production. For example, it is common in organic systems to have a three or four-year rotation (with legumes or other crops) for corn, while the conventional rotation often involves planting corn every other year. In situations like this, it is difficult to make yield comparisons between organic and conventional systems without some sort of time adjustment. Although the high variation among rotation systems worldwide makes it impossible to provide a general time–yield adjustment, evaluating potential differences in performance is important. A thorough evaluation of the rotation effect requires knowledge of the plot-to-plot yield differences between organic and conventional production and the rate of decline of both organic and conventional production as a function of the rotation sequence—information that has not yet been experimentally demonstrated. While rotations would undoubtedly differ under a global organic production system, we have no basis for concluding that this system would be unable to provide enough grain to feed the world.
Organic nitrogen fertilizer
In 2001, the global use of synthetic N fertilizers was 82 million Mg (metric ton)19. Our global estimate of N fixed by the use of additional leguminous crops as fertilizer is 140 million Mg, which is 58 million Mg greater than the amount of synthetic N currently in use (Table 4). Even in the US, where substantial amounts of synthetic N are used in agriculture, the estimate shows a surplus of available N through the additional use of leguminous cover crops between normal cropping periods. The global estimate is based on an average N availability or N-fertilizer equivalency of 102.8 kg N ha−1 (S.D. 71.8, n=76, Table A2, Appendix 2). For temperate regions, the average is 95.1 kg N ha−1 (S.D. 36.9, n=33) and for tropical regions, the average is 108.6 kg N ha−1 (S.D. 99.2, n=43). These rates of biological N fixation and release can match N availability with crop uptake and achieve yields equivalent to those of high-yielding conventionally grown cropsReference Kramer, Doane, Horwath and Kessel42. In temperate regions, winter cover crops grow well in fall after harvest and in early spring before planting of the main food crop43. Research at the Rodale Institute (Pennsylvania, USA) showed that red clover and hairy vetch as winter covers in an oat/wheat–corn–soybean rotation with no additional fertilizer inputs achieved yields comparable to those in conventional controlsReference Leigh and Johnson24, Reference Pimentel, Hepperly, Hanson, Douds and Seidel25, Reference Radke, Liebhardt, Jahnke, Peters and Power44. Even in arid and semi-arid tropical regions, where water is limiting between periods of crop production, drought-resistant green manures, such as pigeon peas or groundnuts, can be used to fix NReference Giller, McDonagh, Cadisch, Syers and Rimmer26, Reference Dakora and Keya45, Reference Sultan, Gintzburger, Obaton, Robin, Touchane and Guckert46. Use of cover crops in arid regions has been shown to increase soil moisture retentionReference Fahrney, El-Swaify, Lo, Joy and Power47, and management of dry season fallows commonly practiced in dry African savannas can be improved with the use of N-fixing cover crops for both N-fixation and weed control48. Areas in sub-Saharan Africa which currently use only very small amounts of N fertilizer (9 kg ha−1, much of it on non-food crops48) could easily fix more N with the use of green manures, leading to an increase in N availability and yields in these areasReference Giller, McDonagh, Cadisch, Syers and Rimmer26. In some agricultural systems, leguminous cover crops not only contribute to soil fertility but also delay leaf senescence and reduce the vulnerability of plants to diseaseReference Brady and Weil30.
Our estimates of N availability from leguminous cover crops do not include other practices for increasing biologically fixed N, such as intercroppingReference Vandermeer, Carroll, Vandermeer and Rosset49, alley cropping with leguminous treesReference Nair50, rotation of livestock with annual cropsReference Lampkin32, and inoculation of soil with free-living N-fixersReference Dobereiner51—practices that may add considerable N fertility to plant and animal productionReference Giller and Wilson52. In addition, rotation of food-crop legumes, such as pulses, soy, or groundnuts, with grains can contribute as much as 75 kg N ha−1 to the grains that follow the legumesReference Boddey, de Moraes, Alves and Urquiaga33.
These methods can increase the N-use efficiency by plants. Since biologically available N is readily leached from soil or volatilized if not taken up quickly by plants, N use in agricultural systems can be as low as 50%Reference Honeycutt, Clapham and Leach53. Organic N sources occur in more stable forms in carbon-based compounds, which build soil organic matter and increase the amount of N held in the soilReference Pimentel, Hepperly, Hanson, Douds and Seidel25, Reference Drinkwater, Wagoner and Sarrantonio54. Consequently, the amount of N that must be added each year to maintain yields may actually decrease, because the release of organic N fixed in one season occurs over several yearsReference Brady and Weil30.
These results imply that, in principle, no additional land area is required to obtain enough biologically available N to replace the current use of synthetic N fertilizers. Although this scenario of biological N fixation is simple, it provides an assessment, based on available data, for one method of organic N-fertility production that is widely used by organic farmers and is fairly easy to implement on a large scale. This scenario is not intended to be prescriptive for any particular rotation or location, but to demonstrate the possibility of this type of cover-cropping system to fix large quantities of N without displacing food crops or expanding land area. The Farm Systems Trial at the Rodale Institute uses legume cover crops grown between main crops every third year as the only source of N fertility and reports comparable grain yields to those of conventionally managed systems, while using non-legume winter cover crops in other years to maintain soil quality and fertility and to suppress weeds (R. Seidel and P. Hepperly, personal communication, 2006). In practice, a range of methods acceptable in organic agriculture provides critical flexibility in N-managementReference Lampkin32, including many sources other than cover crops. Although some environmental and economic circumstances pose challenges to reliance on leguminous fertilizersReference Peoples and Craswell55, the full potential of leguminous cover crops in agriculture is yet to be utilized. Implementation of existing knowledge could increase the use of green manures in many regions of the worldReference Crews and Peoples56. Future selection for crop varieties and green manures that have higher rates of N fixation, especially in arid or semi-arid regions, and perform well under N-limiting conditions, as well as for improved strains of N-fixing symbionts, combined with reductions in the amount of N lost from legume-based production systems, and increases in the planting of legumes, hold great promise for increasing the role of biological N-fixation in fertility managementReference O'Hara, Howieson, Graham and Leigh57. The capacity for increased reliance on legume fertilizers would be even greater with substantive changes in the food system, such as reduction of food waste and feeding less grain to livestockReference Crews and Peoples56.
Prospects for More Sustainable Food Production
Our results suggest that organic methods of food production can contribute substantially to feeding the current and future human population on the current agricultural land base, while maintaining soil fertility. In fact, the models suggest the possibility that the agricultural land base could eventually be reduced if organic production methods were employed, although additional intensification via conventional methods in the tropics would have the same effect. Our calculations probably underestimate actual output on many organic farms. Yield ratios were reported for individual crops, but many organic farmers use polycultures and multiple cropping systems, from which the total production per unit area is often substantially higher than for single crops48, Reference Piper58. Also, there is scope for increased production on organic farms, since most agricultural research of the past 50 years has focused on conventional methods. Arguably, comparable efforts focused on organic practices would lead to further improvements in yields as well as in soil fertility and pest management. Production per unit area is greater on small farms than on large farms in both developed and developing countriesReference Rosset59; thus, an increase in the number of small farms would also enhance food production. Finally, organic production on average requires more hand labor than does conventional production, but the labor is often spread out more evenly over the growing seasonReference Pimentel, Hepperly, Hanson, Douds and Seidel25,60–Reference Granatstein62. This requirement has the potential to alleviate rural unemployment in many areas and to reduce the trend of shantytown construction surrounding many large cities of the developing world.
The Millennium Ecosystem Assessment17 recommends the promotion of agricultural methods that increase food production without harmful tradeoffs from excessive use of water, nutrients, or pesticides. Our models demonstrate that organic agriculture can contribute substantially to a more sustainable system of food production. They suggest not only that organic agriculture, properly intensified, could produce much of the world's food, but also that developing countries could increase their food security with organic agriculture. The results are not, however, intended as forecasts of instantaneous local or global production after conversion to organic methods. Neither do we claim that yields by organic methods are routinely higher than yields from green-revolution methods. Rather, the results show the potential for serious alternatives to green-revolution agriculture as the dominant mode of food production.
In spite of our optimistic prognosis for organic agriculture, we recognize that the transition to and practice of organic agriculture contain numerous challenges—agronomically, economically, and educationally. The practice of organic agriculture on a large scale requires support from research institutions dedicated to agroecological methods of fertility and pest management, a strong extension system, and a committed public. But it is time to put to rest the debate about whether or not organic agriculture can make a substantial contribution to the food supply. It can, both locally and globally. The debate should shift to how to allocate more resources for research on agroecological methods of food production and how to enhance the incentives for farmers and consumers to engage in a more sustainable production system. Finally, production methods are but one component of a sustainable food system. The economic viability of farming methods, land tenure for farmers, accessibility of markets, availability of water, trends in food consumption, and alleviation of poverty are essential to the assessment and promotion of a sustainable food system.
Acknowledgements
The course, ‘Food, Land, and Society’, at the University of Michigan, provided the incentive for this study. We are grateful to the farmers whose practices inspired this research. We thank P. Hepperly and R. Seidel for discussion and for providing us with data from the Rodale Farming Systems Trial. Members of the New World Agriculture and Ecology Group (NWAEG) provided useful insights. We thank D. Boucher, L. Drinkwater, W. Lockeretz, D. Pimentel, B. Needelman, J. Pretty, B. Schultz, G. Smith, P. Rosset, N. Uphoff, and J. Vandermeer for comments on several versions of this paper. This paper also benefited from the comments and recommendations of anonymous reviewers.
Appendix 1: Yield Ratios
The studies used to estimate the yield ratios for different food categories in Table 1 come from 91 sources in Table A1 describing results from experiments at research stations, comparisons of paired farms, and comparisons before and after the transition to organic production. The data come from 53 countries and 12 US states. Some comparisons during the transition to organic production come from surveys, especially in the data for the developing world. Data range in observation length from one growing season to over 20 years. Despite the observation that yields following conversion from conventional to organic production initially decline and then tend to increase over timeReference Leigh and Johnson24, we did not omit studies of short duration so as not to bias estimates of relative yield. We included data from previous comparisons of organic and conventional production, notably StanhillReference Stanhill23, Lampkin and PadelReference Lampkin and Padel63, and McDonald et al.Reference McDonald, Hobbs and Riha64 for the SRI in the developing world. Over 80% of the examples listed come directly from peer-reviewed journal articles or are cited or figured in them. The remainder come from technical books, conference proceedings, technical reports from universities, government agencies or independent research foundations, or the Web site of a university research station.
Table A1. Yield ratios of organic production:non-organic production, grouped by FAO food categories analyzed in the text. (A) Data from developed countries, where comparisons are between organic and conventional (green-revolution) production methods. (B) Data from developing countries, where comparisons are primarily between organic and non-intensive methods.
(A) Developed countries
(B) Developing countries
For the developing world, there are fewer controlled comparisons of organic versus non-organic methods than for the developed world. Much of our data in Table A1B comes from one source (Pretty and HineReference Pretty and Hine65), which is a compilation from surveys in developing countries of yield comparisons before and after farmers adopted specific agroecological practices. In order to determine whether the survey data biased our results, we tested the hypothesis that the average yield ratio based on survey data and unreported methods differed significantly from the average yield ratio based on experimental data and quantitative comparisons of paired farms. The only food category with a substantial sample size of yield ratios in both categories of studies was grains (n=102). We subdivided grains into rice and all other grains, because more than half of our data concern rice but these data are quite unequally distributed between the two categories of studies. For rice (n=61), a t-test (p=0.55) comparing the average yield ratios from surveys and unreported methods versus experiments and paired farms failed to reject the null hypothesis that the average yield ratios do not differ significantly. For all other grains (n=41), a t-test (p=0.45) also failed to reject the null hypothesis. Thus, we concluded that the survey data have not unduly biased our results for the developing world. (No data for the developed world come from surveys.)
Appendix 2: Nitrogen from Cover Crops
Table A2. Data and sources for nitrogen availability from cover crops; the country where the study occurred is listed when it is known. Multiple entries from the same source represent data for different plant species or varieties. Values with asterisk (*) were calculated based on 66% of the N fixed by a cover crop becoming available for plant uptake during the growing season following the cover cropReference Hoyt and Hargrove34. The other values are the ‘fertilizer-replacement value,’ determined as the amount of N fertilizer needed to achieve yields equivalent to those obtained using N from cover crops. Values from studies in the US were the basis for calculating the N available from cover crops for the United States in Table 4.
