Resetting global expectations from agricultural biofuels

Matt Johnston¹, Jonathan A Foley², Tracey Holloway¹, Chris Kucharik¹ and Chad Monfreda³

¹ Center for Sustainability and the Global Environment (SAGE), University of Wisconsin, Madison, WI 53726, USA
² Institute on the Environment (IonE), University of Minnesota, St Paul, MN 55108, USA
³ Arizona State University, Tempe, AZ 85287, USA

Received 18 July 2008
Accepted 12 December 2008
Published 13 January 2009

Contents

• 1. Introduction
• 2. Materials and methods
• 3. Results
• 4. Discussion
• References

1. Introduction

Aggressive renewable energy policies have helped the biofuels industry grow at a rate few could have predicted. Surplus agricultural feedstocks in the US and Europe have quickly been exhausted, contributing to commodity prices increases of biofuel crops such as corn, soybean and rapeseed. And while biofuel's role in the recent price increase of other agricultural commodities—such as rice and wheat—has been largely refuted, there is still forward-looking concern that the continued development of biofuels will place additional pressure on food supplies during a period of tremendous demand growth from developing countries [1, 2]. Many point to the vast potential of next-generation biofuels, such as cellulosic-ethanol and algae-biodiesel, however, agricultural-based biofuels are the only alternatives to liquid fossil fuels being produced profitably and in large volumes today [3, 4]. This has led to many developing countries—from Malaysia to Malawi—increasing their biofuel production for both export and domestic use [5–7]. While discourse on the energy balance and environmental impacts of agricultural biofuel feedstocks are common [8–10], the potential they hold for additional production has received considerably less attention. Using newly-developed, spatially-explicit global agricultural datasets (Monfreda et al 2008), hereafter referred to as the M3 cropland datasets, we challenge widely-held assumptions about the global production of biofuels from agricultural feedstocks, and show it may be significantly less than previously thought [11]. Furthermore, based on the huge variability in global yields for agricultural biofuel crops, our results demonstrate the need to incorporate regionally-specific crop data in future studies.

The biofuel production potential from different feedstock crops are most commonly compared by their fuel yield per unit area. Media articles, journals, and policy reports are rife with examples of biofuel yield tables demonstrating, for example, the productivity per hectare of maize-ethanol is roughly half that of sugarcane [12–15]. Similarly, it is often reported that rapeseed (canola) will yield nearly twice the biodiesel as soybean per hectare [16–18]. Yield tables have become prolific due to their simplicity in conveying the otherwise complex relationships between the starch, sugar, or oil content of different crops and how they translate to potential fuel volumes—something most people tasked with communicating biofuels to ever diverse audiences can appreciate. However, the frequency at which these yield figures are quoted and re-quoted have led to an air of certainty in the data that was never originally intended. It is often unclear how agricultural yield values were first derived, how representative they are of different regions of the world, and how appropriate current yield estimates are for comparing different crops.

Crop handbooks, manuals, and, increasingly, online databases are some of the primary sources of consolidated information about agricultural crops and their potential application in energy production [19–23]. While these collections are often in an encyclopedic format, they are not intended to be replete. Biofuel yield estimates are collected from numerous disparate sources, often with little control for the geographic location, climate, soil type or agricultural management regime of the crop in question. Some of the yield estimates are from experimental agricultural stations (with exceptionally high yields), while others are from region-specific agricultural censuses, estimating the average yield of some countries. Yield tables that draw from these sources must often choose a single value (from either a single country, or even a single farm) to represent that crop globally. Building on these existing estimates to date, we provide improved, spatially-scalable methods of selecting more representative values. Table 1 lists the yield values from two commonly referenced crop tables and that form the basis for our comparison [16, 18].

Crop yields are influenced by several factors including mean climate and variability, soil conditions, and inputs and management, but are also impacted by social, political, and economic influences [24, 25]. Because research has already documented how historical yield increases have resulted from improved technology and advanced hybrids, better agronomic management practices, and increased inputs [25–32], we know that inconsistencies in adoption of these improvements from one region to the next, coupled with variability in climate and soil conditions, has created a large yield gap for many crops at the global scale. Mean climate and weather-variability impact yields by controlling diurnal temperature ranges, growing season length, the delivery of timely precipitation, and the frequency of extreme conditions such as droughts and floods [33]. Soil properties such as soil depth and water holding capacity, organic matter, bulk density, and pH—many of which are impacted by previous and current land use practices—ultimately influence crop yield potential [33]. Inputs such as irrigation, fertilizers, and pesticides, and management factors such as planting and harvest date, machinery usage, and efficiency of farming operations also contribute largely to global yield variability. As a result, estimating global biofuel feedstock production from these singular yield values—especially ones chosen arbitrarily—is problematic at best.

Here we present a new biofuel yield analysis based on the M3 cropland datasets. These new data represent geographically-specific patterns of the actual, reported yield for the vast majority of world crops. This gives us the first opportunity to consider patterns of biofuel feedstock production in different regions, across global, continental, national and sub-national scales. Our results are still presented as a collection of `single values' for each crop-country combination. However, instead of yield values being selected to represent a region, they are derived from spatially-explicit agricultural census data, and later regionally-specific analyses can always drill down spatially using the source data.

2. Materials and methods

The M3 cropland datasets combine the best available agricultural statistics from around the world—gathered across  ~ 22 000 different census reporting units at the county, state, and country level—with a recent map of global croplands [34] to produce the most detailed global maps of crop area and only global maps of crop yield available [11]. The datasets depict crop areas and yields circa the year 2000 on a 5 min × 5 min ( ~ 10 km × 10 km) latitude–longitude grid, and are the first attempt to comprehensively depict the spatial patterns of yield and crop area for all 175 crops reported in the FAOSTAT database. For example, M3 maps of wheat area coverage and yield are shown in figures 1(a) and (b), respectively.

The crop area and yield statistics were drawn from a variety of agricultural surveys, censuses, and statistical databases, and vetted and standardized to ensure data quality and compatibility. Over 80% of the total area for all crops derives from sub-national sources obtained from  ~ 2300 state-level and  ~ 19 700 county-level political units from 150 countries. The global cropland dataset used to provide a template to distribute these crop area and yield statistics is itself a fusion of national and sub-national cropland area statistics with two satellite-derived land cover classifications (Boston University's MODIS-derived land cover product and the GLC2000 dataset).

The datasets for ten ethanol crops (barley, cassava, maize, potato, rice, sorghum, sugarbeet, sugarcane, sweet potato, and wheat) and ten biodiesel crops (castor, coconut, cotton, mustard, oil palm, peanut, rapeseed, sesame, soybean, sunflower) were converted from metric-tonnes-per-hectare to liters-per-hectare and analyzed across 238 countries, territories and protectorates.

Biodiesel-per-hectare yields for each crop i and country j, BY_ij, were calculated using the following formula:

Current agricultural yields, CY_ij, are recorded in tonnes-per-hectare, provided as by the M3 cropland database [11]. To determine what fraction of the mass of each crop is vegetable oil, the yields were first multiplied by the per cent oil content as recorded by the FAO for each crop i, OC_i [35]. Vegetable oil masses were then divided by the oil densities corresponding to each crop, OD_i [36, 37], to arrive at liters of raw vegetable oil per hectare. Raw vegetable oil volumes were reduced by a processing ratio (PR) of 0.9622 to account for processing into food-grade vegetable oil—a form suitable for food-exports and for refining into biodiesel [38]. Finally, food-grade vegetable oils were multiplied by the refining ratio, RR; a term that reflects the conversion efficiency of processed vegetable oil to refined biodiesel. On average, using current refining equipment setup in a continuous-flow process, RR = 0.98 [38]. Table 2(a) lists the simplified conversion factors to translate from tons of biomass to liters of biodiesel.

We calculated ethanol-per-hectare yields for each crop i and country j, EY_ij, using the following equation:

As with the biodiesel example above, ethanol yield calculations began with the current agricultural yields in tonnes-per-hectare, CY_ij [11]. These mass values were converted to liters using refining factors for each crop, RF_i [39, 40]. Unlike biodiesel, which has similar processing and refining steps once the oil is separated, ethanol processing is more crop-specific. All the refining factors used in this study, listed in table 2(b), include the conversions from extracting, fermenting and distilling the sugars present in the various crops.

Liter-per-hectare fuel yield and area cultivated were analyzed globally for each of the 20 crops in study to provide yield quartiles and distributions. Additionally, quartiles and distributions were calculated individually for every country-crop combination allowing for more detailed region-specific studies in the future. All yield quartile values and distributions were first weighted by crop area so that smaller, unrepresentative plots would not artificially skew the results in one direction or another. To accomplish this, data was sorted by yield from lowest to highest, and then area was aggregated until the point at which each quartile (0%, 25%, 50%, 75%, 100%) of the total cultivated area was reached. The yield values corresponding with these points were selected for each crop, giving the area-weighted quartiles used to construct the statistical box-and-whisker plots.

Potential error in our analyses may arise due to variation in the source quality of agricultural statistics used to create the M3 cropland datasets, as well as potential errors in satellite data input to the crop distributions and/or limitations of the statistical merging of these data inputs. The majority of these errors, however, would affect the spatial distribution of yields and area coverage, which would later be minimized when we aggregate statistics using political boundaries.

Additional sources of potential error could come from the factors used in ethanol and biodiesel mass-to-liter conversions. Ethanol conversions were estimates from Nigam and Agrawal (2004) and a USDA study titled The Economic Feasibility of Ethanol Production from Sugar in the United States, however, additional conversion estimates were identified from five other sources [41–45]. Across the various sources, seven of the ten ethanol crops had differing estimates for mass-to-liter conversions. These conflicting estimates deviated from those used in our analyses by 3%–16%, with an average of 10%. Biodiesel yields are also expected to vary, as FAO oil content percentages are based on global averages and processing and refining ratios assume large-scale, continuous-flow processes that may not be used by all countries.

Finally, we recognize that any agro-industrial sponsored biofuel feedstock projects would likely result in higher than average yields in many countries—if done on a large enough scale. However, these operations would face many of the same obstacles as domestic efforts to improve yields, including: inefficiencies in (or lack of) transportation infrastructure, distribution of soil inputs and hybrid crops, lack of mechanization and irrigation, knowledge-transfer of management practices, as well as cultural and political barriers. Access to capital could help alleviate many of these issues, but challenges remain and it would be risky to count on achieving large yield increases quickly.

3. Results

Compared to earlier biofuel yield tables, our detailed agricultural analysis (figures 2(a) and (b)) shows a strikingly different picture of global biofuel yields [16–18]. For most crops, previous reports have overestimated yields by  ~ 100% or more. Barley, cassava, castor, maize, rapeseed, and sunflower all show that previous global biofuel yields were overestimated by at least 100%, with wheat–ethanol and groundnut–biodiesel estimates having been overestimated by  ~ 150% or more.

Our analysis shows that existing, commonly-accepted biofuel yield estimates become highly unrealistic as agricultural biofuels productions expands across the world. Even if we take an `optimistic' view of global yield data, it is difficult to imagine how these yields could be achieved at a large scale in the near future. Comparing previously-reported biofuel yields to actual median yields in developing countries we see even larger overestimates. For example, even though the Democratic Republic of the Congo (DRC) is the fifth largest grower of cassava globally, yields in that country trail those of its developing country peers, not to mention the  ~ 140% disparity with values chosen for previous yield tables (figure 3(a)). In fact, in many cases quoted yields are even optimistic for developed countries with efficient, advanced agricultural systems when viewed across all cultivated areas—as witnessed by Canadian rapeseed production (figure 3(b)).

As more and more biofuel production begins to be produced outside the few dominant countries of today—primarily the US, Brazil, and EU member countries—regionally-specific yield results will be needed in future policy, economic, and environmental analyses of biofuel growth. To encourage this, we also calculated complete results for 238 countries, states, territories and protectorates for use in more focused national and regional studies. Using these supplemental statistical analyses, publicly available at http://www.sage.wisc.edu/energy, future studies will be able to more accurately assess the potential and impacts of biofuels production from agricultural feedstocks. In aggregate, there are twenty summary tables (10 for ethanol crops, 10 for biodiesel crops) and over 3000 country box plots and histograms, such as those discussed above for sweet potato from the DRC (figure 3(a)) and rapeseed from Canada (figure 3(b)).

4. Discussion

Our results point to the continued existence of significant and geographically disparate agricultural yield gaps for most biofuel crops. This study demonstrates that using a single yield estimate (often taken from a unique location, or an agricultural field trial, and then applied to large regions—or worse globally) can be misleading, and often overestimates the actual yield of agricultural feedstocks by  ~ 100–150%. These gaps in productivity generally follow economic stratifications, but they are equally influenced by biophysical, political and cultural factors. Even when limiting yield values to the prime growing regions for common biofuel crops, existing biofuel yield tables have tended to choose values on the optimistic end of the range. For instance, even if one were to assume that the 4750 l/hectare oil palm-biodiesel yield from Brown were only intended to reflect Malaysian production potential, that value is still in the top 90th percentile of oil palm-agriculture in that country [16]. A more detailed, geographically-specific analysis of (actual and potential) yields is needed.

This paper is by no means the first to suggest the existence of such yield gaps [46–49]. Agro-businesses, academics and researchers have made great efforts to translate the post-WWII productivity gains of the `green revolution' in developed countries to regions such as Central America, South Asia, and Sub-Saharan Africa—some more successfully than others [49–51]. However, existing assessments of yield variation have largely been data limited, and as such focus on particular data-rich crops or specific regions [52–54]. The spatially-explicit nature of the new M3 cropland datasets will allow for quantification of potentially closable gaps in agricultural production across the world and for all crops reported by the FAO.

Cassman [46], perhaps, defines yield potential most elegantly as, `the yield obtained when an adapted cultivar is grown with the minimal possible stress that can be achieved with best management practices.' However, in addition to the myriad of agriculture-related factors influencing yield—from climate and soil conditions to management and mechanization—external political, economic and cultural factors often determine the ultimate success or failure of closing production gaps. Herculean efforts have been made in the past to close yield gaps in developing countries, only to be impeded by the lack of well-functioning transportation infrastructure, distribution of inputs, and access to capital and markets [55–57]. The biofuels revolution has lead to host of projects started under the guise of agricultural or rural development [58, 59]. The impact of important geographic variations in yield may be particularly acute when considering country-specific labor costs, land availability, agricultural management regimes, climate, and other factors. A better spatial understanding of yield potential can help optimize future development efforts and lower financial risk by coordinating with realistic infrastructure projects such as roads and ports, and choosing crops and regions that hold the most biophysical potential for the least investment.

Understanding where, how and for which crops yields can be improved globally will affect all aspects of agricultural production. However, this will be of particular importance to the biofuels industry, which has come under increasing pressure to slow the conversion of new agricultural lands for biofuels production. There is mounting research suggesting that the carbon benefits of biofuels will be eliminated, or worse reversed, if intact, carbon-rich ecosystems are converted to biofuel agriculture [60–62]. However, while a renewed focus on increasing production on existing agricultural lands may help slow this expansion, intensification can have equally devastating environmental effects as extensification depending on how it is implemented. Intense row-crop and plantation agriculture has been linked to eutrophication of rivers, lakes, and coastal areas [63, 64], depletion of soil nutrients and organic matter thereby decreasing soil quality [65, 66], and increased size and frequency of ocean hypoxic zones [67, 68]. Any efforts to increase crop productivity must do so as sustainably as possible—utilizing innovative agricultural initiatives such as more precise application of fertilizer inputs and irrigation through precision agriculture practices, improving nitrogen use efficiency [69, 70], and improved long-term tillage practices to reduce agrochemical runoff and organic matter losses—lest biofuel's realized benefits become further diminished.

The intention of this study is to reset the global expectations surrounding agricultural biofuels to a more realistic starting point, not to provide the final word on the debate. We expect existing trends in the production and consumption of agricultural crops to continue, including: annual increases in yields, the expansion of agricultural frontiers into natural ecosystems, and growth in consumption due to rising affluence in developing countries [9, 32]. In fact, the recent increase in commodity prices will likely encourage renewed focus on research and development, spurring productivity gains beyond the typical  ~ 2% averaged annual gains. And while the advent of next-generation biofuels may one day render the discussion of agricultural-based biofuels moot, it will take years for the technology to become widely distributed in developing countries—further increasing the importance of this study for near-term growth in the biofuels industry.

References

[1]

Ferrett G 2007 Biofuels `crime against humanity' BBC News (October 27) 

[2]

Runge C F and Senauer B 2007 How biofuels could starve the poor Foreign Affairs http://www.foreignaffairs.org/20070501faessay86305/c-ford-runge-benjamin-senauer/how-biofuels-could-starve-the-poor.html 

[3]

Himmel M E, Ding S-Y, Johnson D K, Adney W S, Nimlos M R, Brady J W and Foust T D 2007 Biomass recalcitrance: engineering plants and enzymes for biofuels production Science 315 804–7 
CrossRefPubMed

[4]

Perlack R et al 2005 Biomass as feedstock for a bioenergy and bioproducts industry: the technical feasibility of a billion-ton annual supply ORN Laboratory 

[5]

Cramer J and Koenders B 2007 Biomass will benefit developing countries Government.nl April 27, 2007 

[6]

Ching O T 2008 Malaysia doubles 2008 biodiesel export forecast Business Times (Kuala Lumpur, Malaysia) 

[7]

Junior S B 2008 Malawi leads in ethanol export AfricaNews Blantyre, Malawi 

[8]

Farrell A E et al 2006 Ethanol can contribute to energy and environmental goals Science 311 506–8 
CrossRefPubMed

[9]

UN-Energy 2007 Sustainable Bioenergy: A Framework for Decision Makers 

[10]

OECD/ITF 2007 Round Table on Biofuels: Linking Support to Performance (Paris: Joint Transport Research Centre) p 36 

[11]

Monfreda C, Ramankutty N and Foley J A 2008 Farming the planet. Part 2: the geographic distribution of crop areas and yields in the year 2000 Glob. Biogeochem. Cycles 22 GB1022 

[12]

Elbehri A 2007 Cellulosic biomass for bioenergy: economic and analytical issues Economic Research Service and Farm Foundation Workshop on Biofuels (Washington, DC: Economic Research Service, USDA) 

[13]

IEA 2004 Biofuels for Transport (Paris: International Energy Agency) p 216 

[14]

Johnson F X 2006 Biofuels markets, production potential, and international trade: the case of ethanol from sugarcane & sweet sorghum in southern Africa Africa is Energizing Itself (Vienna: Cane Resources Network for Southern Africa) 

[15]

Thoenes P 2006 Biofuels and Commodity Markets—Palm Oil Focus FAO, Commodities and Trade Division) p 15 

[16]

Brown L R 2006 Plan B 2.0: Rescuing a Planet Under Stress and a Civilization in Trouble ed L Starke (New York: W W Norton) p 352 

[17]

Laws F 2006 Castor oil: new source for biodiesel? Southeast Farm Press  http://southeastfarmpress.com/mag/farming_castor_oil_new/ 

[18]

Worldwatch Institute 2007 Biofuels for Transportation: Global Potential and Implications for Sustainable Agriculture and Energy in the 21st Century ed T W Institute (Washington, DC: The Worldwatch Institute) p 480 

[19]

1992 Alternative Field Crops Manual 

[20]

Duke J A 1983 Handbook of Energy Crops unpublished 

[21]

Duke J A 2001 Handbook of Nuts 2nd edn (New York: CRC Press LLC) p 343 Herbal Reference Library 

[22]

Markle G M et al 1998 Food and Feed Crops of the United States 2nd edn (Willoughby, OH: Meister Publishing) 

[23]

NewCROP 2008 New Crops Resource Online Program Center for New Crops & Plant Products—Purdue University) 

[24]

Naylor R, Falcon W and Zavaleta E 1997 Variability and growth in grain yields, 1950–94: does the record point to greater instability? Popul. Dev. Rev. 23 41–58 
CrossRef

[25]

Stewart W M et al 2005 The contribution of commercial fertilizer nutrients to food production Agron. J. 97 1–6 
CrossRef

[26]

Brown L R 2005 Outgrowing the Earth: The Food Security Challenge in an Age of Falling Water Tables and Rising Temperatures (New York: W W Norton) 

[27]

Duvick D N 1989 Possible genetic causes of increased variability in US maize yields Variability in Grain Yields: Implications for Agricultural Research and Policy in Developing Countries ed P B R Hazel (Baltimore, MD: Johns Hopkins University Press) pp 147–56 

[28]

Duvick D N 1992 Genetic contributions to advances in yield of US maize Maydica 37 69–79 

[29]

Duvick D N and Cassman K G 1999 Post-green revolution trends in yield potential of temperate maize in the North-Central United States Crop Sci. 39 1622–30 
CrossRef

[30]

French J B and Headley J C 1989 Influence of technology and weather on the variability in US maize and wheat yields Variability in Grain Yields: Implications for Agricultural Research and Policy in Developing Countries ed P B R Hazel (Baltimore, MD: Johns Hopkins University Press) pp 270–84 

[31]

Kucharik C J and Ramankutty N 2005 Trends and variability in US corn yields over the 20th century Earth Interact. 9 1–29 
CrossRef

[32]

Mann C C 1999 Crop scientists seek a new revolution Science 283 310–4 
CrossRef

[33]

Ramankutty N et al 2002 The global distribution of cultivable lands: current patterns and sensitivity to possible climate change Glob. Ecol. Biogeogr. 11 377–92 
CrossRef

[34]

Ramankutty N et al 2008 Farming the planet. Part 1: the geographic distribution of global agricultural lands in the year 2000 Glob. Biogeochem. Cycles 22 GB1003 

[35]

FAOSTAT 2007 Technical Conversion Factors for Agricultural Commodities Accessed on November 12 

[36]

Elert G 2007 Density of Cooking Oil The Physics Factbook http://hypertextbook.com/facts/2000/IngaDorfman.shtml Accessed on November 12 

[37]

Oxford 2007 Chemical and Other Safety Information The Physical and Theoretical Chemistry Laboratory Oxford University) Accessed on November 12 

[38]

Van Gerpen J et al 2004 Biodiesel Production Technology National Renewable Energy Laboratory) USDo Energy 
CrossRef

[39]

Nigam R B and Agrawal P K Ethanol manufacture from cane juice and molasses: a diversification fro sugar industry in India, 2004 Proc. Int. Conf. on Biofuels: Perspectives & Prospects (India: Winrock International) 

[40]

Shapouri D H and Salassi D M 2006 The Economic Feasibility of Ethanol Production from Sugar in the United States US Department of Agriculture) 

[41]

Atthasampunna P et al 1987 Production of fuel ethanol from cassava World J. Microbiol. Biotechnol. 3 135–42 
CrossRef

[42]

Dai D et al 2006 Energy efficiency and potentials of cassava fuel ethanol in Guangxi region of China Energy Convers. Manag. 47 1686–99 
CrossRef

[43]

El Bassam N 1998 Energy Plant Species: Their Use and Impact on Environment and Development (London, UK: Earthscan) p 334 

[44]

Stewart G A et al 1979 The Potential for Liquid Fuels from Agriculture and Forestry in Australia (Mentone, Victoria: CSaIR Organisation/Alexander Bros) 

[45]

USDA 2006 The Economics of Bioethanol Production in the EU Global Agriculture Information Network Report ed M K Francis (Brussels: USDA Foreign Agricultural Service) 

[46]

Cassman K G 1999 Ecological intensification of cereal production systems: yield potential, soil quality, and precision agriculture Proc. Natl Acad. Sci. USA 96 5952–9 
CrossRefPubMed

[47]

DeVries F, Rabbinge R and Groot J J R 1997 Potential and attainable food production and food security in different regions Phil. Trans. R. Soc. Lond. B 352 917–28 
CrossRef

[48]

Ewert F et al 2005 Future scenarios of European agricultural land use I. Estimating changes in crop productivity Agric. Ecosyst. Environ. 107 101–16 
CrossRef

[49]

1969 The green revolution Nutr. Rev. 27 133–6 
CrossRefPubMed

[50]

Kalra N et al 2007 An approach to bridging yield gaps, combining response to water and other resource inputs for wheat in northern India, using research trials and farmers' fields data Agric. Water Manag. 93 54–64 
CrossRef

[51]

Traxler G and Byerlee D 1992 Economic returns to crop management research in a post-green revolution setting Am. J. Agric. Econ. 74 573–82 
CrossRef

[52]

Atherton B C et al 1999 Site-specific farming: a perspective on information needs, benefits and limitations J. Soil Water Conserv. 54 455–61 

[53]

Calvino P and Sadras V 2002 On-farm assessment of constraints to wheat yield in the south-eastern Pampas Field Crops Res. 74 1–11 
CrossRef

[54]

Casanova D et al 1999 Yield gap analysis in relation to soil properties in direct-seeded flooded rice Geoderma 91 191–216 
CrossRef

[55]

2007 Globalization, Agriculture and the Least Developed Countries Making Globalization Work for the LDCs (Istanbul: United Nations Ministerial Conference of the Least Developed Countries) 

[56]

Collier P 2007 The Bottom Billion: Why the Poorest Countries are Failing and What Can be Done About It (New York: Oxford University Press) 

[57]

Easterly W 2001 The Elusive Quest for Growth: Economists' Adventrures and Misadventures in the Tropics (Boston: MIT Press) 

[58]

Cadenas A and Cabezudo S 1998 Biofuels as sustainable technologies: perspectives for less developed countries Technol. Forecast. Soc. Change 58 83–103 
CrossRef

[59]

Kojima M and Johnson T 2005 Potential for Biofuels for Transport in Developing Countries ed J U W Bank (Washington, DC: Energy Sector Management Assistance Program) 

[60]

Searchinger T et al 2008 Use of US croplands for biofuels increases greenhouse gases through emissions from land-use change Science 319 1238–40 
CrossRefPubMed

[61]

Fargione J et al 2008 Land clearing and the biofuel carbon debt Science 319 1235–8 
CrossRefPubMed

[62]

Gibbs H K et al 2008 Carbon payback times for crop-based biofuel expansion in the tropics: the effects of changing yield and technology Environ. Res. Lett. 3 034001 
IOPscience

[63]

Bennett E M, Carpenter S R and Caraco N 2001 Human impact on erodable phosphorus and eutrophication: a global perspective Bioscience 5 227–34 
CrossRef

[64]

Carpenter S R 2005 Eutrophication of aquatic ecosystems: bistability and soil phosphorus Proc. Natl Acad. Sci. USA 102 10002–5 
CrossRefPubMed

[65]

Kucharik C J, Brye K R, Norman J M, Foley J A, Gower S T and Bundy L G 2001 Measurements and modeling of carbon and nitrogen cycling in agroecosystems of southern Wisconsin: potential for SOC sequestration during the next 50 years Ecosystems 4 237–58 
CrossRef

[66]

Post W M and Kwon K C 2000 Soil carbon sequestration and land-use change: processes and potential Glob. Change Biol. 6 317–28 
CrossRef

[67]

Donner S D and Kucharik C J 2008 Corn-based ethanol production compromises goal of reducing nitrogen export by the Mississippi River Proc. Natl Acad. Sci. USA 105 4513–8 
CrossRefPubMed

[68]

Rabalais N N, Turner R E and Scavia D 2002 Beyond science into policy: Gulf of Mexico hypoxia and the Mississippi River Bioscience 52 129–42 
CrossRef

[69]

Cassman K G, Dobermann A and Walters D 2002 Agroecosystems, nitrogen-use efficiency, and nitrogen management AMBIO 31 132–40 
CrossRefPubMed

[70]

Tillman D, Cassman K G, Matson P A, Naylor R and Polasky S 2002 Agricultural sustainability and intensive production practices Nature 418 671–7 
CrossRefPubMed