www.biosciencemag.org March 2010 / Vol. 60 No. 3

• BioScience 223

Forum

Effects of US Maize Ethanol onGlobal Land Use and GreenhouseGas Emissions: Estimating Market-mediated Responses

Thomas W. herTel, alla a. Golub, andreW d. Jones, michael o’hare, richard J. Plevin, anddaniel m. Kammen

Releases o greenhouse gases (GHG) rom indirect land-use change triggered by crop-based biouels have taken center stage in the debate over the role o biouels in climate policy and energy security. This article analyzes these releases or maize ethanol produced in the United States. Factoring market-mediated responses and by-product use into our analysis reduces cropland conversion by 72% rom the land used or the ethanol eedstock.Consequently, the associated GHG release estimated in our ramework is 800 grams o carbon dioxide per megajoule (MJ); 27 grams per MJ per year, over 30 years o ethanol production, or roughly a quarter o the only other published estimate o releases attributable to changes in indirect land use. Nonetheless, 800 grams are enough to cancel out the benefts that corn ethanol has on global warming, thereby limiting its potential contribution in the context o Caliornia’s Low Carbon Fuel Standard.Keywords: biouels, ethanol, land-use change, greenhouse gas emissions, market-mediated eects

To illustrate the importance o these numbers, considerthat Caliornia’s LCFS requires motor uel carbon intensity be reduced 10%, or or gasoline, rom 96 g per MJ to 86 gper MJ. I ethanol is blended at 20%—twice the current legallimit—the ethanol’s total global warming index, includingILUC, would have to be 46 g per MJ to meet this target.The size o the ILUC eect remains highly uncertain andclearly requires additional analysis: The estimates that theCaliornia and US regulatory agencies have produced areabout a quarter o the 100 g per MJ reported by Searchingerand colleagues (2008a). However, the agencies’ estimates arestill large enough to make maize ethanol an unattractivecompliance option or mitigating current carbon intensity or meeting uel-use mandates; they are also likely to greatly dampen enthusiasm or other biouels rom ood crops.In this article we use the global economic commodity andtrade model, GTAP-BIO (Hertel et al. 2010), to provide anew, comprehensive analysis o market-mediated changes inglobal land use in response to the expansion o US-grownmaize or ethanol. We nd the increase in cultivated landassociated with US-based maize ethanol to be just two-thso the amount estimated by Searchinger and colleagues(2008a). Still, adding our ILUC values to typical directIn April 2009, the Caliornia Air Resources Board adoptedthe Low Carbon Fuel Standard (LCFS) (Farrell et al. 2007a,2007b, CARB 2009), and in May o the same year, theUS Environmental Protection Agency issued the Energy Independence and Security Act (USEPA 2009a). Theseactions signal that greenhouse gas (GHG) releases romindirect (or induced) land-use change (ILUC) triggered by crop-based biouels have moved rom scientic debate toconsequential public policy. The predominant transporta-tion biouel in the United States is maize ethanol, and itwill remain so or the near uture. To date, the only peer-reviewed estimate o emissions due to ILUC rom the pro-duction o maize ethanol is about 3000 grams (g) o carbondioxide (CO

) equivalent discharge per annual megajoule(MJ) o maize ethanol production capacity (Searchingeret al. 2008), or 100 g per MJ i allocated over 30 years o production.

Direct

releases o GHG also occur during thecultivation and industrial processing o maize ethanol. Es-timates o these, not including ILUC, are about 60 to 65 go CO

equivalent per MJ (CARB 2009), although improve-ments in process technologies (Wang et al. 2007, Plevin andMueller 2008) and arming practices (Kim et al. 2009) may lower this value.

BioScience

60: 223–231. ISSN 0006-3568, electronic ISSN 1525-3244. © 2010 by American Institute o Biological Sciences. All rights reserved. Requestpermission to photocopy or reproduce article content at the University o Caliornia Press’s Rights and Permissions Web site at

www.ucpressjournals.com/ reprintino.asp.

doi:10.1525/bio.2010.60.3.8

224 BioScience •

March 2010 / Vol. 60 No. 3

www.biosciencemag.org

Forum

emissions gives a total carbon intensity or maize ethanolthat will have to be signicantly reduced through betterprocess technology (Wang et al. 2007, Plevin and Mueller2008), a feet capable o using much higher ethanol blendlevels, and an extremely long period o maize cultivation, i maize ethanol is to contribute useully to GHG reductionsin transportation.Vehicle uel policies or global warming stabilization posethe question, “I a unit o energy is obtained rom uel Arather than uel B, how much less greenhouse gas will beemitted?” The answer, in units o grams o CO

equiva-lent per MJ, should include indirect releases anywhere onthe planet—because releases everywhere are mixed, andtheir warming eects are diuse—caused by predictableresponses o the world’s production system to the change inuel use. Among the most important o these is the releaseo CO

when biouels’ demand or eedstock triggers a suc-cession o land-use changes that cause orest and otherecosystems with high carbon stocks (oten ar rom wherethe biouel is grown) to be converted to cultivation (Search-inger et al. 2008a). These stocks are usually burned or decay,although some, typically less than 10% o the total, may be sequestered in timberproducts or as charcoal in soil. In theollowing sections we describe changesdue to expanded ethanol production,triggered by initial diversion o a unit o maize armland in the United States romood to uel, and ollowing through to thevarious changes in agricultural practicesand land conversion around the worldthat result in ILUC.Beore proceeding, it should be notedthat time is very important in ouranalysis, an issue explored in O’Hareand colleagues (2009). I a biouel withdirect GHG emissions lower than thoserom ossil uel is produced over a longenough time horizon, this initial dis-charge will eventually be oset (Fargioneet al. 2008). However, i productionends ater only a ew years, the ILUCswamps any advantage the biouel hasinsoar as global warming is concerned.We estimate the discharge caused by increasing annual production capacity in megajoules, and we emphasize thatthe assumed production period or thebiouel being analyzed proportionally aects the estimated GHG emissions permegajoule. Dividing the initial ILUC by 20 instead o 30, representing a 20-yearproduction period, increases the per-megajoule value implied by our 800 ginitial discharge by a actor o 50%, toabout 40 g CO

per MJ.

Estimation o ecosystems converted and associatedcarbon emissions

We model the expansion o US maize ethanol use rom 2001levels to the 2015 mandated level o 56.7 gigaliters (GL) per year by orcing 50.15 GL o additional ethanol production,with the higher costs passed orward to consumers in theorm o higher uel prices. (As the volume o productionincreases, so too does the ILUC impact [Tyner et al. 2009].)Looking at the average eect over the entire 50.15 GL, thepredicted ILUC is higher than or the rst increment, butlower than or the nal increment to ethanol production.The version o GTAP that we used identies land-coverchanges within 18 agroecological zones (AEZs) denedby rainall and temperature (Lee et al. 2009), as well as 18trading regions. The rst panel o gure 1 summarizes thecontinental pattern o land conversion induced by increasedethanol production. Globally, cropland cover increases by 3.8 million hectares (Mha). In the majority o AEZs, crop-land increases at the expense o both pasture and orests.However, some o this decrease in orested area is compen-sated or elsewhere, as both orestry and cropland increase at

Figure 1. Global land conversion and associated greenhouse gas emissions due to increased maize ethanol production of 50.15 gigaliters per year at 2007 yields, by region.

www.biosciencemag.org March 2010 / Vol. 60 No. 3

• BioScience 225

Forum

the expense o pasture in some AEZs. We estimate that mostcropland conversion arises within the United States, ollowedby its dominant export competitors and trading partners. Incontrast to Searchinger and colleagues (2008a), we estimatear less conversion in some o the large but relatively closedagricultural economies, such as India. This geographic ap-proach to trade is supported by the recent econometric worko Villoria (2009), who rejects the integrated world markethypothesis implicit in the analysis o Searchinger and col-leagues (2008a). It should be noted that, independent o themodeling ramework, international trade plays a key role inthe ILUC impacts o biouels (Searchinger et al. 2009).To examine the global warming implications o theseland conversions, we developed an emission actor or eachtype o transition predicted in each region: orest to crop,pasture to crop, and pasture to orest. These emission ac-tors account or changes in above- and belowground carbonstocks, as well as changes in 30-year carbon sequestrationby ecosystems that are actively gaining carbon (we do notaccount or changes in climate-relevant biophysical land-surace properties, such as albedo or latent heat fux). Apply-ing these actors to the land-use changes predicted by GTAP,we arrive at 870 teragrams o CO2 emissions, which is 800g per MJ o increased annual ethanol production. Aboveg-round biomass loss accounts or most o these emissions inthe rst ew years ater land conversion, whereas oxidationo soil carbon and avoided sequestration can continue ordecades. The second panel o gure 1 shows emissions by region and land conversion type. Carbon sequestered incrop biomass is also shown. The lion’s share o emissionsoccurs in the United States and Canada, where a greater pro-portion o the orest is expected to be cleared or crops.Forest area increases with greater maize production insome places (gure 1), mostly in Europe and Asia, whereclimatic conditions provide a comparative advantage or or-ests over crops displaced by biouel production. Signicantadditional cropland expansion occurs in Arica and LatinAmerica, but mostly rom pasture, which contains much lesscarbon than do orest ecosystems. In Europe, we use a loweremission actor or deorestation because cropland is already reverting to orest, and biouel cropland demand merely slows this process. The result is avoided (slow) sequestrationrather than (rapid) release o aboveground carbon.Using straight-line amortization over 30 years o produc-tion at current uel yields (ollowing Searchinger et al. 2008a)results in ILUC emissions o 27 g CO

per MJ. This is roughly one-ourth the value estimated by Searchinger and colleagues(104 g CO

per MJ). Nonetheless, adding our lower estimateo emissions to the 65 g CO

per MJ direct emissions romtypical US maize ethanol production would nearly eliminatecarbon benet o this biouel relative to typical gasoline (94to 96 g per MJ; Farrell et al. 2006, Wang 2007), which shouldperhaps encourage some ethanol producers to transitionto more climate-riendly technologies (Plevin and Mueller2008). These values suggest a “carbon payback time” (Gibbset al. 2008, Fargione et al. 2008) o 28 years.The GTAP model estimates changes in the economic useo land (i.e., among orest, cropland, and pasture uses). Ingeneral, however, many ecosystems (specic types o or-est, grassland, savannah, or wetland)—each with a uniqueprole o carbon stocks and sequestration rates—within agiven region might be converted to or rom these economicuses. To estimate which ecosystems are likely to be convertedin a given region and the associated carbon emissions, weadapted the model developed by Searchinger and colleagues(2008a), which relies on data compiled by the Woods HoleResearch Institute (described in detail in Searchinger et al.2008b). We describe here only the basic concept and ourmodications to that ramework.The model divides the globe into 11 regions: Europe;developed Pacic; ormer Soviet Union; North Arica/MiddleEast; Canada; United States; Latin America; South and SouthEast Asia; Arica; India, China, Pakistan; and the rest o theworld (ROW). In each region, up to ve ecosystem types areidentied or each o which we estimate above- and below-ground carbon stocks, along with the carbon fuxes associ-ated with converting these ecosystems to cropping or permit-ting these ecosystems to recover rom other uses. In addition,or each region we estimate the historical rates o conversionto agriculture o each ecosystem type. Thus, the ecosystemand carbon data underlying our emission actors is o coarserresolution than the AEZ level at which the GTAP model esti-mates land conversion; our analysis could be rened, how-ever, i higher resolution global data were available.We modied Searchinger and colleagues’ (2008a) approachas ollows:

For ecosystems converted to cropping, we assume that the

•

replacement cropping system stores 5 megagrams (Mg) Cper hectare in the rst year (see table 5.9 in IPCC 2006).We assume that 10% o orest biomass is sequestered in

•

either timber products or charcoal in soil, and that theremaining 90% is oxidized to CO

.We ignore non-CO

•

emissions (IPCC 2006, p. 5.29).

These changes result in slightly lower ILUC emission actorsthan shown in Searchinger and colleagues (2008a; or details,see the supplementary material at

www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=3160

).The Searchinger model treats emissions in Europe and theormer Soviet Union in a special way, assuming that crop-land is already in a process o reversion to orests in thoseregions. Thus, additional cropland resulting rom biouelexpansion merely slows this reversion and avoids the seques-tration that otherwise would have occurred. Searchinger andcolleagues’ data provides estimates o carbon sequestrationin regrowing orests, as well as carbon sequestration rateswithin existing orests and grasslands.Rather than calculate a single emission actor or all con-version to cropland in a particular region, we determinedseparate emission actors or each o the dominant transi-tions predicted by GTAP. In our analysis, three types o con-version dominate: orest to cropland, pasture to cropland,