Category Archives: Corn ethanol

Accentuating the positive?

30 Apr 20
Chris Malins
No Comments

Estimating emissions associated with indirect land use change (ILUC) is a fundamental part of analysing the likely net GHG emissions impacts of biofuel mandates, and in some regulations (e.g. U.S. Renewable Fuel Standard, California Low Carbon Fuel Standard and ICAO’s CORSIA) estimates of ILUC emissions associated with specific feedstocks have been integrated into regulatory lifecycle analysis frameworks. This academic paper, co-authored by Chris Malins of Cerulogy with Richard Plevin and Robert Edwards, examines the development of one particular model for estimating ILUC emissions (GTAP-BIO) and asks how well supported by relevant data and/or analysis various adjustments and innovations introduced to the model over the past decade or so have been.

Particular attention is paid to a series of model amendments that have enhanced the role of intensive responses (increased productivity on existing agricultural land) in the model and reduced the predicted extent of land use changes from biofuel policy. The paper finds that there is a lack of compelling evidence supporting adopted assumptions, and that on more than one occasion model adjustments that have been presented as ‘neutral’ have in fact predictably resulted in reduced output ILUC values. It also notes that while the ‘cropland-pasture’ land category in the U.S. has been made central to the model outcomes there is very little evidence available to confirm that this is a realistic assumption, and that emission factors have been adopted for cropland-pasture conversion that are difficult to justify analytically. Reductions in modelled ILUC estimates ought therefore to be understood as at least as much the result of subjective decisions by the modelling teams involved as the result of any objectively demonstrable improvement in our understanding of the systems being studied.

The paper concludes that it is unclear whether more recent published ILUC estimates are likely to be closer to the ‘real’ average ILUC values for corn ethanol and soy biodiesel than higher values from earlier assessments.

Comparing GTAP ILUC results to observations of ethanol related land use change

29 Mar 19
Chris Malins
No Comments

For over ten years, indirect land use change modeling has been an important part of assessing the environmental impact of U.S. biofuel policy. While several models have been developed to undertake these assessments, notably the FAPRI-FASOM model used by the Environmental Protection Agency (EPA) in determining the lifecycle emissions of fuels supported by the Renewable Fuel Standard (RFS) (U.S. Environmental Protection Agency, 2010), the most prolific use of a single model has been the GTAP computational general equilibrium model, used for regulatory analysis by the California Air Resources Board (California Air Resources Board, 2014), in the Argonne National Laboratory’s GREET model (Argonne National Laboratory, 2017) and in a sequence of independent analyses by researchers. The GTAP model itself does not include land use change emission factors, and thus areal land use change results output from GREET must be coupled to emission factor models to produce ILUC results in terms of carbon dioxide emissions per unit of energy produced.

One notable feature of the ILUC factor results reported using the GTAP model over the past decade is that they have shown a tendency to reduce over time, as shown for in the figure below for corn ethanol.

Point ILUC factor estimates for US corn ethanol obtained with variants of the GTAP-BIO model

A forthcoming academic paper[1] will discuss in detail the underlying analytical reasons for these reductions, and queries whether the changes to the model and emissions factors that have driven these changes have been adequately justified. While assessing modeling choices is an interesting exercise, it is also enlightening to compare reported model results to observed changes in U.S. agriculture. One source for observed data on U.S. land use changes that may be associated with the growth of the corn ethanol industry is a series of papers by researchers at the University of Wisconsin (including Lark, Salmon, & Gibbs, 2015; Wright et al., 2017), which use data from the USDA cropland data layer (CDL). Here, we compare ILUC results for corn that are documented by Qin, Dunn, Kwon, Mueller, & Wander (2016) against the observed land use changes in the U.S. in the period 2008-2012 documented in Wright et al. (2017).

Qin et al. (2016) give one of the lowest corn ILUC numbers available in the literature, as low as 2.1 gCO2e/MJ at the bottom end of the interval given (the range reflects different yield and tillage assumptions), but are these results consistent with observed land use changes? Below, the reported model outputs from the ‘Corn2’ scenario in  Qin et al. (2016)[2] are considered, and compared to agricultural statistics and the results of the analysis presented in (Wright et al., 2017) for the years 2008-2012, during which ethanol productions increased by 4.2 billion gallons. It is important to note up front that it is always difficult to compare observed land use changes to the results of indirect land use change models. It is generally not possible to definitively identify whether a given land use change would have occurred in the absence of biofuel demand. In contrast, in ILUC modeling we know by hypothesis that all land use changes reported are due to biofuel demand. Apparent inconsistencies between observed land use data and ILUC models may have several reasonable explanations. The results noted below cannot refute the results of the modeling, but do perhaps suggest that some model assumptions ought to be reexamined. 

Qin et al. (2016) consider an increase of biofuel demand by 11.59 billion gallons (the ‘shock’), which is about equivalent to the increase in ethanol production in the U.S. between 2004 and 2015 (based on data from the EIA[3]), and assess how much land use change was likely to have been caused by that increase. Given average corn production[4] and ethanol refinery[5] yields for the period 2011-2015, producing a billion gallons of ethanol requires about a million hectares of corn. The net cropland expansion in the U.S. modeled by Qin et al. (2016) is much less than this however, only 160 thousand hectares per billion gallons of additional corn ethanol demand. Almost all of the U.S. land use change (154 thousand hectares per billion gallons) was modeled coming from a category of land identified as ‘cropland pasture’ – defined by the USDA as land that at some previous point was cropped, and is still suitable for cropping, but since then has been use as pasture. A further 112 thousand hectares of land use change per billion gallons are predicted outside the U.S. That the net cropland expansion modeled should be less than the absolute cropland requirement is not surprising to anyone familiar with ILUC modeling. There are several effects that reduce net requirements for additional land – notably these include the use of co-products from ethanol production (distillers’ grains) as animal feed, reducing demand for food[6], and the possibility that agricultural productivity increases (Malins, Searle, & Baral, 2014). While it is not surprising that the net land requirement is less than the gross requirement, it is clear that GTAP is modelling a strong role of responses other than land use change in meeting additional corn demand, given that the net land demand is only about a quarter of the gross. 

Wright et al. (2017), looking at satellite data, identify 1.1 million hectares of land within 50 miles of ethanol refineries being converted from non-crop status to cropping in the period 2008-2012, with most of this land (700 thousand hectares) converted to corn or soybean cropping. A further 600 thousand hectares of land from 50 to 100 miles from ethanol refineries was converted from non-crop statuses, though over half of this was planted with other crops. Further than 100 miles from corn ethanol refineries only 20% of newly converted land was planted with corn or soybeans[7]. Wright et al. (2017) argue that depending on the mix of rotations adopted (how much land remains in corn-soy rotation and how much is given over to continuous corn) then their analysis would be consistent with between 500,000 and 1 million hectares of previously uncropped land being converted for corn production within 100 miles of ethanol refineries over the period considered. If all of this additional corn area could indeed be attributed to the corn ethanol mandate, that represents somewhere between 130 thousand and 260 thousand hectares of land use change to corn production for every billion gallons of additional ethanol demand in that period, comparable to the modelled values. Data from the AFDC[8] shows that during this period use of corn for non-fuel applications reduced, so it is not unreasonable to assume that most conversion of uncropped land to corn production in this period was a response to the corn ethanol mandate. Preliminary analysis by the same group (Lark et al., 2019) estimates that 1.1 million hectares of land conversion since 2005 is associated with the ethanol mandate of the RFS2, about 200 thousand hectares per billion gallons.

Another comparison point comes from related work supported by the National Wildlife Federation (Hendricks, 2018), which aims to provide an economic assessment of the impact of the RFS2 on land use changes in the period 2008-2016  using data from the National Resources Inventory (NRI). This assessment attributes 1.2 million hectares of net new cropland in the period 2009-2016 to the RFS, predominantly coming from conversion of former CRP land. If all of this was associated with the corn mandate, it would be equivalent to about 190 thousand hectare per additional billion gallons of ethanol production – part of this expansion ought to be associated to the biodiesel mandate, and so the overall conclusion is quite comparable to the U.S. land use change modeled by (Qin et al., 2016).

While the realized rates of land use change for corn ethanol production identified in these studies are comparable to those modeled by Qin et al. (2016), there is evidence that the role of the cropland pasture land category as a source of cropland may be dramatically overstated by the GTAP modeling. For example, using the same dataset as considered by Wright et al. (2017), this study finds that 22% of newly cropped land in the period 2008-2012 was converted from “long-term (20 + year) unimproved grasslands”. This compares to the (Qin et al., 2016) result in which at most 5% of converted grasslands could have possibly have fallen into the long-term unimproved category[9]. The cropland pasture category in GTAP does not directly correspond to any single land category in the University of Wisconsin assessments, but the observations certainly suggest that assuming that over 90% of new cropland associated with ethanol demand comes from cropland pasture is likely to be a considerable over-estimate. The economic analysis by (Hendricks, 2018) also gives a very different result to the GTAP modeling. This economic analysis suggests that over 95% of grassland converted to cropping due to RFS was former CRP land. As CRP land is not pastured, it does not fall under the definition of cropland pasture in the agricultural census, and indeed GTAP has a separate category for CRP land introduced at the same time as the cropland pasture category.   

Lark et al. (2015) also document rates of forest land conversion to cropland that appear to be larger than anticipated by Qin et al. (2016)[10], although it is even more difficult for these relatively small forest area changes to draw a firm conclusion about whether this should be attributed to the ethanol mandate.

These discrepancies regarding the source of new land are potentially important in the ILUC analysis, because different land use changes have very different assumed carbon losses in the modeling. In particular, when using the CCLUB emission factors included in the GREET model the conversion of cropland pasture to corn is assumed to result in an increase in carbon sequestration, whereas Lark et al. (2015) argue that the conversion to cropland they identify is likely to be associated with very significant carbon dioxide emissions. This fundamentally different interpretation seems to arise primarily due to the use of a rather questionable modeling assumption in the development of the CCLUB emission factors. In CCLUB, soil carbon under the cropland pasture land use is modeled using DAYCENT and treating cropland pasture as if it has uniformly been cropped before 1951, then used as pasture for 25 years before spending the last 35 years being farmed for a generic crop. This land use history is almost the opposite of the cropland pasture definition (previously cropped but then used as pasture for several years. Combining this difficult to justify modeling choice with the high rates of cropland pasture conversion modeled with GTAP by Qin et al. (2016) leads to a dramatic reduction in the predicted ILUC factor. 

The importance of these emission factor assumptions can be further illustrated by comparing the average carbon loss per hectare assumed for conversion of grassland to cropland in Qin et al. (2016) (for the Corn2 scenario with CCLUB emission factors) with average values estimated by Spawn, Lark, & Gibbs (2012) for grassland conversion in the Lark et al., (2015) results. Spawn et al. (2012) estimate an average loss of about 50 tonnes of carbon per hectare for conversion of new land to corn agriculture. Qin et al. (2016) in contrast assume an average increase by 13 tonnes of carbon stored per hectare.


Recent analysis with the GTAP model has suggested that ILUC emissions from corn ethanol may be much lower than was calculated in either the initial or revised ILUC analysis for the California Air Resources Board, or by the US Environmental Protection Agency. The results from Qin et al. (2016) using GTAP with the CCLUB emission factor model have even been integrated into the GREET lifecycle analysis model. Comparing these GTAP results to land use changes identified by other sources, however, suggests that total U.S. land use changes may be slightly underestimated, and strongly suggests that the role of cropland pasture as a land source may be significantly exaggerated. These features of recent GTAP modeling, combined with highly questionable assumptions about carbon sequestration when cropland pasture is converted to corn cropping have likely led to significant underestimates of likely ILUC emissions.


Argonne National Laboratory. (2017). The Greenhouse gases, Regulated Emissions, and Energy use in Transportation Model. Retrieved August 2, 2018, from

California Air Resources Board. (2014). Appendix I – Detailed analysis for indirect land use change. Sacramento, CA. Retrieved from

Hendricks, N. P. (2018). The Impact of the Renewable Fuels Standard on Cropland Transitions.

Lark, T. J., Hendricks, N. P., Pates, N., Smith, A., Spawn, S. A., Bougie, M., … Gibbs, H. K. (2019). IMPACTS OF THE RENEWABLE FUEL STANDARD ON AMERICA ’ S LAND AND WATER. In American Academy for the Advancement of Science (AAAS) Annual Meeting (pp. 1–13). Washington DC.

Lark, T. J., Salmon, J. M., & Gibbs, H. K. (2015). Cropland expansion outpaces agricultural and biofuel policies in the United States. Environmental Research Letters, 10, 44003. Retrieved from

Malins, C., Searle, S. Y., & Baral, A. (2014). A Guide for the Perplexed to the Indirect Effects of Biofuels Production. International Council on Clean Transportation. Retrieved from

Qin, Z., Dunn, J. B., Kwon, H., Mueller, S., & Wander, M. M. (2016). Influence of spatially dependent, modeled soil carbon emission factors on life-cycle greenhouse gas emissions of corn and cellulosic ethanol. GCB Bioenergy, 8(6), 1136–1149.

Spawn, S. A., Lark, T. J., & Gibbs, H. K. (2012). U.S. cropland expansion released 115 million tons of carbon (2008-2012). In America’s Grasslands Conference 11/15/2017, Fort Worth, TX. Fort Worth. Retrieved from

U.S. Environmental Protection Agency. (2010). Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis, 2010(February 2010), 1109., February 2010

Wright, C. K., Larson, B., Lark, T. J., Gibbs, H. K., Salmon, J. M., Gibbs, H. K., … Gibbs, H. K. (2017). Recent grassland losses are concentrated around U. S. ethanol refineries. Environmental Research Letters, 12(4), 044001.


This article was supported by the National Wildlife Federation.

[1] Malins, C., Plevin, R., Edwards, E. (2019). Accentuating the positive – how robust are reductions in modeled estimates of the indirect land use change from conventional biofuels?

[2] Referred to in the CCLUB module of GREET as the ‘Corn 2013’ scenario.

[3]  and


[5] Calculated from AFDC data on total corn use for ethanol

[6] In the case of feed corn, this would primarily mean reducing demand for animal products and therefore reducing use of corn as animal feed.

[7] Corn and soy are often grown in rotation.


[9] Of a total 1.9 million acres of conversion of cropland pasture plus other grassland in GTAP-CCLUB, 95% was from cropland pasture.

[10] In 2008-12 nineteen thousand hectare of forest conversion for every additional billion gallons of ethanol demand, about half of it identified by (Wright et al., 2017) as within 100 miles of ethanol refineries  compared to 6 thousand hectares per billion gallons in (Qin et al., 2016).

Accentuating the positive – has optimism bias driven reductions in ILUC estimates?

29 Mar 19
Chris Malins
No Comments

Indirect land use change, often abbreviated to ILUC, refers to the expected expansion of agricultural area (and subsequent release of carbon from biomass and soils) when biofuel policies increase demand for agricultural commodities. In 2008, research led by Tim Searchinger using the FAPRI economic model[1] suggested that accounting for these ILUC emissions might eliminate the presumed climate benefit of using corn ethanol, but more recent analyses have suggested that the ILUC effect is much smaller than originally suggested. ILUC cannot be measured directly, because by definition it is an indirect effect where causality is mediated by market mechanisms, and so researchers have developed complex systems of economic modeling to produce scenarios for the way we might expect the agricultural system to react to increased biofuel demand.

In the U.S., one of these models has taken an increasingly dominant role in ILUC analysis – GTAP, the computable general equilibrium economic model of the Global Trade Analysis Project. Results from GTAP are used in the Low Carbon Fuel Standard regulation in California, and have been integrated into the GREET (Greenhouse gases, Regulated Emissions, and Energy use in Transportation) model of the Argonne National Laboratory. With a steady stream of results, updates and model adjustments, the ILUC research using GTAP is probably the most prolific ILUC modeling strand in the world in terms of the number of papers published and ILUC estimates recorded.

In 2009 when the California Air Resources Board first regulated for ILUC emissions, they estimated the ILUC from corn ethanol production at 30 gCO2e/MJ[2] and the ILUC from soy biodiesel at 62 gCO2e/MJ. More recent GTAP estimates, however, have been as low as 10 gCO2e/MJ for both fuels (e.g. Chen et al., 2018; Qin, Dunn, Kwon, Mueller, & Wander, 2016). The figure below shows the downward tendency over time in ILUC estimates reported for soy biodiesel using GTAP. A reduction by 20 gCO2e/MJ or more in estimated ILUC emissions can significantly affect our understanding of how much climate benefit a fuel delivers, and by affecting regulatory treatment of biofuels a lower lifecycle carbon emissions value can have large financial implications.

Point ILUC factor estimates for US corn ethanol obtained with variants of the GTAP-BIO model

Given the very large reported reductions in estimated ILUC in some of the lower results compared to the initial CARB assessments, up to a 60% reduction for corn ethanol and a 90% reduction for soy biodiesel, it is useful to understand why the numbers have fallen so far. A forthcoming paper co-authored by Chris Malins of Cerulogy reviews in detail innovations introduced to the modeling framework since 2009, and concludes that the evidentiary basis for many of the changes that have led to reduced ILUC estimates is weak.

The review looks in particular at changes made to the model in five areas over the past decade, reviewing the strength of evidence for the changes made and the impact of the changes: the intensive yield response; the role of cropland pasture; cropping intensity change; yield at the extensive margin; and the emission factors associated with land use changes.

Intensive yield change

The intensive yield change refers to the amount that agricultural yields can be increased in response to increased demand. A strong intensive yield response means less extra land is needed to produce biofuel feedstock. The GTAP modelers have generally preferred a value of 0.25[3] as a central estimate for this parameter, based on analysis of historical corn yields. There have however always been highly divergent expert opinions on this question. For example, economist Steve Berry argued in a report for the Air Resources Board that there is no robust econometric evidence for the value of 0.25, and that ta value of 0.1 would be more appropriate given the data available.[4]  

While the Air Resources Board ended up using a reduced average yield response value in its 2014 regulatory modeling (compared to the value of 0.25 used in its 2009 work), subsequent studies with GTAP have continued to be based on the parameter of 0.25. Indeed, since 2017 the response has effectively been increased even further, due to a model amendment which implemented differentiated yield response by region, but increased it in more regions than it was reduced in. These asymmetric changes were made despite presenting no evidence that the previous yield response had been too weak on average.

For assessing the ILUC of soy biodiesel, it is also important to note that there is some evidence in the literature that the yield response for soybeans is much weaker than that for corn. Applying differentiated yield responses for other crops would therefore likely have increased ILUC results, especially for soy biodiesel. It is doubly unfortunate for the soy analysis that while regional yield responses have been implemented in a way that strengthens the yield response for all feedstocks, but no action has been taken to differentiate in the modeling the yield response of different crops.     

Cropland pasture

Cropland pasture is a land category in USDA reporting that refers to areas that are currently pastured but that have been used for annual crops in the (relatively) recent past. This land category was added to the GTAP database for the U.S. and Brazil in 2010, presumed to have lower carbon stocks than other land types. The model assumes that conversion of cropland pasture to crop production is easier than conversion of forest or permanent pasture, and so introducing this land category immediately reduced ILUC emissions. Since then, a sequence of amendments to model structure and parameters have continually increased the importance of cropland pasture conversion in the model results.

While cropland pasture has taken on a central role in the GTAP modeling, the evidence that it is so readily converted to cropland is actually rather patchy. Statistical data on cropland pasture area is complicated by the fact that the wording of USDA questionnaires has changed, which USDA believe may have caused a large reduction in reported areas. Despite the lack of evidence to support such a major change to the modeling, a large elasticity of cropland pasture yield to rent was subsequently introduced, also without any direct evidence to support the value chosen. Changes made in 2013 to regionalize the land use changes associated with price increases further reduced the likelihood of forest or permanent pasture conversion in many regions, based on weak evidence, thus again increasing the role of cropland pasture still further. An alternative regional approach suggested in a 2012 paper would have given a more balanced result, but has subsequently been ignored.   

Cropping intensity

Cropping intensity refers to the number of times a given area of land can be harvested in one year. Traditional annual cropping includes only one harvest, but especially in warmer climates it may be possible to grow two or even more crops in a single year. This possibility was not explicitly included in the original GTAP ILUC modeling, although some experts argued that it was taken into account through the yield response. Since 2017, increased cropping intensity has been supported by GTAP modeling, and as one might expect has further reduced predicted ILUC emissions. Most countries do not directly report on cropping intensity, and thus analysis of cropping intensity changes has tended to rely on comparing harvested and planted area data, even though datasets may not be readily comparable. No analysis has been presented that robustly demonstrates a link between demand or price increases and increased cropping intensity – major weaknesses in one of the studies used to justify introducing a strong response are documented in the appendix to a previous paper.[5]  

Extensive yield

In general, we expect that farmers will preferentially plant the best land available to them, and therefore that yields will be higher on land already being farmed than on areas where agriculture expands. In the earlier GTAP modeling it was assumed that new agricultural land would have a yield two thirds that achieved on land already farmed, but this was revised based on work published in 2010. A global assessment was undertaken of expected ‘net primary production’ (NPP) using an ecosystem model, and it was assumed that the difference modeled in NPP between areas under cropping and areas under natural cover would indicate the likely difference between yields on existing land and newly farmed land. Using the new system generally resulted in increased assumed yields on newly farmed land, but questions remain about whether the analytical approach used was appropriate, and whether it had adequate resolution. In particular, it has not been explained why for many regions the new results seemed to run counter to economic logic – notably, why would farmers not have been more successful in identifying and farming the most promising land? These questions have never been resolved.

Emission factors

The economic model provides a prediction for which types of land may be converted due to biofuel demand, but turning this into an emissions estimate requires making assumptions about the carbon stock changes following land use changes. Since 2014, the ILUC estimates included in the GREET model have not by default used the emission factors developed for the California Air Resources Board, instead relying on the Carbon Calculator for Land Use Change from Biofuels Production (CCLUB).[6] This model results in systematically lower ILUC estimates than are obtained when using alternative carbon stock change values.

One major reason for the difference comes back to cropland pasture. The CCLUB modeling assumes that there are significant increases in carbon sequestration when cropland pasture is converted to annual cropping. The basis for this is a somewhat baffling underlying modeling decision to treat cropland pasture as if it was under annual crops from 1976 to 2010, and then converted to corn or soy agriculture in 2011. This directly contradicts the definition of cropland pasture, which of course must have been used for pasture immediately before conversion. This odd modeling choice is compounded by some difficult-to-justify assumptions about increased carbon sequestration in corn and soy agriculture compared to other annual crops. The upshot is that cropland pasture conversion, which in real life almost certainly results in carbon losses, is treated as a carbon gain. For the soy biodiesel pathway, there are also difficult-to-justify modeling choices made that result in underestimated emissions from peat drainage associated with oil palm expansion in Southeast Asia – for instance averaging prevalence of peat soils across administrative districts with no reference to their size or suitability for palm agriculture, so that the city of Jakarta is weighted equally with areas on the oil palm frontier in Kalimantan.  


The more detailed review of these issues paints a picture of a systematic and chronic willingness within the community that has developed the ILUC modeling in GTAP to make modeling decisions that reduce ILUC estimates, even where the evidence is weak. The resulting downward bias in model outcomes is compounded by an apparent corresponding resistance to invest time in model amendments that would increase ILUC estimates, even where the evidence base is relatively strong. Even if strong evidence was available for the modeling changes made, choosing to focus only on changes that reduce ILUC emissions would gradually introduce a downward bias into the results. When the bar is set low for the quality of supporting evidence, this bias manifests itself in the results very quickly.  

The result is a long-term optimism bias pushing reported ILUC emissions ever lower, so that it is impossible to conclude with any confidence to what extent the reported reductions represent a real improvement in our understanding of ILUC and to what extent they reflect a subjective decision by a small group of modelers that the numbers ought to be smaller.  

ILUC emissions are important – they are used in regulatory analysis, and our understanding of ILUC informs our understanding of whether biofuel policies are effective tools to reduce climate change. New, lower ILUC results are often presented as having great policy importance – one recent paper that presented a lower ILUC value for corn ethanol concluded that, “it is important to note the importance of the new results for the regulatory process. …[because] the current estimate values are substantially less than the values currently being used for regulatory purposes.”[7]

On the contrary, reviewing the development of the GTAP model leads not to the conclusion that regulators ought to unquestioningly adopt the most recent analysis, but that it is vital that any regulatory assessment of ILUC emissions should be balanced and should not rely too heavily on the work of any single modeling group. The California process has always involved public consultation and expert workgroups, providing a degree of balance. In contrast, the ILUC results used in the GREET model are not tempered by any formal consultative process. Any future revisions to regulatory ILUC values must involve an open and honest assessment of the evidence base for all model features, and should ensure that possible model changes get equal consideration, regardless of whether they would be expected to increase ILUC emissions outcomes or reduce them.


California Air Resources Board. (2009). Proposed Regulation to Implement the Low Carbon Fuel Standard, Volume I, Staff Report: Initial Statement of Reasons. Sacramento, CA: California Air Resources Board. Retrieved from

California Air Resources Board. (2014). Appendix I – Detailed analysis for indirect land use change. Sacramento, CA. Retrieved from

Chen, R., Qin, Z., Han, J., Wang, M. Q., Taheripour, F., Tyner, W. E., … Duffield, J. (2018). Life cycle energy and greenhouse gas emission effects of biodiesel in the United States with induced land use change impacts. Bioresource Technology, 251, 249–258.

Hertel, T. W., Golub, A. A., Jones, A. D., O’Hare, M., Plevin, R. J., & Kammen, D. M. (2010). Effects of US Maize Ethanol on Global Land Use and Greenhouse Gas Emissions: Estimating Market-mediated Responses. BioScience, 60(3), 223–231.

Qin, Z., Dunn, J. B., Kwon, H., Mueller, S., & Wander, M. M. (2016). Influence of spatially dependent, modeled soil carbon emission factors on life-cycle greenhouse gas emissions of corn and cellulosic ethanol. GCB Bioenergy, 8(6), 1136–1149.

Taheripour, F., Cui, H., & Tyner, W. E. (2017). An Exploration of agricultural land use change at the intensive and extensive margins: implications for biofuels induced land use change. In Bioenergy and Land Use Change (pp. 19–37). Retrieved from

Taheripour, F., & Tyner, W. E. (2013a). Biofuels and Land Use Change: Applying Recent Evidence to Model Estimates. Applied Sciences, 3, 14–38. Retrieved from

Taheripour, F., & Tyner, W. E. (2013b). Induced Land Use Emissions due to First and Second Generation Biofuels and Uncertainty in Land Use Emission Factors. Economics Research International, 2013, 1–12.

Taheripour, F., Zhao, X., & Tyner, W. E. (2017). The impact of considering land intensification and updated data on biofuels land use change and emissions estimatesmen. Biotechnology for Biofuels, 10(1), 1–16.

Tyner, W. E., Taheripour, F., Zhuang, Q., Birur, D. K., & Baldos, U. (2010). Land Use Changes and Consequent CO2 Emissions due to US Corn Ethanol Production: A Comprehensive Analysis. Department of Agricultural Economics, Purdue University, 1–90. Retrieved from


This article was supported by the National Wildlife Federation.


[2] Grams of additional carbon dioxide equivalent emissions due to indirect land use change for every megajoule of biofuel produced.

[3] This is the elasticity of yield to own price, the fraction by which the yield of crops increases for every 100% increase in the price of that crop.





Navigating the maize

13 Jul 17
Chris Malins
No Comments

Update – An updated critique reacting to republished results from USDA is available here:

Earlier in 2017, the United States Department of Agriculture published a study undertaken by the consultants ICF International that included a reassessment of the lifecycle greenhouse gas intensity of corn ethanol. The report concluded that corn ethanol’s greenhouse gas intensity is already lower than the level projected for 2022 by the Environmental Protection Agency in the regulatory impact analysis for the Renewable Fuel Standard 2, a conclusion that has been embraced by advocates for the corn ethanol industry.

In this study, commissioned by the Clean Air Task Force and National Wildlife Federation, we provide a critical review of the ICF report. We find that the report shows a lack of balance, systematically emphasising evidence that could suggest that the performance of corn ethanol is better than previously modeled, while understating or ignoring conflicting evidence. More problematic even than this lack of balance, however, we find that the analysis in the report is riddled with errors of methodology and data, many of them at the most basic level, so as to render some of the results presented essentially meaningless. Given the many issues identified, we conclude that the work presented is wholly inadequate to justify any firm conclusion on whether the corn ethanol emissions estimates made by EPA could or should be revised down.



In the report “Navigating the maize” it is stated that, “The ICF report appears to double count the emissions benefits associated with the production of ethanol co-products”. Following clarification of the ICF approach in subsequent papers, we realise that we had misunderstood the approach used and that this is not the case – co-products are not double counted by ICF.

Our misunderstanding arose from a methodological difference between the ICF analysis and the original EPA regulatory impact assessment that was not clear to us from the original documentation. In the EPA work, agricultural emissions are assessed only on the net increase in corn production to meet the mandate (this is a consequential approach). The effect of distillers’ grains in reducing net corn demand is already included before agricultural emissions are calculated. The ICF reassessment makes reference to the net demand change value from the EPA work, and therefore we had understood that the role of co-products was implicitly included by use of this net demand value. Based on clarifications given in updates to the ICF work  we now understand that ICF in fact analyse emissions for an average corn acre and then apply these based on gross corn demand. The step that makes reference to the EPA net demand value is redundant (this term is effectively added on both the numerator and the denominator of the equation so that it is cancelled out) . Based on this understanding it is indeed appropriate to apply a co-product credit under this methodology. 

Corn on the scales