Comments on “Environmental Outcomes of the US Renewable Fuel Standard”

Business News/Analysis Energy (DOE)Farming/Growing Federal Agency/Executive Branch Federal Regulation Feedstock Feedstocks Field/Orchard/Plantation Crops/Residues Infrastructure Opinions Policy R & D Focus Sustainability Wisconsin

March 22, 2022

Print This Article Share it With Friends

by F. Taheripour, S. Mueller, H. Kwon, M. Khanna, I. Emery, K. Copenhaver, M. Wang (Argonne National Laboratory) The Systems Assessment Center and its collaborators provide a detailed technical review of a recently published article “Environmental Outcomes of the US Renewable Fuel Standard” by Lark et al. (2022). Our review explored modeling approach and data sources for land use changes, types of land conversions, and soil organic carbon changes, among other parameters, in the Lark et al. study, which resulted in significantly high greenhouse gas emissions of US domestic land use changes of corn ethanol presented in that study. READ MORE Download paper

Leading Researchers Contradict Hit Piece on Ethanol’s Environmental Impacts (Renewable Fuels Association)

Argonne Debunks Recent Negative Ethanol Study (Energy.AgWired.com)

Other Related articles

Excerpt from Argonne National Laboratory: Lark et al. (2022) recently published “Environmental Outcomes of the US Renewable Fuel
Standard” and addressed domestic land use change (LUC) of corn ethanol and associated greenhouse gas (GHG) emissions that are potentially caused by the U.S. Renewable Fuel Standard (RFS), as introduced in the 2005 Energy Policy Act and in the 2007 Energy
Independence and Security Act (EISA). To do so, they considered the corn ethanol volume changes and LUC between 2008 and 2016 1.

In their assessment, Lark et al. assumed a business-as-usual (BAU) scenario (representing the goals of RFS1 for ethanol volume, as adopted in the 2005 Energy Policy Act by Congress, between 2008 and 2016), a new scenario (representing the goals of RFS2 for ethanol volume, as
adopted in the 2007 EISA by Congress, between 2008 and 2016) to determine domestic LUC due to the RFS2. With no integrative modeling exercise, the authors simply calculated the average of the annual differences between the goals of RFS1 and RFS2 (5.5 billion gallons [Bgal]) and considered that volume of ethanol as the average annual contribution of RFS2 to new ethanol consumption between 2008 and 2016. Instead of using an integrated, coherent framework, as is the case with equilibrium models, in which changes in crop prices and associated LUCs at the intensive and extensive margin are determined simultaneously, Lark et al. applied a few loosely connected empirical methods to examine the impact of the RFS2 on three crops (corn, soybeans, and wheat). They estimated the short-term increases in commodity prices between 2008 and 2016 induced by the RFS2 and estimated that the prices of corn, soybeans, and wheat would increase by 30%, 20%, and 20%, respectively, due to an increase in the annual consumption of ethanol by 5.5 Bgal.

In the next step, Lark et al. used the Cropland Data Layer (CDL) from the United States Department of Agriculture (USDA) in combination with some other information on returns on cropland to estimate the probabilities of land transitions between cropland, pasture land, and Conservation Reserve Program (CRP) land. Using their projected increases in the prices of corn, soybeans, and wheat in combination with the estimated land transition functions, Lark et al. calculated that the area of corn plantation, adjusted for distiller’s dried grains (DDG), would increase by 2.8 million hectares (Mha) due to the RFS2, and that would lead to an increase in cropland area by 2.1 Mha. They showed that an overwhelming share of land conversion due to the RFS2 would be conversion of CRP land to active cropland. Lark et al. assigned a set of significantly large land use emissions factors to the CRP land conversion and likely double counted the N2O emissions in adding their LUC emissions to the rest of life-cycle analysis (LCA) emissions of corn-based ethanol, leading to the conclusion that the GHG emissions (commonly called carbon intensity) of ethanol are at least 24% higher than those of gasoline.

After a detailed technical review of the modeling practices and data used by Lark et al., we conclude that the results and conclusions provided by the authors are based on several questionable assumptions and a simple modeling approach that has resulted in overestimation of the GHG emissions of corn ethanol. In what follows, we present the general findings of our review.

Our review is organized in nine sections. In the first section we discuss their estimation of land conversions. The second section addresses systematic overestimation of soil organic carbon (SOC) changes by the authors. The issue of double-counting of N2O emissions in the Lark et al. LCA is discussed in the third section. In the fourth section we refer to some inconsistencies in results provided by Lark et al. We then discuss misattribution of ethanol volumes to the RFS2 by those authors in Section 5. The assessment of impacts of yield improvement and DDG offsets on the demand for cropland are addressed in Section 6. The estimation of price impacts of the RFS is discussed in Section 7. Section 8 outlines deficiencies in modeling land transition. Finally, we conclude our findings in Section 9.

1. Land Conversions
Land use changes identified by Lark et al. are likely representing conversion of fallow/idle land to crops rather than conversion of permanent grasslands and thus is unlikely to result in a large carbon debt upon conversion.

...

2. Systemic Overestimation of Soil Organic Carbon Changes
Lark et al. likely overestimated soil carbon loss by a factor of two to eight for land use change by apply carbon response functions that are relevant for conversion of native or undisturbed grassland to cropland and not for CRP and cropland pasture to cropland.

...

3. Double-counting of N2O Emissions and Omissions of Avoided Emissions
Lark et al. appeared to have double-counted the N2O emissions with fertilizer use for corn farming by adding 9 gCO2e/MJ of ethanol to the remaining LCA results of corn ethanol and overlooked that these were already included in the corn farming related emissions as is the case in most LCA calculations, such as those from the Greenhouse gases, Regulated Emissions, and Energy use in Technologies (GREET) model.

...

4. Inconsistencies in Results Obtained by Lark et al.
We note several findings reported by Lark et al. that are difficult to rationalize, as we illustrate in Figure 7. First, consider Panel A of Figure 7, which shows the projected changes in carbon intensity (CI) in cropland by county obtained from the Lark et al. results presented in their Figure 2. Panel A of Figure 7 reveals inexplicably negative CIs. Why are there negative changes in CI per hectare in cropland? When cropland increases, with no improvement in SOC due to land management or high carbon crops, carbon emissions should also increase, and when cropland declines, carbon emissions should decrease. A negative CI implies an error and that the results were not reviewed for accuracy. This figure also reveals extremely large values of the examined ratio up to more than 12 Gg CO2e/ha. These extreme CI values per hectare of land are not explained or justified.

...

5. Misattribution of Ethanol Volumes to the RFS2
Lark et al. attributed 5.5 Bgal of ethanol per year to RFS2 between 2008 and 2016 by comparing the volume under RFS2 and RFS1 without considering other drivers of ethanol production. The expansion in the biofuel industry (including corn ethanol and other biofuels), even in the short time period from 2008 to 2015, occurred due to many drivers, including but not limited to changes in non-RFS biofuels supporting policies (such as ban of MTBE in gasoline blends but needed oxygenate in gasoline blends and tax credits), changes in crude oil price, changes in demand for gasoline, the 10% blend rule and the blend wall issue, changes in livestock industry and its demand for feed crops and other feeds (e.g., DDG and meal products).

6. The Amount of LUC Attributed to Corn Ethanol Without Careful Consideration of Yield Increase and DDG Offsets

...

7. Estimation of Price Effects of the RFS2
The validity of picking the time period of 2006-2010 to assess price impacts with the 5.5 Bgal for RFS2 between 2008 and 2016 is questionable.

...

8. Modeling Land Transition
Lark et al. did not recognize cropland pasture as a sub-category of cropland in their analyses and perhaps treated this type of land as pasture land or fallow land. This misidentification and the method used by the authors to assess land return is likely to have artificially led to the additional demand for active cropland being met largely by CRP land and not by cropland pasture.

...

9. Conclusions
Over the past 15 years many papers have studied LUC due to biofuel production and policy. In the absence of any observed evidence, some early papers published on this topic claimed that producing ethanol in the U.S. would generate major deforestation in the country and elsewhere and that the emissions associated with the conversion of natural land to cropland would cause GHG emissions that would increase the carbon intensity of ethanol to a level higher than the carbon intensity of gasoline. Over time various studies showed that those early papers overstated the magnitude of deforestation due to biofuels. The more advanced analyses, relying on the recent actual observations on land use changes, showed that intensification in crop production due to yield improvement and cultivation of idled cropland, and shifting demand form unwanted feed crops to biofuels by-products have jointly eliminated the need for conversion of natural land to cropland for biofuel production and hence provided significantly lower estimates for land use change emissions due to biofuels.

In their recent publication, Lark et al. have at least clearly confirmed that there is no evidence supporting deforestation and conversion of natural land to crop production due to biofuels or any other driver. Hence, from this perspective they confirmed the findings of other recent publications that there is no evidence of deforestation in the U.S. due biofuels.

However, Lark et al. simply assumed that RFS was responsible for an expansion in ethanol consumption by 5.5 Bgal a year. With this premise and by using a few loosely connected empirical methods, the authors evaluated the impact of the assumed increase in ethanol volume on three crops (corn, soybeans, and wheat). Relying on the short-term increases in the prices of these commodities during the time period of 2006-2010, assuming that these price increases will sustain over 30-year time period, projecting return on cropland and pasture land using problematic outdated projections for future crop prices and many other variables, applying CDL data with low accuracy in detecting land use types, ignoring the fact that reduction in CRP land area was due to CRP funding cut by Congress, Lark et al. projected the assumed increase in ethanol consumption by 5.5 Bgal would lead to an increase in corn area by 2.8 Mha over 30 year time horizon and that increases the area of active cropland by 2.1 Mha. Their projection suggested that most of the expansion in active crop land comes from conversion of one type of unused cropland (CRP) to crop production. Regardless of the accuracy of this projection with respect to the type of unused land and its magnitude, the fact that area of active cropland could increase by cultivation of unused land in the U.S. due to additional demand for biofuels is not new finding and has been addressed and well noted in the existing literature. However, the type of unused cropland that has been cultivated is uncertain. Finally, while the existing literature concludes that marginal cropland is not a rich soil carbon content land, Lark et al. assigned very high emissions factors to CRP land and concluded that emissions due to LUC for corn ethanol was large. In addition, in a misunderstood LCA practice, involving double counting and neglecting various sources of emissions savings due to biofuel production, Lark et al. maintained that the carbon intensity of corn ethanol is larger than that of petroleum gasoline.

As presented above, in this technical review of Lark et al., we address a few key and apparent issues that need more careful examination of Lark et al. We highlight these issues below:

• The Lark et al. modeling approach that followed only short-run changes in 2008-16 in individual crops - corn, soybeans, and wheat - missed the long-run pattern of changes in the mix of crops and the combined effect across all crops produced in the U.S. This short- term analysis generated a higher demand for active cropland and overestimated land conversion from CRP to crop production than what is consistent with observed trends in data.

• The arbitrary choice of working with CDL data for the short-time segment of 2008-16 does not represent the U.S. long-term cropping pattern. The farming sector in some years deviates from its long-term pattern in response to short-term shocks in commodity prices and then returns to its long-term pattern when short-term price shocks disappear.

• The Lark et al. modeling approach is too limited to effectively consider the drivers of ethanol industry and its interaction with other industries including the cropping and livestock industries. The Lark et al. modeling approach projected an increase in the planted area of corn by 2.8 Mha by considering one single factor of 5.5 Bgal ethanol in the U.S. For the same time period, due to all drivers, the average of annual changes in the harvested area for corn may have been -0.25 Mha in the U.S. What can justify the difference between the Lark et al. projection and actual observations? Are there some factors that canceled out the impacts of RFS2? Or has the modeling practice missed some important drivers?

• The RFS has begun in 2005, continued until today, and could be continued in future. Picking an arbitrary time segment out of this long time period can lead to distorted results. Picking the time period of 2008-16 over which there was a major increase in crop prices (not only due to biofuels) and assigning the estimated land conversion for that period to a policy that will remain in place for a long time period can result in biased land use change attributable to the RFS. This biased attribution can certainly cause overestimated LUC magnitude for the RFS.

• The short-term changes in crop rotation does not reflect the long-run pattern of corn-soy rotation. Furthermore, Lark et al. with no justification picked a subset of their selected study regions to calculate changes in crop rotation. This selection eliminated areas with large shares in soybeans or wheat.

• Lark et al. did not recognized cropland pasture as a sub-category of cropland in their analyses and perhaps treated this type of land as pasture land or fallow land. This misidentification and the method used by them to assess land return artificially push the need for additional active cropland to CRP land.

• The CRP land left this program simply because there was no budget to keep them in the program. Assigning a portion expired CRP land to RFS (or any other biofuel) is problematic.

• To estimate probability of land transformation, Lark et al. used outdated and inaccurate projections for future crop prices and several other variables. In addition, in an ad hoc manner, they assumed costs of crop production remain constant over the 10-year projection period for the stream of expected returns on cropland. These made their land transformation projection questionable.

• We tested how much land expansion could be expected, given yield increases over the considered time frame as well as land offsets provided by DDG animal feed in order to meet the Lark et al. assumed 5.5 Bgal of ethanol stimulated by the RFS. Our analysis shows that at high level, yield increase on 2008 year corn acres over-compensate for ethanol demand. Even with ethanol production the 2008 corn footprint would still be down by 4.26 million acres. Ethanol demand may not drive an expansion above the 2008 year corn footprint but other factors including urban development may shift the corn footprint around.

• Our analysis of the cropland expansion data layer presented in Lark et al. supporting information revealed that areas identified by the authors as expansion to cropland may often be short-term fallow/idle lands (less than 10 years). In fact, many parcels identified by Lark et al in their “Cropland Expansion Layer” appear to be prime examples of land on the margin that is toggling between agriculture and fallow/idle state based on crop price signals. This would likely result in a systemic overestimation of SOC emissions for these parcels. Without such observation data to support their estimates, Lark et al. should have considered their results with high uncertainty.

• The authors missed the fact that corn ethanol LCA studies capture the N2O emissions from any change in nitrogen applied to corn in farming GHG emissions. As a result, they may have double-counted N2O emissions in their LUC emissions. They also failed to take
into account emissions savings due to avoided consumption and improvements in livestock industry induced by using biofuel by products.

• Lark et al. projected that in many counties area of cropland would increase largely (up to 2000 hectares for 1 hectare of changes in corn area). What justifies these magnificent changes? These large changes suggest that Lark et al. overestimated the land transformation elasticities.

• Lark et al. projected that the area of corn increases in 1,353 counties and decreases in 349 counties. In addition, their results showed changes in cropland in 126 counties with zero change in corn area. These odd results strongly suggest that the Lark et al. modeling approach may have considered reshuffles of crops among geographic locations of crop production with significant LUC emission implications.

1In their main manuscript, Lark et al. referred to 2008 to 2016 as the eight years of their assessment time period. From this specification, it seems that the authors refer to changes in eight years of 2008, 2009, 2010, …, 2015. However, in various other places of their paper and supplemental information (SI), they referred to 2009-2016 as their study period. In this note we refer to changes in the eight years from 2008 to 2015, unless we quote Lark et al. and perform analyses where they clearly reference 2016 as the end year.

2 First, we accessed data via https://zenodo.org/record/3905243#.YjeNMOrMJPb and downloaded the geodatabase of US_land_conversion_2008-16.gdb.zip. Second, we opened layer "ytc" (which is described on the webpage as areas in cropland expansion "areas converted to crop production between 2008 and 2016" with the year of expansion listed in the polygon) in ArcGIS software. Third, we loaded Lark et al. "cropland expansion" layer into Google Earth Engine (GEE) (See Figures 1-5) using a GEE javascript.

3 Based on data from FAOSTAT. READ MORE