Don't throw out California's ILUC factors yet

Soy biodiesel is generally understood to have high indirect land use change (ILUC) emissions.  For example, California assigns an ILUC factor of 29 gCO2e/MJ to biodiesel in its Low Carbon Fuel Standard (LCFS) program, while the European Commission’s latest study estimates 150 gCO2e/MJ. But a new study led by Argonne National Laboratory (ANL) has a surprising conclusion: ILUC emissions from soy biodiesel are only 4-10 gCO2e/MJ. This study uses the same underlying economic model as California’s 2015 analysis for the LCFS and yet estimates ILUC emissions 66%-85% lower than California does. How is this possible?

Much of this massive reduction in ILUC emissions appears to come from the ANL study’s assumptions around palm oil. Our research has shown that U.S. soy biodiesel demand indirectly leads to increased production of palm oil because diverting soy oil from use in food and chemicals drives greater consumption of imported palm oil as a substitute in those uses. ANL’s analysis reflects this substitution effect and finds that U.S. soy biodiesel demand leads to oil palm expansion in Indonesia and Malaysia. Palm oil is a big deal in ILUC modeling because it is associated with very high greenhouse gas (GHG) emissions from peat drainage and deforestation. Not all oil palm is grown on peat soils, and in fact the proportion of new palm plantations that are established on drained peat is a key factor in determining the magnitude of ILUC emissions for palm oil and for any related biodiesel pathway, like soy.

Here’s where things start to get muddy. California assumes 28%-30% of all new oil palm plantations are established on peat soils, citing a 2010 Joint Research Center (JRC) report, while the ANL study adopts an older 2009 analysis by Winrock International performed for the U.S. Environmental Protection Agency’s (EPA) GHG assessment for the Renewable Fuel Standard (RFS) program. The ANL study describes Winrock’s analysis as finding that only 5% of land conversion occurs on peatland and notes that the difference in these assumptions is a “major discrepancy,” as “peatland loss can drive ILUC-related GHG emissions in some cases or models.”

Who’s right? The ANL study calls for “further efforts [and] more up-to-date data.” The study then cites analyses of oil palm expansion from 2011 and 2012 and refers to them as “new observations [that] may serve to improve models,” even though the authors still choose to use Winrock’s older 2009 results in their model. To see whether the analyses used and cited in the ANL study are actually the most “up-to-date” and “new,” we run through the available information on oil palm expansion on peat and summarize it in the table below.

First of all, the Winrock analysis does not actually report that 5% of oil palm expansion occurs on peat. Indeed, it does not present a value for the fraction of oil palm planted on peat soils. It only analyzes the proportion of total land area with peat soils. In spreadsheets available on the federal docket, Winrock provided EPA with estimates of the prevalence of peat soils in each of 45 states, provinces, and territories in Indonesia and Malaysia. Not only is this data not directly applicable to the fraction of oil palm expansion on peat, but the figure of 5% quoted in the ANL study represents the arithmetic mean across the 45 jurisdictions, treating tiny, peat-free provinces like Jakarta with equal weight as large, peat-heavy jurisdictions. In its analysis for the RFS2, EPA weights these jurisdictions by the amount of land use change historically observed in each province or state; using this approach, 10% of predicted forest to perennial conversion and 7% of total natural land to perennial conversion occurs on peatland for Indonesia and Malaysia combined.

Several other analyses have empirically assessed the share of oil palm plantations on peat, and these analyses provide estimates substantially higher than 5%. The 7% peat share in Winrock’s 2009 analysis represents the very low end of the range of all published studies that have attempted to ascertain the prevalence of oil palm plantations on peatlands, summarized in the table below. It is also one of the oldest studies available on peatland area, and one that does not even assess the prevalence of oil palm plantations on peat.

More recent studies done between 2011 and 2016 (Koh et al., 2011Miettinen et al., 20122016) use a more robust method: manually identifying oil palm plantations on drained peatland in high resolution satellite images. These newer studies are particularly useful because they provide a record of the prevalence of oil palm on peat at several different points in time from 1990 to 2015. The figure below shows an estimate of the share of total oil palm plantation area on peat by comparing the results of these studies with total harvested oil palm data from FAOSTAT. The share of total oil palm area on peat rose from 10% in 1990 to 23% in 2015. This clearly demonstrates that the share of new oil palm plantation area established on peat as opposed to other types of soils has accelerated over this time period.

New and total oil palm plantations on peatland, Indonesia and Malaysia
Figure: Share of new and total oil palm plantations on peatland from 1990-2015 in Indonesia and Malaysia from Miettinen et al. (2012; 2016) and FAOSTAT

The figure also shows the share of new oil palm plantations established on peat, calculated by comparing the change in palm-on-peat area with the change in total harvested palm area for each year that data is available. This is a much more relevant measure for predicting future land-use change as a consequence of biofuel policy because it tells us where new plantations are being established, rather than lumping new plantations in with old ones. By this measure, 33% of new oil palm plantations established over the period 2010-2015 were on peat soils in Indonesia and Malaysia. This is very close to a projection of 32% oil palm expansion on peat for the period 2010-2020 in Miettinen et al. (2012), who used a similar method.

It is also close to California’s assumption of 28%-30%, and is well supported by the available data. So yes, let’s use newer, up-to-date evidence on peat emissions from oil palm expansion. The good news is that California already does.

Table: Literature review of studies estimating the share of oil palm plantations on peat

Source Region Methodology Finding on percent of oil palm on peat soils
Hooijer (2006), as cited in JRC (2010) Sumatra, Kalimantan, Papua (Indonesia) Share of land concessions on peat for both oil palm and timber 27%
Casson et al. 2007, as quoted in CIFOR (2009) Kalimantan (Indonesia) Share of land concessions on peat 17%
Riau, Sumatra (Indonesia) 13-50%
EPA (2010) Indonesia and Malaysia Peat soil prevalence (not an estimation of oil palm on peat) 7%
Koh et al. (2011) Indonesia and Malaysia Share of oil palm plantations on peat using satellite imagery 11%
EPA (2012) Indonesia and Malaysia Projection for 2022 based on a multi-factor model 12%
Indonesia Share of oil palm plantations on peat using overlaid maps of oil palm and peat 22%
Malaysia 13%
Miettinen at al. (2012) Indonesia and Malaysia Projection of new oil palm plantations on peat for 2010-2020 based on satellite imagery 32%
Calculated from Miettinen et al. (2016) Indonesia and Malaysia Share of oil palm plantations on peat using satellite imagery; compared with harvested oil palm area from FAOSTAT 23%
Share of new oil palm plantations on peat (change in area) from 2010-2015 using same data as above 33%

Alternative fuels Life-cycle analyses