GTAP

The computable general equilibrium model of the Global Trade Analysis Project (GTAP for short) has been used for ILUC modelling by the California Air Resources Board in the LCFS, alongside GLOBIOM in analysis for CORSIA, by the Argonne National Laboratory for ILUC estimates include in its GREET model (Argonne National Laboratory, 2017) and in a series of publications authored by academics at Purdue University in the U.S. There

10.3

14.4 13.4

6.6

54.3 55.8

53.8 51.8

0 10 20 30 40 50 60

Maize Wheat Sugarcane Sugarbeet Palm oil Soy oil Rapeseed

oil Sunflower oil gCO2e/MJ

Economic models for assessing ILUC

are probably more iterations of ILUC modelling published using the GTAP model than any other model. One factor contributing to this regular stream of publications is that fact that the main GTAP modelling team is based in a university department rather than another type of institution. The ILUC values used for regulatory compliance in the California Low Carbon Fuel Standard are illustrated in Figure 10. These results are not from a single scenario but are the average emissions across thirty scenarios for each feedstock calculated for the Air Resources Board.

Figure 10 ILUC results obtained with GTAP for California Air Resources Board Source: California Air Resources Board (2014)

Note: Values adjusted from a 30 year to a 20-year amortisation19

The hierarchy of emissions values is similar to that delivered by the MIRAGE modelling of Laborde (2011); palm oil has the highest assessed value followed by soy oil, ethanol crops have lower values. One difference is the relatively low value calculated for rapeseed oil (referred to as canola in North America). California Air Resources Board (2014) provides relatively little detail of the results of the modelling, but the land use change and land use change emissions results by region and AEZ are provided for one of the thirty scenarios as part of the package for version 52 of the AEZ-EF (agro-ecological zone emission factor) model used in the LCFS modelling (Plevin, Gibbs, et al., 2014)²⁰. These results (Figure 11) show that the area of land use increase predicted is significantly larger in the soy scenario, but also that the location of expected land use changes are quite distinct. While both predict the bulk of land use changes to occur in North America, the rapeseed scenario shows expansion in Canada while the soy scenario shows expansion in the U.S. itself. The soy scenario also shows larger area increases in South America and Southeast Asia. The lower amount of modelled palm

19 One-off emissions from land conversion are simply adjusted to the 20-year amortisation by multiplying by 1.5 (as the emissions are divided over a shorter period). Peat emissions are

considered to persist over the period considered, and therefore we treat the estimated share of annualised peat emissions in the ILUC values for palm oil as independent of amortisation period.

20 https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=4346 0

10 20 30 40 50 60 70 80 90 100

Maize Sugarcane Soy oil Rapeseed oil Palm oil

gCO2e/MJ

oil expansion in the rapeseed scenario is a significant contributor to the lower resulting ILUC value.

Figure 11 Location of cropland expansion for U.S. rapeseed oil and soy oil scenarios with GTAP.

Source: own calculations based on soy and canola scenarios 6 from California Air Resources Board (2014). Note that the soy result has been normalised here to match the rapeseed oil shock size of 400 million gallons.

It is also important to recognise that ILUC modelling results can be quite different depending on the region in which the demand shock is modelled. Figure 12 shows that the predictions for location of cropland expansion in response to rapeseed demand are completely different between GTAP modelling for U.S. demand and GLOBIOM modelling for EU demand. In GLOBIOM, the main locations for cropland expansion as EU demand for rapeseed biodiesel increases are within the EU itself and to a lesser extent in North America, sub-Saharan Africa and Southeast Asia. In GTAP, the biggest land use changes are in Canada and sub-Saharan Africa with relatively little expansion in Southeast Asia. At least equally importantly for the results, the GTAP modelling for U.S. demand predicts only a tenth as much net land expansion per unit of energy produced. This implies a much stronger productivity response in the GTAP modelling.

0 50000 100000 150000 200000 250000 300000

Rapeseed Soy

Hectares

Sub-Saharan Africa South East Asia South Asia Oceania North America

Middle East and North Africa Latin America

European Union East Asia

Commonwealth of Independent States

Economic models for assessing ILUC

Figure 12 Location of cropland expansion for EU rapeseed oil biodiesel scenario with GLOBIOM and U.S. rapeseed oil scenario with GTAP

Source: own calculations based on canola scenario 6 from California Air Resources Board (2014) and results from Valin et al. (2015)

Like GLOBIOM, GTAP has also been used in the development of ILUC factors for CORSIA – these estimates are shown in Figure 13. GTAP modelling has been subjected to considerable scrutiny both in the context of its use in the LCFS and its development independent of the California Air Resources Board. Malins et al. (2020) raises a number of questions relating to the model, and in particular is rather critical of the strong role that agricultural intensification has been given within the GTAP framework (and the consequent reductions in modelled ILUC emissions).²¹ It is suggested that the evidentiary basis for changes to the modelling framework since 2009 has sometimes been weak, and that a lower standard of evidence may have been applied before introducing changes that increase intensive responses (and therefore reduce ILUC emissions in model results) than before introducing changes that could increase modelled ILUC results.²²

21 A response to (Malins et al., 2020) is provided by (Taheripour et al., 2021).

22 One example of this is that GTAP does not include a mechanism to model the conversion of unmanaged land to agricultural use. This was identified as a limitation in the GTAP framework over a decade ago and was part of the EPA’s argument for preferring the FAPRI-FASOM modelling system. The MIRAGE model has a GTAP-like structure and includes a mechanism to allow conversion of unmanaged land, but this mechanism has never been copied across, nor (to the best of our knowledge) has any alternative mechanism been suggested.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

GLOBIOM GTAP

Sub-Saharan Africa South East Asia South Asia Oceania North America

Middle East and North Africa Latin America

European Union East Asia

Commonwealth of Independent States

Figure 13 ILUC results obtained with GTAP for CORSIA Source: ICAO CAEP (2019)