New Uncertainties in Land Use Changes Caused by the Production of Biofuels

Abstract

The environmental benefits of biofuels continue to be debated. Recent attention focuses on biofuel-induced land use changes and their impact on greenhouse gas emissions. We develop an international, multi-commodity, partial equilibrium model and measure the impact of US and EU biofuel production on land use as well as the associated greenhouse gas emissions. We find these measures to be sensitive to changes in assumptions, specifically relating to yields. We note the possibility of offsetting cross-commodity effects. We identify the context and time path as previously unrecognized sources of variation and potential error in greenhouse gas calculations.

Appendices

Appendix

37.1.1 Model Documentation

This document summarizes the commodity models for Argentina, Brazil, Canada, China, European Union, Indonesia, India, Japan, Malaysia, Mexico, and the four developing country aggregates. These models are solved simultaneously, along with the FAPRI–MU.⁵ model and a rest-of-world aggregate trade, for world market-clearing balances for the commodities covered. The following tables detail selected components of the model (Tables 37.5, 37.6, 37.7, and 37.8).

Table 37.5 Countries modeled explicitly and four developing country aggregates

Table 37.6 Selected commodities

Table 37.7 Land aggregates

Table 37.8 Other selected quantities and prices

37.1.2 Summary of Equations of a Country or Developing Country Aggregate

Select equations for a representative country are summarized below. Cost indices and the link from prices to expected prices and returns are omitted here for brevity. The abbreviation “CC” denotes a relevant commodity or land use and “OC” indicates the other commodities or land uses in the same set. Here, soybean oil and meal prices (SO and SM) are used as key prices for vegetable oil and oilseed meal aggregates (Table 37.9).

Table 37.9 Selected country equations

37.1.3 Elasticities

Elasticities are required for all land use, yield, feed demand, stock, livestock product supply, and domestic use equation for all the countries. Due to the large number of elasticities in the model, only simple averages are reported here (Tables 37.10, 37.11, 37.12, and 37.13).

Table 37.10 Average food demand elasticities

Table 37.11 Average feed demand elasticities

Table 37.12 Average crop area elasticities

Table 37.13 Average aggregate area elasticities

37.1.4 Yield Trend Estimates

Yield trend estimates, apart from price effects, are relevant to the projections reported in the paper. The table reports the simple average trend rates of yield growth by crop (see Table 37.14).

Table 37.14 Average yield trends

Greenhouse Gas (GHG) Emission Calculations

Primary sources are the IPCC and the EPA. Unless otherwise specified, references to EPA are to Assessment and Standards Division Office of Transportation and Air Quality U.S. Environmental Protection Agency, Draft Regulatory Impact Analysis: Changes to Renewable Fuel Standard Program 2009

http://www.epa.gov/OMS/renewablefuels/420d09001.pdf

with supplemental data files.

37.2.1 Emissions from Crop Production

37.2.1.1 Emissions from Agricultural Chemical Production

Emissions from agricultural chemical production were based on EPA methods. National fertilizer and agricultural chemical data was taken from the FAO’s FERISTAT database, which reports fertilizer use by crop and country for a single year between 1988 and 2004. GREET factors are applied to obtain the total per unit lifecycle emissions of the respective agricultural chemicals. The agricultural chemicals per hectare emissions are regressed on yield for country-and-commodity pairs with data. We calculate an elasticity of emissions to yield at mean values. The emission from each hectare of a crop in a country is equal to the base emission level for that country and that emission rate increases according to the yield change from the base value and the crop-specific elasticity. A lower limit is set at zero; we do not allow a negative relationship between yields and emissions.

37.2.1.2 Crop Land Uses

Crop- and country-specific GHG emission calculations and include emissions associated with synthetic fertilizers and crop residues for all crops and methane emissions in the case of rice. These emissions can rise with yields depending on cross-country comparison of emissions to yields.

37.2.1.3 Crop N₂O Emissions

For international N₂O emissions we considered both direct and indirect emissions from synthetic fertilizer application, and crop residue N in a method synonymous with EPA. Rate of N application was obtained from FAO’s FERISTAT. From this direct N20 emissions were calculated assuming an appropriate rate of N₂O volatilization. Indirect emissions were adjusted for leaching and runoff and utilized a leaching and runoff emissions factor. We regress the synthetic fertilizer emissions on yield (using only country-and-commodity pairs with data, before inserting averages). We calculate an elasticity of emissions to yield at mean values. The emission from each hectare of a crop in a country is equal to the base emission level for that country and that crop increased according to the yield change from the base and the crop-specific elasticity. Emissions from crop residue are also included.

37.2.1.4 Rice CH₄ Emissions

Emissions from rice cultivation followed the EPA methods. The default IPCC emission rates were used and scaled for each cropping regime: irrigated, rainfed lowland, upland and deepwater by country by day. Rice cultivation season lengths were taken from the International Rice Research Institute (IRRI) to obtain annual emissions factors. Rice emissions of this type are not linked to yields.

37.2.1.5 Total Crop Land Emissions

Crop land N₂O (fertilizer and residue), agricultural chemical emissions, and rice CH₄ emissions are multiplied by the corresponding areas, converted into CO₂-e terms, and added. For developing country aggregates of the model, we take a weighted average of those countries within the aggregate for which emission data are available. The weight assigned each country for aggregating emissions is usually the harvested area, but production data for sugar and palm oil (PSD data). We use average values in cases where there are no data.

37.2.1.6 Agricultural Fuel Use

Lifecycle emissions are inclusive of both direct fuel use and heat and electricity consumption. Original data comes of agricultural fuel use emissions are from IEA [19] “CO₂ Emissions from Fuel Combustion” from which EPA Lifecycle GHG/Tailpipe CO₂ factors are applied. The resulting 2005 international agricultural sector emissions are then divided by 2005 national agricultural area to derive an average per area emission factor as found in EPA. Totals for developing countries are weighted averages of component countries based on FAOSTAT data.

37.2.1.7 Land Conversion

Calculated GHG emissions from land conversion include only emissions caused by deforestation. Emissions per hectare converted from forest use to select other uses are drawn from EPA. We use the average emissions of land conversion among developing countries represented on the table for developing country aggregates in the model. Our land uses do not match exactly the uses given in the original data. Land conversion from forest is calculated by taking the difference between initial and final forest area. If the net change is negative, then the former forest land is allocated among other broad land uses. The allocation is based on the weighted share of each land use with a positive net change. Thus, this measure is exclusively an estimate based on net effects, so deforestation in one place that is offset by an increase in forest area elsewhere would give a zero deforestation number in these calculations. This presumably overlooks some amount of normal turn-over in forest area as regards absolute numbers. However, when we consider the implications of a change from baseline paths to alternative scenario paths the focus moves to the changes in land use, this error might be less critical to scenario analysis.

37.2.1.8 Forest Sequestration

The main concern associated with sequestration is the annual difference between carbon sequestered in forestland compared to that of cropland. This is calculated in similar fashion to the EPA document using above- and below-ground sequestration data that decompose the long-term deforestation emission effects. These data give forest sequestration rates for some countries in the model. We calculate a ratio of the sequestration rate to the dry mass for those cases where there are data for both. This is applied to the dry mass on forest land for other countries to give an estimate of the sequestration rate of those countries. However, this assumes the same ratio would apply for above- and below-ground sequestration rates. For this, IPCC estimates of the dry mass on forest land are used. We aggregate these data by country according to FAO forest type data (Global Forest Resource Assessment [10]). For developing country aggregates of the commodity model, we use the countries with the most forest area and ignore some of the countries with much smaller forest area. We ignore any differences in sequestration rates of new or old forests.

37.2.2 Livestock Production

GHG emissions caused by enteric fermentation and manure management are country- and commodity-specific, but these are not tied to any measure of feed use changes relative to the base data. Rates of emissions are from EPA, but are on a per head per year basis. To convert to a per unit of livestock product basis, FAOSTAT data representing meat production and animal inventories are used. The inventories at one point of time in the year are taken as estimates of the average number of head in that year, although this may ignore seasonal differences. The ratio of output to animal inventory is used to convert the original per head emission rate into a per unit of output emission rate. Chicken is used to represent all poultry. For each of the four developing country aggregates, a weighted average rate is used. The weights are production of the animal output in question (e.g. beef, pork, and poultry).

37.2.3 Fuel Use

Transportation, storage, and tailpipe emissions of ethanol and biodiesel emissions are included. Tailpipe emissions of different GHGs are given by EPA and transportation and storage for ethanol are provided in a supplemental file. We apply the same transportation and storage emissions to biodiesel. We assume at present that ethanol and biodiesel displace gasoline and diesel at a rate of one BTU for one BTU.