Methodology

Overview of methods


Estimates of tropical commodity-driven deforestation and associated GHG emissions were calculated by a team of researchers at Stockholm Environment Institute’s Trase Initiative, Chalmers University of Technology, and the Senckenberg Biodiversity and Climate Research Centre. The methodology used to provide estimates presented in this guide were reviewed by a technical advisory committee of experts in forests and climate change.

Calculations for the estimates of GHG emissions from tropical commodity-driven deforestation used in this guide drew heavily from methods employed in two recent papers:

Pendrill, Florence, Martin U. Persson, Javier Godar and Thomas Kastner. (2019). Deforestation displaced: trade in forest-risk commodities and the prospects for a global forest transition. Environmental Research Letters 14(5). https://doi.org/10.1088/1748-9326/ab0d41

Pendrill, Florence, Martin U. Persson, Javier Godar, Thomas Kastner, Daniel Moran, Sarah Schmidt and Richard Wood. (2019). Agricultural and forestry trade drives large share of tropical deforestation emissions. Global Environmental Change 56:1-10. https://doi.org/10.1016/j.gloenvcha.2019.03.002 

The analysis linking deforestation risk to agricultural and forestry production, trade and consumption includes three main steps:

Forest loss was detected via spatial datasets on tree cover loss, and then attributed to expanding land uses (cropland, pastures and forest plantations) in proportion to their relative area of expansion. This attribution loss was implemented at the national level, except for Brazil and Indonesia, where loss was attributed at the subnational level. Subnational-level data is useful for these two countries which hold a large share of remaining tropical forests, but also account for a large share of tropical forest loss.
 
The model for attributing forest loss to these three land-use categories is based on two main premises: (a) where cropland expands, it first expands into pastures, and then into forests, and (b) where pastures and forest plantation areas expand, they primarily replace forest land. These premises are consistent with data from additional studies that describe, quantify and assess the predominant deforestation-related land-use transitions across the tropics.
 
Next, forest loss attributed to cropland expansion was allocated to individual crops in relative proportion to their expansion area. For example, if areas planted to soybeans accounted for half of the total cropland expansion in a country, half of the country’s cropland deforestation was attributed to the country’s soybean production. National-level data on cropland and pasture areas in 2000-2018 were taken from the Food and Agriculture Organization Corporate Statistical database (FAOSTAT). The sub-attribution of cropland deforestation to individual crops was based on harvested area data from FAOSTAT.


Carbon emissions resulting from land cover changes were quantified by estimating net carbon stock changes (changes in the amount of carbon that has been sequestered and stored in a forest) for these previously forested areas. Emissions from peatland drainage were also included. Loss of carbon stock was quantified using satellite remote sensing techniques and existing literature on carbon losses in the tropics.


An international trade model was used to link the deforestation footprint involved in the production of each commodity, tracing any deforestation involved from production to countries of consumption throughout international supply chains. This analysis used production data from FAOSTAT and bilateral trade data. The calculations aimed to track products along supply chains, including re-exports and processing, up to the point where they are physically consumed either as food as livestock feed or in industrial processes.


Uncertainty in attribution of deforestation-related GHG emissions to countries, commodities, and trade flows


The data and claims presented here are based on multiple types of scientifically rigorous evidence with large sample sizes; yet, there is always some inherent uncertainty embedded in the methods of any model. Overall, we present this data with a high level of confidence and describe remaining uncertainties below. Given these uncertainties, the values provided in this guide are relatively consistent with other credible studies that have produced similar estimates, with a few differences due to methodological choices. There are several sources of uncertainty associated with the data presented in this guide:


Data availability and use of model to attribute deforestation levels. Ideally, attribution of forest loss and associated carbon emissions to agricultural and forestry production would be based on spatially explicit (e.g., remotely-sensed) data. However, existing spatial analyses of land cover and use following forest loss in the tropics are limited both geographically and temporally. Although pan-tropical data on forest loss and land cover exist, quality and consistency of land classifications across datasets is still too poor for combining these to sufficiently assess post-forest land. Therefore, a model was used to attribute detected forest loss. This model does not describe all possible land-use transitions, but rather reflects the predominant land-use transitions related to tropical deforestation.


Agricultural statistics. The area of forest plantations in the FAO Forest Resources Assessment is only quantified every five years (e.g., 2010, 2015) and linearly interpolated between. Data for the years 2016 and 2017 are extrapolated based on the 2010 to 2015 trend, so uncertainty increases slightly for those years. FAOSTAT data vary considerably in quality between different countries and years. The subnational data for Brazil and Indonesia also vary in quality between administrative units and years. Production data are obtained from FAOSTAT, which are in turn based on statistics reported by national statistical offices. However, as the data are directly related to national-level food security for each respective country, reasonable data quality can be assumed.


Trade model. For data on bilateral trade flows, two values are typically reported for each flow from country A to country B: export data reported by country A, and import data from country B. Sometimes these two values do not match, for legitimate reasons. The presented data give priority to reported import data, but in a similar study which gave priority to reported export data, the overall findings did not change.