What's Changed
- Using 1-km drivers of forest loss map (2022 version): The emissions model now uses the 1-km drivers of tree cover loss map (under review at Environmental Research Letters), led by Michelle Sims, Radost Stanimirova, and Google Deepmind, run using TCL through 2022. This drivers of loss map replaces the 10-km map (Curtis et al. 2018). Although the 1km drivers map was run using TCL through 2022, we applied it to the entire TCL/emissions timeseries (2001-2023). The 1km drivers map is not published yet and will soon be updated to use TCL through 2023. As part of switching to the 1-km drivers map, we had to decide which carbon pools would be emitted by each driver (7 instead of the previous 5) and what soil emission factors and fire/peat factors to use for each driver. These decisions are documented in the C++ code.
- Outputting separate emissions for each GHG: The emissions model now outputs emissions for CO2, CH4, and N2O separately; previously it output CO2 and non-CO2 (CH4 and N2O combined). Separate emissions by gas is necessary for compliance with the Greenhouse Gas Protocol Land Sector and Removals Guidance/Standard.
- Deadwood and litter are emitted as non-CO2 gases during fires for non-land use change drivers: IPCC 2019 guidelines state that deadwood and litter loss result in non-CO2 emissions (but not CO2 emissions) when land use change doesn't occur. We were not doing this previously, so now this is fixed.
- Cleaned up emissions C++ code: C++ emissions decisions trees are more readable. They also no longer output driver-specific emissions tile sets, which we weren't using for anything.
Note that everything before the emissions part of the model did not change and was not rerun (model extent, removal factors, gain year count, gross removals, carbon pools in year of loss). Only gross emissions and net flux are affected by this change.