Study WP5.8 Using Machine-Learning to Evaluate and Understand our Capability to Model Tropical Wetland Methane Emissions


This Study is led by Rob Parker and Cristina Ruiz Villena from NCEO (University of Leicester). Additional contributors to this Study are Nic Gedney from the Met Office and Paul Palmer from the University of Edinburgh.

The main CCI ECVs used in this Study are Greenhouse Gas (methane), Land Surface Temperature, Soil Moisture, Land Cover, and possible Vegetation.

It is estimated that this Study will run from January 2024 until July 2024.

The models used are JULES (land surface) and GEOS-Chem (atmospheric). This Study aims to develop an emulator for JULES wetland methane, use its explainability to show which factors matter in the model, drive the emulator with CCI Earth Observation data to generate wetland fluxes and compare those to a methane inversions performed on GOSAT/TROPOMI ESA CCI data. Current ensembles of JULES simulations with different driving data, temperature dependency, vegetation and wetland masks show massively different methane fluxes. This Study aims to address this.

Results and conclusions

Results and conclusions will be provided once the Study is complete.