Skip to Content

Estimation of global methane soil sink using-synthesized datasets and knowledge-guided machine learning

Active Dates 9/1/2023-8/31/2025
Program Area Environmental Systems Science
Project Description
The objective of this proposed research is to estimate the spatial and temporal variability in global methane soil sinks using a knowledge-guided machine learning (KGML) framework. This novel framework combines process-based and machine-learning models, and synthesizes multi-source direct and indirect measurements of soil methane oxidation to improve model training, interpretability, and accuracy across spatial and temporal scales. Natural methane oxidation by microbes in upland soils is the second largest sink in the global methane budget, but its importance has been widely underestimated. The magnitude and long-term trends of global methane soil sinks are highly uncertain due to overlooked microbial processes and contradicting studies. Accurately quantifying global methane soil sinks is extremely important to reduce biases in current and future global methane budgets. In this proposed KGML framework, process-based models will be used as scientific foundations to develop the KGML hierarchical structure and to generate millions of synthetic data for pretraining. We will build separate machine-learning submodules for soil thermal, hydrological, and biogeochemical processes, and an overarching model structure to link the submodules. The key biogeochemical constraints (e.g. soil methane substrate, temperature, and moisture influences) will be carefully embedded into the cost function using known principles and empirical functions as knowledge-guided losses. The developed KGML framework will be trained/validated with direct measurements of soil methane oxidation fluxes from FLUXNET-CH4 and chamber measurements. Using global soil moisture and temperature data, we will further optimize the model to capture temporal and spatial heterogeneity. The finely-constrained KGML model will finally be extrapolated to the global scale and be used to generate new global methane soil sink products at daily and 4-km resolution from 1984 to 2022.
Award Recipient(s)
  • University of Colorado Boulder (PI: Oh, Youmi)