Land surfaces dissipate energy through latent (LE) and sensible (H) heat fluxes that modulate atmospheric temperature and humidity, which in return affect land surface vegetation and soil processes. Within this two-way land-atmosphere coupling, surface energy partitioning (LE versus H) plays a central role in connecting the land and atmospheric states and fluxes.
In this study, we combined machine learning (ML) and causal inference approaches to characterize and reduce the uncertainties (including parametric, structural, and forcing uncertainties) of CMIP6 simulated evaporative fraction (defined as LE / (LE + H)) using 64 FLUXNET sites observations that cover five major biomes.
We found that accounting for biases in surface forcing variables, the simulated Evaporative Fraction (EF) in CMIP6 models could be substantially improved (R increase from 0.47 to 0.66). Leaf area index, vapor pressure deficit, and precipitation were the most important variables leading to prediction improvement. Furthermore, the ML-based parameterization showed promise to further reduce model biases (R improved from 0.66 to 0.80) in spite of the limited improvement at evergreen broadleaf forest sites where model bias may be dominated by structural inaccuracies.