Land Surface Energy Partitioning Dominates Dry-Season Water Availability Uncertainties in Earth System Models

Accurately characterizing dry-season water availability (W_d) is critical for projecting terrestrial carbon exchange and global water security. W_d is commonly calculated as the minimum value of precipitation minus evapotranspiration within each calendar year. However, Earth System Model (ESM) projected W_d contains substantial uncertainties and can disagree on even the sign. Based on a newly proposed framework, we disentangle the uncertainty sources in ESM-based W_d projections. Results demonstrate that ESM-based W_d uncertainties are dominated by land surface energy partitioning (summarized by evaporation fraction, denoted as EF) instead of precipitation or available energy. As such, EF alone can explain more than 83% of inter-ESM variability in historical and future W_d projections. Compared against data-driven benchmarks, ESMs tend to overestimate dry-season EF—suggesting that W_d is likely to be underestimated in ESMs. Our analysis indicates that the ET resistance parameterization is the central error source in ESM-based EF, which should be constrained to enhance the reliability of EF, and by extension, W_d projections.