An emergent constraint exposes widespread underestimation of drought impacts by Earth System Models

The response of vegetation canopy conductance (g_c) to changes in moisture availability (ϒ_gc) during drought is a major source of uncertainty in climate projections. Representing ϒ_gc accurately in Earth System Models (ESMs) is particularly problematic because no regional scale g_c observations exist with which to evaluate it. Here, we overcome this challenge by deriving an emergent constraint on ϒ_gc across ESMs from Phase 6 of the Coupled Model Intercomparison Project (CMIP6). We leverage an ensemble of satellite, reanalysis and station-based estimates of surface temperatures, which are physically and statistically linked to ϒ_gc due to the local cooling effect of g_c. We find that models systemically underestimate ϒ_gc by ~50%, particularly in semi-arid grasslands, croplands, and savannas. Based on the mediating effect of g_c on carbon, water and energy fluxes through land-atmosphere interactions, the underestimation of modeled ϒ_gc in these regions contributes to biases in temperature, transpiration and gross primary production. Our results provide a novel benchmark to improve model representation of vegetation dynamics and land-atmosphere feedbacks in these regions, thus improving forecasting ability of climate extremes under future climate change scenarios.