Evaluating the drought response of CMIP5 models using global gross primary productivity, leaf area, precipitation, and soil moisture data

Realistic representation of vegetation’s response to drought is important for understanding terrestrial carbon cycling. We evaluated nine Earth system models from the historical experiment of the Coupled Model Intercomparison Project Phase 5 for the response of gross primary productivity (GPP) and leaf area index (LAI) to hydrological anomalies. Hydrological anomalies were characterized by the standardized precipitation index (SPI) and surface soil moisture anomalies (SMA). GPP and LAI in models were on average more responsive to SPI than in observations revealed through several indicators. First, we find higher mean correlations between global annual anomalies of GPP and SPI in models than observations. Second, the maximum correlation between GPP and SPI across 1–24month time scales is higher in models than observations. And finally, we found stronger excursions of GPP to extreme dry or wet events. Similar to GPP, LAI responded more to SPI in models than observations. The over-response of models is smaller if evaluated based on SMA instead of SPI. LAI responses to SMA are inconsistent among models, showing both higher and lower LAI when soil moisture is reduced. The time scale of maximum correlation is shorter in models than the observation for GPP, and themarkedly different response time scales amongmodels for LAI indicate gaps in understanding how variability of water availability affects foliar cover. The discrepancy of responses derived from SPI and SMA among models, and between models and observations, calls for improvement in understanding the dynamics of plant-available water in addition to how vegetation responds to these anomalies.