Water availability limits ecosystem productivity across biomes, most especially in arid environments. The need to understand water constraints on ecosystem function is urgent, and has spurred the development of empirical vegetation responses to rainfall variation, either using measurements across different environments (spatial models) or observed historical responses within single sites (temporal models). These relationships demonstrate a large difference between spatial and temporal responses of vegetation to precipitation anomalies. However, available soil water is what supports plant function, not precipitation, and variation in precipitation does not fully translate to proportional variation in soil moisture. We hypothesize that the large difference between spatial and temporal sensitivities of vegetation to precipitation is in part due to the effect of decoupled soil moisture dynamics. Newly available soil moisture estimates from the SMAP mission provide a unique opportunity to test this hypothesis.
We build both spatial and temporal models to characterize the relationship between water availability (soil water and rainfall) and multiple vegetation function proxies. Across-space sensitivity is always larger than across-time sensitivity and such discrepancy of vegetation function to precipitation is larger than that to soil moisture. This suggests that future projections using spatial models would overestimate the vegetation sensitivity but models based on vegetation-soil water relationship would alleviate this overestimation. In addition, we evaluate whether an ensemble of land surface models (TRENDY-V9) could capture these patterns appropriately. Finally, multiple hypotheses for explaining the discrepancy between across-space and across-time sensitivity are examined including local adaption, differences in soil moisture retention and additional constraints. In summary, our results would advance the understanding of the complex ecohydrological processes along the plant-soil continuum and provide an approach to assess the performance of space-for-time-substitution methods and process-based models to project the fate of drylands under climate change.