Plants as sensors: vegetation response to rainfall predicts subsurface water storage capacity in Mediterranean climates

Methods are lacking to characterize critical zone (CZ) structure at spatial scales relevant to earth system and dynamic global vegetation models. This knowledge gap results in poor quantification of CZ plant-available water storage capacity, hindering realistic prediction of the response of plants and streamflow to anticipated changes in the hydrological cycle. Here, we exploit the phase offset between water and energy delivery in rain-dominated Mediterranean climates to use plants as sensors to infer belowground water storage capacity. We hypothesize that if the magnitude of stored plant-available subsurface water is the primary control on dry season plant water use, then (remotely sensed) measures of transpiration may be used to infer rooting zone storage capacity. We encapsulate this idea within an ecohydrological modeling framework that describes how the stochastic properties of rainfall interact with storage capacity on intra-annual timescales to control annual variations in plant-available water storage, and thus dry season plant water use. The model reveals that where storage capacity is high relative to mean annual rainfall, plant-available water storage is not replenished in all years, and so storage and thus plant water use are sensitive to annual total rainfall. Where storage capacity is low, storage is typically replenished but can be depleted rapidly between storm events, resulting in plant insensitivity to annual total rainfall but sensitivity to spring rainfall patterns. Both high and low storage capacity result in relatively highly variable stored water for summer, and thus predicted highly variable summer transpiration; in contrast, variability is minimized at intermediate storage capacity. The model captures these diverse responses of stored water (and consequently summer vegetation water use) -- as mediated by water storage capacity -- to precipitation dynamics. Consequently, we show that a simple model inversion can be used to estimate rooting zone water storage capacity. We validate model inversion predictions using direct observations of plant-available water storage capacity in soils and weathered bedrock at two intensively monitored sites in the Northern California Coast Ranges. The model accurately predicts the magnitude of the combined dry season soil and rock moisture loss that supports transpiration, in contrast to existing soils maps, which underestimate plant-available water storage by up to a factor of three. Strongly contrasting weathering profiles and hence porosity structures at the study sites demonstrate the method is robust across diverse modes of storage and runoff generation.