Inferring geographic and climatic variation in plant hydraulic traits from flux data

Plant water stress has major implications for vegetation productivity, mortality, and global carbon cycle variations. Yet, its representation in models constitutes a major source of uncertainty. This is related to uncertainties in the representation of water stress exposure due to unknown belowground rooting structure and water storage, and uncertainties in the representation of interactive responses to atmospheric dryness (vapour pressure deficit, VPD) and belowground moisture. Recent theoretical developments have given rise to several hydraulically explicit models for predicting plant physiological responses to VPD and belowground moisture at the leaf level. The P-hydro model that we have recently developed, accounts for the simultaneous acclimation of stomatal conductance and photosynthetic capacity using an eco-evolutionary optimality approach. Based on three first principles (balance of water demanded by transpiration and that supplied from the soil, coordination of the carboxylation-limited and light-limited photosynthesis rates, and maximization of net profit after accounting for the costs of maintaining photosynthetic and hydraulic capacities), it correctly predicts the responses of stomatal conductance, assimilation rates, leaf water potentials, and photosynthetic capacities to changing hydroclimatic environments. However, the hydraulic strategies of plants depend critically on their belowground rooting environments, such as soil properties and rooting depth. Here, we implement the P-hydro model for modelling ecosystem-level fluxes and couple it with a simple model of soil water balance within the rsofun modelling framework. The water-balance model accounts for variations in root-zone water storage capacity of plants, thus allowing us to characterize both above-ground and belowground hydraulic strategies. We investigate (1) the power of P-hydro in simulating gross primary production and evapotranspiration at globally distributed FLUXNET sites under conditions of simultaneous atmospheric and belowground dryness, benchmarked against a non-hydraulically explicit version of the model (P-model), (2) apply a Bayesian data assimilation approach to infer plant and soil hydraulic traits (plant conductivity and vulnerability, root zone water storage capacity, and cost parameters of the optimality model), (3) assess the environmental dependencies of the inferred traits and the generalisability of the model for global simulations. The P-hydro model, together with the inferred trait relationships, promises a simple yet robust approach to predicting the global environmental dependencies of ecosystem productivity.