Transpiration dominates the terrestrial evapotranspiration (ET) fluxes. Stomatal control on transpiration irectly connects the water, energy and carbon cycles. Physiologically, this stomatal control responds to hydroclimatic stress through the influence of vapor pressure deficit (VPD) and soil water available on an integrated set of plant hydraulic processes. While predicted changes in soil water availability under climate change are relatively heterogeneous and uncertain, increasing temperatures are driving large increases in VPD. In most Earth system and land surface models, the ET response to each of the two stresses is evaluated through independent empirical relations, while neglecting plant hydraulics. However, because soil moisture and VPD are highly correlated, it is possible these empirical relationships are not correctly disentangling responses to the two hydrologic stressors. Thus, it is imperative to ensure that empirical models of ET can correctly capture the response to soil moisture and VPD. Here, we compare the response of predicted ET to hydrologic stressors for both plant hydraulic and empirical models across 40 globally-representative FLUXNET sites. To ensure that the models are optimally parametrized, and because plant water stress responses are highly variable between species, a Markov Chain Monte Carlo method is used to optimally parametrize each model each site.
Results/Conclusions
We show that the plant hydraulic model can be interpreted as having a functional form similar to the empirical model, but with different soil water availability stress response, and with a different VPD-sensitivity parameter m. For the hydraulic model, the VPD-sensitivity varies with leaf water potential, whereas in an empirical formulation, m is assumed constant per site. The two models perform similarly when long-term average periods are considered. However, during high VPD periods (e.g. greater than the 75th percentile for that site), the hydraulic model is able to match observed ET considerably better than the empirical model at a majority of sites. We show that this can be attributed to a hydraulic constraint, which imposes an inverse relationship between the VPD and soil water availability due to the joint control of leaf water potential on both factor. That is, in effect, common empirical models have too low a VPD sensitivity that is compensated for by increased sensitivity to soil water. This suggests that most land surface and earth system models will underestimate the response of ET (and productivity) to future, hotter droughts.