全球变化知识资源中心

Recent studies have begun to incorporate spatially variable plant traits into ecohydrological models, but temporal trait variability remains under-studied. Because of its potential to influence ecosystem function, representing stress-induced temporal trait variability into models should be a research priority. We present a new data-model integration approach to identify temporal variability in plant traits and generate stochastic-in-time model parameterizations. The data-conditioned stochastic parameterization was developed within the CLM 4.5 model utilizing global trait data as prior information and tested for a desert shrubland site. A synthetic experiment demonstrated that the framework successfully uncovered time-varying trait values. Using in-situ ecohydrological observations, we found the specific leaf area (SLA) for a common broadleaf-evergreen-shrub to be temporally dynamic and significantly correlated with seasonal water availability. We constructed a regression model based on the data-conditioned SLA estimates and soil wetness and used it to generate stochastic SLA parameters for a 40-year hindcast simulation. The stochastic-in-time SLA parameters resulted in greater productivity and water use efficiency than a standard static parameter. Our stochastic-in-time method can help evaluate stress-induced trait plasticity that extends our understanding beyond sparse spatial plant trait database and improve our ability to simulate carbon and water fluxes under global change.