Bayesian inference of hydraulic properties in and around a white fir using a process-based ecohydrologic model

Abstract

We present a parameter estimation study of the Soil-Tree-Atmosphere Continuum (STAC) model, a process-based model that simulates water flow through an individual tree and its surrounding root zone. Parameters are estimated to optimize the model fit to observations of sap flux, stem water potential, and soil water storage made for a white fir (Abies concolor) in the Sierra Nevada, California. Bayesian inference is applied with a likelihood function that considers temporal correlation of the model errors. Key vegetation properties are estimated, such as the tree's root distribution, tolerance to drought, and hydraulic conductivity and retention functions. We find the model parameters are relatively non-identifiable when considering just soil water storage. Overall, by utilizing multiple processes (e.g. sap flow, stem water potential, and soil water storage) during the parameter estimation, we find the simulations of the soil and tree water properties to be more accurate when compared to observed data.