Multi-objective parameter estimation for simulating canopy transpiration in forested watersheds

Abstract

A Jarvis based [Philos. Trans. R. Soc. London, Ser. B 273 (1976) 593] model of canopy stomatal conductance was evaluated in context of its application to simulating transpiration in a conifer forest covered watershed in the Central Sierra Nevada of California, USA. Parameters influencing stomatal conductance were assigned values using Monte Carlo sampling. Model calibration was conducted by evaluating predicted latent heat fluxes against thermal remote sensing estimates of surface temperature. A fuzzy logic approach was used to select or reject simulations and form a restricted set of ensemble parameter solutions. Parameter estimates derived from the ensembles were evaluated using theory on how stomatal conductance regulates leaf water potential to prevent runaway cavitation. Canopy level parameters were found to be sufficient for predicting hydraulically consistent transpiration when soils were well watered. A rooting length parameter controlling the amount of plant available water was a sufficient addition to the parameter set to predict hydraulically consistent transpiration when soil moisture stress was occurring. Variations in maximum stomatal conductance among different hillslopes within the watershed were explained by a light threshold parameter. The results demonstrate that the Jarvis model can be reliably parameterized using thermal remote sensing data for estimating transpiration in meso-scale watersheds.