A simple surface conductance model to estimate regional evaporation using MODIS leaf area index and the Penman‐Monteith equation

Abstract

Listen

Translate article

We introduce a simple biophysical model for surface conductance, Gs, for use with remotely sensed leaf area index (Lai) data and the Penman‐Monteith (PM) equation to calculate daily average evaporation, E, at kilometer spatial resolution. The model for Gs has six parameters that represent canopy physiological processes and soil evaporation: gsx, maximum stomatal conductance; Q50 and D50, the values of solar radiation and atmospheric humidity deficit when the stomatal conductance is half its maximum; kQ and kA, extinction coefficients for visible radiation and available energy; and f, the ratio of soil evaporation to the equilibrium rate corresponding to the energy absorbed at the soil surface. Model parameters were estimated using 2–3 years of data from 15 flux station sites covering a wide range of climate and vegetation types globally. The PM estimates of E are best when all six parameters in the Gs model are optimized at each site, but there is no significant reduction in model performance when Q50, D50, kQ, and kA are held constant across sites and gsx and f are optimized (linear regression of modeled mean daily evaporation versus measurements: slope = 0.83, intercept = 0.22 mm/d, R2 = 0.80, and N = 10623). The average systematic root‐mean‐square error in daytime mean evaporation was 0.27 mm/d (range 0.09–0.50 mm/d) for the 15 sites. The average unsystematic component was 0.48 mm/d (range 0.28–0.71 mm/d). The new model for Gs with two parameters yields better estimates of E than an earlier, simple model Gs = cLLai, where cL is an optimized parameter. Our study confirms that the PM equation provides reliable estimates of evaporation rates from land surfaces at daily time scales and kilometer space scales when remotely sensed leaf area indices are incorporated into a simple biophysical model for surface conductance. Developing remote sensing techniques to measure the temporal and spatial variation in f is expected to enhance the utility of the model proposed in this paper.