Sensitivity of a clumped model of evapotranspiration to surface resistance parameterisations: Application in a semi-arid environment

This paper explores the sensitivity of a three-source (vegetation, P, bare soil, BS, and soil under plant, S) evapotranspiration clumped model (CM) developed for sparsely vegetated areas, to the parameterisation of the surface resistances in semi-arid climate. We analysed the sensitivity of the CM to: (i) the location and depth of the soil water content (θ) measurements, (ii) the accuracy in the measurements of the variables involved in the parameterisation of the surface resistances, with special attention to the in-canopy water vapour saturation deficit (D0) and to the average plant leaf area index (L), and (iii) the simplification of the parametric equations of the soil surface resistances. The sensitivity of the CM was tested in a stand of Anthyllis cytisoides located in a semi-arid area of the Southeast of Spain. In this stand, θ was measured at three different depths (0.02 m, 0.05 m and 0.15 m), D0 was calculated through iterations of the CM, taking into account the water vapour fluxes of the different evaporating sources, and L   was measured with a destructive direct method. These variables were included in parametric equations for estimating the surface resistances of P (gsp), BS and S (rsp, rsbs and rss, respectively). Evapotranspiration estimates were compared to Eddy Covariance system measurements. Results showed that for estimating rsbs and rss the θ   should be measured as superficially as possible, whereas when estimating rsp the θ should be measured at a depth where the effect of the extremely low values of the superficial layer of soil is excluded or attenuated (in our study area this depth was 0.15 m or the integrated θ of the first 0.15 m). Moreover, the CM was more sensitive to the accuracy in the estimation of D0 than to any other variable, while errors in the measurement of L and θ had similar effects in the CM evapotranspiration estimates (a 50% underestimation of D0 gave rise to an underestimation of λE of 70%, whereas an overestimation of 50% in D0 led to an overestimation of λE of 16%; errors of ±50% in L and θ gave rise to errors of λE of ±16%). Finally, the CM showed a higher sensitivity to the parameterisation of each surface separately, with a specific parametric equation for each soil surface resistance, than to the use of specific values of θ measured in each soil surface.