Estimating evapotranspiration using an improved two-source energy balance model coupled with soil moisture in arid and semi-arid regions

The availability of soil moisture greatly influences the surface energy balance and evapotranspiration (ET) in arid and semi-arid regions. The soil moisture-based two-source energy balance model used the soil resistance and canopy resistance to constrain the soil evaporation and plant transpiration. However, the soil resistance is expressed as an empirical model with soil moisture, which needs to be calibrated to obtain the unknown parameters for different soil types, and the canopy resistance does not consider the influence of soil moisture. In this study, we introduced a dry soil layer (DSL) based soil resistance and an optimal stomatal conductance model (OSM) based canopy resistance, both of which are related to soil moisture, into the two-source energy balance (TSEB) model to constrain the soil evaporation (E) and plant transpiration (T). The new model (TSEB_SM) was evaluated at five sites with a wide range of soil moisture and different vegetation covers, which are located in arid and semi-arid regions. The results showed that the TSEB_SM performed better than the TSEB for LE/ET estimation, with about 20 % RMSE reduction in daily ET at irrigated cropland and alpine meadow sites, and a reduction of 31 %, 54 %, and 33 % at desert steppe, shrub forest, and forest sites, respectively. The trends of the E and T partitioned by the TSEB and TSEB_SM models followed the temporal trend of LAI, and there is generally better agreement of the E and T between the TSEB_SM model and the water use efficiency (uWUE) method. The sensitivity analysis for the resistances in the TSEB_SM model elucidated that soil resistance and canopy resistance had more noticeable effect on ET simulation, especially under soil water stressed conditions. Moreover, the results of the two models under different soil moisture conditions showed that the TSEB_SM model had a similar performance in LE/ET as the TSEB model under soil moisture sufficient conditions, but it could effectively reduce the overestimation of LE/ET by the TSEB model under water limited conditions (<30 %). Finally, the uncertainty analysis of model input parameters showed that the uncertainty of soil moisture had more significant influence on the model than LAI. Combined with reliable soil moisture from remote sensing, the TSEB_SM model has important significance for monitoring evapotranspiration in arid and semi-arid regions.