Developing the soil texture effects on the surface resistance to bare soil based on MOD16 algorithm to estimate evapotranspiration over the Tibetan Plateau

Evapotranspiration (ET), composed of evaporation (ETs) and transpiration (ETc) and intercept water (ETw), plays an indispensable role in the water cycle and energy balance of land surface processes. A more accurate estimation of ET variations is essential for natural hazard monitoring and water resource management. For the cold, arid, and semi-arid regions of the Tibetan Plateau (TP), previous studies often overlooked the decisive role of soil properties in ETs rates. In this paper, an improved algorithm for ETs in bare soil and an optimized parameter for ETc over meadow based on MOD16 model are proposed for the TP. The nonlinear relationship between surface evaporation resistance (r_s^s) and soil surface hydration state in different soil texture is redefined by ground-based measurements over the TP. Wind speed and vegetation height were integrated to estimate aerodynamic resistance by Yang et al. (2008). The validated value of the mean potential stomatal conductance per unit leaf area (C_L) is 0.0038m s^-1. And the algorithm was then compared with the original MOD16 algorithm and a soil water index–based Priestley-Taylor algorithm (SWI–PT). After examining the performance of the three models at 5 grass flux tower sites in different soil texture over the TP, East Asia, and America, the validation results showed that the half-hour estimates from the improved-MOD16 were closer to observations than those of the other models under the all-weather in each site. The average correlation coefficient(R²) of the improved-MOD16 model was 0.83, compared with 0.75 in the original MOD16 model and 0.78 in SWI-PT model. The average values of the root mean square error (RMSE) are 35.77W m^-2, 79.46 W m^-2, and 73.88W m^-2 respectively. The average values of the mean bias (MB) are -4.08W m^-2, -52.36W m^-2, and -11.74 W m^-2 overall sites, respectively. The performance of these algorithms are better achieved on daily (R²=0.81, RMSE=17.22W m^-2, MB=-4.12W m^-2; R²=0.64, RMSE=56.55W m^-2, MB=-48.74W m^-2; R2=0.78, RMSE=22.3W m^-2, MB=-9.82W m^-2) and monthly (R2=0.93, RMSE=23.35W m^-2, MB=-2.8W m^-2; R2=0.86, RMSE=69.11W m^-2, MB=-39.5W m^-2; R2=0.79, RMSE=62.8W m^-2, MB=-9.7W m^-2) scales. Overall, the results showed that the newly developed MOD16 model captured ET more accurately than the other two models. The comparisons between the modified algorithm and two mainstream methods suggested that the modified algorithm could produce high accuracy ET over the meadow sites and has great potential for land surface model improvements and remote sensing ET promotion for the ET region.