Effects of climate and terrestrial storage on temporal variability of actual evapotranspiration

Knowledge of the temporal variability in actual evapotranspiration (E) is essential for a better understanding of the interaction and feedback between atmospheric and land surface hydrologic processes under various natural and anthropogenic conditions. Recently, Zeng and Cai (2015) proposed a decomposition framework of the E variance, based on water balance and the Budyko hypothesis. On the basis of a long-term (1960-2008) land surface dataset, this study applies the theoretical framework to assess the effects of climate and terrestrial storage factors on the interannual and intra-annual variance in E across China. An error decomposition framework is developed to quantify the key factors in the error of the predicted E variance. The results show that the prediction of the E variance is more accurate in arid climates than in humid climates, and the corresponding error is primarily controlled by the variability of precipitation (P) and runoff (R). Climate is the primary source for the E variance, and the dominant sources shift from potential evaporation (PET) in humid climates to P in arid climates. The interactions between P and PET tend to dampen the E interannual variance and enhance the E intra-annual variance, and this effect is especially significant in humid climates. Terrestrial storage change is more capable of accommodating climatic fluctuations at the intra-annual scale than at the interannual scale, and for some arid regions it is the dominant factor influencing the E variance. The response of terrestrial storage to P is more significant than its response to PET, especially for regions with strong human impact. Neglecting the effects of terrestrial storage would possibly underestimate or overestimate the E variance in both humid and arid climates, due to the interactions between climate and the change in the terrestrial storage.