Evapotranspiration and runoff reflects the hydrologic response to land use and climate, influencing the precipitation partition and water availability globally. In the past decades, global and regional land surface models had been developed and used to evaluate water energy balances, however, scholars are interested in more simple, robust approaches.

The Budyko framework is a simple but effective tool for assessing impacts of climate forcing and land surface factors on water and energy cycles. The original Budyko framework was appropriate for large basins and long-term averages. However, deviations could be also observed. In order to eliminate these deviations, several approaches had been developed, among which the Fu’s equation is a widely used one. Fu’s equation employs a parameter to account for the differences in underlying surface, such as vegetation, soil properties etc. Hence, how to estimate  become essential, and researchers have developed several approaches to estimate  in recent years, but all of which are lack of universality.

Dr. XU Xianli, from the Institute of Subtropical Agriculture, Chinese Academy of Sciences (ISA) and his colleagues developed a neural network (NN) model using a data set of 256 basins (224 small basins and 32 large basins). Firstly, the  for each basin was optimized using least squares technique based on the annual values of P, ETP, and ET (P-Q), and then the stepwise multiple linear regression technique was used to selected influencing variables, including NDVI, elevation, slope gradient, slope aspect, drainage area, basin center latitude, basin center longitude and compound topographic index. Thereafter, the selected variables were used as the inputs to train, validate and test the NN model. Finally, estimated ET from the Budyko framework with estimated  was compared with the observations in the 256 basins and a remote sensing-based ET in ~36,600 global basins to evaluate the proposed model. Results demonstrated that the Budyko framework with NN estimated  reproduced observed annual ET well for the 256 basins, and the predicted mean annual ET for ~36,600 global basins is in good agreement (R² = 0.72) with a global satellite-based ET product.

This study is supported by “100 talents program” (Y323025111 and Y251101111) and Western Development Project (KZCX2-XB3-10) of the Chinese Academy of Sciences and the Key Project of the National Twelfth Five-Year Research Program of China (2010BAE00739).

The related findings have been published on Geophysical research letters (X. Xu et al., Geophysical research letters, 40 (2013), 6123–6129). http://onlinelibrary.wiley.com/doi/10.1002/2013GL058324/full