资源环境科技发展态势分析平台

Total basin discharge plays an important role in the hydrological cycle for the regions. Estimation of total basin discharge cannot only help us manage water resources well but also provide a better understanding about water resources variability and hydrologic cycle. In this study, the Gravity Recovery and Climate Experiment (GRACE) data combined with precipitation and evapotranspiration (ET) is used to present first estimate of total basin discharge (similar to 60.2 km(3)/year) based on water balance method over the entire Yarlung Tsangpo River basin during 2003-2014. The estimated basin discharge shows a good correlation with observed runoff (NSE = 0.70; NRMSE = 0.37; BIAS = -6.1%). An artificial neural network (ANN) model is also proposed to build the relationship between terrestrial water storage anomalies (TWSAs) with the other hydrological data available (e.g., precipitation, temperature, ET, and soil moisture storage) during 2003-2010 and then applied to hindcasting TWSA before 2003 in order to generate a longer record of TWSA. The results show that ANN-generated TWSA using soil moisture storage and ET matches best with GRACE-derived TWSA during 2003-2010, showing a correlation coefficient (r) of 0.89 and a normalized root-mean-square error (NRMSE) of 0.40, which indicate that the ANN model is an effective way to generate TWSA beyond the GRACE data record. The comparison between total basin discharge using ANN-generated TWSA with the observed runoff data during 1998-2010 indicates a significant consistency with NSE = 0.66, NRMSE = 0.41, and BIAS = -4.9%. Our findings illustrate that GRACE data and the ANN model can be jointly and effectively used to assess the total basin discharge in the large-scale basins with limited hydrological data.