Modeling the multiple time scale response of hydrological drought to climate change in the data-scarce inland river basin of Northwest China

It is difficult to quantitatively assess the response of hydrological drought (HD) to climate change in the inland river basins of northwest China because of the complicated geographical environment and scarce data. To address this problem, we conducted a hybrid model by integrating the ensemble empirical mode decomposition (EEMD), the long short-term memory (LSTM) model, and the statistical downscaling method and selected the Aksu River Basin (ARB) as a typical representative of data-scarce inland river basin in northwest China to simulate its hydrological drought in the period of 1980–2015 based on reanalysis climate data and hydrological observation data. The coefficient of determination (R²), the mean absolute error (MAE), the Nash–Sutcliffe efficiency coefficient (NSE), and the index of agreement d (d index) all showed high simulation accuracy of the hybrid model (R² = 0.712, MAE = 0.304, NSE = 0.706, and d index = 0.901 of the Aksu River Basin), and the simulated effect of the hybrid model is much better than that of a single long short-term memory model. The simulated results in the Aksu River Basin by the model revealed that hydrological drought in the two subbasins (i.e. the Kumarik River Basin (KRB) and the Toshkam River Basin (TRB)) showed similar cycles on the seasonal scale, the interannual scale, and the interdecadal scale, which are mainly controlled and influenced by regional climate change. On the seasonal scale, the effect of precipitation and temperature on hydrological drought is not significant; on the interannual scale, precipitation is the key factor compared to temperature in inducing hydrological drought change; on the interdecadal scales, the correlations between precipitation, temperature, and hydrological drought were the strongest and most significant.