A hybrid model to predict the hydrological drought in the Tarim River Basin based on CMIP6

Drought simulation and prediction are of great significance to drought early warning. However, it is difficult to predict hydrological drought in data-scarce areas. To address this problem, Tarim River Basin was selected as a typical representative of the data-scarce inland river basin in China, we constructed a hybrid model by combining the complete ensemble empirical mode decomposition with adaptive noise and the long short-term memory method to predict hydrological drought from 2022 to 2100 based on CMIP6. The results show that meteorological drought has quasi-3-month, quasi-5-month, quasi-7-month, quasi-1-year, quasi-2-year, quasi-4-year, quasi-9-year, quasi-17-year and quasi-54-year cycles. Hydrological drought has quasi-3-month, quasi-5-month, quasi-6-month, quasi-1-year, quasi-2-year, quasi-4-year, quasi-9-year, quasi-29-year and quasi-32-year cycles. The components of meteorological drought and hydrological drought have significant correlations on monthly, interannual, and interdecadal scales, with correlation coefficients of 0.282, 0.573, and 0.340, respectively, and p values of 0.000. The hybrid model had a better prediction accuracy (R2 = 0.951, MAE = 0.131, NSE = 0.951, d index = 0.987) than previous studies. The trend of the hydrological drought index in the sustainable development model (SSP1-2.6) shows a trend of increasing severity with a rate of − 0.004/10 years from 2022 to 2100. And from the sustainable development model (SSP1-2.6) to the unbalanced development model (SSP5-8.5), the hydrological drought gradually becomes more serious. This study provides a new mechanism for predicting hydrological drought in data-scarce areas and is of great significance for the early warning of hydrological drought in this area.