Time-varying hydrological simulation based on a SWD-SSC method

The Fenhe River Basin, China The hydrological parameters remained constant in a changing environment may no longer be applicable. The split-sample calibration method based on sliding window division (SWD-SSC) was proposed as a time-varying method, which confronts a challenge in the past due to the small number of samples and discontinuity. The annual LULC input and 11-year-block sliding scheme in distributed SWAT model were used to calibrate parameters. Sensitive parameters related to runoff, vegetation canopy and soil moisture demonstrated how characteristics change over time. Non-linear relationships for parameter variations were constructed by multiple algorithms combined with vegetation indexes, land use, human activity such as social economy factors. SWD-SSC method performed well during the whole periods. When the same parameters sets were applied to all periods, the accuracy will be inferior to SWD-SSC with a decrease by 12–200% in NSE. Random Forests and Back-Propagation neural network show best performance in describing the relationship between hydrological parameters and environmental factors. The impact of climate change and human activities on runoff account for 33.2% and 66.8%. The impact of land use changes on runoff are 10.9% (-2.53 mm) and 8.7% (-1.35 mm) by time-varying parameters fixed parameters, respectively. These findings aid in the comprehension of hydrological processes and enhance simulation accuracy under changing environment, ensuring future regional water safety. [Display omitted] •The time-varying SWD-SSC method was established in hydrological simulation.•A 11-year-block data was taken to calibrate and obtain numerous samples.•Nonlinear relationship between environment and parameters were described.•It has great significance model adaptability in hydrological simulation.