Estimation of spatiotemporally varying parameters for grid-based distributed hydrologic models

Study region: Xiangjiang and Baihe River basins, China. Study focus: Hydrologic models often use either time-varying or spatially heterogeneous parameter methods to improve runoff simulations. However, few methods account for both dimensions simultaneously, limiting model accuracy and reducing insight into the effects of climate change and human activities on rainfall-runoff relationships. To fill this gap, a spatiotemporally varying parameter estimation method, DAKG-SWD-DP, is proposed here. This method involves three steps: (1) division of the dataset into sub-periods using a sliding window-based split-sample calibration (SWD-SSC) method; (2) calibration of spatially heterogeneous parameters for each sub-period using a dimension-adaptive key grid (DAKG) strategy; and (3) optimization of spatiotemporal parameter variations through dynamic programming to consider both simulation accuracy and parameter continuity. New hydrological insights for the region: (1) the DAKG-SWD-DP method significantly improves runoff simulation compared to the constant parameter, DAKG, and SWD-SSC methods. Specifically, the NSE increases by 0.05 in the Xiangjiang River basin and 0.09 in the Baihe River basin compared to the constant parameter method; (2) the DAKG-SWD-DP method outperforms the DAKG-SWD method in capturing the relationships between hydrologic parameters and environmental factors, due to enhanced parameter continuity. Additionally, the DAKG-SWD-DP method efficiently identifies climatic factors as key drivers of parameter variations in both basins, while human activities, such as reservoir construction, are also key drivers in the Baihe River basin.