Coupling the Xinanjiang model and wavelet-based random forests method for improved daily streamflow simulation

Abstract

Abstract Daily streamflow modeling is an important tool for water resources management and flood mitigation. This study compared the performance of the Xinanjiang (XAJ) model and random forests (RF) method in a daily streamflow simulation, and proposed several hybrid models based on the XAJ model, wavelet analysis, and RF method (including XAJ-RF model, WRF model, and XAJ-WRF model). The proposed methods were applied to Shiquan station, located in the Upper Han River basin in China. Five performance measures (NSE, RMSE, PBIAS, MAE, and R) were adopted to evaluate the modeling accuracy. Results showed that XAJ-RF model had a relatively higher level of accuracy than that of the XAJ model and the RF model. Compared to the RF and XAJ-RF models, the performance statistics of WRF and XAJ-WRF were better. The results indicated that the coupled XAJ-RF model can be effectively applied and provide a useful alternative for daily streamflow modeling and the application of wavelet analysis contributed to the increasing accuracy of streamflow modeling. Moreover, 14 wavelet functions from various families were tested to analyze the impact of various mother wavelets on the XAJ-WRF model.