Application of Regularized Dynamic System Response Curve for Runoff Correction Based on HBV Model: Case Study of Shiquan Catchment, China

Error correction is an important postprocess technique in hydrological modeling. The paper aims to investigate the effectiveness and suitability of a recently proposed error correction method, the regularized dynamic system response curve (RDSRC) method in the Hydrologiska Byråns Vattenbalansavdelning (HBV) model. In the HBV model, the effective precipitation can reflect the error of model inputs so it is used for runoff error correction. Three commonly used statistics [Nash-Sutcliffe efficiency (NSE) coefficient, root-mean-squared error (RMSE), and percent bias (PBIAS)] were used to evaluate the model performance. In this study, the RDSRC method was tested by a synthetic case and a real case. Results of the synthetic case showed that the NSE, RMSE, and PBIAS of runoff simulation before correction were 0.943, $182.73{\mathrm{m}}^3/\mathrm{s}$, and $-0.626\%$, respectively, while those statistics after correction by the RDSRC method were $\mathrm{1,0.61}{\mathrm{m}}^3/\mathrm{s}$, and 0.002%, respectively. In the real case, the HBV model was used for daily runoff modeling in the Shiquan catchment in China; meanwhile, the RDSRC method and the AR(2) model were compared for runoff error correction. The corrected daily runoff by the RDSRC method had much better accuracy than the second-order autoregressive [AR(2)] model and the HBV model. After correction of the RDSRC method, the NSE of the HBV model increased from 0.83 to 0.946, whereas the RMSE of the HBV model decreased by 43.6%, from 246.37 to $138.96{\mathrm{m}}^3/\mathrm{s}$. This case study provides two main conclusions: (1) the RDSRC method had better performance in runoff error correction than the AR(2) model, and (2) the RDSRC method is an effective alternative for daily runoff error correction when used for the HBV model.