Evaluation of Bias Correction Methods for Improving Weather Radar-based Rainfall Estimation

Abstract

This study focused on the reflectivity measurement bias (Z-bias) and the rainfall bias (R-bias) in the radar-rainfall estimation process. To correct the Z-bias, the self-consistency constraint method using reflectivity, differential reflectivity, and specific differential phase was carried out. Then, the Mean Field Bias Correction (MFBC) method and the Local Gauge Correction (LGC) method, were conducted to correct the R-bias. All of correction methods were applied to the Radar-AWS Rainrate model to calculate radar-rainfall estimates. The accuracy of the Z-bias correction method in 18 rainfall cases was improved by considering the RMSE (from 7.37 to 7.21 mm/hr) and the Correlation Coefficient (CC) (from 0.83 to 0.84) on average. The accuracy of results in the Changma and local torrential cases was improved by applying the Z-bias correction. The accuracy obtained by the LGC method (by 63.7% in RMSE and 11.7% in CC compared to MFBC) was superior to the MFBC method results because, in the LGC method, different R-biases were applied to each rainfall estimates in each radar pixel.