Real-time bias correction of rainfall nowcasts using biward tracking method

Abstract

In the rainfall nowcasting, various error sources cause high uncertainty and bias of rainfall nowcasts. This study proposes a novel user-friendly and post-processing approach for rainfall nowcast bias correction from the perspective of hydrological users and operators of flood nowcasting system. This study first shows how the novel approach called the biward tracking (BiT) technique is applied effectively for real-time bias correction of the rainfall nowcast targeting 1-hour lead time and then verifies the applicability to the storm events occurred in Korea. The technique presented here consists of two steps, first the bias correction ratio is derived from backward tracking, which is then applied to the rainfall nowcast, searched by applying forward tracking. A simple and effective pattern correlation method is used for real-time storm tracking. As example applications, this study considers the six largest storm events that occurred in Korea from 2019 to 2021. This study uses the rainfall nowcasts generated by the MAPLE radar-based nowcasting system that contains a set of 36 nowcasts with a 10-minute interval covering the entire territory of Korea with a total of 1,050×1,050 grids with 0.5×0.5 km2 grid size, and the rain gauge data observed at over 580 stations. The derived results are then compared with those based on the conventional bias correction method. All the findings in this study confirm that the BiT-based method outperforms the conventional method. That is, the correction ratios determined by the BiT-based method are all found to be within reasonable range, without any serious outliers. Also, the bias-corrected rainfall nowcasts are found to be close to the ground observations. On the other hand, the conventional method produces many overestimated outliers, including more than 36 mm/h, even worse than the uncorrected rainfall nowcasts.