An Optical Flow Derived Nowcasting Approach with Fuzzy Logic and Model Wind Field Vectors

Short term (0~6 hours) radar based quantitative precipitation forecasting (QPF), also known as nowcasting, plays an important role in flood warning and other hydrological applications. It was found that radar based QPF approaches generally show better performances comparing to model-based approaches. A fuzzy logic nowcasting algorithm is proposed in the current work. In this algorithm, the motion field is first derived through an improved Lucas-Kanade optical flow approach. The derived field is then combined with the model wind field calculated from the Rapid Refresh (RAP) numerical weather prediction (NWP) algorithm through a fuzzy logic approach. The historical data is also included into the fuzzy logic approach, which can be used to predict storm’s intensity. The storm’s location and intensity can be predicted through extrapolating the radar observation patterns into the future. The proposed approach was validated with an individual 48-hour hurricane case, Hurricane Irma that caused major economic and ecological damages during 2017. In the evaluation, the performance of the proposed approach is also compared with the results from a cross correlation approach, an optical flow approach, and the Rapid Refresh (RAP) prediction model