Evaluation of an Open-Source Radar-Based Nowcasting Tool in a Tropical Environment

In this presentation, an open-source Python library for radar-based nowcasting is introduced and evaluated. This library, based on optical flow tracking, can identify features in a radar image using corner detectors, tracking these features using an optical flow algorithm. Motion is extrapolated for future times, advecting multiple radar fields linearly or in a semi-Lagrangian scheme. In previous research, this library has shown skill over eleven events that ranged from two to twenty-three hours with cases showing critical success index values greater than 0.5 for up to 60 minutes. This presentation will expand this research and present the results of the evaluation of this nowcasting methodology using a challenging dataset – three months of radar data in a tropical environment.