Locally Optimal Weighting of Global Precipitation Forecasts from Precipitation Nowcasting and Numerical Weather Prediction
Quantitative precipitation forecast (QPF) is an important goal of numerical weather prediction (NWP). Precipitation nowcasting based on spatiotemporal extrapolation is also known to be a computationally feasible QPF method. Since nowcasting statistically outperforms NWP at shorter lead times, optimal forecasts have been achieved by merging precipitation forecasts from NWP and nowcasting systems. The optimal weights, which are essential for merged forecasts, has been usually assumed to be spatially-uniform. This study explores to improve merged precipitation forecasts further by optimizing the weights locally. Here, we propose a spatially-localized version of the threat score (local threat score; LTS), and local weights are optimized to maximize the LTS at each location.
We test the locally-optimal weighting with global precipitation forecasts. As the NWP component, the Local Ensemble Transform Kalman Filter with the Nonhydrostatic ICosahedral Atmospheric Model, so-called NICAM-LETKF, is used. As the nowcasting component, the RIKEN’s nowcasting system called GSMaP_RNC is used. The results show that merging precipitation forecasts is beneficial for global QPF. Merged precipitation forecasts the local weights provides better QPF than using spatially-uniform weights, especially in the extratropics.
Mr. Kenta Kurosawa RIKEN Center for Computational Science