Abstract
Statistical verification of short term numerical weather prediction (NWP) experiments using the Japan Meteorological Agency (JMA) non-hydrostatic model (NHM) and the advanced research WRF (the weather research and forecasting model), referred to as WRF-ARW, was conducted around Japan and Southeast Asia. Two seasons (July 2007 and January 2008) that include typical weather systems in each region were selected for the verification. The results of the verification were compared between the two models with the same initial/boundary conditions, the same domain size (3000 km × 3000 km), and the same horizontal resolution (20 km). Forecast results were verified by the surface precipitation estimated by satellites and vertical profiles based on sonde observations.
The threat scores for precipitation around Japan determined by the two models are almost the same. In the rainy season (July 2007), they are not far from the typical values for summer determined by the JMA operational mesoscale model. WRF tends to predict more rain than NHM and the bias scores of NHM are closer to unity than those of WRF-ARW in most precipitation intensity ranges. In the Southeast Asia region, the threat scores of the two models are only about half of those for the rainy season in Japan. In the dry season (July 2007), the threat scores determined by WRF are slightly better than those by NHM, while in the rainy season (January 2008) NHM is slightly better than WRF-ARW for weak to moderate rains. Statistical scores against sonde are of almost the same magnitude, though the root mean square errors for wind of NHM are slightly better than those of WRF-ARW.
