Diagnostic Analysis of Multimodel Rainstorm Forecast for Cases Based on MODE Method

It is very important to analyze the performance of model rainstorm forecast for understanding intensity and position deviation precipitation improving ability. This paper uses classical scoring MODE (Method Object-Based Diagnostic Evaluation) method evaluate different products. The observation data used in article mainly include CMA (China Meteorological Administration) multi-source merged analysis, ECMWF (European Centre Medium-Range Weather Forecasts), CMA-Meso (Mesoscale CMA) SCMOC (National Center grid guidance product) data. At same time, possible correction heavy rainfall area explored by using high low-level circulation configuration model. main conclusions are as follows: ① spatial verification shows that number, intensity, location objects match observed best, which obviously better than forecasts. There significantly fewer objects, each single object larger, often fails report small convective precipitation. ② reason TS score it reduces number avoids “double punishment” phenomenon caused error, leads failure local ③ Analyzing relationship between situation forecasts model, found consistent with upper system, especially strong rising vertical velocity 700 hPa value specific humidity 850 hPa. When coincides air system but not ground convergence potential, has a large deviation. surface can be reference improve forecast.