Abstract:

Using Grapes_3km, mesoscale regional model precipitation forecast data over East, North, and Northeast China and 1 h precipitation data set of Land surface Data assimilation System of China Meteorological Administration (CLDAS) from May to September of 2020, the short-term heavy precipitation forecast (≥20 mm·h^-1) around the Bohai Sea under the influence system of 850 hPa shear line outside the subtropical high, Northeast Cold Vortex, 500 hPa upper trough, typhoon, 500 hPa upper trough outside typhoon and 850 hPa shear line was tested and evaluated based on the neighborhood method.The results show that the FSS (fractions skill score) and hit ratio of each regional model are very low in the forecast of short-term heavy precipitation around the Bohai Sea when the neighborhood is 5 km.With the increase of the neighborhood, the test results give the score results that are valuable when there is displacement deviation in the forecast starting from the 50 km neighborhood.Grapes_3km and East China regional model FSS receives the minimum forecast skills in the 250 km and 300 km neighborhood, respectively, for the short-time heavy precipitation under the Category 5 influence system, whereas the other regional models do not receive the minimum forecast skills in the neighborhood.For the short-time heavy precipitation under other affected systems, the FSS of the four models in each neighborhood does not receive the minimum prediction skill, and the FSS of the North China model for the type 2, 3, and 4 affected systems is higher than that of the other models.The FSS of Grapes_3km for the type 1 affected system is higher than that of the other models.According to the test results of hit and false alarm ratios, Grapes_3km, the North China model and the East China model have better prediction effects for short-term heavy precipitation under type 1, 3, and 5 influence systems, type 2, 3, and 4 influence systems, and type 5 influence systems, respectively.

Key words: Influence system, Fractions skill score, Probability of detection, False alarm ratio