Abstract:

Based on precipitation data observed at regional meteorological stations from June to September 2020, we verified the precipitation forecasts from 7 numerical models.We divided the summer precipitation in Liaoning province into 5 types according to the precipitation intensity and selected 36-hour and 48-hour forecast times to test the distribution errors and correlations of the precipitation field, as well as the location of the rain-bands.The results show that the 36-hour forecast is better than the 48-hour forecast, and the amount of 48-hour forecast precipitation is generally larger than that of the real observations.In early summer, late summer, and early autumn, the forecast deviation of each model is relatively small, while in midsummer, the prediction deviation of the model is large, especially in those stations where the heavy precipitation occurs.According to the 36-hour correlation coefficient between the precipitation predictions and observations and scatter distribution map, the ECMWF (European Centre for Medium-Range Weather Forecasts) model has the best forecast performance in those stations where the precipitation is classified as weak and weaker.The NCEP (National Centers for Environmental Prediction) model and WRF_3KM (Weather Research and Forecasting) model followed; For the moderate, heavier, and heavy precipitation, the NCEP model has the best forecast performance, ECMWF, and WRF_3KM followed.Different models have different levels of deviation in the forecast of rain-band location, among which ECMWF has the best performance.

Key words: Numerical model, Precipitation forecast, Verification, Rain-band index