| Popis: |
The forecast of rainstorm in flood season has always been the key and difficult point in the meteorological forecasting operation. First, we used the daily precipitation data from 2 227 meteorological stations in Zhejiang Province from 2016 to 2021 during the flood season (April to October), and divided the precipitation region by applying the K-means clustering algorithm, which employed the Euclidean distance as the similarity measure. Then, the regional correction method is formed by combing the spatial-temporally improved bias correction method and divided regions. Finally, we applied this method to perform the regional correction and validation on the Zhejiang Multi-Model Objective Consensus Forecast (OCF), compared with the overall correction not combined with divided regions.The results are as follows. (1) The K-means clustering algorithm can divide Zhejiang Province into 7 precipitation-similar regions, which show distinct regional characteristics closely related to the topographic features of Zhejiang Province. (2) According to validation during the 2021 flood season, the regional correction performed better than the overall correction in the OCF forecasts.Its main advantages lie in effectively reducing the false alarm of precipitation in the clear-rain forecast and substantially increasing the hit rate (POD) for the heavy rain and above, especially for the rainstorm and above from 0.25 to 0.41. (3) The typical validation show that, for both systematic and convective precipitation, the regional correction can significantly improve the intensity and falling area of precipitation for the rainstorm and above. Especially for the systematic precipitation, the regional correction demonstrated more remarkable effects, which can predict heavy rainstorms. |
