Arid Zone Research ›› 2022, Vol. 39 ›› Issue (1): 64-74.doi: 10.13866/j.azr.2022.01.07

Previous Articles     Next Articles

Spatial error characteristics of rainstorm forecasts of large-scale numerical model over the northeastern side of Tibetan Plateau

ZHANG Junxia(),KONG Xiangwei(),LIU Xinwei,WANG Yong   

  1. Lanzhou Central Meteorological Observatory, Lanzhou 730020, Gansu, China
  • Received:2021-04-24 Revised:2021-06-02 Online:2022-01-15 Published:2022-01-24
  • Contact: Xiangwei KONG E-mail:18419611389@163.com;xiangwei580@163.com

Abstract:

Rainstorm targets were identified from May to September in 2019-2020 over the northeastern side of Tibetan Plateau by using the Contiguous Rain Area technique. There were 18 rainstorm targets in European Center for Medium-Range Weather Forecast (ECMWF) and 11 in Global/Regional Assimilation and Prediction System-Global Forecast System (GRAPES-GFS) mode. The spatial error characteristics—including displacement, intensity, and pattern error—of the two models for rainstorm forecasts were analyzed quantitatively, and the applicability of the two models in rainstorm forecasts over the northeastern side of Tibetan Plateau is here summarized. The results of the study showed that: (1) The pattern errors of rainstorm forecasts in ECMWF and GRAPES-GFS accounted for the largest proportion of the total errors. ECMWF had the smallest intensity error ratio for rainstorm forecast, followed by displacement errors, while the GRAPES-GFS model had the smallest displacement errors, followed by intensity errors. (2) The rainstorm areas in the two models shifted more northwestward compared with observations. The forecast of ECMWF for the rainstorm center was southwestward, while that of GRAPES-GFS was only westward. (3) The forecast of rainstorm area produced by the two models was obviously too limited, meaning that it could easily fail to predict rainstorm events. The GRAPE-GFS model underestimated both maximum precipitation and rainfall intensity by more than 40%, while ECMWF under-forecasted about 11.49% for the average rainfall intensity and overestimated about 1.47% for maximum precipitation. (4) The two models showed a better forecasting ability for rainstorm in the southeastern Gansu and southwestern Shaanxi areas, and a poorer one in the northern Shaanxi, Ningxia, and other northern areas.

Key words: Contiguous Rain Area, spatial error, rainstorm, the northeastern side of Tibetan Plateau