A Study of Forecast Errors in ECMWF Operational Model

In order to have a better uderstanding on ECMWF model and make proper use in operation, the distribution of monthly mean errors and its seasonal variation is studied with one year data of the predicted charts from May 1984 to April 1985. ECMWF numerical forecast is best in middle latitude. In the area north of 75oN, 500 hPa height is forecasted too high and 500—1000 hPa thickness is too warm in spring, summer and autumn, especially in summer, but the latter becomes colder than that observed in winter. In the area south of 30oN, 500 hPa height forecast is too low and 500—1000 hPa thickness is too cold in summer and autumn, with the coldest error center situated on the southeast slope of Xizang Plateau; but in winter and spring, value of error is not so large as in summer and autumn. This seasonal variation seems to correspond with the cloud and precipitation amount or the activities of tropical low over there. Over the Xizang Plateau, the 500 hPa high forecast error is almost negative in every month of the year, with the largest negative value appeared in summer, but it approaches to zero in winter. Comparing the values of systematic error with standard deviation, we may find that in low latitude area, especially in the vicinity of Xizang Plateau, EC forecast charts need modification with systematic errors.