Performance of the United Kingdom Meteorological Office Global Model in Predicting the Movement of Tropical Cyclones

Abstract

A detailed evaluation of the performance of the United Kingdom Meteorological Office Global Model (UKMO) in predicting the movement of 15 tropical cyclones (TCs) that occurred over the western North Pacific during 1987 is presented. The evaluation is based on the methodology used by Chan et al. That is, in addition to the usual mean forecast error, the following error measures are employed: systematic zonal and meridional errors, cross-track, and along-track error components relative to the climatology and persistence (CLIPER) track, and the M score. To identify further the strengths and weaknesses of the model, the forecasts are also evaluated according to four storm-related parameters: latitude, longitude, intensity, and 12-h intensity change. The analyses for the entire sample show that the skill of the UKMO generally increases with forecast intervals, as in the case of other numerical prediction models. However, the short-term forecasts are worse than those of CLIPER, and a substantial error exists in the initial position of the tropical cyclone in the model. The UKMO also has a tendency to overpredict recurvature for westward-moving TCs, and acceleration for recurved TCs. From the analyses of the subsamples stratified based on the storm-related parameters, the UKMO is found to have the best performance for TCs north of approximately 20°N and east of approximately 140°E. Tropical cyclones in the lower latitudes or in close proximity to large landmasses are usually poorly predicted by the model. The more intense the cyclone, the better the UKMO forecast. However, the model makes good predictions for TCs that are weakening.