Operational forecasting of tropical cyclones: past, present, and future

Forecast errors for two warning centres are presented and it is argued that statistical/synoptic methods of forecasting tropical cyclone movement have now been developed as far as is profitable. The effects of uncertainties in the 'best track' and of the large-scale circulation on the determination of forecast errors is discussed. Data are presented on the relationship between the movement of a tropical cyclone and its size and latitude. The need to look to global models for improved forecasts for periods of 48 hours or more is illustrated, but it is shown that current arrangements for the dissemination of observations and current methods of objective analysis of tropical cyclones are both totally inadequate for making useful forecasts for 48 hours or less. It is suggested that synoptic and off-time high resolution observations should be assimilated into a near continuous objective analysis in a domain nested within a global model before improved forecasts for 36 hours and beyond can be achieved.