抄録
Current numerical weather prediction models are capable of making skilful forecasts in the middle and high latitudes for approximately one week. In the tropics, however, the situation is rather different and the predictability is limited to 2-3 days. A number of factors are responsible for the poor model performance in the tropics, namely the lack of adequate data, problems with analysis of even the primary variables and inadequate parameterization of physical processes which leads to large systematic errors in the forecast models. In this paper the status of tropical numerical weather prediction at a number of operational and research institutes is reviewed and attempts to overcome some of the problems listed above are summarized.
