A review of ensemble forecasting techniques with a focus on tropical cyclone forecasting

Abstract

This paper presents a general review of ensemble forecasting techniques, with a focus on short-range and tropical cyclone predictions. The basic ideas and terminology of ensemble forecasting are introduced, and using four measures to evaluate an ensemble (ensemble mean forecast, consistency, spread versus skill, and inclusiveness), various potential utilities (e.g. dynamical probabilistic forecasts) are illustrated. Since the perturbation methodologies designed for medium-range forecasts of mid-latitude synoptic-scale systems - singular vectors, bred modes, and so on - are already quite mature, they are only briefly described here. The general problems encountered in applying ensemble forecasting techniques to short-range and tropical cyclone forecasts are diagnosed, and some recent studies on these topics reviewed. In general, the perturbation methodologies used for short-range ensembles to date can have a skill comparable to or slightly higher than their corresponding high-resolution control forecast. However, the complicated problem of the relationship between initial condition errors and model deficiencies persists. A similar situation also applies to ensembles designed for tropical cyclone forecasting. An additional difficulty is the different error characteristics encountered in the tropics, mainly the result of the strong convection in the area and the mutual interaction with the ocean. Studies from several research groups use quite different perturbation methodologies, but the results are encouraging. Most of them performed ensemble forecasting of tropical cyclone motion, but extensions to tropical cyclone intensity forecasts are also being developed.