Home | About Us | Contacts | Help | Feedback | SEARCH Global | Australia | NSW | Vic. | Qld | WA | SA | Tas. | ACT | NT | Ant. | Weather & Warnings | Hydrology | Climate | Numerical Prediction | About Services | Learn About Meteorology | Registered User Services |

Global Guide
Main Menu

Global Guide to Tropical Cyclone Forecasting:
CHAPTER 3: TROPICAL CYCLONE MOTION

3.1 INTRODUCTION

Forecasting the future location of a tropical cyclone is universally considered to be the most important function by tropical cyclone warning centres (McBride and Holland, 1987). This level of importance is reflected in the large number of forecast techniques that have been developed using a wide range of approaches, from empirical through statistical and dynamical. Until recently, technique development, together with improved observing methods in some basins, had resulted in a slow, but steady decrease of forecast error of approximately 1% per year (Neumann, 1981). Over the past five years, a marked increase in basic research on tropical cyclone motion has occurred, including several concomitant field experiments in the western North Pacific region (Elsberry and Abbey, 1992). At the same time the resolution and initialisation of tropical cyclones in numerical models have been considerably improved. Sharp improvements in forecast skill have resulted, with some cyclone regions having record low levels of forecast errors in 1991/1992 (F. Woodcock, personal communication, 1992).

Substantial improvements are still possible and to be expected, however. Mean forecast errors increase almost linearly to 72 hours, are approximately 30% of the cyclone displacement over the same time, and, when normalised by climatology and persistence, have a small initial skill that increases with time (Pike and Neumann, 1987; Leslie et al., 1990). The distribution is skewed towards large forecast errors, so that the median error is significantly better than the mean. Eliminating these large errors would produce a major improvement in forecasting.

The ultimate limit on forecast errors seems to be much lower than the current level. Bell (1979) compared the best track analyses from several western North Pacific cyclone warning centres (during a period of both aircraft and satellite observations) and found that the uncertainty in the initial location lay around 40 km. Simple extrapolation of this uncertainty level using CLIPER, or the commonly noted linear error growth rate, provides forecast errors that are half those of current forecast techniques. This simple analysis indicates that the current forecast errors may be twice the level dictated by predictability limits.

Further improvements in track prediction will come from several areas, including: research, improved observations, numerical models and statistical techniques. Continued research is required to elucidate the fundamental processes and to provide a solid foundation for future technique development. Improvements in observations are essential, especially those required to better locate tropical cyclones and define their outer circulation and near environment. Improvements in numerical models can be expected to provide further improvements in forecast skill. In particular, improvements will come from better model resolution, the use of synthetic observations to fill in gaps, better assimilation of the cyclone and its environment, and the use of coupled atmosphere/ocean models. Coupling numerical model and statistical forecasts will continue to provide further improvements.

An interesting and powerful new approach is to combine some estimate of the forecast skill with the forecasts. This can be done by so-called Monte Carlo methods of integrating a numerical model many times, whilst introducing slight changes to the initial analyses (Palmer et al., 1992). Alternatively, the forecasts from several centres can be combined in a manner which indicates the degree of forecast uncertainty (Leslie and Holland, 1991).

This chapter describes the approaches used for analysing and forecasting tropical cyclone position. It is divided into four main sections covering position analysis, forecast techniques, the approach to forecasting, and verification methods. Because of the strong operational content of the motion chapter in A Global View of Tropical Cyclones (Elsberry, 1987) some overlap exists on the description of forecast techniques.

Contents Chapter 3.2