A systematic approach to tropical cyclone (TC) track forecasting in the western North Pacific is being developed for use by the Joint Typhoon Warning Center (JTWC) to improve the quantitative accuracy and qualitative (i.e., interpretative) utility of official TC track forecasts. A key requirement for improving the accuracy of official forecasts is to give the forecaster the ability to identify periods when dynamical models produce particularly erroneous track forecasts that should either be ignored or significantly modified by the forecaster.
During 1997 in the western North Pacific approximately 100 cases occurred in which the 72-h track forecast error for either the Navy Operational Global Atmospheric Prediction System (NOGAPS) or the Geophysical Fluid Dynamics (Navy version) (GFDN) TC model exceeded 360 n mi. A detailed evaluation and comparison of the forecast fields of two models and satellite imagery determined that for the vast majority of cases the cause of these large track forecast errors could be readily discerned. The causes could be characterized in terms of the conceptual models of the meteorological knowledge base of the systematic approach. Six forecast scenarios leading to large track errors have been identified: (i) false or excessive Direct Cyclone Interaction with either tropical or mid-latitude circulations; (ii) false or erroneous Indirect Cyclone Interaction with other TCs or weak disturbances; (iii) excessive or insufficient weakening in response to vertical wind shear; (iv) premature or erroneous recurvature through a thin subtropical ridge; (v) premature formation of a reverse monsoon trough; and (vi) failed or intermittent extratropical transition.
For the duration of each errroneous forecast scenario, both model forecast fields and forecast tracks tended to have consistent characteristics. Examples include anomalous growth or weakening of the model TC circulation, and characteristic track forecast shapes such as looping or immediate and unusual departures from the past motion of the TC. In some scenarios, large deviations changes in the temporal continuity of track forecasts occurred at the beginning and end of the erroneous forecast period. Tabulations of these characteristics will be presented, as will the conceptual models that are expected to assist forecasters in detecting and compensating for periods of erroneous forecasts by NOGAPS or GFDN.