Tropical Cyclone Track Forecasts Using NCEP-GFS with Initial Conditions from Three Analyses

Abstract

Tropical cyclone track forecast experiments were conducted using the National Centers for Environmental Prediction Global Forecast System with the initial conditions from three numerical weather prediction centers, to distinguish between tropical cyclone track forecast errors attributable to the initial state uncertainty and those attributable to the model imperfection. The average position error was reduced by replacing the initial conditions from the European Centre for Medium-range Weather Forecasts. The northward recurvature of Lupit (2009) was not reproduced with initial conditions from the Japan Meteorological Agency. It was consistent with the preceding study, indicating sensitivity to the initial state. The sensitivity to the model and the initial state was obtained for Parma (2009), as opposed to the conclusion of the previous study, where Parma was discovered to be insensitive to the initial state, and the error was assumed to come from the model difference. Insensitivity to the initial vortex structures in the predicted tracks for Parma indicates that the error in the steering flow formed by the environmental field around tropical cyclone contributes to the northward bias.