Comparison of Performance of Various Multiple-model Ensemble Techniques in Forecasting Intensity of Tropical Cyclone

Abstract

ABSTRACT: the objective of this study was to explore an optimal multiple-model ensemble technique to aid the forecasting of tropical cyclone (TC) intensity. The maximum winds of TCs as forecast by the models of the european Centre for Medium-Range Weather Forecasts, the Japan Meteorological Agency and the national Centers for environmental Prediction for the period from July 2010 to October 2011 were studied. Performance of various multiple-model ensemble techniques, including equally weighted ensemble, weighted ensemble based on initial forecast error, weighted ensemble based on 12-hour forecast error, bias-corrected equally weighted ensemble and bias-corrected weighted ensemble based on initial forecast error, was verified against the TC intensities post-analysed by the hong Kong Observatory. Results showed that the equally weighted ensemble technique generally outperformed the best of the individual models and other multiple-model ensemble techniques. The mean absolute errors of the equally weighted ensemble technique were the lowest at 12, 24 and 36-hour forecasts, and the error spreads were generally the smallest from 12 to 72-hour forecasts.