A hybrid-model-based probabilistic forecast procedure for the seasonal frequency of tropical cyclones

For effective prevention of climatic disasters from tropical cyclones (TCs), forecast information on TC activity needs to be developed in a form that can be practically exploited in the public. This study aims to construct a probabilistic forecast procedure for the seasonal number of TCs occurring over the western North Pacific based on a hybrid modeling technique. First, a numerical ensemble model, global seasonal forecasting system version 6 (GloSea6), is employed to the hybrid model, where the ensemble outputs are used for dynamical inputs to the statistical module. Then, a Poisson regression modeling is applied to the statistical module and the predictive probability distribution is produced. Lastly, a probabilistic forecast procedure using five forecast categories such as “Below Normal”, “Normal or Below Normal”, “Normal”, “Normal or Above Normal”, and “Above Normal” is proposed based on the probability distribution. All procedure is demonstrated and verified using GloSea6’s hindcasts over the 24 years (1993–2016) during six months (from June to November). This study contributes to disaster management by enhancing the usability and interpretability of seasonal tropical cyclone forecasts, thereby supporting anticipatory decision-making under conditions of climatic uncertainty.