Statistical forecast model for tropical cyclone formation in the Australian region

The relationship between tropical cyclones (TC) in the Australian region and large-scale climate modes and signals such as El Niño - Southern Oscillation, Indian Ocean Dipole or the Australasian monsoon is the subject of ongoing interest. This study introduces indices describing pre-seasonal patterns of environmental parameters affecting TC formation within the Australian region. These indices are used to develop a seasonal forecast model for TC development in the Australian region and its subregions. The atmospheric and oceanic parameters are examined to predict the probability of a number of TCs occuring during the following season (Nov-Apr). The final forecast model uses a Bayesian approach with a Markov chain Monte Carlo method performed with a slice sampler. Finally stepwise regression allows to find the best subset of independent predictors for TC frequency in each subregion. The skill of the model is estimated with the technique of cross-validation from a series of independent hindcasts over the available data.