Regional and sub-basin tropical cyclone activity in the CMCC seasonal prediction system 3.5

Abstract Reliable sub-basin seasonal forecasts of tropical cyclone (TC) activity are fundamental in helping stakeholders make informed decisions and minimize economic and societal losses. While several institutions routinely release TC seasonal forecasts for traditionally studied basins, only a few provide global coverage, limiting confidence for other densely populated regions. Here, we apply a probabilistic clustering approach to identify track patterns across five basins (North Atlantic, Eastern and Western North Pacific, South Indian, and South Pacific) over the period 1993–2016 in retrospective seasonal forecasts produced by the Euro-Mediterranean Center on Climate Change Seasonal Prediction System 3.5. Despite a hemispheric bias in TC frequency—overestimated in the Southern Hemisphere and underestimated in the Northern Hemisphere—the spatial distribution of TC tracks is skilfully represented across all basins. Moreover, predicted and observed year-to-year variability are significantly correlated ( p < 0.1) for most clusters in the North Atlantic, Eastern Pacific and South Indian, while limited to one third of clusters in the Western and South Pacific. Conversely, skill for subtropical clusters is absent across all basins. An analysis of the North Atlantic reveals that cluster skill is mainly attributable to the representation of the ENSO-TC teleconnection, suggesting it may represent a primary driver of cluster predictability across basins. These findings demonstrate that sub-basin forecast skill can improve basin-wide performance in dynamical models, underscoring the value of regional forecasts for refining seasonal outlooks.