Abstract

On average, two tropical storms or cyclones enter the Mozambique Channel between the African mainland and Madagascar each year. Their impact can be devastating. The tropical cyclone Idai, which hit land in Mozambique in 2019, was one of the deadliest storms on record in the Southern Hemisphere. The tracks and strengths of tropical systems are difficult to predict more than a few days ahead. However, their relationship with large-scale features such as the Madden--Julian Oscillation have sparked a hope that the forecast horizon can be extended to the subseasonal time scale, beyond five days. This is the first systematic study of subseasonal skill in predicting tropical storms and cyclones in the Mozambique channel. The ability of a state-of-the-art subseasonal prediction model to predict Idai and 39 other tropical systems is assessed. The model only beats a climatological forecast in predicting local pressure minima at lead times of 1--3 days. Local wind speed maxima forecasts do not outperform a climatological forecast at any lead times. Yet, a few ensemble members usually hint of possible developments up to two weeks ahead. Key environmental precursors of tropical cyclogenesis are also skillfully predicted at lead times of up to two weeks, although the skill deteriorates substantially during storms. In summary, one cannot yet expect dynamical models to predict tropical systems directly. This might change, but for now the model output from dynamical prediction systems should be combined with statistical models and artificial intelligence algorithms to obtain higher subseasonal prediction skill.