SEMINAR ANNOUNCEMENT

Extreme precipitation events cause substantial societal impact, yet they are challenging to predict. Recently, Machine Learning (ML) has revolutionised weather prediction by reducing the computational resources and time required to generate forecasts. However, their skill in accurately predicting weather, especially extreme precipitation events in the tropics, is not adequately evaluated. Here, we evaluate four global ML Weather Prediction (MLWP) models and two Numerical Weather Prediction (NWP) models for 58 extreme precipitation events across the global tropics. One MLWP model, namely AIFS, has shown reasonably good skill in the short-range (3-day lead) prediction of tropical cyclone (TC) induced heavy precipitation. However, all models failed to predict the orographic precipitation events, such as the Wayanad event of 2024. The limitations in the training data and the inadequately resolved orography are suggested to be the reasons for the poor skill in predicting orographic precipitation.