Details of various data used in this study.

Accurate forecasting of extreme rainfall events (EREs) at a regional scale with higher lead times is challenging due to the uncertainties in weather model predictions. This study introduces a novel technique to nowcast heavy- and extreme-rainfall events by analyzing early microphysical signatures in mesoscale convective clouds. The method primarily utilizes the cloud top temperature (T) - cloud effective radius (re) profiles derived using remote sensing. We estimate the probability of the occurrence of heavy- and extreme-rainfall events using a logistic regression model with attributes extracted from the T-re profile and cloud droplet size distribution. Our analysis indicates that the T-re profiles for normal-, heavy-, and extreme-rainfall events exhibit distinct microphysical characteristics, with a prominent diffusional zone during EREs. Applying this model to nowcast recent EREs in the southern Western Ghats (Kerala, India) demonstrates an overall skill score of 93% and a lead time of at least six hours, underscoring the effectiveness of the approach for nowcasting EREs at a regional scale.