Hurdle-RMIL: Addressing Zero Inflation and Long-Tailed Imbalance in Infrared Rainfall Retrieval
The Hurdle-IMDL framework improves satellite-based infrared rainfall retrieval by addressing data imbalances. It uses a 'divide-and-conquer' strategy to manage 'zero inflation' (excessive no-rain data) and the 'long tail' (rare, heavy rain). This approach significantly reduces the systematic underestimation of extreme weather events compared to existing models.