Evaluating Rainfall Erosivity Dynamics under Tropical Cyclone Conditions: An Ensemble Ran-Boost Perspective

Tropical cyclones are intense storms that dump a lot of rain and often lead to serious soil erosion, especially in coastal regions that are already vulnerable. Understanding how much damage this rain can do to the land is key if we want to be better prepared for these kinds of disasters. However, the usual methods we’ve used—like physics models or statistics—don’t always capture those really short bursts of intense rain very well. That means we might be underestimating the real risk of erosion. Therefore, to overcome such limitation, this study introduces Ran-Boost, a hybrid ensemble model combining the designed improved Random Forest and Extreme Gradient Boosting algorithms to improve erosivity estimation during TC events. The Ran-Boost model looks at weather data from coastal areas that get hit by cyclones a lot. It pulls together things like wind speed, pressure, sea and land temperatures to figure out how weather shifts over time and across different spots. From what we’ve seen, Ran-Boost does a better job than older models, especially when it comes to predicting how heavy rain leads to erosion, whether it’s hour-by-hour or during certain storms. That said, it’s not great at catching the worst erosion during the biggest cyclones—probably because we just don’t have enough detailed local data on those really intense downpours. Still, it’s a solid step forward in using machine learning for this kind of thing, and it shows that mixing different models might help us plan for disasters more effectively.