Modélisation De La Précipitation Dans La Partie Sud-Ouest De Madagascar En Fonction De L’Humidité, De L’OLR, De La Chaleur Latente Et De La SST Par Forêt Aléatoire

Abstract

This study aims to model precipitation in the southwest of Madagascar using explanatory climatic variables : humidity, Outgoing Longwave Radiation (OLR), latent heat, and Sea Surface Temperature (SST), through a Random Forest model. The goal is to understand the impact of these variables on precipitation variability and to predict them based on local climatic conditions. The Standardized Precipitation Index (SPI) was used to characterize climatic periods, distinguishing between normal years, dry years, and wet years from 1979 to 2018. The model’s performance was evaluated with an R² = 0.865, indicating a good explanatory capacity, and a SMAPE = 32.58%, which is deemed acceptable for climate forecasting. The results show that variables such as humidity, latent heat, and SST have a significant influence on precipitation in this region, highlighting the importance of these factors for understanding the local climate dynamics.