In the realm of climate predictions and extremes, machine learning (ML) techniques serve as versatile tools not only for capturing climate patterns but also for climate change impact assessments. Generally, each ML method comes with a substantial predictive skill, which demands for comparing performances of different models when selecting the most robust one. Among ML models, decision tree is a basic approach with a tree structure, whose performance was enhanced by implementing boosting techniques and ensemble learning. The improvement led to several tree-based boosting techniques including Adaptive Boosting (AdaBoost), eXtreme Gradient Boosting (XGBoost), Light Gradient-Boosting Machine (LightGBM), and Category Boosting (CatBoost). In this study, performances of these ML models were compared for predicting extreme drought under climate change impacts in the upper part of Reno basin, northern Italy.
