Flood susceptibility assessment is a critical aspect of flood management, particularly in regions highly prone to flooding, such as the Teesta River Basin in the Indian subcontinent. This study investigates the effectiveness of standalone advanced machine learning ensemble models—LogitBoost (LB), Extreme Gradient Boosting (XGBoost), and Gradient Boosting Machine (GBM)—in predicting flood susceptibility, and introduces a novel hybrid stacking ensemble model, the Hybrid Tri-Boost Stack (LB-XGBoost-GBM). The research leverages spatial data augmentation, utilizing Cartosat-1 satellite imagery to enhance the predictive accuracy of flood susceptibility models. Feature selection was rigorously performed using the Boruta algorithm to identify the most influential flood predictors. The models are evaluated using performance metrics, including confusion matrices and Area under the Receiver Operating Characteristic Curve (AUC-ROC) curves, to validate their predictive capabilities. The findings underscore the superior performance of hybrid ensemble method (accuracy of 99.30%, 91.50%, and 91.40% on training, testing, and validation datasets, respectively), which integrates the strengths of the standalone models and highlight the importance of spatial data augmentation in improving model performance, resulting in more accurate and robust flood susceptibility predictions. The outcomes of this study are expected to significantly contribute to the development of effective flood risk mitigation strategies, offering valuable insights for policymakers and disaster management authorities in flood-prone areas.