Urban flood hazard assessment using FLA-optimized boost algorithms in Ankara, Türkiye

Abstract

This study presents a comprehensive analysis of flood hazard mapping in Ankara, the capital of Türkiye, highlighting the critical vulnerability of this major urban center to climate-related disasters. By applying advanced boosting algorithms—specifically, XGBoost, GradientBoost, and CatBoost—along with hyperparameter optimization through the Fick’s law algorithm (FLA), this research introduces an innovative methodology aimed at improving the reliability and accuracy of flood hazard assessments in Ankara’s urban landscape. The analysis utilizes an extensive dataset that integrates topographic, meteorological, hydrological, and anthropogenic variables to provide critical insights into the dynamics of urban flooding with a focus on Ankara’s vulnerability. This approach is novel in that it incorporates FLA for hyperparameter optimization, marking a significant advancement in flood hazard modeling and achieving higher model accuracy and generalizability. Notably, among the various determinants of flood hazard identified, elevation emerges as the most influential factor affecting flood risk in Ankara. This finding underscores the complex relationship between urban geography and flood hazards, and highlights the need for targeted urban planning and infrastructure development strategies to effectively mitigate flood risk. The implications of this research extend beyond the local setting, contributing valuable insights to the global discourse on climate change adaptation and urban resilience. By combining cutting-edge machine learning techniques with in-depth geographic analysis, this study offers a scalable and innovative model for flood hazard assessment and management, providing a critical tool for cities around the world facing similar challenges.