Flood risk analysis with geospatial artifical intelligence techniques

Abstract

In recent years, extreme precipitation events, rapid build-up in urbanization, and irregular land use have significantly increased flood risk. In order to mitigate risks and increase urban resilience, there is a need for the integration of innovative approaches to classical disaster management methods. In this context, Geographic Information Systems (GIS) offer robust spatial analysis workflow and data processing for efficient decision-making capability. This study uses geospatial artificial intelligence (GeoAI) methods to develop a flood risk analysis model in an open-source Python environment. The proposed methodology is applied in the Marmara Region of Türkiye as a case study highlighting flood risk. There are parameters that increase the risk of this disaster. Many factors, especially precipitation regime, dreinage density and distance to these waterways, population density in the region, topographic structure of the land, water flow direction and accumulation, affect the risk of floods and inundations. In this multi-parameter compound, determining the flood and inundation risks in the region is essential for effective disaster management. With the Python-based methods used, the dependency on GIS tools has been reduced and an automatable analysis process has been presented. As a result of the analyses conducted in the Python environment, the areas with high flood risk in the Marmara Region have been presented as an integration of the criteria determined for the hazard and vulnerability map. In this regard, spatial data processing, modeling and analysis are carried out in an integrated manner through open-source libraries. The vulnerability of the Marmara Region in this context, with its potential to work in harmony with which are XGBoost and Random Forest machine learning algorithms, positively affects flood management and offers an innovative perspective. In this context, the flood risk map of the Marmara Region is produced for eleven cities using open-source and governmental data gathered from official institutions to serve as an accessible guide for decision-makers. This study aims to estimate the flood hazard in the Marmara Region using Python-based multi-criteria analysis methods which is Analytic Hierarchy Process (AHP) method is used. Additionally in flood vulnerability context XGBoost and Random Forest machine learning algorithms are used. It is aimed to obtain a holistic high-resolution flood and inundation risk map by combining hazard and vulnerability evaluations. A hybrid model was adopted by examining flood hazard and flood vulnerability components together. A comprehensive flood risk estimation was carried out in the Marmara Region by creating maps using a machine learning-based method in the production of flood vulnerability maps and integrating these maps using the Python-based AHP method in the production of flood hazard maps. Hydrological data was integrated with spatial analysis. Historical disaster data and real-time meteorological parameters were associated using artificial intelligence algorithms. This comprehensive approach aims to develop strategies for the early detection of flood events that may occur in the Marmara Region by revealing how both Hazard and Vulnerability elements complement each other.