Leveraging GIS and Remote Sensing for Urban Flood Risk Prediction

Thesis Abstract

Urban flooding presents a significant challenge to sustainable urban development, exacerbated by rapid urbanization, climate variability, and inadequate drainage infrastructure. Conventional flood risk assessment methods often lack the spatial precision and timeliness necessary for effective urban flood management, emphasizing the urgent need for advanced technological approaches. This study aims to leverage Geographic Information Systems (GIS) and Remote Sensing (RS) technologies to enhance the prediction and mapping of urban flood risks through an integrated, data-driven framework. The specific objectives are (i) to identify and map key hydro-meteorological and topographical variables influencing flood susceptibility in the city, (ii) to develop a spatial flood risk model using GIS and RS datasets, and (iii) to validate and evaluate the predictive accuracy of the model in real-world scenarios. The research employs a mixed-methods approach, combining quantitative spatial analysis with a qualitative assessment of model performance. The study area comprises a metropolitan city with recent episodes of urban flooding, with a population of approximately 2 million residents. The sample size includes 150 flood-prone zones identified through preliminary spatial analysis. Data collection utilizes satellite imagery from Sentinel-2 and Landsat 8 for the last five years, Digital Elevation Models (DEMs), precipitation records, land use and land cover data, and historical flood incident reports sourced from municipal agencies. GIS-based spatial analysis involves topographic wetness indices, flood hazard zoning, and proximity analysis. Remote sensing techniques such as Normalized Difference Vegetation Index (NDVI) and land surface temperature extraction are incorporated to assess environmental conditions influencing flood risk. Data analysis is conducted using a combination of statistical and spatial modeling techniques. Multiple regression analysis is applied to identify relationships between predictor variables and flood occurrences, while Machine Learning algorithms such as Random Forest are employed to develop the flood risk prediction model. The model's accuracy is assessed through receiver operating characteristic (ROC) curves, confusion matrices, and kappa statistics, with cross-validation conducted on segmented data subsets. Theoretically, the study is grounded in the Pressure and Release (PAR) model to understand vulnerabilities, alongside Tobler’s First Law of Geography to justify spatial autocorrelation considerations. Expected findings include significant correlations between elevation, land cover, drainage density, and recent precipitation patterns with flood susceptibility zones. The combined GIS-RS model is anticipated to produce high-resolution flood risk maps with predictive accuracy exceeding 85%. These maps can identify hot spots for targeted flood mitigation interventions, thereby facilitating more effective urban planning and emergency management. Moreover, the research is expected to reveal critical environmental and infrastructural factors that influence flood dynamics in the urban context. The study contributes to existing knowledge by demonstrating an integrated, replicable GIS-RS framework for urban flood risk prediction tailored to city-specific conditions, also highlighting the added value of remote sensing datasets in temporal and spatial monitoring. It advances the application of machine learning in spatial environmental modeling and offers a methodological template for similar urban settings globally. The primary conclusion underscores the importance of utilizing geospatial technologies for proactive flood risk management, emphasizing policy recommendations for integrating GIS-RS systems into urban planning processes. The study advocates for continuous environmental monitoring and the development of real-time flood prediction systems to reduce urban flood vulnerabilities, encouraging policymakers and urban managers to adopt technologically informed, data-driven strategies for sustainable urban resilience.