Design and evaluation of a drone-based flood risk mapping system

Thesis Abstract

Flooding presents a significant hazard in urban and peri-urban environments, causing extensive socio-economic disruption and environmental degradation. Effective flood risk mapping is essential for disaster preparedness, emergency response, urban planning, and climate resilience strategies. Traditional flood risk assessment methods, primarily reliant on ground-based surveys and remote sensing data from satellites, often face limitations related to temporal resolution, spatial coverage, and operational costs, which hinder timely and high-resolution hazard identification. This study aims to design and evaluate a drone-based flood risk mapping system to enhance the accuracy, timeliness, and specificity of flood hazard assessments. The specific objectives include developing a drone deployment protocol tailored for flood-prone areas, integrating multispectral imagery and LiDAR data for detailed topographical and land use analysis, and assessing the system's effectiveness compared to conventional methods. The methodology adopted a mixed-methods research design, combining quantitative spatial analysis with qualitative validation. The target population comprised flood-prone urban districts within a metropolitan city of approximately 2 million inhabitants, with a sample size of 15 selected flood-prone zones identified through preliminary hazard and vulnerability assessments. A fleet of ten quadcopter drones equipped with high-resolution RGB, multispectral, and LiDAR sensors was deployed for data collection during the peak rainy season, capturing post-rainfall flooding extents over these zones. Data collection instruments included drone-mounted sensor payloads, GIS software for data processing, and field-based ground truth validation using differential GPS. The validity and reliability of the data were ensured through calibration of sensors, repeated flight missions, and cross-validation with existing satellite imagery (Sentinel-2 and Landsat 8). Analytical techniques involved GIS spatial analysis, supervised classification algorithms (Random Forest), and statistical validation using accuracy assessment metrics such as the Kappa coefficient and Receiver Operating Characteristic (ROC) analysis. The anticipated findings are that the drone-based system will produce high-resolution flood inundation maps with greater temporal responsiveness and accuracy compared to traditional satellite-based assessments. The multispectral data will facilitate differentiation between floodwaters and land cover, while LiDAR-derived elevation models will enable precise topographical analysis critical for flood modeling. The system is expected to identify flood extents with an overall classification accuracy exceeding 85%, significantly improving hazard delineation in rapid assessment scenarios. Additionally, the study will reveal that drone-based mapping reduces operational costs by approximately 40% and can deliver real-time data within two hours post-flight, thus enabling more responsive disaster management. This research contributes to knowledge by demonstrating the feasibility, efficiency, and enhanced spatial precision of drone technology in flood risk mapping. It extends existing remote sensing paradigms by integrating multispectral and LiDAR sensors into a unified drone deployment framework, offering a scalable solution for local authorities and disaster risk management agencies. The study also elucidates the operational challenges and provides a standardized protocol for drone deployment in flood scenarios, thereby filling critical gaps in current hazard assessment methodologies. The main conclusion emphasizes that drone-based flood risk mapping offers a viable and superior alternative to traditional remote sensing techniques, especially in rapidly changing flood events requiring immediate hazard delineation. Based on the findings, it is recommended that urban disaster management agencies incorporate drone technology into their flood monitoring systems, develop dedicated operational protocols, and invest in capacity building for drone pilots and GIS specialists. Future research should explore the integration of real-time sensor data with predictive flood modeling algorithms to further improve early warning capabilities and community resilience strategies in flood-prone areas.