Development of an Open-Source Urban Flood-Modeling Framework using GIS and Remote Sensing

Thesis Abstract

Urban flood risk is a growing challenge in rapidly urbanizing regions where conventional flood models are often data-intensive and siloed within proprietary software. This study addresses the need for accessible, interoperable, and reproducible flood modelling tools that can operate in data-constrained environments and support evidence-based urban planning. The aim is to design, implement, and evaluate an open-source urban flood-modeling framework that integrates GIS and remote sensing data to simulate pluvial and fluvial flood processes, generate decision-support outputs, and facilitate participatory planning. Specific objectives include (1) to synthesize a modular architecture that couples hydrological, hydraulic, and exposure-damage components within an open-source stack; (2) to implement data assimilation workflows using freely available datasets (SRTM digital elevation model, Landsat/ Sentinel-2 imagery, OpenStreetMap, and local rainfall records) to produce high-resolution flood inundation maps; (3) to develop and validate the framework against historical flood events across three case-city blocks representing diverse urban morphologies; (4) to evaluate framework performance in terms of accuracy, computational efficiency, and usability by non-technical stakeholders; and (5) to provide governance-relevant outputs and documentation to support open data policies and disaster risk reduction planning. The research adopts a design-research methodology, combining a software engineering perspective with quantitative and qualitative evaluation. The population comprises urban catchments with documented flood events, city planners, and GIS analysts in the case study city. A purposive sample of three urban districts (high-density informal settlements, mixed-use corridors, and low-lying residential neighborhoods) will be studied, with 30 representative sub-catchments per district selected for model calibration and validation. Data collection will employ open data sources (digital elevation models at 10 m resolution, multispectral satellite imagery from Sentinel-2, climatic rainfall time series from local meteorological stations) and targeted semi-structured interviews (n=12) with planners and engineers to assess usability and decision-support value. Instrumentation includes a calibration dataset of 25 historical flood events, model validation metrics (Nash-Sutcliffe efficiency, Kling-Gupta efficiency, root mean square error, and flood extent overlap), and a usability questionnaire anchored in the System Usability Scale. Methodologically, the framework will integrate a hydrological runoff model (SCS-CN or green–grey drainage analogs where data permit), a hydraulic routing module (HEC-RAS-inspired open-source equivalents), and a flood-extent estimator using GIS-based raster inundation mapping. Data assimilation will employ Bayesian updating to reconcile model outputs with observed flood extents from post-event remote sensing-derived flood masks. Spatial analysis will leverage regression-based sensitivity analysis and machine-learning-assisted feature importance (random forest) to identify key drivers of flood susceptibility in different urban morphologies. Model validation will use cross-city transferability tests and scenario analysis under climate-frequency amplification. The theoretical underpinnings will draw on the Theory of Change for disaster risk reduction and the open science framework, supplemented by the Urban Resilience theory to explain how open-source tools influence governance and community adaptation. Expected findings include (i) a modular, extensible open-source framework with documented workflows and code repositories that achieve flood-inundation predictions within 0.5–1.5 m spatial accuracy for most test events, (ii) demonstrable improvements in reproducibility and transparency relative to proprietary models, (iii) insights into the relative contribution of rainfall variability, urban morphology, and drainage capacity to flood risk, and (iv) evidence that user-oriented outputs (interactive maps, damage indicators, and scenario dashboards) enhance planning discussions among non-technical stakeholders. The study contributes to knowledge by operationalizing an end-to-end, openly accessible flood-modeling solution that bridges hydrology, hydraulics, GIS, and remote sensing, and by providing a validated methodology for data-constrained cities to perform flood risk assessments with transparent uncertainty quantification. The main conclusion anticipates that open-source integration can democratize urban flood modelling, enabling municipal authorities to test adaptation strategies and engage communities. Recommendations include expanding data-sharing frameworks, developing multilingual user guides, integrating real-time data streams, and extending the framework to support multi-hazard risk assessments in rapidly urbanizing environments.