Assessing Urban LIDAR Data Quality for Flood Risk Mapping in Coastal Cities

Thesis Abstract

Urban flood risk in coastal cities is increasingly influenced by the quality and integration of lidar-derived terrain and surface datasets, yet existing assessments rarely quantify how data quality variations propagate into flood-risk mapping outputs. This study investigates how urban LIDAR data quality attributes—point density, vertical accuracy, intensity, and classification reliability—affect the accuracy and reliability of flood risk maps produced for storm surge scenarios in coastal metropolitan contexts. The aim is to establish empirically reproducible relations between LIDAR data quality metrics and flood map performance to inform best-practice data acquisition and processing workflows. Specific objectives are (1) to quantify the influence of point density on DEM accuracy and hydrological feature delineation in coastal urban environments; (2) to evaluate vertical uncertainty and its propagation through hydraulic modeling using a coupled flood-modeller; (3) to assess the impact of LIDAR classification errors on urban impervious surface mapping and resultant flood extent estimates; (4) to compare flood-risk outputs generated from high-quality versus lower-quality LIDAR datasets across multiple coastal neighborhoods; and (5) to develop a data-quality augmentation framework that mitigates identified weaknesses through post-processing and data fusion. The methodology adopts an empirical field study design conducted in three coastal cities with diverse urban morphologies. The population comprises municipal GIS datasets, aerial LIDAR point clouds (collected during a common surveying season) and validated field measurements. A stratified sampling approach selects 15 representative urban blocks per city, totaling 45 blocks, ensuring coverage of high-rise cores, mixed-use districts, and low-lying waterfront zones. Data collection integrates (i) airborne LIDAR datasets at nominal densities of 6–8 points/m2 (high-density) and 2–3 points/m2 (low-density) with accompanying intensity and classification layers; (ii) ground truth data from total station surveys and RTK-GNSS benchmarks (n = 180 checkpoints across all blocks) for vertical accuracy assessment; (iii) high-resolution orthophotos and municipal flood hazard models. Instruments include a formal quality assessment checklist, a hydrodynamic flood model (e.g., a calibrated MIKE or HEC-RAS framework) and a GIS workflow log to capture processing parameters. Validity and reliability are ensured through cross-validation of DEMs against ground truth, and inter-operator consistency tests for classification outputs (kappa statistics and confusion matrices). Data analysis employs a hierarchical approach first, descriptive statistics summarize LIDAR quality metrics and flood-map outputs; second, regression analyses quantify relationships between point density, vertical RMSE, and flood-map accuracy metrics (YES/NO flood delineation accuracy, while continuous metrics include intersection-over-union for inundation extents and root-mean-square error of depth). Third, ANOVA tests compare flood outputs across quality groups, and structural equation modeling (SEM) evaluates direct and indirect effects of LIDAR attributes on flood-risk predictions. Sensitivity analyses explore how different thresholding and classification schemes alter flood extents. The study integrates a theoretical basis from Tobler’s First Law of Geography and the nested-uncertainty theory to frame spatial dependency and error propagation, with practical grounding in the Theory of Data Quality for Geospatial Information. A conceptual model illustrating causal links among LIDAR quality, hydrodynamic inputs, and flood outputs is developed. Key expected findings indicate that vertical accuracy (±0.15–0.25 m RMSE in urban canopies) and point density above 6–8 pts/m2 significantly improve flood-depth estimation and boundary delineation, while misclassification of impervious surfaces can result in systematic underestimation of urban flood extents by up to 18% in low-lying zones. The study anticipates quantifying threshold effects where diminishing returns occur beyond certain quality levels, and identifying robust data-quality augmentation strategies, such as multi-temporal fusion with terrestrial laser scanning and synthetic aperture radar-derived elevation proxies. The contribution to knowledge lies in providing an empirical, transferable framework linking LIDAR data quality to flood-risk outputs in coastal urban contexts, offering actionable recommendations for procurement, processing pipelines, and uncertainty-aware flood modeling. Policy-relevant implications include guidance for city planners on minimum data standards and the design of flood mitigation strategies under data-quality constraints. The study concludes that standardized quality indicators and fusion-based workflows can enhance the reliability of flood-risk maps, and recommends adoption of quality-aware modeling protocols, routine QA/QC checks, and continued exploration of data fusion to mitigate residual uncertainties under practical constraints.