Comparative Analysis of Satellite-Derived Elevation Models for Urban Planning

Thesis Abstract

This study addresses the critical need for accurate terrain representations to inform urban planning decisions, as elevation models influence flood risk assessment, infrastructure siting, and cadastral accuracy in rapidly urbanizing contexts. The aim is to compare satellite-derived elevation models (SDEMs) from multiple sources to determine their suitability for urban planning applications across varying land cover and shadow conditions. Specific objectives are to (i) evaluate vertical accuracy and LiDAR-consistency of Sentinel-DEM, SRTM, TanDEM-X, and ALOS PALSAR-derived products; (ii) assess the influence of land cover, urban canyon effects, and terrain complexity on model performance; (iii) develop a composite calibration framework to harmonize SDEMs for planning tasks; and (iv) propose guidelines for model selection tailored to distinct urban planning scenarios. The research adopts a comparative cross-sectional design conducted in the city of Lagos, Nigeria, leveraging a stratified sampling framework to capture diverse urban morphologies. The population comprises publicly available SDEMs and high-resolution reference datasets, including a city-scale LiDAR point cloud (covering 400 km2 with 1–2 m native product resolution) and orthophotography, collected within a two-year window. A total of 5,000 ground control points (GCPs) derived from terrestrial surveys and high-precision GNSS benchmarks are used to validate each SDEM, with 1,000 points reserved for independent validation. Data collection instruments include differential GNSS receivers for GCPs, LiDAR-derived height surfaces, and corresponding metadata from each SDEM provider. Statistical techniques encompass quantitative accuracy assessment using RMSE, MAE, and bias stratified by land cover class, followed by multivariate regression and ANOVA to identify factors driving discrepancies. A machine learning-based calibration model, employing random forest and gradient boosting, is developed to fuse SDEMs into a harmonized surface, with model performance evaluated through cross-validation and information criteria. The analysis also incorporates spatial analysis methods such as terrain ruggedness index, urban canyon modeling, and hydrological impact assessment to demonstrate practical implications for drainage planning and flood risk mapping. Expected findings indicate that high-resolution SDEMs (e.g., TanDEM-X) exhibit superior vertical accuracy in dense urban cores but are sensitive to data gaps in shadowed areas, while coarser products (e.g., SRTM) show systematic underestimation in low-relief zones and in tall-building canyons. The study anticipates that a calibrated fusion approach improves absolute accuracy by 15–25% across land cover types and reduces inter-model bias, enabling more reliable urban planning inputs. The contribution to knowledge lies in (i) providing a robust, transferrable framework for evaluating and harmonizing SDEMs in African urban contexts, (ii) identifying the most impactful error sources associated with different sensor modalities, and (iii) delivering a practical calibration schema and decision-support guidelines for planners handling heterogeneous elevation data. The theoretical lens draws on Tobler’s First Law of Geography to justify the spatially varying accuracy and on the Theory of Geographic Information Literacy to frame the interpretation of elevation data for decision-makers. The main conclusion expected is that no single SDEM suffices for all urban planning tasks; instead, an evidence-based fusion of models, guided by land cover and urban morphology, yields the most reliable elevation surfaces for planning purposes. Recommendations include adopting the proposed fusion framework in municipal GIS workflows, prioritizing TanDEM-X and LiDAR-derived products for dense urban cores, and employing the calibration model to harmonize coarser SDEMs for regional-scale analyses, with further work suggested on extending the framework to dynamic urban growth monitoring and climate resilience assessments.