Date of Award

2025

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy in Civil Engineering (PhD)

Administrative Home Department

Department of Civil, Environmental, and Geospatial Engineering

Advisor 1

David Watkins

Committee Member 1

Alex Mayer

Committee Member 2

Mary Ellen Miller

Committee Member 3

Gary Campbell

Abstract

Through the integration of high-resolution geospatial data, hydrologic modeling, and spatial statistics, this research addresses three interrelated topics that are critical to sustainable, equitable, and climate-resilient urban planning. The overall objective of this dissertation is to provide data-driven and machine learning-based geospatial analysis frameworks that enhance our understanding of environmental inequities, flood exposure, and the potential for wetland restoration, with a focus on urban contexts. The study presented in Chapter Two examines the distribution and accessibility of urban green spaces in Detroit, Michigan, using high-resolution geospatial data and geospatial analysis methods, including geographically weighted regression (GWR) and network-based analyses. This study correlates urban green space access inequities with social justice indicators and offers strategies for urban planners to address these inequities. The case study finds that 87% (53%) of buildings lack a park or recreation area within a quarter-mile (half-mile) walking distance, and neighborhoods with higher social vulnerability scores tend to have significantly lower green space availability.

The study presented in Chapter Three assesses urban flood susceptibility using machine learning models to evaluate the influence of topography, land cover, and infrastructure on flood susceptibility in urban environments. Random Forest (RF) and Extreme Gradient Boosting (EGB) models are trained using high-resolution spatial predictors and validated against both FEMA flood insurance claims and Detroit open data flood complaints. The RF method achieves accuracies of 81% and 82% for the flood complaint and FEMA datasets, respectively, and the EGB model achieves accuracies of 57% and 69%. Variable importance analysis reveals that infrastructure-related layers are the most influential in predicting flood complaints, whereas terrain-based variables are dominant in the FEMA model. Additionally, this study reveals significant differences in how flood impacts are distributed in relation to socioeconomic vulnerability.

In Chapter Four, a spatial framework is developed and implemented to identify high-priority wetland restoration sites based on terrain, land cover, and hydrologic potential. A methodology that integrates high-resolution remote sensing data with stochastic depression analysis is used to map potential wetland restoration sites, and the water storage capacity of the identified depressions was quantified. This analysis also incorporates impervious surface and land cover masking, as well as proximity to infrastructure to filter out locations where wetland construction would be infeasible, providing a refined set of viable potential wetland restoration sites.

Collectively, this research contributes to the field of civil, environmental, and geospatial engineering by integrating high-resolution spatial data with machine learning and hydrologic modeling to provide frameworks that can be utilized by urban planners and decision-makers. The spatial frameworks developed through this research provide scalable models that can be adapted for use in other regions facing similar challenges, including infrastructure stress, climate impacts, and environmental inequities. By linking physical and social dimensions of urban systems, this dissertation advances data-driven approaches for building more resilient, sustainable, and equitable urban environments.

Available for download on Friday, July 31, 2026

Share

COinS