Date of Award
Spring 2024
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Civil, Construction, and Environmental Engineering
First Advisor
Walter McDonald
Abstract
Green Stormwater Infrastructure (GSI) has been increasingly utilized to improve urban stormwater management strategies. However, the performance and utility of GSI decrease over time if the infrastructure is not properly maintained. In recent history, the intrinsic operations and maintenance costs associated with the complex networks of new infrastructure have placed a burden on municipalities, ultimately prohibiting many from reaching the full extent of their stormwater management goals. One way for cities to achieve cost savings is through automated monitoring that can quickly assess the condition of GSI assets; however, existing cost-effective technologies are limited. Drones and satellites may be able to meet this gap through large-scale, high-resolution data that can potentially provide information on overland site conditions that can augment or replace in-person GSI inspections. The goal of this study is to apply machine learning classification methods to remotely sensed satellite and drone data to classify the land cover of GSI for use in maintenance and operations efforts. To do this, high-resolution drone and satellite imagery collected in 2022-2023 was utilized to classify 12 GSI sites in Milwaukee, WI into 4 landcover categories (healthy plants, unhealthy plants, dead plants and organic material, and inorganic material) using various supervised machine learning classification models. Results found that classification methods yielded accurate results when classifying both drone imagery (74% - 95%) and high-resolution satellite imagery (60% - 78%). Similar models were then used in the summer of 2023 to identify the maintenance needs of 93 GSI sites in Milwaukee, WI with 73% accuracy. Overall, this study provides a comprehensive overview of the integration of remote sensing and machine learning methods as a pathway for GSI monitoring data collection. In doing so, it highlights the strengths of each data source, the boundary of applicability for each technology, and the need for continued research and development.