The advancement of green infrastructure best management practices (BMP’s) in urban cities has led to increased operations and maintenance costs. One way for municipalities to achieve cost-savings is through automated monitoring that can quickly assess operations and maintenance needs; however, existing cost-effective technologies to do so are limited. Remote sensing data from drones and satellites may be able to meet this gap through high spatial and spectral resolution data. This data can be analyzed to gain information on BMPs that can augment or replace information gained from in-person surveys of BMP sites. The goal of this study is to apply machine learning algorithms to remotely sensed satellite and drone data to classify imagery of BMP sites into categories that can be used to assess maintenance needs. To do so, high resolution drone imagery ( < 5 cm) and satellite imagery ( < 30 cm) collected in 2022-2023 were utilized to classify 9 BMPs in Milwaukee, WI into 4 categories (healthy plants, unhealthy plants, dead plants and organic material, and inorganic material) using machine learning algorithms. Results found that supervised machine learning algorithms were able to yield accurate results when classifying both drone imagery (>85%) and high-resolution satellite imagery (>65%). These models will be used in summer and fall 2023 to identify and check BMP sites within Milwaukee, WI that may need maintenance actions based upon the model classifications. Overall, this study demonstrates the potential for utilizing machine learning algorithms and high-resolution remote sensing data for BMP classification.
