Lee, Cheng-Chun (2021-03). Automated Methods for Evaluating Stormwater Drainage Systems at the Neighborhood Scale Using Mobile Lidar Data. Doctoral Dissertation.
Thesis
Assessing drainage conditions at the neighborhood level can help public works agencies to develop maintenance plans and mitigation strategies to guard against pluvial floods. Drainage condition assessments can also inform property owners about possible drainage problem areas. Current drainage condition assessment methods have two important shortcomings: (a) they depend on manual visual inspection, which is a time-consuming and labor-intensive process, and (b) they ignore areas outside the street right-of-way (e.g., adjacent front yards), despite the interdependency between public drainage system (e.g., roadside channel) and adjacent private properties. To address these shortcomings, this dissertation aims to develop automated methods for assessing the drainage conditions of roadside channels and adjacent land in residential areas by using mobile lidar (Light Detection and Ranging). Mobile lidar is increasingly used for evaluating infrastructure systems due to its ability to provide high-density and high-quality spatial measurements. This dissertation is organized into three technical papers. The first paper provides an automated process for inspecting and evaluating roadside channel systems using data obtained from mobile lidar. The Cloth Simulation Filtering algorithm was employed to split lidar point clouds into bare earth and object datasets. Six key geometrical attributes of roadside channels were computed and analyzed based on the bare earth dataset using statistical and heuristic methods. These geometrical attributes were compared to design and performance manuals to determine deficiencies and inform maintenance decisions. In the second paper, roadside topography was modeled to evaluate surface drainage conditions by incorporating semantic segmentation and flow direction determination using mobile lidar data. The semantic segmentation model identifies the topographic features of lidar images by labeling each pixel as roadside channel, road, or adjacent land. Through the flow direction determination technique, major end points that are away from the roadside channels were identified as problematic low points that could be vulnerable to water ponding. In the third paper, the developed methods were applied to two communities in Harris county (Sunnyside community) and Aransas county (Rockport community) in Texas, with a total street length of 4.67 centerline miles. The six geometrical attributes for channel conditions and two attributes for off-channel conditions were evaluated and compared in the case studies. Overall, this dissertation shows that the developed automated process can evaluate roadside channels and model the roadside topography effectively. The developed methods provide actionable information to both property owners (to address potential off-channel ponding issues) and the municipal authorities (to address channel issues) to mitigate against localized flooding, water ponding, and stormwater-related hazards.
