Professor Southern Methodist University, Texas, United States
Abstract Submission: This research aims to provide insights on how non-traditional and crowd-sourced data can improve understanding of the impacts of urban flooding on human behavior. The first data source is Inrix traffic congestion data, which can be compared to historical background conditions to identify where slowdowns due to flooding are most significant. The second source is cell phone trace data that can provide insights on how phone users’ trip patterns (speed, routes, and travel modes) and other activities vary between flooded and non-flooded conditions. We also leverage crowd-sourced data (311, Twitter, Waze user reports, and Federal Emergency Management Agency claims) to map public reports about floods and assess their correlations with other data sources.
These data are combined with hydrologic data and models, including locations and depths of flood-prone roadway depressions and river flood conditions from gages and the National Weather Service’s National Water Model. All of this information is used to identify which observed changes (e.g., traffic disruptions) are likely to be caused by flooding and how floods affect human behavior and urban traffic.
Patterns in the data are initially identified visually and analyzed statistically. Subsequent work will develop machine learning models to predict how conditions and human behavior may change for particular storm threats and under climate change scenarios. The models will then feed into an agent-based model to identify where and how flood and traffic management should be adapted to minimize disruptions. Insights on urban vulnerability and risk will also be obtained to identify environmental justice implications. Preliminary findings on data patterns and analytics will be presented at the conference.
Learning Objectives/Expected Outcome (Optional) : Understand the value of crowd-sourced and non-traditional data sources in the water sector
