Decision Support for Flood Risk Mitigation:

Automated Data Collection and Visualization Tools

Project Summary

The team will develop automated data collection tools and interactive decision support systems to tackle the grand challenge of increasing coastal flood risks. They will address the challenge of collecting structure data by using machine learning and image analysis to create an automated tool to extract estimates of characteristics such as foundation heights from Google Street View, aerial, and satellite imagery. They also plan to address the need for better risk communication by developing a set of visualization tools that present a much richer set of information about local flood risk and the effectiveness of different risk mitigation options. Individual property owners will be able to use these tools to decide what measures to take to protect their assets, and local planners and policy makers will be able to develop “adaptation timelines” that prioritize infrastructure projects using estimates of current and future risk.

Research Question #1: Can structural characteristics relevant to flood risk (e.g., foundation heights, foundation type, square footage) be extracted from structure imagery using automated machine learning and image analysis methods?  Currently, data sets of risk-relevant structural parameters are expensive to collect manually using surveys or are scattered across innumerable jurisdictions and agencies. Existing data in Louisiana are obsolete, and improved records that account for post-flood reconstruction would be highly valued by public agencies, individual homeowners, and private firms such as insurers. They propose a creative solution of automated data collection and processing to avoid difficulties of crowdsourcing or compiling data from many jurisdictions.

Research Question #2: How is the accuracy of estimates of flood damage impacted by using obsolete data on structural characteristics?   At the planning level, flood risk is often expressed as expected annual damage (EAD), an average measure describing how much damage an area could expect to experience each year. The team will conduct a thorough analysis of how estimates of EAD vary when data from the automated data collection tool is utilized in the CLARA model, as a means of estimating the value of having better information about structural characteristics. They will also apply existing evaluation criteria used by Louisiana to explore whether policy makers would make different decisions about funding risk mitigation projects if they had access to better data.

Research Question #3: How can geospatial data about the probability distribution of flood depths be effectively integrated with information about structural characteristics to improve risk communication and decision making about risk mitigation?  Traditionally, flood risk data viewers present data in individual layers—for example, a map of 100-year flood depths (the depths with a 1% chance of occurring or being exceeded in a given year) or EAD in some aggregated spatial unit. There exists no publicly usable tool that consolidates flood depth data, structural characteristics, and estimated damage with information about the cost and effectiveness of risk mitigation options such as elevating one’s home. The proposed tool will do this, based on empirical evidence from research on risk perception and communication (cf. Kellens, Terpstra et al. (2013).

Research Question #4: How should risk communication tools differ depending on their intended audience and level of spatial aggregation?   Technical analysts at state agencies, local planning boards, and individual homeowners may have very different visualization needs. The team will apply best practices for risk communication to investigate how flood risk data should be visualized differently at different spatial scales, for different planning problems, and for differing levels of sophistication.

Principal Investigator Bio:

David Johnson Assistant Professor of Industrial Engineering and Political Science


David R. Johnson
Assistant Professor of Industrial Engineering and Political Science
Ed Delp
Distinguished Professor of Electrical and Computer Engineering and Professor of Biomedical Engineering
Mohammad Jahanshahi
Assistant Professor of Civil Engineering
Ningning Nicole Kong
Assistant Professor of Library Science, GIS Specialist
Torsten Reimer
Associate Professor of Communication

David Johnson holds a Ph.D. in Policy Analysis from the Pardee RAND Graduate School, with concentrations in Quantitative Methods and Economics. He previously earned a B.S. in Mathematics from North Carolina State University and a MASt in Mathematics from the University of Cambridge, where he was a Gates Cambridge Scholar. Dr. Johnson came to Purdue from the RAND Corporation, where he worked as an Associate Mathematician.

His interdisciplinary research focuses broadly on decision-making under deep uncertainty. He has presented and published on a variety of climate change issues including coastal flood risk management, renewable energy policy, assessment of infrastructure risks, and water scarcity and quality management. He is also working to identify robust policies for nonstructural mitigation measures like elevating or flood-proofing homes.