Visual Analytics for Improved Allocation of Hazard Mitigation Grants

The Challenge

Disasters are becoming more common each year. According to NPR reporting, fewer than 20% of US counties experienced a disaster in the early and mid 1900s compared to more than 50% in 2019¹. In response, the Federal Emergency Management Agency (FEMA) has initiated several types of hazard mitigation grants to reduce the risk of damage to people and property from future disasters. These grants help fund activities including building protective headwalls, watertight enclosures, and elevated structures to prevent erosion and mitigate flood damage; acquiring property in disaster-vulnerable communities and converting it to open spaces; and constructing tornado and storm shelters. 

FEMA spends about $1.6B annually on hazard mitigation grants for state, local, tribal, and territorial (SLTT)-based projects. An investigation conducted by NPR found wide disparities in who received federal dollars and the level of effort in how the money was requested. “Federal aid isn’t necessarily allocated to those who need it most; it’s allocated according to cost-benefit calculations meant to minimize taxpayer risk,” which often hurts minorities and worsens wealth inequality¹. 

The Solution

Our analysis can support better prioritization and allocation of FEMA’s hazard mitigation grant funds by identifying vulnerable communities that are also most in need. We analyzed, at the county-level, FEMA and non-FEMA data related to patterns of disaster risk, economic vulnerability, resiliency, and current distribution of hazard mitigation funds. Our analysis found that distribution of these funds remains inefficient, uneven, reactive, and inadequate to the amount needed to help communities deal with emerging and accelerated natural hazards induced by climate change, especially flooding.

We started our analysis broadly, focusing on states with the highest risks for natural disaster damage within FEMA’s National Risk Index dataset². Using this dataset, we were able to determine the states with the highest expected annual losses due to several natural disasters. For example, the states with the highest expected annual losses for hurricanes included Virginia, Texas, North Carolina, Mississippi, Louisiana, Florida, and Alabama with Virginia having an annual expected loss of $1.7B.  

To better identify communities in need, we ingested the Baseline Resilience Indicators for Communities (BRIC) dataset from the Hazards & Vulnerability Research Institute at the University of South Carolina³. Within the states identified using the National Risk Index, we identified the bottom 10 counties with the lowest infrastructural scores. The infrastructural score from BRIC was selected as the best measure for approximating disaster preparedness. For example, within Alabama, the 10 communities most at risk for hurricanes from an infrastructural standpoint were Lowndes, Greene, Bullock, Cleburne, Choctaw, Washington, Bibb, Chilton, Wilcox, and Winston, with infrastructural resiliency scores ranging from 0.12 to 0.15. 

RiskAI

Our specific research also supports the much broader project of a RiskAI model. FEMA provides access to large, authoritative data sources (National Flood Insurance Program flood hazard data, Repetitive Loss Program data, National Risk Index, etc.) which can help educate the nation and drive data services for US citizens to make risk-informed decisions. In addition, FEMA has access to emerging technologies around data science, including machine learning, that allow the process of ingesting these sources of data and the process of uncovering solutions easier. To help marry these two components, Ardent has developed the RiskAI model that takes advantage of authoritative data sources as well as commercial and citizen-based data to help build a localized risk model for use by homeowners, business owners, local planners, and any entity trying to understand their local risk posture.

Read our RiskAI Whitepaper below or download here.

References

¹ NPR: How Federal Disaster Money Favors the Rich
² FEMA’s National Risk Index dataset identifies communities most at risk to 18 natural hazards and includes risk metrics such as expected annual losses.
³ The BRIC dataset contains six broad categories of community disaster resilience at the county level: social, economic, community capital, institutional, infrastructural, and environmental. The resilience scores range from 0 to 1 with 0 representing a vulnerable community and 1 representing a resilient community.

This case study was created by Ardent’s Data Science and Analytics (DSA) Practice. The team is led by Tino Dinh, Principal. This summary was written by Jonathan Zimmerman, Data Analyst. The accompanying dashboard was created by Katherine Kelso, Data Analyst, Todd Glasco, Data Analyst, and Jonathan Zimmerman, Data Analyst.