From Measurement to Mitigation: Wildfire Hazard, Social Vulnerability, and Federal Funding in the Wildland-Urban Interface

Abstract

This dissertation focuses on the methods used to quantify wildfire risk to communities and on the effects of the operationalization of risk on the distribution of federal funding for community-level mitigation efforts. The first chapter evaluates methods used to map the extent of the wildland-urban interface (WUI), a designation that is widely used to identify communities exposed to wildfire potential. The second and third chapters center on the federal Community Wildfire Defense Grant (CWDG) program, the criteria used by the program to determine eligibility and evaluate applications, and the program's distributional outcomes. Together, these chapters address the three determinants of wildfire risk — hazard, exposure, and vulnerability — and their use in the targeting of federal funding for risk mitigation. The first chapter compares three existing methods for mapping the WUI — zonal, point-based, and raster — and develops an additional cadastral method that identifies human habitation based on a longitudinal dataset combining remotely-sensed building footprints with property records. I tabulate the number of people and structures within the areas classified as WUI by these mapping methods, stratified by a measure of wildfire hazard, to estimate regional and temporal patterns in exposure to high wildfire hazard. While the methods produce a wide range of estimates, the broader trends identified are similar. The largest number of people and structures within the WUI are located in the Southern and Eastern regions, but exposure to high wildfire hazard is greatest in Southern California and Northern California. Additionally, I find that WUI growth over the period from 2000–2020 was most extensive in the Great Basin and Southwest regions. The differences between the mapping methods are significant — the raster method classifies an area 2.4 times larger than the cadastral method as WUI — and vary spatially, suggesting that local factors such as data quality should be considered when choosing a mapping approach for a given application. The second chapter evaluates two aspects of the CWDG evaluation criteria: the spatial units commonly used to aggregate hazard and the low-income threshold used to prioritize communities. I construct an alternate set of aggregation units — place catchments — that may better characterize wildfire hazard within communities that are not designated as Census places. The program's low-income criterion is contrasted against the CDC's Social Vulnerability Index (SVI) to assess the degree to which the income-based threshold proxies for different aspects of community vulnerability. I find that the place catchment approach produces a different spatial distribution of hazard classes in many regions and shows better agreement with fine-scale hazard classes (produced across a 30-meter grid) than the commonly-used method of backfilling gaps between Census places with county-level hazard metrics. The low-income criterion prioritizes many at-risk communities that are not considered socially vulnerable: over 60% of the least vulnerable communities (lowest SVI decile) are considered low-income. Additionally, a significant number of vulnerable communities exposed to wildfire hazard — particularly those whose vulnerability derives from housing characteristics, access to transportation, race, and language — are not prioritized by the low-income criterion. The third chapter delves into the quantitative scoring rubric used to evaluate CWDG applications and the distribution of program funds based on hazard and vulnerability. To enable this analysis, I construct a dataset that matches grant applications and awards to community characteristics. I find that grant awards are largely determined by application scores, with a small amount of undocumented selection occurring among applications with marginal scores. While several of the scoring components are highly influential on grant awards, the scores assigned to the qualitative sections of grant applications are highly correlated and do not appear to capture distinct application characteristics. Communities facing the highest levels of wildfire hazard make up the majority of grant applications and awards, and higher-hazard communities that apply are more likely to be funded. However, higher levels of social vulnerability are not associated with increased odds of grant approval, and the distribution of vulnerability within the applicant and award pools is similar to the overall vulnerability distribution among at-risk communities. These findings indicate that, while the program succeeds in prioritizing grant funding in accordance with its statutory mandate, the program's evaluation criteria do not produce distributional outcomes that favor socially-vulnerable communities.