Rare and highly destructive wildfires drive human migration in the U.S

Abstract

The scale of wildfire impacts to the built environment is growing and will likely continue under rising average global temperatures. We investigate whether and at what destruction threshold wildfires have influenced human mobility patterns by examining the migration effects of the most destructive wildfires in the contiguous U.S. between 1999 and 2020. We find that only the most extreme wildfires (258+ structures destroyed) influenced migration patterns. In contrast, the majority of wildfires examined were less destructive and did not cause significant changes to out- or in-migration. These findings suggest that, for the past two decades, the influence of wildfire on population mobility was rare and operated primarily through destruction of the built environment.

Fig. 1. A relatively small proportion of wildfires cause widespread structure loss.

a Boxplots show the distribution of annual structure damage per destructive wildfire among all wildfires reported by the ICS dataset in the U.S. between 1999 and 2020 that destroyed 1 or more structures (N = 5406). The left whisker indicates the minimum value to the 25th percentile, the right whisker indicates the 75th percentile to the maximum value, the left side of the box indicates the 25th percentile, the right side of the box indicates the 75th percentile, and the line within the box indicates the median. Red dots indicate extreme events that destroyed more structures than maximum values. The dotted blue line indicates a global median of structure damage across all years (2 structures destroyed). While the majority of destructive wildfires affected a relatively small number of structures (90% impacted fewer than 14), a small number of events had an outsized contribution to the total number of structures destroyed. b Figure shows the probability distribution of structures destroyed per wildfire event among the top decile of most destructive wildfires that include spatial details the ICS dataset in the contiguous U.S. from 1999 to 2020 (N = 529). Within this top decile of wildfires (those that destroyed between 14 to 18,804 structures), the count of structures destroyed per event is highly right skewed. The figure shows how we stratified events for subsequent analysis into the less destructive portion of the decile distribution (green line), more destructive portion of the decile distribution (gold line), and the single most destructive event in the distribution, the Camp Fire (red point). We also analyzed the full decile of events (blue line). c Map shows the geographic distribution of wildfires with destruction levels and points of origin reported in the ICS dataset from 1999 to 2020 in the contiguous U.S. (N = 32,296). Each point on the map represents a wildfire point of origin, where the color indicates level of structure loss caused by the fire. Blue dots indicate fires that caused no structure loss; yellow dots indicate the majority of destructive wildfires (90%) that destroyed 1–13 structures; and red dots indicate the most destructive wildfires (top 10%), which destroyed between 14 and 18,804 structures. We focused our analysis on the latter group, analyzing only the most destructive wildfires. Sources: Wildfire data are from the U.S. National Incident Management System/Incident Command System and state boundaries are from the U.S. Census Bureau.

Fig. 2. Out-migration effects of wildfire structure loss are observed only following the most destructive events.

a Figures show evolving, unweighted out-migration probabilities (left) and in-migration probabilities (right) among three subsets of destructive wildfires: (1) full top decile distribution (14–18,804 structures destroyed, N = 519 wildfires), (2) less destructive portion of the top decile (14–257 structures destroyed, N = 463 wildfires), and (3) more destructive portion of the top decile (258–7010 structures destroyed, N = 55 wildfires). Control tract migration probabilities are shown in blue, purple, and green. Vertical dashed line indicates the quarter in which the wildfire occurred. b Figures show evolving, unweighted out-migration probabilities (left) and in-migration probabilities (right) before and after the 2018 Camp Fire. Control tract migration probabilities are shown in blue, purple, and green. Vertical dashed line indicates the quarter in which the wildfire occurred. c For each wildfire event, we selected three rings of control tracts for each cluster of burned tracts (shown in red). Figure shows control selection for the 2000 Cerro Grande Fire in New Mexico. The buffer from the outer edge of burned tracts to 5 miles away is shown in blue; the buffer from 5 to 25 miles is shown in purple; and the buffer between 25 and 50 miles away from the edge of treated tracts is shown in green (left). These buffers are then intersected with spatially overlapping tracts (right). Sources: Migration data are from the Federal Reserve Bank of New York/Equifax Consumer Credit Panel, wildfire data are from the U.S. National Incident Management System/Incident Command System, and tract boundaries are from the U.S. Census Bureau.