How Will Increased Population Affect Wildfire Incidence: is ignition frequency in the Sierra Nevada related to population density?

Executive Summary

Uncovering relationships between wildfire ignitions and explanatory variables such as housing density or population density could lead to powerful new planning and analysis tools for projecting future fire protection workloads. In this study, linear regression of ignition frequency on population and housing densities at the census block level shows a statistically significant positive effect of human presence on ignition frequency in the Sierra Nevada. The regression equations explain about 30% of the variation in ignition frequency, but continued investigations should result in a refined model with increased explanatory and predictive power.

The following hypothetical example (illustrated on the right) shows one possible application of ignition prediction tools in local land use analysis: Let us say that a county land use planner is concerned about the potential impact that a proposed development will have on the existing fire suppression infrastructure. Assume the development is to take place within a 1,000 acre, sparsely populated area with a current average parcel size of 50 acres (12.8 houses per square mile). Further, assume the new subdivision subdivides the land into 1 acre parcels (average parcel size = 1 acre or 640 houses per square mile). The planner applies the housing density regression equation and predicts that annual fire frequency will change from .57 fires per year to 1.65 fires per year, for an additional 1.08 fires per year (a 189% increase). If an average of 2 persons were to occupy each parcel, the planner might use the population regression equation, which predicts a 381% increase in fire starts, or about 1.8 more wildfires annually. The planner can use this information to consult with fire officials about the implications for service or the potential for mitigation and prevention measures.

Fire risk is one of the key factors that impact fire protection costs and losses. FRAP will expand this analysis to develop methods to project how population growth and development patterns will affect future ignition patterns.

Analysis

This study started with ignition data spanning the years 1921 to 1993 (Jones, 1995): agency fire reports from State Responsibility Areas, National Forests, and National Parks located within the study region of Sierra Nevada Ecosystem Project (SNEP). These ignition data are represented spatially as the geographical centers (centroids) of public land survey system (PLSS) sections. For uniformity, we extracted a data set spanning the years 1970 to 1993 (96% of the database records) (Figure 1). We then intersected the ignition data from the SNEP study area with 1990 census block data obtained from the Teale Data Center.

Linear regression traced the relationship between ignition frequency and 843 different housing density classes. Figure 2 shows a histogram of the regression variables. The regression line explains 29% of the variation (adjusted R square=0.29) and the relationship is significant, with a t value for the slope coefficient of 3.97 (t0.05=1.96) Table 1. Here is the equation:

Ignition frequency (fires/year/1000 acres) = 0.545737 + 0.001733 * [Average Housing Density (housing units/square mile)].

Looking at ignition frequency and 1110 different population density classes the linear regression shows slightly more explained variation with an adjusted R square of 0.32 and a significant t value of 3.01 (t0.05=1.96). Table 2 contains the regression results and Figure 3 is a histogram of ignition frequency and population densities. The equation is:

Ignition frequency (fires/year/1000 acres) = 0.419240 + 0.001438 * [Average Population Density (persons/square mile)].

With a 95% confidence interval, an increase of one housing unit per square mile means an additional 0.001733 (+ 0.000185) ignitions per year per 1000 acres. An increase of one person per square mile means an additional 0.001438 (+ 0.000122) ignitions per year per 1000 acres. Inclusion of other potential causal factors in the next round of analysis could explain more of the variance.

Discussion

Further investigations will 1) recheck and correct errors in census block data, 2) obtain new ignition data that include fire cause, 3) use lower resolution census tract data instead of block data, 4) include census blocks or tracts that have no recorded ignitions to avoid potential bias, 5) use the regression equations with California population projections to project future ignition patterns, and 6) incorporate other possible factors such as transportation,development patterns, recreational activities, and biological and biophysical features.

References

Jones, R. 1995. Personal communication. SNEP Program, Centers for Water and Wildland Resources, University of California, Davis.

FOR MORE INFORMATION

Contact Jim Spero via e-mail at james.spero@fire.ca.gov or by phone at (916) 227-2686.

Last edited on June 23, 1997 by Greg Greenwood