Scenarios for future wildfire risk in California:
links between changing demography, land use,
climate, and wildﬁre
Benjamin P. Bryant
and Anthony L. Westerling
Over 21,000 future California residential wildﬁre risk scenarios were developed on a monthly 1/8° grid, using statistical wildﬁre
models. We explore interactions between two global emissions scenarios, three climate models, six spatially explicit population
growth scenarios derived from two growth models, and a range of parameters deﬁning properties’ vulnerability to loss.
Scenarios are evaluated over two future time periods relative to historic baselines. We also explore effects of spatial resolutions
for calculating household exposure to wildﬁre on changes in estimated future property losses. Our goal was not to produce one
authoritative set of future risk scenarios but rather to understand what parameters are important for robustly characterizing
effects of climate and growth on future residential property risks. By end of century, variation across development scenarios
accounts for far more variability in statewide residential wildﬁre risks than does variation across climate scenarios. However,
the most extreme increases in residential ﬁre risks result from combining high-growth/high-sprawl scenarios with the most
extreme climates considered here. Case studies for the Bay Area and the Sierra foothills demonstrate that, while land use
decisions profoundly inﬂuence future residential wildﬁre risks, effects of diverse growth and land use strategies vary greatly
around the state. Copyright © 2014 John Wiley & Sons, Ltd.
Additional supporting information may be found in the online version of this article at the publisher’s web site.
Keywords: ﬁre; wildﬁre; risk; climate; scenario; WUI; wildland–urban interface; spatial
1.1. Climate change and residential wildﬁre risk
Wildﬁres in California routinely threaten people and property, destroy homes, force evacuations, expose large populations to unhealthful air,
and result in deaths and injuries. Climate change may affect the size and frequency of wildﬁres in California, and its impacts are likely to vary
substantially across the state (Westerling et al. 2011a; Bowman et al. 2009; Krawchuk et al. 2009; Westerling and Bryant 2008; Westerling
et al. 2006; and Lenihan et al. 2003). And while wildﬁre poses many hazards, its most direct impacts on humans are fundamentally
connected to how people are distributed over the landscape. In previous work (Bryant and Westerling 2009), we considered how changes
in the probability of large ﬁre events interact with changes in land use to affect residential property risks, drawing on a small number of
scenarios for future climate, land use, and growth. In this paper, we expand the number of climate, land use, and growth scenarios considered
and also consider additional uncertainties and a more sophisticated model of expected housing loss due to wildﬁre, to more robustly
characterize future changes in wildﬁre and wildﬁre-related residential property risk in California. A complementary study (Hurteau et al.
2014) applies our results to assess changes in wildﬁre emissions of greenhouse gases and air pollutants.
This paper’s primary aim is to describe how climate change and human development patterns over California may interact to lead to differing
levels of ﬁre-caused risk to residential property, with a greater focus on the relative impacts of different climate, population growth, and land use
scenarios, as well as parameters related to ﬁre management. This study used climate scenarios derived from three global climate models (GCMs)
from the Intergovernmental Panel on Climate Change (IPCC)’s Fourth Assessment forced with medium-high and low emissions pathways
(Special Report on Emissions Scenarios (SRES) A2 and B1, respectively) (IPCC 2000, 2007). Our growth scenarios are derived from two
different sets of spatially explicit raster data sets, each describing different 21st century population growth and land use scenarios. One set is
based on work by Theobald (2005) and developed by the US Environmental Protection Agency (US EPA 2008) as the Integrated Climate and
Land Use Scenarios (ICLUS) for the USA and is provided at 100 m resolution. The other set is provided at 50 m resolution and generated using
the UPlan growth model, developed for California by Thorne et al. (2012). As in Bryant and Westerling (2009), the primary results of this study
are in the form of statistics on aggregate statewide relative risk, where the reference period is deﬁned based on year 2000 development patterns and
late 20th century (1961–1990) simulated climate. This paper also presents spatial distributions of changes in wildﬁre probabilities and expected
losses to illustrate how these impacts can vary throughout the state.
* Correspondence to: A. L. Westerling, Sierra Nevada Research Institute, University of California, Merced, CA 95343, U.S.A. E-mail:email@example.com
a Pardee RAND Graduate School, the RAND Corporation, Santa Monica, U.S.A.
b Sierra Nevada Research Institute, University of California, Merced, U.S.A.
Environmetrics 2014; 25: 454–471 Copyright © 2014 John Wiley & Sons, Ltd.
Special Issue Paper Environmetrics
Received: 10 March 2014, Revised: 9 April 2014, Accepted: 10 April 2014, Published online in Wiley Online Library: 24 July 2014
(wileyonlinelibrary.com) DOI: 10.1002/env.2280