Impact of anthropogenic climate change on wildfire
across western US forests
John T. Abatzogloua,1 and A. Park Williamsb
a

Department of Geography, University of Idaho, Moscow, ID 83844; and bLamont–Doherty Earth Observatory, Columbia University, Palisades, NY 10964

Increased forest fire activity across the western continental United
States (US) in recent decades has likely been enabled by a number of
factors, including the legacy of fire suppression and human settlement, natural climate variability, and human-caused climate change.
We use modeled climate projections to estimate the contribution
of anthropogenic climate change to observed increases in eight fuel
aridity metrics and forest fire area across the western United States.
Anthropogenic increases in temperature and vapor pressure deficit
significantly enhanced fuel aridity across western US forests over the
past several decades and, during 2000–2015, contributed to 75%
more forested area experiencing high (>1 σ) fire-season fuel aridity
and an average of nine additional days per year of high fire potential.
Anthropogenic climate change accounted for ∼55% of observed increases in fuel aridity from 1979 to 2015 across western US forests,
highlighting both anthropogenic climate change and natural climate
variability as important contributors to increased wildfire potential in
recent decades. We estimate that human-caused climate change contributed to an additional 4.2 million ha of forest fire area during 1984–
2015, nearly doubling the forest fire area expected in its absence.
Natural climate variability will continue to alternate between modulating and compounding anthropogenic increases in fuel aridity, but
anthropogenic climate change has emerged as a driver of increased
forest fire activity and should continue to do so while fuels are
not limiting.
wildfire

  climate change   attribution   forests

W

idespread increases in fire activity, including area burned
(1, 2), number of large fires (3), and fire-season length
(4, 5), have been documented across the western United States
(US) and in other temperate and high-latitude ecosystems over
the past half century (6, 7). Increased fire activity across western
US forests has coincided with climatic conditions more conducive to wildfire (2–4, 8). The strong interannual correlation
between forest fire activity and fire-season fuel aridity, as well as
observed increases in vapor pressure deficit (VPD) (9), fire danger
indices (10), and climatic water deficit (CWD) (11) over the past
several decades, present a compelling argument that climate
change has contributed to the recent increases in fire activity. Previous studies have implicated anthropogenic climate change (ACC)
as a contributor to observed and projected increases in fire activity
globally and in the western United States (12–19), yet no studies
have quantified the degree to which ACC has contributed to observed increases in fire activity in western US forests.
Changes in fire activity due to climate, and ACC therein, are
modulated by the co-occurrence of changes in land management
and human activity that influence fuels, ignition, and suppression.
The legacy of twentieth century fire suppression across western
continental US forests contributed to increased fuel loads and fire
potential in many locations (20, 21), potentially increasing the
sensitivity of area burned to climate variability and change in recent decades (22). Climate influences wildfire potential primarily
by modulating fuel abundance in fuel-limited environments, and
by modulating fuel aridity in flammability-limited environments
(1, 23, 24). We constrain our attention to climate processes that
promote fuel aridity that encompass fire behavior characteristics of landscape ignitability, flammability, and fire spread via fuel

www.pnas.org/cgi/doi/10.1073/pnas.1607171113

desiccation in primarily flammability-limited western US forests by
considering eight fuel aridity metrics that have well-established
direct interannual relationships with burned area in this region
(1, 8, 24, 25). Four metrics were calculated from monthly data for
1948–2015: (i) reference potential evapotranspiration (ETo),
(ii) VPD, (iii) CWD, and (iv) Palmer drought severity index
(PDSI). The other four metrics are daily fire danger indices calculated for 1979–2015: (v) fire weather index (FWI) from the
Canadian forest fire danger rating system, (vi) energy release
component (ERC) from the US national fire danger rating system,
(vii) McArthur forest fire danger index (FFDI), and (viii) Keetch–
Byram drought index (KBDI). These metrics are further described
in the Materials and Methods and Supporting Information. Fuel
aridity has been a dominant driver of regional and subregional
interannual variability in forest fire area across the western US in
recent decades (2, 8, 22, 25). This study capitalizes on these relationships and specifically seeks to determine the portions of the
observed increase in fuel aridity and area burned across western
US forests attributable to anthropogenic climate change.
The interannual variability of all eight fuel aridity metrics averaged over the forested lands of the western US correlated significantly (R2 = 0.57–0.76, P < 0.0001; Table S1) with the logarithm of
annual western US forest area burned for 1984–2015, derived from
the Monitoring Trends in Burn Severity product for 1984–2014 and
the Moderate Resolution Imaging Spectroradiometer (MODIS)
for 2015 (Supporting Information). The record of standardized fuel
aridity averaged across the eight metrics (hereafter, all-metric
mean) accounts for 76% of the variance in the burned-area record,
with significant increases in both records for 1984–2015 (Fig. 1).
Correlation between fuel aridity and forest fire area remains
highly significant (R2 = 0.72, all-metric mean) after removing the
linear-least squares trends for each time series for 1984–2015,
supporting the mechanistic relationship between fuel aridity and
Significance
Increased forest fire activity across the western United States
in recent decades has contributed to widespread forest mortality, carbon emissions, periods of degraded air quality, and
substantial fire suppression expenditures. Although numerous
factors aided the recent rise in fire activity, observed warming
and drying have significantly increased fire-season fuel aridity,
fostering a more favorable fire environment across forested
systems. We demonstrate that human-caused climate change
caused over half of the documented increases in fuel aridity
since the 1970s and doubled the cumulative forest fire area
since 1984. This analysis suggests that anthropogenic climate
change will continue to chronically enhance the potential for
western US forest fire activity while fuels are not limiting.
Author contributions: J.T.A. and A.P.W. designed research, performed research, contributed new reagents/analytic tools, analyzed data, and wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
1

To whom correspondence should be addressed. Email: jabatzoglou@uidaho.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1607171113/-/DCSupplemental.

PNAS Early Edition   1 of 6

EARTH, ATMOSPHERIC,
AND PLANETARY SCIENCES

Edited by Monica G. Turner, University of Wisconsin–Madison, Madison, WI, and approved July 28, 2016 (received for review May 5, 2016)

 1984−1999
2000−2015

Forest Fire Area (kha)

1000

R2 = 0.76
P<0.0001
300

100

30
−1.5

−1

−0.5 0
0.5
1
Fuel aridity (σ)

1.5

2

Fig. 1. Annual western continental US forest fire area versus fuel aridity:
1984–2015. Regression of burned area on the mean of eight fuel aridity
metrics. Gray bars bound interquartile values among the metrics. Dashed
lines bounding the regression line represent 95% confidence bounds, expanded to account for lag-1 temporal autocorrelation and to bound the
confidence range for the lowest correlating aridity metric. The two 16-y periods
are distinguished to highlight their 3.3-fold difference in total forest fire area.
Inset shows the distribution of forested land across the western US in green.

forest fire area. It follows that co-occurring increases in fuel aridity
and forest fire area over multiple decades would also be
mechanistically related.
We quantify the influence of ACC using the Coupled Model
Intercomparison Project, Phase 5 (CMIP5) multimodel mean
changes in temperature and vapor pressure following Williams
et al. (26) (Fig. S1; Methods). This approach defines the ACC
signal for any given location as the multimodel mean (27 CMIP5
models) 50-y low-pass-filtered record of monthly temperature
and vapor pressure anomalies relative to a 1901 baseline. Other
anthropogenic effects on variables such as precipitation, wind, or
solar radiation may have also contributed to changes in fuel
aridity but anthropogenic contributions to these variables during
our study period are less certain (22). We evaluate differences
between fuel aridity metrics computed with the observational
record and those computed with observations that exclude the
ACC signal to determine the contribution of ACC to fuel aridity.
To exclude the ACC signal, we subtract the ACC signal from daily
and monthly temperature and vapor pressure, leaving all other
variables unchanged and preserving the temporal variability of
observations. The contribution of ACC to changes in fuel aridity is
shown for the entire western United States; however, we constrain
the focus of our attribution and analysis to forested environments
of the western US (Fig. 1, Inset; Methods).
Anthropogenic increases in temperature and VPD contributed
to a standardized (σ) increase in all-metric mean fuel aridity averaged for forested regions of +0.6 σ (range of +0.3 σ to +1.1 σ
across all eight metrics) for 2000–2015 (Fig. 2). We found similar
results with reanalysis products (all-metric mean fuel aridity increase of +0.6 σ for two reanalysis datasets considered; Methods),
suggesting robustness of the results to structural uncertainty in
observational products (Figs. S2–S4 and Table S2). The largest
anthropogenic increases in standardized fuel aridity were present
across the intermountain western United States, due in part to
2 of 6   www.pnas.org/cgi/doi/10.1073/pnas.1607171113

larger modeled warming rates relative to more maritime areas (27).
Among aridity metrics, the largest increases tied to the ACC signal
were for VPD and ETo because the interannual variability of these
variables is primarily driven by temperature for much of the study
area (28). By contrast, PDSI and ERC showed more subdued ACC
driven increases in fuel aridity because these metrics are more
heavily influenced by precipitation variability.
Fuel aridity averaged across western US forested areas showed a
significant increase over the past three decades, with a linear trend
of +1.2 σ (95% confidence: 0.42–2.0 σ) in the all-metric mean for
1979–2015 (Fig. 3A, Top and Table S1). The all-metric mean ACC
contribution since 1901 was +0.10 σ by 1979 and +0.71 σ by 2015.
The annual area of forested lands with high fuel aridity (>1 σ)
increased significantly during 1948–2015, most notably since 1979
(Fig. 3A, Bottom). The observed mean annual areal extent of forested land with high aridity during 2000–2015 was 75% larger for
the all-metric mean (+27% to +143% range across metrics) than
was the case where the ACC signal was excluded.
Significant positive trends in fuel aridity for 1979–2015 across
forested lands were observed for all metrics (Fig. 3B and Table
S1). Positive trends in fuel aridity remain after excluding the
ACC signal, but the remaining trend was only significant for
ERC. Anthropogenic forcing accounted for 55% of the observed
positive trend in the all-metric mean fuel aridity during 1979–
2015, including at least two-thirds of the observed increase in
ETo, VPD, and FWI, and less than a third of the observed increase in ERC and PDSI. No significant trends were observed
for monthly fuel aridity metrics from 1948–1978.
The duration of the fire-weather season increased significantly
across western US forests (+41%, 26 d for the all-metric mean)
during 1979–2015, similar to prior results (10) (Fig. 4A and Table
S2). Our analysis shows that ACC accounts for ∼54% of the increase in fire-weather season length in the all-metric mean (15–
79% for individual metrics). An increase of 17.0 d per year of high
fire potential was observed for 1979–2015 in the all-metric mean
(11.7–28.4 d increase for individual metrics), over twice the rate of
increase calculated from metrics that excluded the ACC signal
(Fig. 4B and Table S2). This translates to an average of an additional 9 d (7.8–12.0 d) per year of high fire potential during 2000–
2015 due to ACC.
Given the strong relationship between fuel aridity and annual
western US forest fire area, and the detectable impact of ACC on
fuel aridity, we use the regression relationship in Fig. 1 to model

Fig. 2. Standardized change in each of the eight fuel aridity metrics due
to ACC. The influence of ACC on fuel aridity during 2000–2015 is shown
by the difference between standardized fuel aridity metrics calculated
from observations and those calculated from observations excluding the
ACC signal. The sign of PDSI is reversed for consistency with other aridity
measures.

Abatzoglou and Williams

 −1

% area (+1 σ)

−2

Trend (σ/37 yr)

B

80

Observations
No ACC
Burned area

750

500

60
40

250

20
1950 1960 1970 1980 1990 2000 2010
year
Observations
No ACC
**
1.5
**
**
**
**
1 *
*
**
**
*
0.5
0

SI FWI RC FDI ETo WD BDI PD ean
V m
E F
K
C

PD

Fig. 3. Evolution and trends in western US forest fuel aridity metrics over
the past several decades. (A) Time series of (Upper) standardized annual fuel
aridity metrics and (Lower) percent of forest area with standardized fuel
aridity exceeding one SD. Red lines show observations and black lines show
records after exclusion of the ACC signal. Only the four monthly metrics
extend back to 1948. Daily fire danger indices begin in 1979. Bold lines indicate averages across fuel aridity metrics. Bars in the background of A show
annual forested area burned during 1984–2015 for visual comparison with
fuel aridity. (B) Linear trends in the standardized fuel aridity metrics during
1979–2015 for (red) observations and (black) records excluding the ACC
signal (differences attributed to ACC). Asterisks indicate positive trends at
the (*) 95% and (**) 99% significance levels.

the contribution of ACC on western US forest fire area for the
past three decades (Fig. 5 and Fig. S5). ACC-driven increases in
fuel aridity are estimated to have added ∼4.2 million ha (95%
confidence: 2.7–6.5 million ha) of western US forest fire area
during 1984–2015, similar to the combined areas of Massachusetts
and Connecticut, accounting for nearly half of the total modeled
burned area derived from the all-metric mean fuel aridity. Repeating this calculation for individual fuel aridity metrics yields
ACC contributions of 1.9–4.9 million ha, but most individual
fuel aridity metrics had weaker correlations with burned area
and thus may be less appropriate proxies for attributing burned
area. The effect of the ACC forcing on fuel aridity increased
during this period, contributing ∼5.0 (95% confidence: 4.2–5.9)
times more burned area in 2000–2015 than in 1984–1999 (Fig. 5B).
During 2000–2015, the ACC-forced burned area likely exceeded
the burned area expected in the absence of ACC (Fig. 5B).
A more conservative method that uses the relationship between
detrended records of burned area and fuel aridity (2) still indicates a
substantial impact of ACC on total burned area, with a 19% (95%
Abatzoglou and Williams

A
200

Observations
No ACC

150

100

50

B
50
40
30
20
10
0
1980

1990

2000

2010

year
Fig. 4. Changes in fire-weather season length and number of high fire
danger days. Time series of mean western US forest (A) fire-weather season
length and (B) number of days per year when daily fire danger indices
exceeded the 95th percentile. Baseline period: 1981–2010 using observational records that exclude the ACC signal. Red lines show the observed
record, and black lines show the record that excludes the ACC signal. Bold
lines show the average signal expressed across fuel aridity metrics.

PNAS Early Edition   3 of 6

EARTH, ATMOSPHERIC,
AND PLANETARY SCIENCES

1000

0

fire weather season
length (% of mean)

1

confidence: 12–24%) reduction in the proportion of total burned
area attributable to ACC (Fig. S5).
Our attribution explicitly assumes that anthropogenic increases
in fuel aridity are additive to the wildfire extent that would have
arisen from natural climate variability during 1984–2015. Because
the influence of fuel aridity on burned area is exponential, the
influence of a given ACC forcing is larger in an already arid fire
season such as 2012 (Fig. 5A and Fig. S5C). Anthropogenic increases in fuel aridity are expected to continue to have their most
prominent impacts when superimposed on naturally occurring
extreme climate anomalies. Although numerous studies have
projected changes in burned area over the twenty-first century due
to ACC, we are unaware of other studies that have attempted to
quantify the contribution of ACC to recent forested burned area
over the western United States. The near doubling of forested
burned area we attribute to ACC exceeds changes in burned area
projected by some modeling efforts to occur by the mid-twentyfirst century (29, 30), but is proportionally consistent with midtwenty-first century increases in burned area projected by other
modeling efforts (17, 31–33).
Beyond anthropogenic climatic changes, several additional
factors have caused increases in fuel aridity and forest fire area
since the 1970s. The lack of fuel aridity trends during 1948–1978
and persistence of positive trends during 1979–2015 even after
removing the ACC signal implicates natural multidecadal climate
variability as an important factor that buffered anthropogenic

days with
high fire danger

2

Forest fire area (ha x 1000)

Fuel aridity (σ)

A

 Fig. 5. Attribution of western US forest fire area to ACC. Cumulative forest
fire area estimated from the (red) observed all-metric mean record of fuel
aridity and (black) the fuel aridity record after exclusion of ACC (No ACC).
The (orange) difference is the forest fire area forced by anthropogenic increases in fuel aridity. Bold lines in A and horizontal lines within box plots
in B indicate mean estimated values (regression values in Fig. 1). Boxes in B
bound 50% confidence intervals. Shaded areas in A and whiskers in B bound
95% confidence intervals. Dark red horizontal lines in B indicate observed
forest fire area during each period.

effects during 1948–1978 and compounded anthropogenic effects
during 1979–2015. During 1979–2015, for example, observed
Mar–Sep vapor pressure decreased significantly across many US
forest areas, in marked contrast to modeled anthropogenic increases (Fig. S6) (34). Significant declines in spring (Mar–May)
precipitation in the southwestern United States and summer
(Jun–Sep) precipitation throughout parts of the northwestern
United States during 1979–2015 (Fig. S7 A and B) hastened increases in fire-season fuel aridity, consistent with observed increases in the number of consecutive dry days across the region
(10). Natural climate variability, including a shift toward the cold
phase of the interdecadal Pacific Oscillation (35), was likely the
dominant driver of observed regional precipitation trends (36)
(Fig. S7 B and D).
Our quantification of the ACC contribution to observed increases in forest fire activity in the western United States adds to
the limited number of climate change attribution studies on
wildfire to date (37). Previous attribution efforts have been restricted to a single GCM and biophysical variable (14, 16). We
complement these studies by demonstrating the influence of
ACC derived from an ensemble of GCMs on several biophysical
metrics that exhibit strong links to forest fire area. However, our
attribution effort only considers ACC to manifest as trends in
4 of 6   www.pnas.org/cgi/doi/10.1073/pnas.1607171113

mean climate conditions, which may be conservative because climate models also project anthropogenic increases in the temporal
variability of climate and drought in the western United States (34,
38, 39). In focusing exclusively on the direct impacts of ACC on
fuel aridity, we do not address several other pathways by which
ACC may have affected wildfire activity. For example, the fuel
aridity metrics that we used may not adequately capture the role of
mountain snow hydrology on soil moisture. Nor do we account for
the influence of climate change on lightning activity, which may
increase with warming (40). We also do not account for how fire
risk may be affected by changes in biomass/fuel due to increases in
atmospheric CO2 (41), drought-induced vegetation mortality (42),
or insect outbreaks (43).
Additionally, we treat the impact of ACC on fire as independent from the effects of fire management (e.g., suppression
and wildland fire use policies), ignitions, land cover (e.g., exurban development), and vegetation changes beyond the degree to
which they modulate the relationship between fuel aridity and
forest fire area. These factors have likely added to the area
burned across the western US forests and potentially amplified
the sensitivity of wildfire activity to climate variability and change
in recent decades (2, 22, 24, 44). Such confounding influences,
along with nonlinear relationships between burned area and its
drivers (e.g., Fig. 1), contribute uncertainty to our empirical attribution of regional burned area to ACC. Our approach depends on
the strong observed regional relationship between burned area and
fuel aridity at the large regional scale of the western United States,
so the quantitative results of this attribution effort are not necessarily applicable at finer spatial scales, for individual fires, or to
changes in nonforested areas. Dynamical vegetation models with
embedded fire models show emerging promise as tools to diagnose
the impacts of a richer set of processes than those considered here
(41, 45) and could be used in tandem with empirical approaches
(46, 47) to better understand contributions of observed and projected ACC to changes in regional fire activity. However, dynamic
models of vegetation, human activities, and fire are not without
their own lengthy list of caveats (2). Given the strong empirical
relationship between fuel aridity and wildfire activity identified
here and in other studies (1, 2, 4, 8), and substantial increases in
western US fuel aridity and fire-weather season length in recent
decades, it appears clear from empirical data alone that increased
fuel aridity, which is a robustly modeled result of ACC, is the
proximal driver of the observed increases in western US forest fire
area over the past few decades.
Conclusions
Since the 1970s, human-caused increases in temperature and
vapor pressure deficit have enhanced fuel aridity across western
continental US forests, accounting for approximately over half of
the observed increases in fuel aridity during this period. These
anthropogenic increases in fuel aridity approximately doubled
the western US forest fire area beyond that expected from natural climate variability alone during 1984–2015. The growing
ACC influence on fuel aridity is projected to increasingly promote wildfire potential across western US forests in the coming
decades and pose threats to ecosystems, the carbon budget,
human health, and fire suppression budgets (13, 48) that will
collectively encourage the development of fire-resilient landscapes (49). Although fuel limitations are likely to eventually
arise due to increased fire activity (17), this process has not yet
substantially disrupted the relationship between western US
forest fire area and aridity. We expect anthropogenic climate
change and associated increases in fuel aridity to impose an increasingly dominant and detectable effect on western US forest
fire area in the coming decades while fuels remain abundant.
Abatzoglou and Williams

 We focus on climate variables that directly affect fuel moisture over forested
areas of the western continental United States, where fire activity tends to be
flammability-limited rather than fuel- or ignition-limited (1) (study region
shown in Fig. 1, Inset). There are a variety of climate-based metrics that have
been used as proxies for fuel aridity, yet there is no universally preferred
metric across different vegetation types (24). We consider eight frequently
used fuel aridity metrics that correlate well with fire activity variables, including annual burned area (Fig. 1 and Table S1), in western US forests.
Fuel aridity metrics are calculated from daily surface meteorological data
(50) on a 1/24° grid for 1979–2015 for the western United States (west of
103°W). Although we calculated metrics across the entire western United
States, we focus on forested lands defined by the climax succession vegetation stages of “forest” or “woodland” in the Environmental Site Potential
product of LANDFIRE (landfire.gov). Forested 1/24° grid cells are defined by
at least 50% forest coverage aggregated from LANDFIRE. We extended the
aridity metrics calculated at the monthly timescale (ETo, VPD, CWD, and
PDSI) back to 1948 using monthly anomalies relative to a common 1981–
2010 period from the dataset developed by the Parameterized Regression
on Independent Slopes Model group (51) for temperature, precipitation,
and vapor pressure, and by bilinearly interpolating NCEP–NCAR reanalysis
for wind speed and surface solar radiation. We aggregated data to annualized time series of mean May–Sep daily FWI, KBDI, ERC, and FFDI; Mar–Sep
VPD and ETo; Jun–Aug PDSI; and Jan–Dec CWD. We also calculated the
aridity metrics strictly from ERA-INTERIM and NCEP–NCAR reanalysis products for 1979–2015 covering the satellite era (Supporting Information).
Days per year of high fire potential are quantified by daily fire danger indices
(ERC, FWI, FFDI, and KBDI) that exceed the 95th percentile threshold defined
during 1981–2010 from observations after removing the ACC signal. Observational studies have shown that fire growth preferentially occurs during high
fire danger periods (52, 53). We also calculate the fire weather season length
for the four daily fire danger indices following previous studies (10).
The ACC signal is obtained from ensemble members taken from 27 CMIP5
global climate models (GCMs) regridded to a common 1° resolution for 1850–
2005 using historical forcing experiments and for 2006–2099 using the
Representative Concentration Pathway (RCP) 8.5 emissions scenario (Table
S3 and Supporting Information). These GCMs were selected based on
availability of monthly outputs for maximum and minimum daily temperature (Tmax and Tmin, respectively), specific humidity (huss), and surface
pressure. Saturation vapor pressure (es), vapor pressure (e), and VPD were
calculated using standard methods (Supporting Information). A variety of
approaches exist to estimate the ACC signal (26). We define the anthropogenic signals in Tmax, Tmin, e, es, VPD, and relative humidity by a 50-y lowpass-filter time series (using a 10-point Butterworth filter) averaged across the
27 GCMs using the following methodology: For each GCM, variable, month,
and grid cell, we converted each annual time series to anomalies relative to a
1901–2000 baseline. We averaged annual anomalies across all realizations
(model runs) for each GCM and calculated a single 50-y low-pass-filter annual

1. Littell JS, McKenzie D, Peterson DL, Westerling AL (2009) Climate and wildfire area
burned in western U.S. ecoprovinces, 1916-2003. Ecol Appl 19(4):1003–1021.
2. Williams AP, Abatzoglou JT (2016) Recent advances and remaining uncertainties in
resolving past and future climate effects on global fire activity. Curr Clim Chang
Reports 2:1–14.
3. Dennison P, Brewer S, Arnold J, Moritz M (2014) Large wildfire trends in the western
United States, 1984–2011. Geophys Res Lett 41:2928–2933.
4. Westerling AL, Hidalgo HG, Cayan DR, Swetnam TW (2006) Warming and earlier
spring increase western U.S. forest wildfire activity. Science 313(5789):940–943.
5. Westerling AL (2016) Increasing western US forest wildfire activity: Sensitivity to
changes in the timing of spring. Philos Trans R Soc B Biol Sci 371(1696):20150178.
6. Kasischke ES, Turetsky MR (2006) Recent changes in the fire regime across the North
American boreal region - Spatial and temporal patterns of burning across Canada and
Alaska. Geophys Res Lett 33(9):L09703.
7. Kelly R, et al. (2013) Recent burning of boreal forests exceeds fire regime limits of the
past 10,000 years. Proc Natl Acad Sci USA 110(32):13055–13060.
8. Abatzoglou JT, Kolden CA (2013) Relationships between climate and macroscale area
burned in the western United States. Int J Wildland Fire 22(7):1003–1020.
9. Seager R, et al. (2015) Climatology, variability, and trends in the U.S. vapor pressure
deficit, an important fire-related meteorological quantity. J Appl Meteorol Climatol
54(6):1121–1141.
10. Jolly WM, et al. (2015) Climate-induced variations in global wildfire danger from 1979
to 2013. Nat Commun 6:7537.
11. Dobrowski SZ, et al. (2013) The climate velocity of the contiguous United States
during the 20th century. Glob Change Biol 19(1):241–251.
12. Flannigan MD, Krawchuk MA, de Groot WJ, Wotton BM, Gowman LM (2009) Implications of changing climate for global wildland fire. Int J Wildland Fire 18(5):483–507.

Abatzoglou and Williams

time series for each of the 12 mo for 1850–2099. We averaged each month’s
low-pass-filtered time series across the 27 GCMs and additively adjusted so that
all smoothed records pass through zero in 1901. The resultant ACC signal
represents the CMIP5 modeled anthropogenic impact since 1901 for each
variable, grid cell, and month (Supporting Information).
We bilinearly interpolated the 1° CMIP5 multimodel mean 50-y low-pass
time series to the 1/24° spatial resolution of the observations and subtracted
the ACC signal from the observed daily and monthly time series. We consider
the remaining records after subtraction of the ACC signal to indicate climate
records that are free of anthropogenic trends (26).
Annual variations in fuel aridity metrics are presented as standardized
anomalies (σ) to accommodate differences across geography and metrics. All
fuel aridity metrics are standardized using the mean and SD from 1981 to
2010 for observations that excluded the ACC signal. Although the selection
of a reference period can bias results (54), our findings were similar when
using the full 1979–2015 time period or the observed data (without removal
of ACC) for the reference period. The influence of anthropogenic forcing on
fuel aridity metrics is quantified as the difference between metrics calculated with observations and those calculated with observations that excluded the ACC signal. Area-weighted standardized anomalies and the
spatial extent of western US forested land that experienced high (>1 σ)
aridity are computed for each aridity metric. Annualized burned area as well
as aggregated fuel aridity metrics calculated with data from ref. 50 and the
two reanalysis products are provided in Datasets S1–S3.
We use the regression relationship between the annual western US forest
fire area and the all-metric mean fuel aridity index in Fig. 1 to estimate the
forcing of anthropogenic increases in fuel aridity on forest fire area during
1984–2015. Uncertainties in the regression relationship due to imperfect
correlation and temporal autocorrelation are propagated as estimated
confidence bounds on the anthropogenic forcing of forest fire area. This
approach was repeated using a more conservative definition of the regression relationship, where we removed the linear least squares trend for
1984–2015 from both the area burned and fuel aridity time series before
regression to reduce the possibility of spurious correlation due to common
but unrelated trends (Fig. S5). Statistical significance of all linear trends and
correlations reported in this study are assessed using both Spearman’s rank
and Kendall’s tau statistics. Trends are considered significant if both tests
yield P < 0.05.
ACKNOWLEDGMENTS. We thank J. Mankin, B. Osborn, and two reviewers
for helpful comments on the manuscript and coauthors of ref. 26 for help
developing the empirical attribution framework. A.P.W. was funded by Columbia University’s Center for Climate and Life and by the Lamont-Doherty
Earth Observatory (Lamont contribution 8048). J.T.A. was supported by
funding from National Aeronautics and Space Administration Terrestrial
Ecology Program under Award NNX14AJ14G, and the National Science
Foundation Hazards Science, Engineering and Education for Sustainability
(SEES) Program under Award 1520873.

13. Flannigan M, et al. (2013) Global wildland fire season severity in the 21st century. For
Ecol Manage 294:54–61.
14. Yoon J, Kravitz B, Rasch P (2015) Extreme fire season in California: A glimpse into the
future? Bull Am Meteorol Soc 96:S5–S9.
15. Barbero R, Abatzoglou JT, Larkin NK, Kolden CA, Stocks B (2015) Climate change
presents increased potential for very large fires in the contiguous United States. Int J
Wildland Fire 24(7):892–899.
16. Gillett NP, Weaver AJ, Zwiers FW, Flannigan MD (2004) Detecting the effect of climate change on Canadian forest fires. Geophys Res Lett 31(18):L18211.
17. Westerling AL, Turner MG, Smithwick EAH, Romme WH, Ryan MG (2011) Continued
warming could transform Greater Yellowstone fire regimes by mid-21st century. Proc
Natl Acad Sci USA 108(32):13165–13170.
18. Krawchuk MA, Moritz MA, Parisien MA, Van Dorn J, Hayhoe K (2009) Global pyrogeography: The current and future distribution of wildfire. PLoS One 4(4):e5102.
19. Moritz MA, et al. (2012) Climate change and disruptions to global fire activity.
Ecosphere 3(6):1–22.
20. Marlon JR, et al. (2012) Long-term perspective on wildfires in the western USA. Proc
Natl Acad Sci USA 109(9):E535–E543.
21. Parks SA, et al. (2015) Wildland fire deficit and surplus in the western United States,
1984–2012. Ecosphere 6(12):1–13.
22. Higuera PE, Abatzoglou JT, Littell JS, Morgan P (2015) The changing strength and
nature of fire–climate relationships in the northern Rocky Mountains, U.S.A., 19022008. PLoS One 10(6):e0127563.
23. Pausas JG, Ribeiro E (2013) The global fire–productivity relationship. Glob Ecol
Biogeogr 22(6):728–736.
24. Littell JS, Peterson DL, Riley KL, Liu Y, Luce CH (2016) A review of the relationships
between drought and forest fire in the United States. Glob Change Biol 22(7):
2353–2369.

PNAS Early Edition   5 of 6

EARTH, ATMOSPHERIC,
AND PLANETARY SCIENCES

Methods

 25. Williams AP, et al. (2015) Correlations between components of the water balance and
burned area reveal new insights for predicting forest fire area in the southwest
United States. Int J Wildland Fire 24(1):14–26.
26. Williams AP, et al. (2015) Contribution of anthropogenic warming to California
drought during 2012–2014. Geophys Res Lett 42(16):6819–6828.
27. Sheffield J, et al. (2013) North American Climate in CMIP5 experiments. Part I: Evaluation of historical simulations of continental and regional climatology. J Clim 26(23):
9209–9245.
28. Hobbins MT (2016) The variability of ASCE standardized reference evapotranspiration: A rigorous, CONUS-wide decomposition and attribution. Trans Am Soc Agric Biol
Eng 59(2):561–576.
29. Mann ML, et al. (2016) Incorporating anthropogenic influences into fire probability
models: Effects of human activity and climate change on fire activity in California.
PLoS One 11(4):e0153589.
30. Yue X, Mickley LJ, Logan JA, Kaplan JO (2013) Ensemble projections of wildfire activity and carbonaceous aerosol concentrations over the western United States in the
mid-21st century. Atmos Environ (1994) 77:767–780.
31. Pechony O, Shindell DT (2010) Driving forces of global wildfires over the past millennium and the forthcoming century. Proc Natl Acad Sci USA 107(45):19167–19170.
32. Littell JS, et al. (2010) Forest ecosystems, disturbance, and climatic change in Washington State, USA. Clim Change 102(1-2):129–158.
33. Rogers BM, et al. (2011) Impacts of climate change on fire regimes and carbon stocks
of the U.S. Pacific Northwest. J Geophys Res Biogeosci 116(G3):G03037.
34. Williams AP, et al. (2014) Causes and implications of extreme atmospheric moisture
demand during the record-breaking 2011 wildfire season in the southwestern United
States. J Appl Meteorol Climatol 53(12):2671–2684.
35. Dong B, Dai A (2015) The influence of the Interdecadal Pacific Oscillation on temperature and precipitation over the globe. Clim Dyn 45(9-10):2667–2681.
36. Deser C, Knutti R, Solomon S, Phillips AS (2012) Communication of the role of natural
variability in future North American climate. Nat Clim Chang 2(11):775–779.
37. National Academies of Sciences, Engineering, and Medicine (2016) Attribution of
Extreme Weather Events in the Context of Climate Change (The National Academies Press, Washington, DC).
38. Swain DL, Horton DE, Singh D, Diffenbaugh NS (2016) Trends in atmospheric patterns
conducive to seasonal precipitation and temperature extremes in California. Sci Adv
2(4):e1501344.
39. Polade SD, Pierce DW, Cayan DR, Gershunov A, Dettinger MD (2014) The key role of
dry days in changing regional climate and precipitation regimes. Sci Rep 4:4364.
40. Romps DM, Seeley JT, Vollaro D, Molinari J (2014) Climate change. Projected increase
in lightning strikes in the United States due to global warming. Science 346(6211):
851–854.
41. Knorr W, Jiang L, Arneth A (2016) Climate, CO2 and human population impacts on
global wildfire emissions. Biogeosciences 13(1):267–282.
42. Williams AP, et al. (2013) Temperature as a potent driver of regional forest drought
stress and tree mortality. Nat Clim Chang 3(3):292–297.
43. Hart SJ, Schoennagel T, Veblen TT, Chapman TB (2015) Area burned in the western
United States is unaffected by recent mountain pine beetle outbreaks. Proc Natl Acad
Sci USA 112(14):4375–4380.
44. Van Wagtendonk JW (2007) The history and evolution of wildland fire use. Fire Ecol
3(2):3–17.
45. Bowman DMJS, Murphy BP, Williamson GJ, Cochrane MA (2014) Pyrogeographic
models, feedbacks and the future of global fire regimes. Glob Ecol Biogeogr 23(7):
821–824.
46. Parisien M-A, et al. (2014) An analysis of controls on fire activity in boreal Canada:
Comparing models built with different temporal resolutions. Ecol Appl 24(6):1341–1356.
47. Krawchuk MA, Moritz MA (2014) Burning issues: Statistical analyses of global fire
data to inform assessments of environmental change. Environmetrics 25(6):472–481.

6 of 6   www.pnas.org/cgi/doi/10.1073/pnas.1607171113

48. Millar CI, Stephenson NL (2015) Temperate forest health in an era of emerging
megadisturbance. Science 349(6250):823–826.
49. Smith AMS, et al. (2016) The science of firescapes: Achieving fire-resilient communities. Bioscience 66(2):130–146.
50. Abatzoglou JT (2013) Development of gridded surface meteorological data for ecological applications and modelling. Int J Climatol 33(1):121–131.
51. Daly C, et al. (2008) Physiographically sensitive mapping of climatological temperature and precipitation across the conterminous United States. Int J Climatol 28(15):
2031–2064.
52. Stavros EN, Abatzoglou J, Larkin NK, McKenzie D, Steel EA (2014) Climate and very
large wildland fires in the contiguous Western USA. Int J Wildland Fire 23(7):899–914.
53. Riley KL, Abatzoglou JT, Grenfell IC, Klene AE, Heinsch FA (2013) The relationship of
large fire occurrence with drought and fire danger indices in the western USA, 1984–
2008: The role of temporal scale. Int J Wildland Fire 22(7):894–909.
54. Sippel S, et al. (2015) Quantifying changes in climate variability and extremes: Pitfalls
and their overcoming. Geophys Res Lett 42(22):9990–9998.
55. Littell JS, Gwozdz RB (2011) Climatic water balance and regional fire years in the
Pacific Northwest, USA: linking regional climate and fire at landscape scales. The
Landscape Ecology of Fire (Springer, Dordrecht, The Netherlands), pp 117–139.
56. Morton DC, et al. (2013) Satellite-based assessment of climate controls on US burned
area. Biogeosciences 10(1):247–260.
57. Stocks BJ, et al. (1989) Canadian forest fire danger rating system: An overview. For
Chron 65(4):258–265.
58. Westerling AL, Gershunov A, Brown TJ, Cayan DR, Dettinger MD (2003) Climate and
wildfire in the western United States. Bull Am Meteorol Soc 84(5):595–604.
59. Flannigan MD, et al. (2016) Fuel moisture sensitivity to temperature and precipitation: Climate change implications. Clim Change 134(1-2):59–71.
60. Flannigan MD, Van Wagner CE (1991) Climate change and wildfire in Canada. Can J
Res 21(1):66–72.
61. Dowdy AJ, Mills GA, Finkele K, de Groot W (2010) Index sensitivity analysis applied to
the Canadian Forest Fire Weather Index and the McArthur Forest Fire Danger Index.
Meteorol Appl 17(3):298–312.
62. Mitchell KE, et al. (2004) The multi‐institution North American Land Data Assimilation
System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system. J Geophys Res Atmos 109(D7):D07S90.
63. Allen RG, Pereira LS, Raes D, Smith M (1998) Crop evapotranspiration-Guidelines for
computing crop water requirements-FAO Irrigation and drainage paper 56. FAO,
Rome 300(9):D05109.
64. Willmott CJ, Rowe CM, Mintz Y (1985) Climatology of the terrestrial seasonal water
cycle. J Climatol 5(6):589–606.
65. Andrews PL, Loftsgaarden DO, Bradshaw LS (2003) Evaluation of fire danger rating
indexes using logistic regression and percentile analysis. Int J Wildland Fire 12(2):
213–226.
66. Cohen JE, Deeming JD (1985) The National Fire-Danger Rating System: basic equations. Gen Tech Rep:16.
67. McArthur AG (1967) Fire behaviour in eucalypt forests (Forestry and Timber Bureau
Leaflet 107).
68. Griffiths D (1999) Improved formula for the drought factor in McArthur’s Forest Fire
Danger Meter. Aust For 62(3):202–206.
69. Wallace JM, Hobbs PV (2006) Atmospheric Science: An Introductory Survey (Academic,
Amsterdam), 2nd Ed.
70. Eidenshink JC, et al. (2007) A project for monitoring trends in burn severity. Fire Ecol
3(1):3–21.
71. Roy DP, Boschetti L, Justice CO, Ju J (2008) The collection 5 MODIS burned area
product—Global evaluation by comparison with the MODIS active fire product.
Remote Sens Environ 112(9):3690–3707.
72. van Vuuren DP, et al. (2011) The representative concentration pathways: An overview.
Clim Change 109(1):5–31.

Abatzoglou and Williams