This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 166.7.29.235

This content was downloaded on 25/02/2016 at 18:17

Please note that terms and conditions apply.

How will climate change affect wildland fire severity in the western US?

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 035002

(http://iopscience.iop.org/1748-9326/11/3/035002)

Home Search Collections Journals About Contact us My IOPscience

iopscience.iop.org/page/termshttp://iopscience.iop.org/1748-9326/11/3http://iopscience.iop.org/1748-9326http://iopscience.iop.org/http://iopscience.iop.org/searchhttp://iopscience.iop.org/collectionshttp://iopscience.iop.org/journalshttp://iopscience.iop.org/page/aboutioppublishinghttp://iopscience.iop.org/contacthttp://iopscience.iop.org/myiopscience

Environ. Res. Lett. 11 (2016) 035002 doi:10.1088/1748-9326/11/3/035002

LETTER

Howwill climate change affect wildland fire severity in thewestern US?

SeanAParks1, CarolMiller1, JohnTAbatzoglou2, LisaMHolsinger1,Marc-Andr Parisien3 andSolomonZDobrowski4

1 Aldo LeopoldWilderness Research Institute, RockyMountain Research Station, USDAForest Service, 790 East Beckwith Ave.,Missoula,MT59801,USA

2 Department ofGeography, University of Idaho, 875 PerimeterDrMS3021,Moscow, ID 83844,USA3 Northern Forestry Centre, Canadian Forest Service, Natural Resources Canada, 5320 122nd Street, Edmonton, Alberta T5H3S5,

Canada4 Department of ForestManagement, College of Forestry andConservation, University ofMontana, 32CampusDrive,Missoula,MT

59812,USA

E-mail: sean_parks@fs.fed.us

Keywords:wildland fire,fire severity, fire regime, climate change, disequilibrium

Supplementarymaterial for this article is available online

AbstractFire regime characteristics inNorth America are expected to change over the next several decades as aresult of anthropogenic climate change. Although some fire regime characteristics (e.g., area burnedandfire season length) are relatively well-studied in the context of a changing climate, fire severity hasreceived less attention. In this study, we used observed data from1984 to 2012 for thewesternUnitedStates (US) to build a statisticalmodel offire severity as a function of climate.We then applied thismodel to several (n=20) climate change projections representingmid-century (20402069)conditions under the RCP 8.5 scenario.Model predictions suggest widespread reduction infireseverity for large portions of thewesternUS.However, ourmodel implicitly incorporates climate-induced changes in vegetation type, fuel load, andfire frequency. As such, our predictions are bestinterpreted as a potential reduction infire severity, a potential thatmay not be realized due human-induced disequilibriumbetween plant communities and climate. Consequently, to realize thereductions infire severity predicted in this study, landmanagers in thewesternUS could facilitate thetransition of plant communities towards a state of equilibriumwith the emerging climate throughmeans such as active restoration treatments (e.g.,mechanical thinning and prescribed fire) and passiverestoration strategies likemanaged natural fire (under suitable weather conditions). Resisting changesin vegetation composition and fuel load via activities such as aggressive fire suppressionwill amplifydisequilibrium conditions andwill likely result in increased fire severity in future decades because fuelloadswill increase as the climatewarms andfire danger becomesmore extreme. The results of ourstudy provide insights to the pros and cons of resisting or facilitating change in vegetation compositionand fuel load in the context of a changing climate.

Introduction

Fire regimes in North America are expected to changeover the next several decades as a result of anthro-pogenic climate change (Dale et al 2001). Fire activity(i.e., annual area burned and fire frequency) isexpected to increase in many regions (Krawchuket al 2009, Littell et al 2010) and new research showsthat fire seasons are now starting earlier and ending

later compared to previous decades (Jolly et al 2015).However, the effect of climate change on one veryimportant fire regime characteristicfire severityisnot well-studied or understood (Flannigan et al 2009,Hessl 2011). In the context of this paper, we defineseverity as the degree of fire-induced change tovegetation and soils one year post-fire (Key andBenson 2006, Miller and Thode 2007). For example, astand-replacing fire in upper-elevation conifer forest is

OPEN ACCESS

RECEIVED

30October 2015

REVISED

29 January 2016

ACCEPTED FOR PUBLICATION

1 February 2016

PUBLISHED

23 February 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

2016 IOPPublishing Ltd

http://dx.doi.org/10.1088/1748-9326/11/3/035002mailto:sean_parks@fs.fed.ushttp://dx.doi.org/10.1088/1748-9326/11/3/035002http://crossmark.crossref.org/dialog/?doi=10.1088/1748-9326/11/3/035002&domain=pdf&date_stamp=2016-02-23http://crossmark.crossref.org/dialog/?doi=10.1088/1748-9326/11/3/035002&domain=pdf&date_stamp=2016-02-23http://creativecommons.org/licenses/by/3.0http://creativecommons.org/licenses/by/3.0http://creativecommons.org/licenses/by/3.0

considered high severity because the site has drasticallychanged one year post-fire compared to pre-fireconditions, whereas a surface fire in a grass-dominatedecosystem is considered low severity because thevegetation is nearly fully recovered one-year post fire.

The severity at which a site burns influences vege-tation response and successional trajectory (Barrettet al 2011), faunal response (Smucker et al 2005), car-bon emissions (Ghimire et al 2012), and erosion ratesand sedimentation (Benavides-Solorio and MacDo-nald 2005). Furthermore, human safety and infra-structure are influenced by the severity at which a siteburns (Miller and Ager 2013), and managementresponses to fire and allocation of firefighting resour-ces are also influenced by the expected fire severity(e.g., Calkin et al 2011). As such, there is a need to bet-ter understand how fire severity will respond to achanging climate (e.g.,Miller et al 2009).

At fine temporal scales, fire severity depends onfactors that are highly variable over time, such as firespread rate and direction (e.g., heading versus backingfire) and weather (Finney 2005, Birch et al 2015). Atbroader temporal scales, however, climate (in terms ofclimatic normals) is a major influence through itsinteractive effect on productivity (and hence amountof biomass) and moisture availability (i.e., wet versusdry ecosystems) (Parks et al 2014b, Whitmanet al 2015). Consequently, because fire regimes areintrinsically defined by the characteristics of fires thatoccur over extended periods of time (years to cen-turies) (Morgan et al 2001), evaluations of fire severityover gradients of observed and predicted climatic nor-mals allows for a formal assessment of how fire sever-itymay respond to climate change.

We seek to quantify how fire severity in thecontiguous western United States (US) (hereafter thewestern US)may respond to climate change. We usestatistical relationships between observed climaticnormals and fire severity (Parks et al 2014b, Kaneet al 2015) to conduct a formal evaluation of future fireseverity patterns. Because the relationship between cli-mate and fire regimes is known to be weak in areas ofhigh human impact (Parks et al 2014b), we used datafrom areas with low anthropogenic influence tobuild a statistical model of fire severity as a functionof climatic normals over the 19842012 time period.We then predicted contemporary (19842012) andfuture (mid-century; 20402069) fire severity usingclimate data from numerous global climate models(GCMs) for the western US. As far as we know, thisstudy is the first to examine how fire severity mayrespond to a changing climate over such a broadspatial extent. The results of this study will advanceour understanding of fire regimes in the western USin the context of a changing climate and will assistpolicy makers and landmanagers to better manage forresilient landscapes.

Methods

Consistent with major fire severity mapping efforts(Key and Benson 2006, Eidenshink et al 2007), wedefine fire severity as the degree of fire-induced changeto vegetation and soils. We built a statistical model offire severity as a function of climate by first partition-ing our study area (the western US; figures 1(a) and(b)) into 500 km2 hexagonal polygons (i.e., hexels).Within each hexel, we summarized fire severity usingthe delta normalized burn ratio (dNBR) (Key andBenson 2006), a satellite index (resolution: 30 m) thatdifferences pre- and post-fire Landsat TM, ETM+,and OLI images and has a high correspondence tofield-based measures of severity such as the compositeburn index (CBI; R20.65) (van Wagtendonket al 2004, Parks et al 2014a). The CBI is a post-fireassessment in which individual rating factors in eachof several vertically arranged strata (soil and rock, litterand surface fuels, low herbs and shrubs, tall shrubs,and trees) are assessed on a continuous 03 scaleindicating the magnitude of fire effects. A rating of 0reflects no change due to fire, whereas 3 reflects thehighest degree of change. Factors assessed include soilchar, surface fuel consumption, vegetation mortality,and scorching of trees. Ratings are averaged for eachstratum and then across all strata to arrive at an overallCBI rating for an entire plot. The CBI indicates that,as dNBR values increase, there is generally an increasein char and scorched/blackened vegetation and adecrease in moisture content and vegetative cover(Key and Benson 2006). Measurements of fire severity(dNBR and CBI) are generally conducted one yearafter fire, so any regrowth that occurs within one yearwill result in reduced severity compared to assess-ments conducted immediately post-fire; this is parti-cularly relevant for species that recover quickly afterfire (e.g., resprouting shrubs, grasses).

Fire severity (i.e., dNBR) data were obtained fromthe Monitoring Trends in Burn Severity project(Eidenshink et al 2007) for all fires 400 ha for the19842012 time period. Raw dNBR values obtainedfrom MTBS were adjusted using the dNBR offset(Key 2006), which accounts for differences due to phe-nology or precipitation between the pre- and post-fireimages by subtracting the average dNBR of pixelsoutside the burn perimeter. This adjustment can beimportant when comparing severity among fires(Parks et al 2014a). Amean dNBRwas calculated usingall pixels of all fires that intersected each 500 km2

hexel; pixels classified as nonfuel were excluded in thecalculation of the mean. We square-root transformedmean dNBR values to linearize the relationship to theCBI (figure S1).

We summarized climate normals within eachhexel using five variables with known links to fireregimes (e.g., Littell and Gwozdz 2011, Abatzoglouand Kolden 2013, Parks et al 2015b): actual evapo-transpiration (AET), water deficit (WD), annual

2

Environ. Res. Lett. 11 (2016) 035002

precipitation (PPT), soil moisture (SMO), and snowwater equivalent (SWE). Gridded monthly temper-ature and PPT data were obtained from the para-meter-elevation regression on independent slopesmodel (PRISM; Daly et al 2002), which uses weatherstation data and physiographic factors to map climateat a spatial resolution of 800 m. In addition, dailyand sub-daily surfacemeteorological variables (4 kmresolution) describing temperature, humidity, winds,solar radiation, and precipitation were produced fol-lowing Abatzoglou (2013). These data were collec-tively used to compute climatic water balancefollowing Dobrowski et al (2013) to estimate AET,SWE, SMO, and WD. This water balance modeloperates on a monthly time-step and accounts foratmospheric demand (via the PenmanMonteithequation), soil water storage, and includes the effect oftemperature and radiation on snow hydrology via asnow melt model. Each variable was averaged withineach hexel for the years 19842012, thereby matchingthe years of the fire severity data. We similarly sum-marized these five climate variables representing mid-21st century (20402069) conditions using 20 global

climate models (GCMs) for the RCP8.5 emissions sce-nario (table S1). These tables were statistically down-scaled to the same grid as observed data using themultivariate adapted constructed analogs approach(Abatzoglou andBrown 2012).

Because the relationship between climate and fireis weaker in landscapes that are highly influenced byhumans (Parks et al 2014b), we built our model usingdata from a subset of hexels with low human influence(figure 1(b)). We selected only those hexels that werecomprised of at least 50% designated wilderness ornational park or had an average human footprint(Leu et al 2008)2.5 (on a scale of 110). We furtherlimited our dataset to include only those hexels with atleast 400 ha of total burned area from 1984 to 2012.These selection criteria resulted in 544 hexels that,despite representing a small proportion of our studyarea (8.7%), are climatically representative of much ofthe western US, with the notable exception of the wetregions of the PacificNorthwest (figure S2).

Using data from the subset of 544 hexels, wemodeled fire severity (dNBR) as a function ofcontemporary climate (19842012) using boosted

Figure 1. Study area of the westernUS for whichwe predicted changes infire severity under a future climate.Map showing ecoregionboundaries (TheNatureConservancy 2009) and forested areas (in gray) (a) and showing designatedwilderness areas and nationalparks (in gray) as well as the 544 hexels (in blue) used to build themodels offire severity as a function of climate (b). Ecoregion namesand boundaries provided for context.

3

Environ. Res. Lett. 11 (2016) 035002

regression trees (BRT) (gbm package) in the R statis-tical environment (R Development Core Team 2007).BRT is a nonparametric machine-learning approachthat does not require a priori model specification ortest of hypothesis (Death 2007). The BRT algorithmfits the best possible model to the data structure,including complex interactions among variables. Itdoes so by building a large number of regression trees,whereby, through a forward stage-wise model-fittingprocess, each term represents a small tree built on theweighted residuals of the previous tree. The stage-wiseprocedure reduces bias, whereas variance is decreasedthrough model averaging. The BRT method alsoemploys bagging, the use of a random subset of sam-ples, which typically improves model predictions.Comparisons to other modeling techniques indicatethat BRT models consistently produce robust pre-dictive estimates (Elith et al 2006). We followed therecommendations of Elith et al (2008) for selectingBRT options; we set the bagging fraction to 0.5, learn-ing rate to 0.005, and tree complexity to three. Weused a custom script from Elith et al (2008) to deter-mine the necessary number of trees, thereby reducingthe potential for overfitting. We evaluated the modelfit using the (a) correlation between predicted andobserved fire severity and (b) ten-fold cross-validatedcorrelation between predicted and observed fireseverity.

We used the model to predict contemporary(19842012) fire severity (dNBR) for all hexels in thewesternUS. However, interpreting dNBR and changesin dNBR under a changing climate is challengingbecause dNBR units have no direct ecological inter-pretation. As such, we rescaled these predictions tocorrespond to the ecologically relevant compositeburn index (hereafter inferred CBI) that ranges from0 to 3 (Key and Benson 2006): the lowest predictedseverity was given an inferred CBI of 0.1, which is thethreshold for unchanged (Miller and Thode 2007),and the highest predicted severity was given an infer-red CBI of 3.0.Wewere then able to infer the CBI of allremaining predictions because the square-root trans-formation of dNBR linearized the relationship to CBI(figure S1). Consequently, we generated a map repre-senting the inferred CBI for thewesternUS under con-temporary climate.

We then predicted fire severity for the mid-21stcentury (20402069) as projected by each GCM usingthe BRT model. We inferred CBI as previously descri-bed using the linear relationship between dNBR andCBI of the observed predictions to make the infer-ences. Note that the predictions for all hexels in thewestern US were clamped to avoid predicting outsideof the observed range of severity values; all predictions>3 and

Discussion

Our models based on contemporary fireclimaterelationships predict a widespread reduction in fireseverity for large portions of the western US by themid-21st century. Only a very small proportion of thewestern US is predicted to experience an increase inseverity. Our prediction contrasts with those based onthe direct influence of climate on fuel moisture andassociated fire danger indices that occur at seasonaltime scales (Fried et al 2004, Nitschke and Innes 2008).Our use of broad-scale climate as a proxy for vegeta-tion composition and fuel load instead emphasizes theindirect influence that climate has on fire regimes(Miller and Urban 1999, Higuera et al 2014). Specifi-cally, the predicted decrease in fire severity can beattributed to climatic conditions associated withhigher WDs (figures 5(a) and (b)), lower productivity,and less burnable biomass (Zhao and Running 2010,Stegen et al 2011).

Our approach and findings are based on an impli-cit assumption that vegetation composition and fuelload will track changes in climate. Indeed, this is acommon assumption that underlies numerous cli-mate change studies, including those that use distribu-tion models to project shifts in habitat ranges (Engleret al 2011) and fire activity (Krawchuk et al 2009,Mor-itz et al 2012). Specifically, our predictions of overalllower fire severity implicitly assume that vegetationcomposition and burnable biomass will reflect lowerproductivity associated with warmer and drier cli-mates (e.g., increased WD; figure 5(b)). As such, ourpredictions are best interpreted as a potential reduc-tion in fire severity, a potential thatmay not be realizedwhere there is disequilibrium between climate andvegetation. Disequilibrium dynamics are the result ofmany factors and signals that directional changes inclimate may not result in immediate changes in vege-tation composition and fuel load (Sprugel 1991, Sven-ning and Sandel 2013). For example, leading-edgedisequilibrium can arise when species are dispersal

limited or dont reach reproductive maturity for manyyears (Svenning and Sandel 2013). Trailing-edge dis-equilibrium can arise because some species are long-lived and have deep roots, thereby facilitating survivaland persistence under substantial inter-annual anddecadal fluctuations in climate even though seedlingsof the same species are unable to survive (Grubb 1977,Jackson et al 2009). To compound this, human-induced disequilibrium has also substantially affectedmost ecosystems in the western US (and globally)(Parks et al 2015b), in that natural disturbances such asfire have been excluded by factors such as livestockgrazing, fire suppression, and landscape fragmenta-tion (Marlon et al 2008). Both climate- and human-induced disequilibrium underlie present-day con-cerns about restoration of fire-adapted ecosystemsafter a century of fire exclusion (Stephens et al 2013,Hessburg et al 2015).

Consequently, our predictions are more likely tohold up in the presence of an active disturbanceregime that catalyzes climatically driven changes invegetation composition and fuel load (Flanniganet al 2000, Turner 2010). Disturbance catalysts are cri-tical components for maintaining a dynamic equili-brium between vegetation and climate and appear toalready be occurring with increasing frequency insome regions. For example, many studies have con-cluded that fire activity has increased in recent years(Westerling et al 2006, Kelly et al 2013) andwidespreadtree mortality has been attributed to drought andinsect outbreaks (Allen et al 2010, Bentz et al 2010).In areas recently affected by these disturbances, thepost-fire species and vegetation densities may be moretailored to the emerging climate (Overpeck et al 1990,Millar et al 2007). Although generally consideredundesirable, disturbance-facilitated conversions fromforest to non-forest vegetation are likely to occur insome situations (Stephens et al 2013, Coop et al inpress), especially when compounded by human-induced disequilibrium.

Figure 3.Variable importance in the BRTmodel (a) and partial dependence plots showing the relationship between dNBR and thetwomost influential variables (WDandPPT) (b), (c). Note that the partial dependence plots do not reflect interactions betweenvariables and therefore simplify the relationships.

5

Environ. Res. Lett. 11 (2016) 035002

Most forested regions in the western US are cur-rently experiencing a fire deficit (Marlon et al 2012,Parks et al 2015b) because human activities and infra-structure (e.g., fire suppression and roads) exclude fireas an important disturbance agent. Consequently,human-induced disequilibrium between vegetationand climate, coupled with a changing climate, hasimportant implications for future fire severity. Weposit that such amplified disequilibrium will likelyresult in increased fire severity in future decades as fuelloads increase, fire seasons lengthen, and fire dangerbecomes more extreme (Collins 2014, Jolly et al 2015).

This supposition is consistent with the findings ofother studies that found a climate-induced increase infire severity when assuming static vegetation (Friedet al 2004, Nitschke and Innes 2008). Continuing toresist catalysts of vegetation change only increases theprobability of undesirable effects given that fire isinevitable (North et al 2009, Calkin et al 2015). Analternative to this unsustainable cycle is to activelyfacilitate transition of ecosystems to conditions thataremore suited to the future climate bymeans ofman-aged wildland fire or other restoration treatments(Millar et al 2007).

Figure 4.Predictedfire severity under observed (a) andmid-century climate (b).Mean change infire severity among the 20 predictions(one prediction for eachGCM) (c).

6

Environ. Res. Lett. 11 (2016) 035002

Our study complements and expands our under-standing of controls on fire regimes and how they mayrespond to a changing climate in the western US. Spe-cifically, predicted increases in fire activity (Littellet al 2010, Moritz et al 2012) imply that less biomasswill be able to accumulate between successive fires,resulting in less biomass available for combustion anda reduction in fire severity. Furthermore, predictedincreases in WD (figure 5(b)) are expected to increasewater stress and decrease productivity in the generallywater-limited western US (Chen et al 2010, Williamset al 2013), ultimately reducing the amount of biomassavailable to burn and resultant fire severity. It shouldbe noted, however, that temperature-limited ecosys-tems (i.e., alpine environments) will likely experiencean increase in productivity (and fire severity) under awarmer climate (Grimm et al 2013, Goulden andBales 2014).

Our study relied on observed and predicted cli-matic normals (i.e., multi-decadal averages) to predictpotential changes in fire severity. This is in contrast toother climate change fire studies that used annually orseasonally resolved climate (observed and GCM pro-jections) and fire data to make predictions of potentialchanges in fire activity (i.e., fire frequency or areaburned) (Littell et al 2010, Stavros et al 2014). The lat-ter approach is often used because of the noted impor-tance of climatic extremes on fire regimes (e.g.,Westerling et al 2006). Although we could have builtourmodel of fire severity using annually resolved data,we posit, for the purpose of predicting future fireseverity, using long term averages (e.g., 19842012) ismore appropriate for at least three reasons. First,although several studies have shown that fire severityresponds to annual, seasonal, or daily variability in

climate or weather, the relative influence of this varia-bility can be fairly weak (Dillon et al 2011, Birchet al 2015). This is in contrast to broad temporal scaleswhere the relationship between fire severity and cli-mate has been found to be much stronger (Parkset al 2014b, Kane et al 2015). Second, because modelsbuilt at a fine temporal resolution aremore focused onthe direct influence of climatic variability on fireweather and fuel moisture, they generally fail to incor-porate climate- or fire-induced changes in vegetationcomposition or fuel load (Allen et al 2010, Parkset al 2015a). We suggest that predictions based on cli-matic normals implicitly incorporate such changes(Kelly and Goulden 2008, Marlon et al 2009). Lastly,GCMs may not adequately simulate annual climaticvariability and thus are better suited for predictinglong term trends (Stoner et al 2009).

Our model used broad scale data and the predic-tions of widespread reduced fire severity underfuture climate should be interpreted accordingly. Forexample, fire severity and climate vary at scalesfiner than the spatial resolution of the hexel usedin this study (Schoennagel et al 2004). As such, ouranalysis does not likely capture finer-scale changes infire severity that could occur. For example, in alpineenvironments where localized upward shifts intreeline under a warmer climate are expected to con-tribute to increases in biomass (Higuera et al 2014),fire severity might be expected to increase. Althoughour model of fire severity (dNBR) as a functionof climate performed reasonably well (see sectionResults), we acknowledge that further error may beintroduced due to error in the relationship betweenCBI and dNBR. However, we posit that the improvedecological interpretation attained by converting dNBR

Figure 5.Plot of observedfire severity as a function of observed (19842012)water deficit (WD; themost influential variable in theBRTmodel) for the 544 hexels used to build themodel (a). The red line shows themodel fit according to a generalized additivemodel.Map of predicted increase inWD from contemporary tomid-century (20402069) climatic conditions (b); values depict themulti-model average change between time periods. According to this simple relationship, increasedWDdue to climate changewill result indecreased fire severity. Note that the relationship flattens outwhich suggests aweaker response in dry ecosystems.

7

Environ. Res. Lett. 11 (2016) 035002

to CBI outweighs any increased error in ourpredictions.

Our measure of fire severity relied on dNBR (aunitless ratio) and CBI (a composite rating) and, con-sequently, there is no definable unit of measurement(e.g., grams of carbon consumed m2). Instead weinfer changes in CBI, which integrates several strata(e.g., soil and shrubs) and scales severity from 0 to 3.This is admittedly a somewhat vague framework forassessing potential changes in fire severity, but takesadvantage of the widespread availability of satellite-inferred metrics of fire severity and their documentedcorrelation to the CBI. We suggest future researchefforts involving fire severity and climate change aimto use more definitive and quantitative units of mea-surement. On a similar note, fire severity has ecologi-cal significance beyond what can be inferred fromdNBR and is the result of many complex physical, bio-logical, and ecological factors (Morgan et al 2014). Forexample, in ecosystems that are ill-adapted to fire (e.g.,the Mojave Desert), dNBR values may be irrelevant, asany and all fires might be considered severe (Brooksand Matchett 2006). Accordingly, although we useddNBR and CBI as a convenient and standardized wayto assess fire severity, predictions for some ecoregionsshould be carefully interpreted.

Our model does not consider plant physiologicalresponses to a CO2 enriched atmosphere (e.g.,improved water use efficiency and plant productivity)that could lead to increases in fire severity (Drakeet al 1997, Keenan et al 2013). Given that todays atmo-spheric CO2 concentration is the highest its been forat least 650 000 years (Siegenthaler et al 2005), thiscould be a particularly important consideration forextreme water limited ecosystems such as grasslands,where woody plant encroachment could cause chan-ges in biomass amount and structure (Morganet al 2007, Norby and Zak 2011). Consequently, otherresearch approaches using tools such as dynamic glo-bal vegetation models may predict different outcomes(Thonicke et al 2001).

Although we relied on data from protected areasand other areas of low human influence and thusunderrepresented certain climatic environments (seeBatllori et al 2014), these data represent a surprisinglybroad range of ecosystem types in the western US ran-ging from warm desert (Death Valley National Park(NP) to dry conifer forest (Gila Wilderness) to coldforest (Yellowstone NP) (figure S2). As such, we sug-gest that under-represented climates have only a mar-ginal effect on our results (see figure S2). Indeed, ouranalysis (figure S2) indicates that the data we used tobuild the model adequately represents the climatesof most of the western US with the most notableexception being those in the Pacific Northwest wherefires were historically and are currently infrequent(Agee 1993).

Conclusions

Our study predicts an overall decrease in fire severityfor much of the western US by mid-century(20402069) due to changing climatic conditions.These predictions are best interpreted as potentialdecreases in severity that may not be realized unlessvegetation composition and fuel load change inparallel with climate. Disequilibrium between plantcommunities and climate will only escalate, particu-larly in forested areas, unless natural disturbances andmanagement activities (i.e., prescribed fire andrestoration treatments) act as catalysts of vegetationchange and push plant communities towards a state ofequilibrium with climate. A high degree of disequili-brium between plant communities and climate isgenerally considered undesirable because the resultmay be an uncharacteristically severe wildland fire thatcauses abrupt ecosystem state shifts from, for example,forest to non-forest vegetation (e.g., Coop et al 2016).

Our findings support a passive managementapproach to ecosystem restoration (Arno et al 2000),whereby natural disturbance regimes are used to facil-itate the transition of plant communities towards astate of equilibrium with the emerging climate. Activerestoration treatmentsmay also aid in facilitating thesechanges in certain situations (Millar et al 2007, Ste-phens et al 2010), but the current pace and scale ofsuch treatments is insufficient to make a meaningfulimpact across the vast forested regions of the westernUS (North et al 2012). In addition, legal (e.g., desig-nated wilderness) and logistical constraints (e.g., steepslopes) make certain activities (mechanical thinning)infeasible across a large proportion of land in the wes-tern US (North et al 2014). Achieving landscape resi-lience in a changing climate will likely requireincreased use of managed wildland fire, especiallywhen weather conditions are not extreme (Northet al 2015), and in fact, resisting change via activitiessuch as aggressive fire suppression may be counter-productive in the long-run (Calkin et al 2015). Assuch, the results of this study provide insights to policymakers and land managers in the western US as to thepros and cons of resisting or facilitating change invegetation composition and fuel load in the context ofa changing climate.

Acknowledgments

We acknowledge National Fire Plan Funding from theRocky Mountain Research Station. We thank twoanonymous reviewers for thoughtful comments thatsignificantly improved thismanuscript.

References

Abatzoglou J T 2013Development of gridded surfacemeteorological data for ecological applications andmodellingInt. J. Climatol. 33 12131

8

Environ. Res. Lett. 11 (2016) 035002

http://dx.doi.org/10.1002/joc.3413http://dx.doi.org/10.1002/joc.3413http://dx.doi.org/10.1002/joc.3413

Abatzoglou J T andBrownT J 2012A comparison of statisticaldownscalingmethods suited for wildfire applications Int. J.Climatol. 32 77280

Abatzoglou J T andKoldenCA2013Relationships between climateandmacroscale area burned in thewesternUnited States Int.J.Wildland Fire 22 100320

Agee J K 1993 Fire Ecology of PacificNorthwest Forests (Washington,DC: Island Press)

AllenCD et al 2010A global overview of drought and heat-inducedtreemortality reveals emerging climate change risks forforests Forest Ecol.Manage. 259 66084

Arno S F, ParsonsD J andKeane RE 2000Mixed-severity fireregimes in the northern rockymountains: consequences offire exclusion and options for the futureWilderness Science ina Time of Change Conf. (RockyMountain Research Station,Missoula,MT:USDAForest Service)RMRS-P-15-VOL-5

Barrett K,McGuire AD,HoyEE andKasischke E S 2011 Potentialshifts in dominant forest cover in interior Alaska driven byvariations infire severity Ecol. Appl. 21 238096

Batllori E,Miller C, ParisienM-A, Parks SA andMoritzMA2014 IsUS climatic diversity well represented within the existingfederal protection network? Ecol. Appl. 24 1898907

Benavides-Solorio JDD andMacDonald LH2005Measurementand prediction of post-fire erosion at the hillslope scale,Colorado front range Int. J.Wildland Fire 14 45774

Bentz B J, Regniere J, Fettig C J,Hansen EM,Hayes J L, Hicke J A,Kelsey RG,Negron J F and Seybold S J 2010Climate changeand bark beetles of theWesternUnited States andCanada:direct and indirect effectsBioscience 60 60213

BirchD S,Morgan P, KoldenCA, Abatzoglou J T,DillonGK,HudakAT and SmithAMS2015Vegetation, topographyand daily weather influenced burn severity in central IdahoandwesternMontana forestsEcosphere 6 17

BrooksML andMatchett J R 2006 Spatial and temporal patterns ofwildfires in theMojaveDesert, 19802004 J. Arid Environ. 6714864

CalkinDE, ThompsonMP, FinneyMA andHydeKD2011A real-time risk assessment tool supporting wildlandfiredecisionmaking J. Forestry 109 27480

CalkinDE, ThompsonMP and FinneyMA2015Negativeconsequences of positive feedbacks inUSwildfiremanagement Forest Ecosyst. 2 110

Chen P-Y,WelshC andHamannA 2010Geographic variation ingrowth response ofDouglas-fir to interannual climatevariability and projected climate changeGlob. Change Biol. 16337485

Collins BM2014 Fire weather and large fire potential in thenorthern SierraNevadaAgric. ForestMeteorol. 189-190 305

Coop JD, Parks S A,McClernan S R andHolsinger LM2016Influences of prior wildfires on vegetation response tosubsequentfire in a reburned southwestern landscape Ecol.Appl. in press

Dale VH et al 2001Climate change and forest disturbances: climatechange can affect forests by altering the frequency, intensity,duration, and timing offire, drought, introduced species,insect and pathogen outbreaks, hurricanes, windstorms, icestorms, or landslidesBioscience 51 72334

DeathG2007 Boosted trees for ecologicalmodeling and predictionEcology 88 24351

DillonGK,HoldenZA,Morgan P, CrimminsMA,Heyerdahl EK and LuceCH2011 Both topography andclimate affected forest andwoodland burn severity in tworegions of thewesternUS, 19842006Ecosphere 2 130

Dobrowski S Z, Abatzoglou J, SwansonAK,Greenberg J A,Mynsberge AR,HoldenZA and SchwartzMK2013Theclimate velocity of the contiguousUnited States during the20th centuryGlob. Change Biol. 19 24151

Drake BG,Gonzalez-MelerMA and Long S P 1997More efficientplants: a consequence of rising atmospheric CO2?Annu. Rev.Plant Physiol. PlantMol. Biol. 48 60939

Eidenshink J, Schwind B, BrewerK, ZhuZL,Quayle B andHoward S 2007A project formonitoring trends in burnseverity Fire Ecol. 3 321

Elith J, Leathwick J R andHastie T 2008Aworking guide to boostedregression trees J. Animal Ecol. 77 80213

Elith J et al 2006Novelmethods improve prediction of speciesdistributions fromoccurrence dataEcography 29 12951

Engler R et al 2011 21st century climate change threatensmountainflora unequally across EuropeGlob. Change Biol. 17 233041

FinneyMA2005The challenge of quantitative risk analysis forwildlandfire Forest Ecol.Manage. 211 97108

FlanniganMD,KrawchukMA, deGrootW J,Wotton BMandGowmanLM2009 Implications of changing climate forglobal wildlandfire Int. J.Wildland Fire 18 483507

FlanniganMD, Stocks B J andWottonBM2000Climate changeand forestfires Sci. Total Environ. 262 2219

Fried J S, TornMS andMills E 2004The impact of climate changeonwildfire severity: a regional forecast for northernCaliforniaClim. Change 64 16991

Ghimire B,WilliamsCA, CollatzG J andVanderhoofM2012 Fire-induced carbon emissions and regrowth uptake inwesternUS forests: documenting variation across forest types,fireseverity, and climate regions J. Geophys. Res.Biogeosci. 117G03036

GouldenML andBales RC 2014Mountain runoff vulnerability toincreased evapotranspirationwith vegetation expansion Proc.Natl Acad. Sci. USA 111 140715

GrimmNB et al 2013The impacts of climate change on ecosystemstructure and function Front. Ecol. Environ. 11 47482

Grubb P J 1977Themaintenance of species-richness in plantcommunities: the importance of the regeneration nicheBiol.Rev. 52 10745

Hessburg P F et al 2015Restoringfire-prone Inland Pacificlandscapes: seven core principles Landscape Ecol. 30 180535

Hessl A E 2011 Pathways for climate change effects on fire:models,data, and uncertainties Prog. Phys. Geogr. 35 393407

Higuera P E, Briles C E andWhitlockC 2014 Fire-regimecomplacency and sensitivity to centennial-throughmillennial-scale climate change in RockyMountainsubalpine forests, Colorado,USA J. Ecol. 102 142941

Jackson ST, Betancourt J L, BoothRK andGray S T 2009 Ecologyand the ratchet of events: Climate variability, nichedimensions, and species distributionsProc. Natl Acad. Sci.USA 106 1968592

JollyWM,CochraneMA, Freeborn PH,HoldenZA, BrownT J,WilliamsonG J andBowmanDMJS 2015Climate-inducedvariations in global wildfire danger from1979 to 2013Nat.Commun. 6 7537

KaneVR, Lutz J A, Cansler AC, PovakNA,Churchill D J,SmithDF, Kane J T andNorthMP2015Water balance andtopography predict fire and forest structure patterns ForestEcol.Manage. 338 113

KeenanTF,HollingerDY, Bohrer G,DragoniD,Munger JW,SchmidHP andRichardsonAD2013 Increase in forestwater-use efficiency as atmospheric carbon dioxideconcentrations riseNature 499 3247

Kelly A E andGouldenML2008Rapid shifts in plant distributionwith recent climate changeProc. Natl Acad. Sci. USA 105118236

Kelly R, ChipmanML,Higuera P E, Stefanova I, Brubaker LB andHuF S 2013Recent burning of boreal forests exceeds fireregime limits of the past 10 000 yearsProc. Natl Acad. Sci.USA 110 1305560

KeyCH2006 Ecological and sampling constraints on defininglandscapefire severity Fire Ecol. 2 3459

KeyCH andBensonNC2006 Landscape assessment (LA)FIREMON: Fire EffectsMonitoring and Inventory System edDLutes et al (Fort Collins, CO:Department of Agriculture,Forest Service, RockyMountain Research Station) RMRS-GTR-164, 55

KrawchukMA,MoritzMA, ParisienM-A,VanDorn J andHayhoeK2009Global pyrogeography: the current and futuredistribution ofwildfirePlos One 4 e5102

LeuM,Hanser S E andKnick ST 2008The human footprint in thewest: a large-scale analysis of anthropogenic impacts Ecol.Appl. 18 111939

9

Environ. Res. Lett. 11 (2016) 035002

http://dx.doi.org/10.1002/joc.2312http://dx.doi.org/10.1002/joc.2312http://dx.doi.org/10.1002/joc.2312http://dx.doi.org/10.1071/WF13019http://dx.doi.org/10.1071/WF13019http://dx.doi.org/10.1071/WF13019http://dx.doi.org/10.1016/j.foreco.2009.09.001http://dx.doi.org/10.1016/j.foreco.2009.09.001http://dx.doi.org/10.1016/j.foreco.2009.09.001http://dx.doi.org/10.1890/10-0896.1http://dx.doi.org/10.1890/10-0896.1http://dx.doi.org/10.1890/10-0896.1http://dx.doi.org/10.1890/14-0227.1http://dx.doi.org/10.1890/14-0227.1http://dx.doi.org/10.1890/14-0227.1http://dx.doi.org/10.1071/WF05042http://dx.doi.org/10.1071/WF05042http://dx.doi.org/10.1071/WF05042http://dx.doi.org/10.1525/bio.2010.60.8.6http://dx.doi.org/10.1525/bio.2010.60.8.6http://dx.doi.org/10.1525/bio.2010.60.8.6http://dx.doi.org/10.1890/ES14-00213.1http://dx.doi.org/10.1016/j.jaridenv.2006.09.027http://dx.doi.org/10.1016/j.jaridenv.2006.09.027http://dx.doi.org/10.1016/j.jaridenv.2006.09.027http://dx.doi.org/10.1016/j.jaridenv.2006.09.027http://dx.doi.org/10.1186/s40663-015-0033-8http://dx.doi.org/10.1186/s40663-015-0033-8http://dx.doi.org/10.1186/s40663-015-0033-8http://dx.doi.org/10.1111/j.1365-2486.2010.02166.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02166.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02166.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02166.xhttp://dx.doi.org/10.1016/j.agrformet.2014.01.005http://dx.doi.org/10.1016/j.agrformet.2014.01.005http://dx.doi.org/10.1016/j.agrformet.2014.01.005http://dx.doi.org/10.1641/0006-3568(2001)051[0723:CCAFD]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2001)051[0723:CCAFD]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2001)051[0723:CCAFD]2.0.CO;2http://dx.doi.org/10.1890/0012-9658(2007)88[243:BTFEMA]2.0.CO;2http://dx.doi.org/10.1890/0012-9658(2007)88[243:BTFEMA]2.0.CO;2http://dx.doi.org/10.1890/0012-9658(2007)88[243:BTFEMA]2.0.CO;2http://dx.doi.org/10.1890/ES11-00271.1http://dx.doi.org/10.1111/gcb.12026http://dx.doi.org/10.1111/gcb.12026http://dx.doi.org/10.1111/gcb.12026http://dx.doi.org/10.1146/annurev.arplant.48.1.609http://dx.doi.org/10.1146/annurev.arplant.48.1.609http://dx.doi.org/10.1146/annurev.arplant.48.1.609http://dx.doi.org/10.4996/fireecology.0301003http://dx.doi.org/10.4996/fireecology.0301003http://dx.doi.org/10.4996/fireecology.0301003http://dx.doi.org/10.1111/j.1365-2656.2008.01390.xhttp://dx.doi.org/10.1111/j.1365-2656.2008.01390.xhttp://dx.doi.org/10.1111/j.1365-2656.2008.01390.xhttp://dx.doi.org/10.1111/j.2006.0906-7590.04596.xhttp://dx.doi.org/10.1111/j.2006.0906-7590.04596.xhttp://dx.doi.org/10.1111/j.2006.0906-7590.04596.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02393.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02393.xhttp://dx.doi.org/10.1111/j.1365-2486.2010.02393.xhttp://dx.doi.org/10.1016/j.foreco.2005.02.010http://dx.doi.org/10.1016/j.foreco.2005.02.010http://dx.doi.org/10.1016/j.foreco.2005.02.010http://dx.doi.org/10.1071/WF08187http://dx.doi.org/10.1071/WF08187http://dx.doi.org/10.1071/WF08187http://dx.doi.org/10.1016/S0048-9697(00)00524-6http://dx.doi.org/10.1016/S0048-9697(00)00524-6http://dx.doi.org/10.1016/S0048-9697(00)00524-6http://dx.doi.org/10.1023/B:CLIM.0000024667.89579.edhttp://dx.doi.org/10.1023/B:CLIM.0000024667.89579.edhttp://dx.doi.org/10.1023/B:CLIM.0000024667.89579.edhttp://dx.doi.org/10.1029/2011JG001935http://dx.doi.org/10.1029/2011JG001935http://dx.doi.org/10.1073/pnas.1319316111http://dx.doi.org/10.1073/pnas.1319316111http://dx.doi.org/10.1073/pnas.1319316111http://dx.doi.org/10.1890/120282http://dx.doi.org/10.1890/120282http://dx.doi.org/10.1890/120282http://dx.doi.org/10.1111/j.1469-185X.1977.tb01347.xhttp://dx.doi.org/10.1111/j.1469-185X.1977.tb01347.xhttp://dx.doi.org/10.1111/j.1469-185X.1977.tb01347.xhttp://dx.doi.org/10.1007/s10980-015-0218-0http://dx.doi.org/10.1007/s10980-015-0218-0http://dx.doi.org/10.1007/s10980-015-0218-0http://dx.doi.org/10.1177/0309133311407654http://dx.doi.org/10.1177/0309133311407654http://dx.doi.org/10.1177/0309133311407654http://dx.doi.org/10.1111/1365-2745.12296http://dx.doi.org/10.1111/1365-2745.12296http://dx.doi.org/10.1111/1365-2745.12296http://dx.doi.org/10.1073/pnas.0901644106http://dx.doi.org/10.1073/pnas.0901644106http://dx.doi.org/10.1073/pnas.0901644106http://dx.doi.org/10.1038/ncomms8537http://dx.doi.org/10.1016/j.foreco.2014.10.038http://dx.doi.org/10.1016/j.foreco.2014.10.038http://dx.doi.org/10.1016/j.foreco.2014.10.038http://dx.doi.org/10.1038/nature12291http://dx.doi.org/10.1038/nature12291http://dx.doi.org/10.1038/nature12291http://dx.doi.org/10.1073/pnas.0802891105http://dx.doi.org/10.1073/pnas.0802891105http://dx.doi.org/10.1073/pnas.0802891105http://dx.doi.org/10.1073/pnas.0802891105http://dx.doi.org/10.1073/pnas.1305069110http://dx.doi.org/10.1073/pnas.1305069110http://dx.doi.org/10.1073/pnas.1305069110http://dx.doi.org/10.4996/fireecology.0202034http://dx.doi.org/10.4996/fireecology.0202034http://dx.doi.org/10.4996/fireecology.0202034http://dx.doi.org/10.1371/journal.pone.0005102http://dx.doi.org/10.1890/07-0480.1http://dx.doi.org/10.1890/07-0480.1http://dx.doi.org/10.1890/07-0480.1

Littell J S andGwozdz RB 2011Climatic water balance in regionalfire years in the PacificNorthwest, USA: linking regionalclimate and fire and at landscape scalesThe Landscape Ecologyof Fire edDMcKenzie et al (Netherlands: Springer) vol 213,p 177-139

Littell JS,OneilEE,McKenzieD,HickeJA,LutzJA,NorheimRAandElsnerMM2010Forestecosystems,disturbance,andclimaticchangeinWashingtonState,USAClim.Change10212958

Marlon J R, Bartlein P J, Carcaillet C,GavinDG,Harrison S P,Higuera P E, Joos F, PowerM J and Prentice I C 2008Climateand human influences on global biomass burning over thepast twomillenniaNat. Geosci. 1 697702

Marlon J R et al 2009Wildfire responses to abrupt climate change inNorthAmerica Proc. Natl Acad. Sci. USA 106 251924

Marlon J R et al 2012 Long-termperspective onwildfires in thewesternUSA Proc. Natl Acad. Sci. USA 109E53543

Millar C I, StephensonNL and Stephens S L 2007Climate changeand forests of the future:managing in the face of uncertaintyEcol. Appl. 17 214551

Miller C andAger AA2013A review of recent advances in riskanalysis for wildfiremanagement Int. J.Wildland Fire 22 114

Miller C andUrbanDL 1999Amodel of surfacefire, climate andforest pattern in the SierraNevada, California Ecol.Modelling114 113

Miller J D, SaffordHD,CrimminsMandThodeAE 2009Quantitative evidence for increasing forestfire severity in theSierraNevada and Southern cascademountains, CaliforniaandNevada, USAEcosystems 12 1632

Miller J D andThodeAE 2007Quantifying burn severity in aheterogeneous landscapewith a relative version of the deltanormalized burn ratio (dNBR)Remote Sens. Environ. 1096680

Morgan J A,MilchunasDG, LeCainDR,WestM andMosier AR2007Carbon dioxide enrichment alters plant communitystructure and accelerates shrub growth in the shortgrasssteppeProc. Natl Acad. Sci.USA 104 147249

Morgan P,HardyCC, SwetnamTW,RollinsMGand LongDG2001Mapping fire regimes across time and space:understanding coarse and fine-scale fire patterns Int. J.Wildland Fire 10 32942

Morgan P, Keane RE,DillonGK, Jain TB,HudakAT,Karau EC,Sikkink PG,Holden ZA and Strand EK 2014Challenges ofassessing fire and burn severity usingfieldmeasures, remotesensing andmodelling Int. J.Wildland Fire 23 104560

MoritzMA, ParisienM-A, Batllori E, KrawchukMA,VanDorn J,GanzD J andHayhoeK 2012Climate change and disruptionsto global fire activity Ecosphere 3 49

NitschkeCR and Innes J L 2008Climatic change and fire potentialin South-Central British Columbia, CanadaGlob. ChangeBiol. 14 84155

NorbyR J andZakDR2011 Ecological lessons from free-air CO2enrichment (FACE) experimentsAnnu. Rev. Ecol., Evol.Systematics 42 181203

NorthM, BroughA, Long J, Collins B, Bowden P, YasudaD,Miller J and SugiharaN2014Constraints onmechanizedtreatment significantly limitmechanical fuels reductionextent in the SierraNevada J. Forestry 113 408

NorthM,Collins BMand Stephens S 2012Using fire to increase thescale, benefits, and futuremaintenance of fuels treatmentsJ. Forestry 110 392401

NorthM,HurteauMand Innes J 2009 Fire suppression and fuelstreatment effects onmixed-conifer carbon stocks andemissionsEcol. Appl. 19 138596

NorthMP, Stephens S L, Collins BM,Agee J K, Aplet G,Franklin J F and Ful PZ 2015Reform forestfiremanagement Science 349 12801

Overpeck J T, RindD andGoldberg R 1990Climate-inducedchanges in forest disturbance and vegetationNature 343 513

Parks SA,DillonGKandMiller C 2014aAnewmetric forquantifying burn severity: the relativized burn ratioRemoteSens. 6 182744

Parks SA,Holsinger LM,Miller C andNelsonCR 2015aWildlandfire as a self-regulatingmechanism: the role of previous burnsandweather in limiting fire progression Ecol. Appl. 25147892

Parks SA,Miller C, ParisienM-A,Holsinger L,Dobrowski S Z andAbatzoglou J T 2015bWildland fire deficit and surplus of thewesternUnited States, 19842012Ecosphere 6 275

Parks SA, ParisienM-A,Miller C andDobrowski S Z 2014b Fireactivity and severity in thewesternUS vary along proxygradients representing fuel amount and fuelmoisture PlosOne 9 e99699

RDevelopment Core Team2007R: A Language and Environment forStatistical computing (Vienna, Austria: R foundation forcomputing)

Schoennagel T, VeblenTT andRommeWH2004The interactionoffire, fuels, and climate across rockymountain forestsBioscience 54 66176

SiegenthalerU et al 2005 Stable carbon cycle-climate relationshipduring the late Pleistocene Science 310 13137

Smucker KM,Hutto R L and Steele BM2005Changes in birdabundance after wildfire: Importance offire severity and timesincefireEcol. Appl. 15 153549

Sprugel DG1991Disturbance, equilibrium, and environmentalvariability: what is Natural vegetation in a changingenvironment?Biol. Conservation 58 118

Stavros EN, Abatzoglou J,McKenzieD and LarkinN2014Regionalprojections of the likelihood of very largewildland fires undera changing climate in the contiguousWesternUnited StatesClim. Change 126 45568

Stegen J C, SwensonNG, Enquist B J,White E P, PhillipsO L,Jorgensen PM,WeiserMD,Mendoza AMandVargas PN2011Variation in above-ground forest biomass across broadclimatic gradientsGlob. Ecol. Biogeogr. 20 74454

Stephens S L, Agee J K, Fule PZ,NorthMP,RommeWH,SwetnamTWandTurnerMG2013Managing forests andfire in changing climates Science 342 412

Stephens S L,Millar C I andCollins BM2010Operationalapproaches tomanaging forests of the future inMediterranean regionswithin a context of changing climatesEnviron. Res. Lett. 5 024003

Stoner AMK,HayhoeK andWuebbles D J 2009Assessing generalcirculationmodel simulations of atmospheric teleconnectionpatterns J. Clim. 22 434872

Svenning J-C and Sandel B 2013Disequilibrium vegetationdynamics under future climate changeAm. J. Bot. 100126686

TheNature Conservancy 2009TNCTerrestrial Ecoregions,Arlington, VA,USA (http://maps.tnc.org/files/metadata/TerrEcos.xml)

ThonickeK, Venevsky S, Sitch S andCramerW2001The role offiredisturbance for global vegetation dynamics: coupling fire intoa dynamic global vegetationmodelGlob. Ecol. Biogeogr. 1066177

TurnerMG2010Disturbance and landscape dynamics in achangingworldEcology 91 283349

vanWagtendonk JW, Root RR andKeyCH2004Comparison ofAVIRIS and Landsat ETM+ detection capabilities for burnseverityRemote Sens. Environ. 92 397408

Westerling AL,HidalgoHG, CayanDRand SwetnamTW2006Warming and earlier spring increase westernUS forestwildfire activity Science 313 9403

Whitman E, Batllori E, ParisienM-A,Miller C, Coop JD,KrawchukMA,ChongGWandHaire S 2015The climatespace of fire regimes in north-westernNorth AmericaJ. Biogeogr. 42 173649

WilliamsAP et al 2013Temperature as a potent driver of regionalforest drought stress and treemortalityNat. Clim. Change 32927

ZhaoMandRunning SW2010Drought-induced reduction inglobal terrestrial net primary production from2000 through2009 Science 329 9403

10

Environ. Res. Lett. 11 (2016) 035002

http://dx.doi.org/10.1007/s10584-010-9858-xhttp://dx.doi.org/10.1007/s10584-010-9858-xhttp://dx.doi.org/10.1007/s10584-010-9858-xhttp://dx.doi.org/10.1038/ngeo313http://dx.doi.org/10.1038/ngeo313http://dx.doi.org/10.1038/ngeo313http://dx.doi.org/10.1073/pnas.0808212106http://dx.doi.org/10.1073/pnas.0808212106http://dx.doi.org/10.1073/pnas.0808212106http://dx.doi.org/10.1073/pnas.1112839109http://dx.doi.org/10.1073/pnas.1112839109http://dx.doi.org/10.1073/pnas.1112839109http://dx.doi.org/10.1890/06-1715.1http://dx.doi.org/10.1890/06-1715.1http://dx.doi.org/10.1890/06-1715.1http://dx.doi.org/10.1071/WF11114http://dx.doi.org/10.1071/WF11114http://dx.doi.org/10.1071/WF11114http://dx.doi.org/10.1016/S0304-3800(98)00119-7http://dx.doi.org/10.1007/s10021-008-9201-9http://dx.doi.org/10.1007/s10021-008-9201-9http://dx.doi.org/10.1007/s10021-008-9201-9http://dx.doi.org/10.1016/j.rse.2006.12.006http://dx.doi.org/10.1016/j.rse.2006.12.006http://dx.doi.org/10.1016/j.rse.2006.12.006http://dx.doi.org/10.1016/j.rse.2006.12.006http://dx.doi.org/10.1073/pnas.0703427104http://dx.doi.org/10.1073/pnas.0703427104http://dx.doi.org/10.1073/pnas.0703427104http://dx.doi.org/10.1071/WF01032http://dx.doi.org/10.1071/WF01032http://dx.doi.org/10.1071/WF01032http://dx.doi.org/10.1071/WF13058http://dx.doi.org/10.1071/WF13058http://dx.doi.org/10.1071/WF13058http://dx.doi.org/10.1890/ES11-00345.1http://dx.doi.org/10.1111/j.1365-2486.2007.01517.xhttp://dx.doi.org/10.1111/j.1365-2486.2007.01517.xhttp://dx.doi.org/10.1111/j.1365-2486.2007.01517.xhttp://dx.doi.org/10.1146/annurev-ecolsys-102209-144647http://dx.doi.org/10.1146/annurev-ecolsys-102209-144647http://dx.doi.org/10.1146/annurev-ecolsys-102209-144647http://dx.doi.org/10.5849/jof.14-058http://dx.doi.org/10.5849/jof.14-058http://dx.doi.org/10.5849/jof.14-058http://dx.doi.org/10.5849/jof.12-021http://dx.doi.org/10.5849/jof.12-021http://dx.doi.org/10.5849/jof.12-021http://dx.doi.org/10.1890/08-1173.1http://dx.doi.org/10.1890/08-1173.1http://dx.doi.org/10.1890/08-1173.1http://dx.doi.org/10.1126/science.aab2356http://dx.doi.org/10.1126/science.aab2356http://dx.doi.org/10.1126/science.aab2356http://dx.doi.org/10.1038/343051a0http://dx.doi.org/10.1038/343051a0http://dx.doi.org/10.1038/343051a0http://dx.doi.org/10.3390/rs6031827http://dx.doi.org/10.3390/rs6031827http://dx.doi.org/10.3390/rs6031827http://dx.doi.org/10.1890/14-1430.1http://dx.doi.org/10.1890/14-1430.1http://dx.doi.org/10.1890/14-1430.1http://dx.doi.org/10.1890/14-1430.1http://dx.doi.org/10.1890/ES15-00294.1http://dx.doi.org/10.1371/journal.pone.0099699http://dx.doi.org/10.1641/0006-3568(2004)054[0661:TIOFFA]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2004)054[0661:TIOFFA]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2004)054[0661:TIOFFA]2.0.CO;2http://dx.doi.org/10.1126/science.1120130http://dx.doi.org/10.1126/science.1120130http://dx.doi.org/10.1126/science.1120130http://dx.doi.org/10.1890/04-1353http://dx.doi.org/10.1890/04-1353http://dx.doi.org/10.1890/04-1353http://dx.doi.org/10.1016/0006-3207(91)90041-7http://dx.doi.org/10.1016/0006-3207(91)90041-7http://dx.doi.org/10.1016/0006-3207(91)90041-7http://dx.doi.org/10.1007/s10584-014-1229-6http://dx.doi.org/10.1007/s10584-014-1229-6http://dx.doi.org/10.1007/s10584-014-1229-6http://dx.doi.org/10.1111/j.1466-8238.2010.00645.xhttp://dx.doi.org/10.1111/j.1466-8238.2010.00645.xhttp://dx.doi.org/10.1111/j.1466-8238.2010.00645.xhttp://dx.doi.org/10.1126/science.1240294http://dx.doi.org/10.1126/science.1240294http://dx.doi.org/10.1126/science.1240294http://dx.doi.org/10.1088/1748-9326/5/2/024003http://dx.doi.org/10.1175/2009JCLI2577.1http://dx.doi.org/10.1175/2009JCLI2577.1http://dx.doi.org/10.1175/2009JCLI2577.1http://dx.doi.org/10.3732/ajb.1200469http://dx.doi.org/10.3732/ajb.1200469http://dx.doi.org/10.3732/ajb.1200469http://dx.doi.org/10.3732/ajb.1200469http://maps.tnc.org/files/metadata/TerrEcos.xmlhttp://maps.tnc.org/files/metadata/TerrEcos.xmlhttp://dx.doi.org/10.1046/j.1466-822X.2001.00175.xhttp://dx.doi.org/10.1046/j.1466-822X.2001.00175.xhttp://dx.doi.org/10.1046/j.1466-822X.2001.00175.xhttp://dx.doi.org/10.1046/j.1466-822X.2001.00175.xhttp://dx.doi.org/10.1890/10-0097.1http://dx.doi.org/10.1890/10-0097.1http://dx.doi.org/10.1890/10-0097.1http://dx.doi.org/10.1016/j.rse.2003.12.015http://dx.doi.org/10.1016/j.rse.2003.12.015http://dx.doi.org/10.1016/j.rse.2003.12.015http://dx.doi.org/10.1126/science.1128834http://dx.doi.org/10.1126/science.1128834http://dx.doi.org/10.1126/science.1128834http://dx.doi.org/10.1111/jbi.12533http://dx.doi.org/10.1111/jbi.12533http://dx.doi.org/10.1111/jbi.12533http://dx.doi.org/10.1038/nclimate1693http://dx.doi.org/10.1038/nclimate1693http://dx.doi.org/10.1038/nclimate1693http://dx.doi.org/10.1038/nclimate1693http://dx.doi.org/10.1126/science.1192666http://dx.doi.org/10.1126/science.1192666http://dx.doi.org/10.1126/science.1192666

IntroductionMethodsResultsDiscussionConclusionsAcknowledgmentsReferences