Forest Ecology and Management 379 (2016) 61–69

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier .com/locate / foreco

Sensitivity of reconstructed fire histories to detection criteriain mixed-severity landscapes

http://dx.doi.org/10.1016/j.foreco.2016.08.0090378-1127/� 2016 Published by Elsevier B.V.

⇑ Corresponding author at: Climate & Ecosystem Dynamics Research Laboratory,Geography Department, University of Guelph, 50 Stone Road E, Guelph, ON N1G2W1, Canada.

E-mail address: vanessa.stretch@gmail.com (V. Stretch).

Vanessa Stretch a,⇑, Ze’ev Gedalof a, Jaclyn Cockburn a, Michael F.J. Pisaric b

aGeography, University of Guelph, 50 Stone Road E, Guelph, ON N1G 2W1, CanadabGeography, Brock University, 1812 Sir Isaac Brock Way, St. Catharines, ON L2S 3A1, Canada

a r t i c l e i n f o a b s t r a c t

Article history:Received 29 February 2016Received in revised form 3 August 2016Accepted 5 August 2016

Keywords:DendrochronologyFire scarStand establishmentEvent detectionRegime classificationRocky Mountains

In heterogeneous forest landscapes prone to wildfires, accurate classification of the fire regime beyonddirect observations and records is difficult. This is in part due to the methods used to reconstructhistorical fires in complex, heterogeneous landscapes with varying fire severities. Mixed-severity fireregimes, defined as variations in wildfire severity over time and/or space, have important implicationsfor ecosystem functioning and forest management. Fire event detection is used to reduce uncertaintiesin historical tree-ring proxy records. The number of fire events considered in fire regime classificationvaries based on detection criteria (filters) that are researcher-selected, such as number of trees or plotsrecording fire. Here we analyzed the sensitivity of fire regime classification to common detection criteriain a mixed-severity fire regime in the eastern foothills of the Rocky Mountains, Alberta, Canada. We foundthat detection criteria bias records toward high-severity events and against potentially ecologicallysignificant low-severity fires in mixed-severity regimes, ultimately classifying them as higher severity.We conclude that detection criteria methods must address the scale that is relevant to the ecologicalor management questions being addressed.

� 2016 Published by Elsevier B.V.

1. Introduction

Mixed-severity fire regimes represent variability in the ecolog-ical effects of fire on forest ecosystems over both space and time. Asingle fire event may result in varying sized patches of high-,moderate-, or low-severity effects with a range of tree mortality,if there is high variability in microsite (plot-level) topography(Moritz et al., 2011; Perry et al., 2011; Hessburg et al., 2016).Mixed-severity fire regimes can also be described at a range ofspatial scales, from the site- or stand-level up to the regional- orlandscape-level (usually desired for management) (Peterson,1998). Fire can also be mixed in severity over time, depending onfire weather, time-since-last-fire, and rates of fine fuel production(Agee, 1993, 2003; Arno et al., 1995; Perry et al., 2011; Hessburget al., 2016). For example, high-severity fires may be interspersedwith low- or moderate-severity events. Together the variabilityof severities over both space and time make understanding anddetecting mixed-severity fire regimes difficult.

A variety of evidence is retained from differing fire events overspace and time, thus complicating the detection and classificationof mixed-severity fire regimes. For example, at the stand-level,low-frequency, high-severity fire often kills the majority of vegeta-tion (Turner and Romme, 1994; Peterson, 1998; Meyn et al., 2007).Subsequent regeneration in the years following fire may result ineven-aged tree cohorts and previous fire evidence may be lost.Reconstructing these events is complicated by the criteria usedto identify post-fire cohort dates. Post-fire cohort detection isresearcher-defined, usually with consideration of species time-to-coring height. However, this detection is complicated furthermoreby the uncertainty in the cause of establishment (e.g. another formof high-severity disturbance, such as an insect outbreak). Higherfrequency, low-severity fires kill understorey vegetation and/orleave fire scars on mature trees (McBride, 1983; Gutsell andJohnson, 1996). Reconstructing these regimes is complicated bydifferential preservation of fire scars by species, as well asminimum numbers of replicate scars found on other trees. Thesechallenges are typically addressed by using post-sampling fireevent detection criteria (filters) to increase certainty that theseevents represent fire activity.

In areas where fire regimes are well understood, targetsampling for fire evidence of known severities results in successfulfire regime reconstructions (Farris et al., 2010, 2013). For example,

62 V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69

in ponderosa pine (Pinus ponderosa) forests in the western US, firescarred trees are purposely selected for analysis (Baker and Ehle,2001). It was well established that this region has been character-ized by low-severity, surface fire regimes. However, in understud-ied areas where an assumption of the fire regime is made (i.e.assuming a regime based on forest type or a neighbouring regime),sampling for a range of unknown fire evidence may result in moreaccurate fire regime representation. For example, North America’sboreal forests, which were historically believed to be representedby a single low-frequency, high-severity fire regime resulting in amosaic of even-aged stands, is now understood as a spatiallydynamic fire regime (Bergeron et al., 2001). Individual high-severity fires can also result in complex mosaics at different spatialscales (Margolis et al., 2007, 2011; Margolis and Swetnam, 2013;O’Connor et al., 2014). Sampling solely for high-severity fire evi-dence (i.e. stand establishment) in boreal forests would result ina simplification of the fire regime. For example, Marcoux et al.(2015) found evidence of a high-severity fire regime in southeast-ern British Columbia, Canada that resulted from 20th-century sur-face fire suppression and harvesting that homogenized foreststructure. However, the authors found evidence of a historicallymixed-severity fire regime in this location. We hypothesize thattargeted sampling areas of a single, known severity may bias theregime classification to either end of the classification continuumin mixed-severity regimes (Heinselman, 1973; Agee, 1993;Swetnam et al., 1999; Van Wagner et al., 2006).

Canadian boreal forests have historically been managed withthe assumption of infrequent, high-severity, stand-replacing firesbeing the dominant disturbance agent (Johnson, 1992). However,Amoroso et al. (2011) found evidence for mixed-severity fireregimes in the Rocky Mountain Foothills, on the western edge ofCanada’s boreal forest. A probable explanation for this regime isthe degree of topographic variability that may result in mixed-severity fire effects, as both even-aged stands and fire scars werefound (Amoroso et al., 2011). Similarly, recent studies found evi-dence of mixed-severity fire in the montane cordillera directlyadjacent to the Foothills (Chavardes and Daniels, 2016; Marcouxet al., 2013, 2015). There is a need for a broader search formixed-severity fire regimes in these areas.

Here, we present a case study from the eastern foothills of theRocky Mountains, Alberta, Canada, and test the sensitivity of com-mon fire regime classification methods to mixed-severity regimes.Our objectives were: (1) to incorporate high- and low-severity fireevidence to reconstruct fire regimes in a complex region withlittle-to-no known fire histories in order to classify the fire regimeand (2) to analyze the impact of post-sampling detection criteriaon the final fire regime classification, in order to ensure the ecolog-ical integrity of forest management.

2. Methods

2.1. Case study area

In this case study we analyze two sites (ZS and PC) in the north-ern portion of the Hinton Wood Products Forest Management Area(HWP FMA) in the eastern foothills of the Rocky Mountains nearHinton, Alberta, Canada (Fig. 1). HWP manages approximatelyone million hectares of land extending east from the Rocky Moun-tains into the boreal plains ecozone. Here the eastern foothills areconsidered part of the Alberta upland ecoregion (EcologicalStratification Working Group, 1996), which is classified as thewestern edge of the boreal forest in Canada. The forests aredominated by lodgepole pine (Pinus contorta) and black spruce(Picea mariana), with lesser amounts of white spruce (Picea glauca),

trembling aspen (Populus tremuloides), balsam fir (Abies balsamea),and western larch (Larix occidentalis).

The HWP FMA has had minimal fire activity since the 1990s andis therefore classified as extreme to high risk for wildfire in currentmanagement planning. It is believed that about 44,000 ha (<5% ofthe managed area) have burned since the 1930s, with most eventsbeing small (<1–10 ha) in size (Hinton Wood Products, 2010).Harvesting (pulp and timber extraction) activities have been ongo-ing since 1956, with a current sustainable harvesting managementplan in place that involves mimicking the natural wildfire distur-bance regime by using variable cutblock sizes and shapes andretaining remnant stands, as well as burning harvest wastein situ. Prior to 1956, there was minimal known activity in thearea; it was used mostly as hunting and camping grounds fornearby settlements.

2.2. Site and sampling selection

Using a combination of random and condition-based samplingtechniques, we reconstructed wildfire activity at two sites.Working in a large potential sampling area (almost one millionhectares), and with the hopes of incorporating multiple lines of fireevidence, we surveyed the landscape at more than 30 sites lookingfor signs of historical evidence of fire occurrence since at least themid-1900s to suggest mixed-severity before selecting our final twostudy sites. These sites were selected because they are of interestto the forest management agency and have a complex harvestinghistories, are well-forested and accessible, and are topographicallydiverse. These characteristics would be sufficient to preserveevidence of mixed-severity fire should it exist here.

To select specific plots for sampling within each study site(�10,000 ha), we used a condition-based selection approach. Tofacilitate a multi-century reconstruction, we used Alberta Vegeta-tion Index (AVI) stand age polygons to eliminate stands less than100 years old (Environment and Sustainable ResourceDevelopment, 2008). By excluding these stands, we are not biasingthe record against recent high-severity fires; most of the excludedstands were harvested, and large, stand-replacing fires were rela-tively rare during the past 100 years (Hinton Wood Products,2010). We then randomly selected six accessible stand polygonsfrom each cardinal direction (24 plots at each site), in order toassure sampling was dispersed across the site (Fig. 1). Samplingplots were established in the centre of each polygon and perma-nently staked (plot sizes ranged from 0.0019 to 0.0716 ha).

2.3. Tree-ring data collection and analysis

To capture the high-severity fire record, canopy and subcanopytrees (if present) were sampled for tree establishment estimation(Barrett and Arno, 1988; Baker and Ehle, 2001; Heyerdahl et al.,2012). We used an n-tree sampling design that resulted in variableplot sizes. This was done by coring to pith within the lower 30 cm(basal coring) of the 10 closest canopy and 10 closest subcanopytrees (with a DBH > 5 cm) to plot centre.

Because post-fire cohorts on their own do not represent all fireactivity (Lorimer, 1985), and to capture the low-severity firerecord, cross-sectional and partial cross-sectional sampling wascarried out at sites where visual fire evidence existed (McBride,1983; Lorimer, 1985; Rist et al., 2011). Fire-scarred trees withinthe plot were sampled (identified by charring, catfaces, and basaltriangular scars); as well, downed wood was cut into to examinefor evidence of buried fire scars and sampled if appropriate(Agee, 1993, 1998; Swetnam et al., 1999; Niklasson andDrakenberg, 2001). To reduce sampling redundancy, at most fivesamples were selected from both dead and live trees.

Fig. 1. Study area, Eastern Foothills of the Rocky Mountains, Alberta, Canada. (A) Hinton Wood Products Forest Management Area (HWP FMA) outlined in white. (B) Twostudy sites north of Hinton. Black dots represent sampling plots at each site. Shading represents differences in elevation across the region. DEM source: Shuttle RadarTopography Mission, NASA.

V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69 63

All samples were returned to the laboratory and processedusing standard dendrochronological techniques (Stokes andSmiley, 1996). Samples were visually crossdated, measured inWinDendro (v. 2008a) (Regent Instruments Inc., 2009), and thecrossdating was confirmed using COFECHA (v. 6.04P) (Grissino-Mayer, 2001a). Crossdating of dead material was accomplishedusing COFECHA and suggested dates were then confirmed usingvisual crossdating techniques. We excluded any scars that wethought were fire upon sampling but were uncertain were fire inthe laboratory (lacked a distinctive charred scar and tip). Theremaining fire scars were assigned to event years from the cross-dated chronology. Geometric pith estimation techniques wereused on any tree-ring samples with missing piths (Applequist,1958; Duncan, 1989). Because samples were taken within thelower 30 cm of the trunk-ground interface (Barrett and Arno,1988; Amoroso et al., 2011), and with consideration of the timeto coring height of lodgepole pine and black spruce, no correctionfactors were applied to pith dates (Johnson and Fryer, 1989;Nyland, 1998; Antos and Parish, 2002; Axelson et al., 2009). Firescar dates were then entered into FHAES (Brewer et al., 2014)and FHX2 (Grissino-Mayer, 2001b) for compilation and analysisof mean fire return intervals (mFRIs). High-severity fire, even-aged cohorts were identified by 60% of trees in the canopy or sub-canopy having pith dates within 10 years of each other, with nolive samples older than the earliest cohort date. No exact fire yearsfor high-severity events were assigned in this analysis, but ratherthe number of likely events was tabulated.

2.4. Fire regime classification

In order to classify the fire regime at a regional scale, weanalyzed evidence from each plot to classify the two study sites.

This allowed us to evaluate spatial patterns in classification withineach site. We classified each plot based on the severity evidencefound. Low-severity fire evidence was tabulated from the fire scarrecord, and high-severity estimates were calculated based oncohort establishment detection at each plot.

Different filters for detection of low-severity fires were ana-lyzed to evaluate their effect on regime classifications. The filtersassessed here included all fire scar evidence at all plots (non-filtered), a minimum of two trees recording fire at each plot, and10% of trees recording fire at each plot. Both minimum andpercentage filters were assessed to analyze differences in eventdetection stemming from varying sample sizes at each plot. Filterswere also spatially scaled up to include all plots recording any firescar event and minimum two plots recording an event, as well asacross sites (regional scale).

To examine the influence of detection criteria on high-severityevents, all individual tree establishment dates at each study sitewere combined to show regional pulses in establishment.Establishment dates were binned at different bin widths to furtherdemonstrate the under- or over-representation of establishmentpulses that may be related to high-severity fire at the site-level.Bins of 5-, 10-, 20-, and 25-years were analyzed here.

3. Results

3.1. Records of low-severity fire

In total, 188 fire scars were dated from 29 plots containing low-severity fire evidence. The fire scar record spans 1808 to present(Fig. 2). Of these, 77% of scars were found at the PC site with theremainder at the ZS site. From these fire scars, 76 fire events wererecorded on at least one tree. Fire events were detected in 71% of

64 V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69

PC plots (n = 54 events); while only 50% of plots in ZS had scars(n = 38 events).

3.1.1. Plot-scale comparisonsTo examine event detection from fire scars within each plot, we

applied two filters (10% of trees at a plot & 2 trees at a plot record-ing fire). We found more fire events were detected when applyinga 10% of trees at a plot recording fire filter (with n ranging from 15to 25 trees) than the 2 trees at a plot filter, with a single fire scarconstituting an event by this criterion (Table 1).

3.1.2. Site-scale comparisonsTo detect potential larger-scale events, fire scar events that

were recorded at two or more plots were tabulated. Thirty-four

Fig. 2. Fire demography diagram. Chronology lengths are shown as solid black lines. Phistogram at the bottom represents regional fire scar frequency (synchrony).

events were recorded at two or more plots across sites, with 25in PC plots and four in ZS plots (Table 1).

3.1.3. Cross-scale comparisonsCombining the large-scale and small-scale detection filters, 27

fire scar events across both sites were detected at two plots, with10% of trees at each plot recording fire (24 events in the PC siteand one in the ZS site). Four events were recorded at two or moreplots across sites with each plot having at least two trees recordingthe event (all of these events were in the PC site) (Table 1).

We calculated mFRIs for each of our filters to analyze the effectof detection filters (Reed and Johnson, 2004). When calculatingcomposite mFRIs for sites (all plots at a site) and the region (allplots across both sites), we averaged the number of years betweenfire events. Scaling up with this method can decrease the overall

lus signs (+) indicate fire scars and dark triangles represent cohort initiation. The

Table 1Number of fire events and mean fire return intervals (mFRIs) calculated for different event detection filters.

Filter PC # events mFRI (years) ZS # events mFRI (years) Regional # events mFRI (years)

All trees at plots 54 2.2 38 5.2 76 2.610% trees/plot 49 2.4 14 7.2 56 2.12 trees/plot 17 4.0 0 N/A 17 4.02 plots 25 2.2 4 17.3 34 2.62 plots & 10% trees 24 2.3 1 N/A 27 2.02 plots & 2 trees 4 8.7 0 N/A 4 8.7

Note: Site and regional mFRIs are calculated as averages of composite intervals from relevant plots (Agee, 1993). mFRIs here are biased by the different sample sizes (numberof plots) and areas of each filter. Fire scar records span 1808 to present, with the pith to initial scar and most recent fire-free period excluded in the mFRI. The PC and ZSrecords are different lengths.

V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69 65

mFRI at larger scales, however our purpose was to analyze theinfluence of all individual fire events on the fire regime and denotehow often any fire activity was occurring anywhere on the land-scape (Agee, 1993). The mFRI for at least one plot detecting a fireevent at site PC was 2.2 years and 5.2 years at site ZS. Whenfiltering to at least two plots recording, mFRIs at PC remainedthe same (2.2 years), but increased at ZS to 17.3 years (see Table 1for all calculated mFRIs). Overall, mFRIs do not change substan-tially with different filters in this landscape as fire return intervalsare variable (Fig. 3), but the number of detected fire events variesgreatly.

3.2. Stand establishment

From the groupings of establishment (Fig. 4), multiple pulsesare evident at smaller bin sizes, which become masked as bin sizeincreases. Larger bin sizes show two main establishment pulses atthe regional-level, one at the end of the 1800s that is predomi-nantly lodgepole pine and another beginning in the 1920s that ispredominantly black spruce. These patterns are seen at both sites,demonstrating larger-scale pulses in establishment.

Canopy cohorts were detected at 14 plots (n = 24, 58%) at the PCsite and at nine plots (n = 24, 38%) at the ZS site. Subcanopycohorts were detected at four plots (n = 9, 44%) and three plots

(A) (B)

Fig. 3. Variability in mFRIs based on different detec

(n = 11, 27%), respectively (Fig. 2 and Table 2). For those plots withsubcanopies, significant differences between canopy and sub-canopy ages were detected at 18 plots (n = 20, 90%).

3.3. Fire regime classifications

Based on the inclusion of all fire evidence, the number of plotsexhibiting different fire regime evidence was tabulated in order toconceptualize fire regimes across the landscape, as well as at thesite-level (Table 2). The majority of fire evidence was low-severity (fire scars), found at 29 plots (60%). Overall, cohort detec-tion at plots revealed evidence of high-severity fire at 26 plots(54%), with moderate-severity evidence inferred from a subcanopycohort at seven plots (15%). Overall, 44% of plots had evidence ofmixed-severity fire events, resulting in an overall mixed-severityfire regime classification. The remaining landscape is classified as21% low-severity fire regime and 13% high-severity fire regime(with the remaining 23% of the landscape showing no fire evidenceover the length of the record – approximately the last 150 years).More mixed-severity plot-level classifications were found at PC(n = 14, 58%) than at ZS (n = 7, 29%).

However, when applying detection criteria/filters within plots,plot-level fire regime classification substantially changes (Fig. 5and Table 3). Low-severity regimes decrease in number when

tion criteria/filters. (A) ZS site and (B) PC site.

(A) (B)

Fig. 4. Regional binned stand establishment estimated dates. (A) ZS site and (B) PC site. Establishment dates span 1735–1993 (n = 745).

Table 2Tabulation of fire severity evidence for fire regime classification.

Fire evidence # PC Plots # ZS Plots # Regional plots

Low-severity 16 13 29High-severity canopy 14 9 23High-severity subcanopy 4 3 7High-severity subcanopy only 0 0 0High-severity & low-severity 14 6 20Low-severity only 3 7 10High-severity canopy only 3 4 7

66 V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69

10% of trees at a plot and 2 trees per plot filters are applied (16 to10 and two, respectively). Filters also decrease the amount of over-all fire evidence by classifying some plots as having no evidence offire.

4. Discussion

4.1. Fire histories from multiple lines of evidence

Analyzing the stand establishment data, we found more plot-level high-severity fire evidence in site PC than in site ZS. Site PCis characterized as a drier area based on vegetation compositionand site characteristics (e.g. higher-elevation, plateau). We alsodetected subcanopy cohorts at the plot-level, which may indicatemoderate- or mixed-severity fire. When analyzing the site-levelhigh-severity fire evidence, we found two pulses in establishment,one in the late 1800s and one in the 1920s (Fig. 4). However, whenthe data are filtered into larger establishment bin widths, the

pulses merge into a single pulse at 25-year bins. Binning at differ-ent widths can thus overlook potential small, higher-severity fireevents happening at the site- or plot-scale. Overall, we found thatthe study sites experience high-severity fire at both the plot- andsite-level, but have not experienced much high-severity fire sinceharvesting activities began (in the mid 1950s).

Our fire scar data demonstrates that this region experiencesvariable, but frequent low-severity fire as well (Fig. 2 and Table 1).Site PC, which is located on a higher-elevation plateau, experienceslow-severity fire more frequently than site ZS, based on mFRIs(Table 1). We also found that due to the high number of low-severity fire events, when applying detection criteria filters, includ-ing or excluding a few events does not alter the mFRI significantlybut biases it toward more infrequent fire (Table 1). The largest dis-crepancy in mFRI occurs at ZS, which has overall less fire evidence(across all severities). Applying spatial filters for event detectionreduces the number of fire events, ultimately biasing the mFRIstoward a more infrequent fire regime, as well.

An important implication of our event detection analysis is thatwe are confident that all events indicate a fire that affected at leastone tree (scar) or stand (cohort). In the case of scars, we werecareful to exclude scars that were not unambiguously fire (e.g.mechanical or insect scars) (Gutsell and Johnson, 1996), meaningthat from an ecological perspective they represent a disturbanceevent that creates structural complexity. Many traditionalapproaches to fire reconstructions have focused on events that scarmultiple trees, and exclude these single scar events, which biasregime classification toward more infrequent fire. However, froma conservation perspective, they are potentially quite ecologicallyimportant. For example, changes in fire activity at multiple scales

Fig. 5. Fire event detection filters & classification of fire regime. (A) All events at allplots filter, (B) two trees per plot filter, and (C) 10% of trees per plot filter. Greenplots have no detected fire events, yellow plots have low-severity regimes, red plotshave high-severity regimes, pink plots have a subcanopy cohort (moderate-severityfire) only, and orange plots have mixed-severity regimes. (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web version ofthis article.)

Table 3Tabulation of plot-level fire regime classification based on fire event detection filters.

Filter Classification # PCplots

# ZSplots

# Regionalplots

All events at allplots

No fire evidence 5 5 10High-severity 3 5 8Low-severity 5 11 16Mixed-severity 11 3 14High-severity(subcanopy)

0 0 0

Two trees perplot

No fire evidence 8 12 20High-severity 5 9 14Low-severity 2 0 2Mixed-severity 9 0 9High-severity(subcanopy)

0 3 3

10% of trees perplot

No fire evidence 5 7 12High-severity 4 8 12Low-severity 5 5 10Mixed-severity 10 1 11High-severity(subcanopy)

0 3 3

V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69 67

can result in changes in biodiversity by altering species composi-tion, age distributions, and habitat (Odion and Sarr, 2007). There-fore, it is imperative to select criteria that represent fire at arelevant ecological or management scale.

4.2. Classifying fire regimes

Senici et al. (2013) discuss how spatial or landscape differencescan impact fire frequencies, creating different regimes at local spa-tial scales. Here, we found that applying common event detection

filters resulted in different plot-level fire regime classifications(Fig. 5 and Table 3). As a whole, this area, as well as each site, isclassified as mixed-severity based on multiple lines of fire evi-dence, but at the plot-level there is high spatial variability in theregime, which is further complicated by researcher-selected fireevent detection. As can be seen in Fig. 5 and Table 3, applying com-mon event detection filters biases fire regimes towards high-severity. This is particularly troubling when we have collectedmultiple lines of mixed-severity fire evidence.

Another important consideration for fire regime classification isthe non-static nature of fire regimes over time (Lertzman et al.,1998; Beaty and Taylor, 2001; Carcaillet et al., 2001). For example,a high-severity, stand-replacing fire regime may experience inter-mittent small, low-severity, surface fires; this ‘new’ regime mayreturn to the stand-replacing regime after the next stand-replacing fire. Here we found that while no large, landscape-levelfires have been present in this region since at least the mid-1800s (Fig. 4), many small, low-severity fires have occurred since.We also found subcanopy trees established in cohorts, potentiallyindicating moderate-severity events. However, a few hundredyears may pass between high-severity, landscape-scale fire eventsin this boreal landscape. Two temporal regimes operating simulta-neously are possible (Barrett et al., 1991). For example, Masters(1990) found that while a climate-dominated fire regime, charac-terized by infrequent fires, has existed in the Kootenays of south-ern British Columbia, Canada since the early 1500s, a morefrequent regime post-1919 resulting from anthropogenic sources,now co-exists.

The difficulty with the mixed-severity classification is that itcan encompass a range of definitions (Agee, 2005). It can be a staticregime that experiences moderate-severity fires with varyingeffects across space, it can be a landscape that experiences differ-ent severity regimes at the site-/plot-level, or it can be temporallya stand or landscape that moves from low- to moderate- to high-severity regimes over time as relationships with fire controls (i.e.fuels, ignitions, climate, etc.) change (Lertzman et al., 1998; Arnoet al., 2000; Agee, 2005; Amoroso et al., 2011; Halofsky et al.,2011). The latter is most likely in highly topographically diverselandscapes. The region presented here may be a combination ofthese, as we have presented evidence of changes in fire activityand establishment over time and variations in fire evidence andseverity across space. We conclude that in mixed-severity regimessuch as these, defining fire events using stricter criteria biasesresults against low-severity fire in general, and mixed-severity fire

68 V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69

in particular. According to Whitlock et al. (2010), fire reconstruc-tions and regime classifications should involve a reconceptualiza-tion of the fire regime concept as more variable to encompassthe scale of concern. We propose here that fire reconstructions insimilar regions be comprised of multiple lines of evidence and ana-lyzed using different detection criteria in order to assure that theclassification captures the scale that is most appropriate to thequestions being addressed (Arno et al., 2000; Fulé et al., 2003;Agee, 2005; Falk et al., 2007; Conedera et al., 2009; Whitlocket al., 2010; Halofsky et al., 2011). These analyses are crucial forstudies of mixed-severity fire regimes.

5. Conclusions

Historical fire reconstructions in landscapes with unknown firehistories are ultimately difficult. We have demonstrated here howmultiple lines of evidence collected using a condition-based ran-dom sampling of sites and plots can provide a detailed reconstruc-tion. We analyzed the sensitivity of detection criteria to regimeclassification using stand establishment and fire scar data. We con-clude that post-sampling analysis (fire event detection via binningand filtering) is crucial to fire regime classification. mFRIs that arecritical to ecologically understanding and managing these land-scapes can be greatly affected by the selection of fire event detec-tion criteria. Specifically, strict detection criteria bias fire recordsagainst low and mixed-severity events that may be ecologicallymeaningful.

Acknowledgements

This research was funded by a Natural Sciences and EngineeringResearch Council (NSERC) Collaborative Research and Develop-ment Grant with Hinton Wood Products and the Foothills ResearchInstitute (Z. Gedalof), and a NSERC Industrial Postgraduate Scholar-ship and MITACs Accelerate Internship (V. Stretch). We would liketo thank our collaborators D. Andison (Foothills Research Institute)and R. Bonar (Hinton Wood Products), as well as the commentsfrom two anonymous reviewers.

References

Agee, J., 1993. Fire Ecology of Pacific Northwest Forests. Island Press, Washington,DC, pp. 113–136.

Agee, J., 1998. The landscape ecology of western forest fire regimes. Northwest Sci.72 (17), 24–34.

Agee, J.K., 2003. Historical range of variability in eastern Cascades forests,Washington, USA. Landscape Ecol. 18, 725–740.

Agee, J., 2005. The complex nature of mixed severity fire regimes. In: Taylor, L.,Zelnik, J., Cadwallader, S., Hughes, B. (Eds.), Mixed Severity Fire Regimes:Ecology and Management Symposium Proceedings, Spokane, WA, 17–19November 2004. Association of Fire Ecology MISCO3. Washington StateUniversity, Pullman, WA.

Amoroso, M., Daniels, L., Bataineh, M., Andison, D., 2011. Evidence of mixed-severityfires in the foothills of the Rocky Mountains of west-central Alberta, Canada.For. Ecol. Manage. 262 (12), 2240–2249.

Antos, J., Parish, R., 2002. Dynamics of an old-growth, fire-initiated, subalpine forestin southern interior British Columbia: tree size, age, and spatial structure. Can. J.For. Res. 32 (11), 1935–1946.

Applequist, M., 1958. A simple pith locator for use with off-center increment cores.J. Forest. 56 (2), 141.

Arno, S.F., Scott, J.H., Hartwell, M.G., 1995. Age Class Structure of Old GrowthPonderosa Pine/Douglas-Fir Stands and Its Relationship to Fire History. UnitedStates Department of Agriculture, Forest Service Research Paper INT-RP-481.

Arno, S., Parsons, D., Keane, R., 2000. Mixed-severity fire regimes in the northernRocky Mountains: consequences of fire exclusion and options for the future.Wilderness Science in a Time of Change, Conference, vol. 5, pp. 225–232.

Axelson, J., Alfaro, R., Hawkes, B., 2009. Influence of fire and mountain pine beetleon the dynamics of lodgepole pine stands in British Columbia, Canada. For. Ecol.Manage. 257 (9), 1874–1882.

Baker, W., Ehle, D., 2001. Uncertainty in surface-fire history: the case of ponderosapine forests in the western United States. Can. J. For. Res. 31 (7), 1205–1226.

Barrett, S., Arno, S., 1988. Increment-Borer Methods for Determining Fire History inConiferous Forests, General technical report INT-244. USDA Forest Service,Intermountain Research Station, Ogden, UT.

Barrett, S., Arno, S., Key, C., 1991. Fire regimes of western larch-lodgepole pineforests in Glacier National Park, Montana. Canadian Journal of Forest Research21 (12), 1711–1720.

Beaty, R., Taylor, A., 2001. Spatial and temporal variation of fire regimes in a mixedconifer forest landscape, Southern Cascades, California, USA. J. Biogeogr. 28 (8),955–966.

Bergeron, Y., Gauthier, S., Kafka, V., Lefort, P., Lesieur, D., 2001. Natural firefrequency for the eastern Canadian boreal forest: consequence for sustainableforestry. Can. J. For. Res. 31, 384–391.

Brewer, P.W., Velásquez, M.E., Sutherland, E.K., Falk, D.A., 2014. Fire HistoryAnalysis and Exploration System (FHAES) version 2.0.0b, [computer software]<http://www.fhaes.org>.

Carcaillet, C., Bergeron, Y., Richard, P., Fréchette, B., Gauthier, S., Prairie, Y., 2001.Change of fire frequency in the eastern Canadian boreal forests during theHolocene: does vegetation composition or climate trigger the fire regime? J.Ecol. 89 (6), 930–946.

Chavardes, R.D., Daniels, L.D., 2016. Altered mixed-severity fire regime hashomogenised montane forests of Jasper National Park. Int. J. Wildland Fire 25(4), 433–444.

Conedera, M., Tinner, W., Neff, C., Meurer, M., Dickens, A., Krebs, P., 2009.Reconstructing past fire regimes: methods, applications, and relevance to firemanagement and conservation. Quatern. Sci. Rev. 28 (5–6), 555–576.

Duncan, R., 1989. An evaluation of errors in tree age estimates based on incrementcores in kahikatea (Dacrycarpus dacrydioides). New Zealand Nat. Sci. 16 (3), 1–37.

Ecological Stratification Working Group, 1996. A National Ecological Framework forCanada. Agriculture and Agri-Food Canada, Research Branch, Centre for Landand Biological Resources Research and Environment Canada, State of theEnvironment Directorate, Ecozone Analysis Branch, Ottawa/Hull, ON.

Environment and Sustainable Resource Development, 2008. Alberta VegetationIndex. Government of Alberta <http://esrd.alberta.ca/forms-maps-services/maps/resource-data-product-catalogue/forest-vegetation-inventories.aspx>.

Falk, D., Miller, C., McKenzie, D., Black, A., 2007. Cross-scale analysis of fire regimes.Ecosystems 10 (5), 809–823.

Farris, C.A., Baisan, C.H., Falk, D.A., Yool, S.R., Swetnam, T.W., 2010. Spatial andtemporal corroboration of a fire-scar-based fire history in a frequently burnedponderosa pine forest. Ecol. Appl. 20 (6), 1598–1614.

Farris, C.A., Baisan, C.H., Falk, D.A., Van Horne, M.L., Fulé, P.Z., Swetnam, T.W., 2013.A comparison of targeted and systematic fire-scar sampling for estimatinghistorical fire frequency in south-western ponderosa pine forests. Int. J.Wildland Fire 22 (8), 1021–1033.

Fulé, P., Crouse, J., Heinlein, T., Moore, M., Covington, W., Verkamp, G., 2003. Mixed-severity fire regime in a high-elevation forest of Grand Canyon, Arizona, USA.Landscape Ecol. 18 (5), 465–486.

Grissino-Mayer, H., 2001a. Evaluating crossdating accuracy: a manual and tutorialfor the computer program COFECHA. Tree-Ring Res. 57 (2), 205–221.

Grissino-Mayer, H., 2001b. Research report: FHX2 – software for analyzingtemporal and spatial patterns in fire regimes from tree rings. Tree-Ring Res.57 (1), 115–124.

Gutsell, S., Johnson, E., 1996. How fire scars are formed: coupling a disturbanceprocess to its ecological effect. Can. J. For. Res. 26 (2), 166–174.

Halofsky, J., Donato, D., Hibbs, D., Campbell, J., Cannon, M., Fontaine, J., Thompson, J.,Anthony, R., Bormann, B., Kayes, L., Law, B., Peterson, D., Spies, T., 2011. Mixed-severity fire regimes: lessons and hypotheses from the Klamath-SiskiyouEcoregion. Ecosphere 2 (4), 1–19.

Heinselman, M., 1973. Fire in the virgin forests of the Boundary Waters Canoe Area,Minnesota. Quatern. Res. 3 (3), 329–382.

Hessburg, P.F., Spies, T.A., Perry, D.A., Skinner, C.N., Taylor, A.H., Brown, P.M.,Stephens, S.L., Larson, A.J., Churchill, D.J., Povak, N.A., Singleton, P.H., McComb,B., Zielinski, W.J., Collins, B.M., Salter, R.B., Keane, J.J., Franklin, J.F., Riegel, G.,2016. Tamm review: management of mixed-severity fire regime forests inOregon, Washington, and Northern California. For. Ecol. Manage. 366, 221–250.

Heyerdahl, E., Lertzman, K.,Wong, C., 2012.Mixed-severity fire regimes in dry forestsof southern interior British Columbia, Canada. Can. J. For. Res. 42 (1), 88–98.

Hinton Wood Products, 2010. Mountain Pine Beetle Forest Management PlanAmendment: Technical Report 1 – Landbase Classification. Hinton WoodProducts, A division of West Fraser Mills, Forest Management Agreement.

Johnson, E., Fryer, G., 1989. Population dynamics in lodgepole pine-engelmannspruce forests. Ecology 70 (5), 1335–1345.

Johnson, E., 1992. Fire and Vegetation Dynamics: Studies From the North AmericanBoreal Forest. Cambridge University Press, Cambridge, UK.

Lertzman, K., Fall, J., Dorner, B., 1998. Three kinds of heterogeneity in fire regimes:at the crossroads of fire history and landscape ecology. Northwest Sci. 72, 4–23.

Lorimer, C., 1985. Methodological considerations in the analysis of forestdisturbance history. Can. J. For. Res. 15 (1), 200–213.

Marcoux, H.M., Gergel, S.E., Daniels, L.D., 2013. Mixed-severity fire regimes: howwell are they represented by existing fire-regime classification systems? Can. J.For. Res. 43 (7), 658–668.

Marcoux, H.M., Daniels, L.D., Gergel, S.E., Da Silva, E., Gedalof, Z., Hessburg, P.F.,2015. Differentiating mixed- and high-severity fire regimes in mixed-coniferforests of the Canadian Cordillera. For. Ecol. Manage. 341, 45–58.

Margolis, E.Q., Swetnam, T.W., Allen, C.D., 2007. A stand-replacing fire history inupper montane forests of the southern Rocky Mountains. Can. J. For. Res. 37(11), 2227–2241.

V. Stretch et al. / Forest Ecology and Management 379 (2016) 61–69 69

Margolis, E.Q., Swetnam, T.W., Allen, C.D., 2011. Historical stand-replacing fire inupper montane forests of the Madrean Sky Islands and Mogollon Plateau,southwestern USA. Fire Ecol. 7 (3), 88–107.

Margolis, E.Q., Swetnam, T.W., 2013. Historical fire–climate relationships of upperelevation fire regimes in the south-western United States. Int. J. Wildland Fire22 (5), 588–598.

Masters, A., 1990. Changes in forest fire frequency in Kootenay National Park,Canadian Rockies. Can. J. Botany 68 (8), 1763–1767.

McBride, J., 1983. Analysis of tree rings and fire scars to establish fire history. Tree-Ring Bull. 43, 51–67.

Meyn, A., White, P., Buhk, C., Jentsch, A., 2007. Environmental drivers of large,infrequent wildfires: the emerging conceptual model. Prog. Phys. Geogr. 31 (3),287–312.

Moritz, M.A., Hessburg, P.F., Povak, N.A., 2011. Native fie regimes and landscaperesilience. In: McKenzie, D., Falk, D., Miller, C. (Eds.), The Landscape Ecology ofFire. Springer, New York, NY, pp. 52–86.

Niklasson, M., Drakenberg, B., 2001. A 600-year tree-ring fire history from NorraKvills National Park, southern Sweden: implications for conservation strategiesin the hemiboreal zone. Biol. Conserv. 101 (1), 63–71.

Nyland, R., 1998. Patterns of lodgepole pine regeneration following the 1988Yellowstone fires. For. Ecol. Manage. 111 (1), 23–33.

O’Connor, C.D., Falk, D.A., Lynch, A.M., Swetnam, T.W., 2014. Fire severity, size, andclimate associations diverge from historical precedent along an ecologicalgradient of the Pinaleño Mountains, Arizona, USA. For. Ecol. Manage. 329, 264–278.

Odion, D., Sarr, D., 2007. Managing disturbance regimes to maintain biologicaldiversity in forested ecosystems of the Pacific Northwest. For. Ecol. Manage. 246(1), 57–65.

Perry, D.A., Hessburg, P.F., Skinner, C.N., Spies, T.A., Stephens, S.L., Taylor, A.H.,Franklin, J.F., McComb, B., Riegel, G., 2011. Ecology of mixed-severity fireregimes in Washington, Oregon, and California. For. Ecol. Manage. 262, 703–717.

Peterson, D., 1998. Large-scale fire disturbance: from concepts to models.Northwest Sci. 72 (Special Issue), 1–3.

Reed, W., Johnson, E., 2004. Statistical methods for estimating historical firefrequency from multiple fire-scar data. Can. J. For. Res. 34 (11), 2306–2313.

Regent Instruments Inc., 2009. WinDENDRO: An Image Analysis System for Tree-Rings Analysis, version 2008a [computer software] <http://www.regentinstruments.com/assets/windendro_about.html>.

Rist, S., Goebel, P., Corace III, R., Hix, D., Drobyshev, I., Casselman, T., 2011. Do partialcross sections from live trees for fire history analysis result in higher mortality 2years after sampling? For. Ecol. Manage. 262, 940–946.

Senici, D., Lucas, A., Chen, H., Bergeron, Y., Larouche, A., Brossier, B., Blarquez, O., Ali,A., 2013. Multi-millennial fire frequency and tree abundance differ betweenxeric and mesic boreal forests in central Canada. J. Ecol. 101 (2), 356–367.

Stokes, M., Smiley, T., 1996. An Introduction to Tree-Ring Dating. University ofArizona Press, Tucson, AZ.

Swetnam, T., Allen, C., Betancourt, J., 1999. Applied historical ecology: using the pastto manage for the future. Ecol. Appl. 9 (4), 1189–1206.

Turner, M., Romme, W., 1994. Landscape dynamics in crown fire ecosystems.Landscape Ecol. 9 (1), 59–77.

Van Wagner, C., Finney, M., Heathcott, M., 2006. Historical fire cycles in theCanadian Rocky Mountain parks. Forest Sci. 52 (6), 704–717.

Whitlock, C., Higuera, P., McWethy, D., Briles, C., 2010. Paleoecological perspectiveson fire ecology: revisiting the fire-regime concept. Open Ecol. J. 3, 6–23.