Recent mountain pine beetle (Dendroctonus ponderosae; pine beetle) outbreaks in the western United States have affected nearly 18 million ha of pine (Pinus spp.) forest and are unprecedented in spatial extent, severity, and duration, yet little is known about wildlife responses to large-scale insect outbreaks. Elk (Cervus canadensis) are important wildlife whose dominant management paradigm on public lands has focused on providing security habitat to increase survival during hunting seasons and to maintain elk presence on public lands to promote hunter opportunity. To assess the effect of pine beetles and associated changes in forest structure on elk security, we used a time series to characterize canopy cover pre- and post-pine beetle outbreak, characterized relative canopy cover among the dominant forest types in the study area post-pine beetle outbreak, and used global positioning system location data from male and female elk to define habitat relationships and security during the archery and rifle hunting seasons. Our study area was within the Elkhorn Mountains of southwest Montana, USA, 2015–2017, which experienced 80% mortality of lodgepole pine (Pinus contorta) forests during a pine beetle outbreak that peaked in 2008. We observed an 8.5% reduction in canopy cover within pine beetle-infested lodgepole pine forests, yet canopy cover remained relatively high among other forest types post-outbreak. The top-ranked habitat security models contained positive relationships with canopy cover, distance to motorized routes, terrain ruggedness, and slope with few notable differences among sexes and seasons. Across sexes and seasons, 75% and 50% of elk use was within areas with average canopy cover values ≥31 ± 6.65 (SD)% and ≥53 ± 5.7% that were an average of ≥2,072 ± 187.93 m and ≥3,496 ± 157.32 m from a motorized route, respectively. Therefore, we recommend fall elk security be defined as areas that meet these criteria for minimum canopy cover and distance from motorized routes in the Elkhorn Mountains and in other landscapes with similar forest characteristics and hunting pressures. Although we observed expected reductions in canopy cover within pine beetle-infested forests, defoliation alone did not appear to negatively affect elk security or reduce canopy cover below our management recommendations. Nonetheless, because of the prevalence of standing dead trees in our study area, we recommend future work that investigates the relationships with pine beetle-infested areas post-blowdown because changes in ground structure and costs of locomotion may affect elk habitat and security. © 2019 The Authors. Journal of Wildlife Management published by Wiley Periodicals, Inc. on behalf of The Wildlife Society.

Recent mountain pine beetle (Dendroctonus ponderosae; pine beetle) outbreaks in the western United States have affected nearly 18 million ha of pine forest and are unprecedented in spatial extent, severity, and duration (Chan-McLeod 2006, Meddens et al. 2012, Ivan et al. 2018). Damage caused by pine beetles results in widespread tree mortality, defoliation, and eventual blowdown of dead trees, and can strongly influence forest community composition and structure, timber production, wildfire dynamics, and wildlife habitat (Jenkins et al. 2008, Klenner and Arsenault 2009, Pfeifer et al. 2011, Simard et al. 2011, Saab et al. 2014). Wildlife responses to pine beetle outbreaks are dynamic and complex, and vary by taxa, sex, season, and the outbreak successional stage (Saab et al. 2014, Ivan et al. 2018). Understanding the effect of pine beetle outbreaks is critical to land and wildlife managers tasked with promoting and conserving wildlife populations where widespread tree mortality has altered wildlife-habitat relationships through changes to forest structure, vegetation communities, and nutrient flow (Chan-McLeod 2006). Although such changes can have profound effects on wildlife and are well studied when associated with other natural or managed disturbances (e.g., wildfire, prescribed fire, and timber management), little is known about wildlife responses to large-scale insect outbreaks (Martin et al. 2006, Saab et al. 2014).

Elk (Cervus canadensis) have a broad distribution across western North America and can influence vegetation and plant communities (Frank and Evans 1997, Hobbs 2003) and provide economic benefits to regional communities through tourism and hunting (Duffield and Holliman 1988). Elk management on public lands has traditionally focused on providing adequate cover and forage while minimizing motorized routes as the dominate attributes of habitat quality (Hillis et al. 1991, Lyon and Canfield 1991). The management goals under this paradigm are to provide security areas to increase elk survival during the hunting seasons and to maintain elk presence on public lands to promote hunter opportunity (Hillis et al. 1991). However, the traditional management paradigm has been complicated by changes in land use practices and the increase in non-traditional private landowners that are more interested in natural amenities than livestock production and may have restricted public hunting opportunities on parcels that are privately owned (Haggerty and Travis 2006). This results in non-hunted parcels that function as security areas but do not follow the traditional management paradigm on public land and challenge management strategies in areas that strive to maintain or reduce elk population sizes through regulated harvest of adult females (Proffitt et al. 2010, 2013).

Providing additional security areas on public lands through policy changes (i.e., road closures) or habitat improvement projects, has been highlighted as an important management objective to retain elk on public lands and reduce the redistribution of elk to private lands with reduced or eliminated hunting opportunity (Proffitt et al. 2013, Ranglack et al. 2017, DeVoe et al. 2019). Although management paradigms have shifted to accommodate changes in land use practices and attitudes towards elk on private lands, managing for security areas on public lands has remained an important management objective (Ranglack et al. 2017). However, the availability of elk security areas with ample canopy cover may be at odds with the increasing prevalence of pine beetle-killed forests across western North America. The loss of canopy cover is one of the first structural changes in forests after a pine beetle infestation and is arguably the biggest change affecting wildlife (Chan-McLeod 2006). Although specific management recommendations for security habitat have been set for patch size and distance to roads (Hillis et al. 1991), management recommendations for canopy cover have been more varied across the western United States, often relying on general descriptive measures (i.e., heavy cover; Christensen et al. 1993). Although recent work has provided quantitative recommendations for canopy cover based on a meta-analysis across southwest Montana, USA (Ranglack et al. 2017), the work did not include analyses of forest systems affected by pine beetles. The loss of cover after a pine beetle infestation may increase elk vulnerability and require updates to forest management guidelines regarding traditional security measures on public lands to maintain desired levels of survival and retain elk on public lands during the hunting seasons. Within the context of unprecedented pine beetle outbreaks throughout western North America forests, understanding how insect outbreaks and other disturbances that reduce canopy cover with respect to wildlife security is an important management objective.

We used a multipronged observational approach to characterize the reductions in canopy cover associated with pine beetle outbreaks and the effect on elk security in the Elkhorn Mountains of southwest Montana. We expected to observe a decline in canopy cover in infested areas pre- and post-pine beetle outbreak and a general reduction in canopy cover in infested forests relative to uninfested forests. With respect to elk security, given the anticipated loss of canopy cover associated with the pine beetle outbreak, we expected elk to use security areas characterized by large roadless areas (i.e., security patches) and rugged terrain, and we expected to observe an increase in selection for covariates associated with the traditional security paradigm for both sexes across the gradient of relative risk associated with the archery and rifle hunting seasons (Ranglack et al. 2017, DeVoe et al. 2019).

STUDY AREA

Our study area was broadly defined by the Elkhorn Mountains (2,600 km²; 1,141–2,866 m elevation), a relatively isolated mountain chain in southwest Montana (Fig. 1). The climate was characterized by short, cool summers (17.4°C mean Jul temperature) and long, cold winters (−4.8°C mean Jan temperature) with mean annual precipitation ranging from 263 mm to 959 mm (PRISM Climate Group 2016). The core of the study area consisted of public lands (46%) predominantly managed by United States Forest Service (USFS) and Bureau of Land Management, which were surrounded by low elevation private lands (54%). We conducted our study between 2015 and 2017, 7–10 years after the peak of the pine beetle infestation in 2008. In total, the pine beetle outbreak killed approximately 190 km² (80%) of lodgepole pine forests in the study area.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

Elkhorn Mountain study area defined by the population-level elk annual range of collared individuals, southwest Montana, USA, 2015–2017. Dark shading in the core of the study area depicts the forests infested by mountain pine beetles. Blue shading depicts lands open to public hunting.

Lower elevations were dominated by a mixture of open sage-grassland (e.g., big sagebrush [Artemisia tridentata], bluebunch wheatgrass [Pseudoroegnaria spicata], Idaho fescue [Festuca idahoensis], rough fescue [Festuca scabrella], bluegrasses [Poa spp.]) and patches of timber (primarily Rocky Mountain juniper [Juniperus scopulorum] or Douglas fir [Pseudotsuga menziesii]). Additionally, cultivated crops (pasture grasses and leguminous forbs) bordered the northern and southwestern portions of the study area. Upper elevations were dominated by dry coniferous forests (e.g., lodgepole pine, Douglas fir, ponderosa pine [Pinus ponderosa]) with small interspersed meadows (USFS et al. 1993). Sympatric ungulates included mule deer (Odocoileus hemionus), white-tailed deer (O. virginianus), moose (Alces alces), and a small population of bighorn sheep (Ovis canadensis). Carnivores included mountain lion (Puma concolor), bobcat (Lynx rufus), coyote (Canis latrans), and American black bear (Ursus americanus). Grey wolves (C. lupus) were transient in the study area.

The Elkhorn Mountains are within Montana Fish, Wildlife and Parks hunting district (HD) 380, one of Montana's most popular hunting areas since the early 1960s (DeSimone and Vore 1992, Montana Fish Wildlife and Parks 2018). Estimated hunter effort has continued to increase since the 1960s with record highs of 3,936 hunters and 31,786 hunter days in 2015. Male elk hunting in this region is particularly popular because of a unique spike regulation that generates a relatively high male:female ratio and a mature male age structure. When this study occurred (2015–2017), anyone with a general hunting license could harvest a spike or antlerless elk during the archery season. During the rifle season anyone with a general hunting license could harvest a spike, and youth hunters could harvest an antlerless elk. For each year of the study, the number of limited either-sex permits was 120. The number of limited B licenses (antlerless elk) valid for either the entire hunting district or a portion of the hunting district was 700 in 2015, 800 in 2016, and 800 in 2017.

METHODS

Canopy Cover Characterizations

We used aircraft survey data collected by the United States Department of Agriculture Forest Health Protection Aviation Program to delineate pine beetle-affected areas as observed by tree canopy discoloration indicating mortality (USFS 2017, 2018). However, the pine beetle delineations generated from aerial flights did not strictly mirror the patchy mosaic on the landscape and occasionally included grassland meadows or other landcover classifications not affected by pine beetles. To reduce the contamination of non-beetle-killed areas, we clipped the polygons delineating pine beetle infestations to include only the areas that overlapped with lodgepole pine or lodgepole pine shade-intolerant mixed forest classifications as defined by the Vegetation Mapping Program (VMap; USFS 2014).

We characterized canopy cover in pine beetle-infested forests using a time series from the Rangeland Analysis Platform (RAP; Jones et al. 2018) spanning 1990–2017, inclusive of the pine beetle outbreak in the Elkhorn Mountains. The RAP uses a combination of remotely sensed and ground-based data to predict per-pixel (30 m × 30 m) percent canopy cover across the western United States. Although developed for rangelands, the RAP model is an effective tool for describing changes in canopy cover through time with a fine-scale (annual) time series not available in other canopy cover layers (i.e., LANDFIRE, National Land Cover Database). To assess changes in canopy cover within pine beetle-infested forests over time, we characterized the percent change from historical canopy cover levels prior to the pine beetle outbreak (1990–2000) within the 3 largest pine beetle polygons in the study area, which totaled 111 km² and accounted for nearly 60% of the lodgepole pine forests infested by pine beetle.

To assess changes in canopy cover between pine beetle-infested forests and uninfested forests, we summarized the percent canopy cover in infested and uninfested lodgepole pine forests and ponderosa pine and Douglas fir forest classifications. In aggregate, these forest types contained 99% of the forested landscape within the study area. We characterized the mean (±SD) percent canopy cover in each forest type using VMap canopy cover estimates from 2013 (USFS 2014). These estimates represented a mean response of canopy cover defoliation associated with pine beetle outbreaks approximately 5 years after peak infestation and were the most recent estimates for the Elkhorn Mountains after the pine beetle infestation. Moreover, the VMap methods used to estimate canopy cover were specifically developed to note changes in canopy cover associated with pine beetle outbreaks (S. R. Brown, USFS, personal communication).

Security Habitat Modeling

During the winters of 2015 and 2017, we captured adult (>1.5 yr) elk by helicopter net-gunning and darting in accordance with an approved animal welfare protocol (Institutional Animal Care and Use Committee Project Number: FWP09-2014, FWP10-2017). We deployed remote-upload collars (Lotek Wireless model LifeCycle, New Market, Ontario, Canada) programmed to transmit 1 location every 23 hours through the Globalstar satellite network, the exact timing of which varied across collars. Average fix success excluding data transmission gaps was 96 ± 3%, indicating negligible landcover-induced fix bias (Johnson and Gillingham 2008). Collars on adult females remained on the animal for life and on adult males were programmed to automatically release after 4 years. We censored all locations with positional dilution of precision >10 to reduce spatial imprecision (D'eon and Delparte 2005).

Our primary objective was to characterize elk selection to define elk security habitat and inform land management practices on public lands. We focused our security habitat modeling on individuals that remained on lands open to public hunting and retained in our dataset only individuals that had >50% of their GPS locations on lands open to public hunting (i.e., public lands or private lands enrolled in Montana Fish, Wildlife, and Parks hunter access program). To further target the periods when hunting pressure most strongly influenced elk behavior, we restricted our analyses to legal hunting hours defined as 0.5 hours before and after sunset.

We developed separate sex-season resource selection functions using a used-available design (Manly et al. 2002) and generalized linear mixed-effect models with a random intercept for each individual to account for autocorrelation within an individual and unequal sample sizes among individuals (Gillies et al. 2006). We fit mixed-effect models using the lme4 R package (Bates et al. 2015) and the inverse link function. Our used dataset was defined by the locations collected from GPS collars, whereas we sampled availability at a 1:15 used to available ratio within each individual's annual 100% minimum convex polygon. The 1:15 ratio ensured a sufficient sample to avoid numerical integration error and convergence issues given the single location per day fix rates of the deployed GPS collars (Northrup et al. 2013). We characterized resource selection separately for the archery and rifle hunting seasons, which reflected varying degrees of perceived risk from low to high. Following the legal definitions, the archery season was the 6-week period starting on the first Saturday in September, and the 5-week rifle season started 5 weeks prior to the Saturday after Thanksgiving.

We evaluated multiple landscape attributes associated with elk vulnerability including 3 definitions of security patches, percent canopy cover, and distance to motorized routes (Table 1). Although commonly used in elk security models, we did not include a covariate that described public or private land ownership because our analyses were restricted to individual elk that predominantly used lands that allowed public hunting. We derived our criteria for defining security patches from Hillis et al. (1991), which has been widely incorporated into USFS security habitat management standards, and defined security patches as contiguous areas of ≥100 ha located ≥0.8 km from an open motorized route. To evaluate the importance of canopy cover in definitions of elk security, the 2 additional security patch covariates included contiguous areas that contained ≥50% coverage of ≥15% and ≥30% canopy cover. In addition to the covariates associated with the traditional management paradigm, we evaluated the importance of terrain covariates associated with security: slope and ruggedness. For ruggedness we used the vector ruggedness measure (VRM), a unitless measure of landscape ruggedness integrating variation in slope and aspect (Sappington et al. 2007). Following a large body of literature from regional studies modeling seasonal habitats and security areas for elk and other ungulates, we used a pseudothreshold (natural log) functional form for canopy cover, distance to motorized routes, and ruggedness, and a quadratic functional form for slope (Sawyer et al. 2007; Proffitt et al. 2010, 2013; Ranglack et al. 2017; DeVoe et al. 2019).

Table 1. Covariates used in male and female elk security habitat modeling, Elkhorn Mountains, southwest Montana, USA, 2015–2017

Covariate	Description	Functional formsa	Reference
Security patch	Contiguous areas of ≥100 ha located ≥0.8 km from an open motorized route.	Binary	Hillis et al. (1991), Christenses et al. (1993), DeVoe et al. (2019)
Security patch with canopy cover	Contiguous areas of ≥100 ha located ≥0.8 km from an open motorized route and with at least half of the security patch containing 15% and 30% canopy cover as 2 different covariates.	Binary	Hillis et al. (1991), Christenses et al. (1993), DeVoe et al. (2019)
Canopy cover	Percent canopy cover.	Ps	USFS 2014, Ranglack et al. (2017), DeVoe et al. (2019)
Distance to motorized route	Distance (m) to motorized 2-track, dirt, or paved road designated as open to public use within each respective hunting season.	Ps	McCorquodale et al. (2003), Ranglack et al. (2017)
Ruggedness	Vector ruggedness measure (VRM): a unitless measure of landscape ruggedness integrating variation in slope and aspect.	Ps	Hillis et al. (1991), Sappington et al. (2007), DeVoe et al. (2015), Lamont et al. (2019)
Slope	Slope in degrees.	Sq	Hillis et al. (1991), Ranglack et al. (2017), DeVoe et al. (2019)

^a Ps = pseudothreshold, Sq = squared or quadratic, binary signifies categorical covariate for which we did not consider functional forms.

We used a tiered approach to progress from univariate to multivariate models evaluating the traditional security metrics and landscape attributes. In each tier, we identified the top-ranked model for each sex and season using corrected Akaike's Information Criterion (AIC_c; Burnham and Anderson 2002). In the first tier we fit univariate models for each of the security patch covariates (with and without canopy cover) to evaluate the importance of canopy cover in defining security patches. In the second tier we fit 5 candidate models containing multiple combinations of covariates associated with the traditional management paradigm on public lands. Lastly, in the third tier, we added slope and terrain ruggedness as a paired combination to evaluate the addition of landscape covariates. With the final model we generated predictive plots for each continuous covariate while holding all other covariates at their mean value. To provide management recommendations for covariates with a pseudothreshold form, we calculated cumulative elk use as the cumulative area under the curve starting with the largest probability value to target the covariate range with the highest relative probability of use for each sex and season. We then identified the range of covariate values that corresponded to 75% and 50% of the area under the curve and used the minimum values within the range to define thresholds for security and preferred security areas, respectively. These values served as the basis for our management recommendations and reflect the narrowest covariate range associated with 75% (security) and 50% (preferred security) of elk use. To minimize the influence of extreme outliers on the area under the curve and corresponding management thresholds, we removed from the available dataset covariate values that were well beyond the distribution of values observed in the used dataset. For quadratic covariates (i.e., slope), we noted the covariate value that maximized the relative probability of use for each sex and season.

We preformed model validation using a k-fold cross-validation within an iterative process where we withheld the locations for each 5 folds, fit an exponential resource selection function (RSF) with the data that were retained, and then predicted the fitted values for the observations that were withheld (Boyce et al. 2002). We then summed the occurrence of used locations within 10 equal-area RSF bins and evaluated the correlation between the frequency of occurrence and the relative RSF score using the Spearman's rank correlation (Boyce et al. 2002). The adjusted frequencies should be highly correlated with the relative RSF if the model performs well (Boyce et al. 2002). We conducted all analyses and data processing in the R environment for statistical computing (R Core Team 2018) in combination with the sf (Pebesma 2018) and raster (Hijmans 2017) packages for spatial analysis and the tidyverse suite (Wickham 2017) for data manipulation and visualization.

RESULTS

Canopy Cover

The reductions in percent canopy cover below historical levels before the pine beetle outbreak (1990–2000) mirrored the pine beetle outbreak timeline with a 2–3-year lag after the initial infestation (Fig. 2). Between the peak pine beetle outbreak in 2008 and the following 2 years, there was an average 8.5 ± 2.5% reduction in canopy cover within the pine beetle-infested areas we evaluated. After the pine beetle outbreak and associated canopy cover declines, canopy cover began to increase in 2010 and 2011, nearly approaching historical averages by 2015, 7 years after peak infestation.

Within forested portions of the study area, forests classified as uninfested lodgepole pine had the highest values for canopy cover post-pine beetle infestation with a mean of 77 ± 14.6% (Fig. 3). Similar to the trends described in the canopy cover time series, there was an 8% reduction in canopy cover in infested lodgepole forests ( $urn:x-wiley:0022541X:media:jwmg21781:jwmg21781-math-0001$ = 69 ± 15%). Canopy cover in pine beetle-infested lodgepole forests remained higher than canopy cover in uninfested Douglas fir and ponderosa pine forests (including both infested and uninfested stands), which had 54 ± 18.9% and 27 ± 8.7% canopy cover, respectively.

.1 Security Habitat Modeling

We captured 59 adult (male = 24, female = 35, total excludes 3 individuals that died during capture) elk during the winters of 2015 and 2017. To evaluate security on lands open to public hunting, we censored 4 males and 12 females that had >50% of their GPS locations on private lands that restricted hunting during the archery and rifle seasons. Among censored elk, a mean of 74 ± 15% of male and female GPS locations were in areas with restricted hunting access. Lastly, we censored 3 individuals (males = 2, females = 1) without any daytime locations within a hunting season. Among the remaining 40 (male = 18, female = 22) individuals, there were 72 animal-years with an average of 30 ± 20.5 GPS locations for each individual-season and an average of 572 ± 235 GPS locations for each sex-season grouping.

In the first tier of model selection, where we evaluated the addition of 15% and 30% as canopy cover thresholds to define security patches, security patches defined without a canopy cover threshold were top-ranked for males in both seasons and for females during the archery season (Table 2). The exception was females during the rifle season where security patches with 30% canopy cover were top-tanked. Nonetheless, there was a large degree of uncertainty in the top-ranked model and a difference of only 0.68 AIC_c units among the 3 univariate models. Because there was not a clear top-ranked model for females during the rifle season, and to have continuity among the second tier sex-season models with respect to security patch definitions, we selected security patches defined without canopy cover as the top model for all sex-seasons (Table 2).

Table 2. Habitat security model selection results for tiers 1–3 comparing used and available locations for male and female elk in the archery and rifle hunting seasons, Elkhorn Mountains, southwest Montana, USA, 2015–2017. Models are arranged by corrected Akaike's Information Criterion (AIC_c) ranking. We also present the number of parameters (K) and model weight (w_i)

Model tier	Sex-season	Modela	K	AIC_c	ΔAIC_c	w_i
Tier 1: Security patch	Female-archery	Security.00	3	6,399.02	0.00	0.67
		Security.30	3	6,401.72	2.70	0.17
		Security.15	3	6,401.86	2.85	0.16
	Female-rifle	Security.30	3	4,570.29	0.00	0.41
		Security.00	3	4,570.86	0.56	0.31
		Security.15	3	4,570.97	0.68	0.29
	Male-archery	Security.00	3	3,916.05	0.00	1.00
		Security.15	3	3,984.58	68.53	0.00
		Security.30	3	3,984.68	68.62	0.00
	Male-rifle	Security.00	3	2,132.94	0.00	0.96
		Security.30	3	2,140.85	7.91	0.02
		Security.15	3	2,140.85	7.91	0.02
Tier 2: Traditional security	Female-archery	Canopy cover + DMOT	4	6,267.63	0.00	1.00
		Canopy cover	3	6,293.88	26.25	0.00
		Canopy cover + security.00	4	6,294.21	26.59	0.00
		DMOT	3	6,366.89	99.26	0.00
		Security.00	3	6,399.02	131.39	0.00
	Female-rifle	Canopy cover + DMOT	4	4,328.85	0.00	1.00
		Canopy cover + security.00	4	4,363.95	35.10	0.00
		Canopy cover	3	4,382.79	53.94	0.00
		DMOT	3	4,527.33	198.49	0.00
		Security.00	3	4,570.86	242.01	0.00
	Male-archery	Canopy cover + DMOT	4	3,902.77	0.00	0.78
		DMOT	3	3,905.31	2.55	0.22
		Canopy cover + security.00	4	3,913.24	10.47	0.00
		Security.00	3	3,916.05	13.29	0.00
		Canopy cover	3	3,974.72	71.95	0.00
	Male-rifle	Canopy cover + DMOT	4	2,096.46	0.00	0.99
		Canopy cover + security.00	4	2,105.41	8.96	0.01
		Canopy cover	3	2,110.71	14.25	0.00
		DMOT	3	2,121.66	25.20	0.00
		Security.00	3	2,132.94	36.48	0.00
Tier 3: Traditional and landscape security	Female-archery	Canopy cover + DMOT + Slope^2 +ruggedness	7	6,233.22	0.00	1.00
		Canopy cover + DMOT	4	6,267.63	34.41	0.00
	Female-rifle	Canopy cover + DMOT + slope^2 +ruggedness	7	4,305.30	0.00	1.00
		Canopy cover + DMOT	4	4,328.85	23.54	0.00
	Male-archery	Canopy cover + DMOT + slope^2 +ruggedness	7	3,808.99	0.00	1.00
		Canopy cover + DMOT	4	3,902.77	93.77	0.00
	Male-rifle	Canopy cover + DMOT + slope^2 +ruggedness	7	2,029.19	0.00	1.00
		Canopy cover + DMOT	4	2,096.46	67.27	0.00

^a Security.00 = security patch with no canopy cover requirement; security.15 and security.30 = security patch with ≥50% coverage of 15% and 30% canopy cover; DMOT = distance to motorized route.

In the second tier we evaluated 5 univariate and multivariate combinations of the traditional security metrics (i.e., distance to motorized routes, canopy cover, and security patches defined without a canopy cover threshold from the first tier). In all sex-seasons the multivariate model containing distance to motorized routes and canopy cover was top-ranked by a minimum of 2.55 ( $urn:x-wiley:0022541X:media:jwmg21781:jwmg21781-math-0002$ = 18.21 ± 15.06) AIC_c units (Table 2). In general, there was relatively more support for multivariate models than for univariate models and the least support for security patches as a single predictor of elk security.

In the third tier, the inclusion of slope and ruggedness improved AIC_c ranking in all sex-seasons by a minimum reduction of 23.5 ( $urn:x-wiley:0022541X:media:jwmg21781:jwmg21781-math-0003$ = 57.74 ± 31.97) AIC_c units (Table 2), resulting in the same model structure in the final model for all sex-seasons, which contained canopy cover, distance to motorized routes, ruggedness, and slope covariates (Appendix A). Across all sex and season combinations, the Pearson's correlation coefficients were ≤0.4.

The top-ranked habitat security models performed well with a strong correlation between area-adjusted frequencies and the relative RSF for all sexes and seasons (female-archery: r_s = 0.93, P < 0.001, female-rifle: r_s = 0.99, P < 0.001, male-archery: r_s = 0.97, P < 0.001, male-rifle: r_s = 0.91, P < 0.001). Additionally, with only a few exceptions the top-ranked models produced similar results across sexes and seasons (Fig. 4). There was a positive pseudothreshold relationship with canopy cover, indicating increased selection for areas that had relatively more canopy cover, and the relationship was stronger for females than for males. Males during the archery season had the weakest relationship with canopy cover and a coefficient estimate that was nearly zero (Appendix A), resulting in a flat prediction line across the observed values of canopy cover (Fig. 4). Although between-season differences in selection for canopy cover were modest, both sexes selected for higher canopy cover values in the rifle season than in the archery season. The stronger relationship for females resulted in higher canopy cover thresholds associated with security and preferred security areas in both seasons. Females during the rifle season selected security areas containing canopy cover values ≥39% and preferred security areas containing canopy cover values ≥60%.

The relationship with distance to motorized routes was positive in all top-ranked season-sex models, indicating increasing selection for areas farther from roads. This relationship was steeper for females than for males and similar across the rifle and archery hunting seasons (Fig. 4). There was strong overlap in covariate values delineating security and preferred security among both sexes and seasons. The most conservative values were for males during the archery season where security areas were defined as being ≥2,303 m from a road and preferred security areas were ≥3,679 m from a road.

The ruggedness covariate had the most varied results among sexes and seasons. Females had a positive association with ruggedness during the archery season and a negative relationship during the rifle season (Fig. 4). In contrast, the relationship was positive for males with little difference between the rifle and archery season, both of which had a stronger relationship relative to females. The stronger relationship for males resulted in relatively high ruggedness thresholds associated with security and preferred security areas, which occurred in areas with ruggedness values ≥0.03 and ≥0.05, respectively. Lastly, the quadratic functional form for slope indicated an average optimal slope value of 18 ± 1.6° with little variation among sexes or seasons (Fig. 4).

DISCUSSION

Our study combined canopy cover characterizations with wildlife habitat modeling after a pine beetle infestation to assess the effect of pine beetle on elk security through reductions in canopy cover. We observed an 8.5% reduction in canopy cover within pine beetle-infested lodgepole pine forests compared to historical levels before the pine beetle outbreak. Nonetheless, canopy cover in pine beetle-infested forests remained relatively high ( $urn:x-wiley:0022541X:media:jwmg21781:jwmg21781-math-0004$ = 69 ± 15%) and had higher cover values than uninfested Douglas fir and ponderosa pine forests by 15% and 42%, respectively. The presence of canopy cover was an important component in defining elk security and was an important predictor of male and female elk distributions in the archery and rifle hunting seasons. However, given the relatively high degree of cover offered by lodgepole pine forests, including those affected by pine beetles, the changes in cover associated with defoliation after the pine beetle infestation were relatively minor and did not result in a meaningful reduction in canopy cover below the thresholds used to define security and preferred security areas. Across sexes and hunting seasons, elk security areas contained average canopy cover values ≥30 ± 6.7% and contained 75% of elk use on the landscape. Preferred security areas, which contained 50% of elk use on the landscape contained canopy cover values ≥53 ± 5.7%. In general, although we observed expected reductions in canopy cover within pine beetle-infested forests, the level of reductions that we observed in our study did not appear to negatively affect elk security after the pine beetle infestation.

In addition to the selection for areas with higher canopy cover, security habitat in the Elkhorn Mountains was characterized by positive associations with increasing distances from motorized routes, relatively rugged slopes, and moderate slope angles of 18°. In general, these findings corroborate with other studies characterizing elk security and habitat (Lyon and Canfield 1991, McCorquodale et al. 2003, Ranglack et al. 2017, DeVoe et al. 2019). Although the traditional definition of an elk security patch that has been broadly incorporated into forest management (i.e., contiguous areas of ≥100 ha located ≥0.8 km from an open motorized route; Hillis et al. 1991) did not receive strong support among our candidate model set, our work lends credence to the importance of canopy cover and distance to motorized routes in defining elk security.

The importance of canopy cover in defining and managing elk security on public lands is varied across the western United States. In areas with expansive forest cover, elk security can be influenced by road management alone, yet where forest cover is limited or patchy, incorporating canopy cover into definitions of elk security is also important (Christensen et al. 1993, Unsworth et al. 1993). In contrast to the definitions for distances to motorized routes and patch size, which have largely followed the Hillis paradigm (Hillis et al. 1991), the varied importance of canopy cover has resulted in a variety of arbitrary cover definitions used to define elk security (Christensen et al. 1993). A recent meta-analysis incorporating data from 325 female elk in southwest Montana that occupied public lands and private lands with restricted hunter access recommended managing for areas with ≥13% canopy cover that are ≥2,760 m from motorized routes when defining elk security (Ranglack et al. 2017). Their recommendations were based on the covariate value that corresponded to half of the change in the relative probability of use over the observed covariate range (Ranglack et al. 2017). Although these methods may be well suited for relationships between use and cover characterized by pseudothreshold curves with a sharp rollover, they are arbitrary in the selection of 0.5 and less appropriate for curves with a broad rollover where half of the total gain across the x- and y-axes may not result in meaningful management recommendations. Rather than provide a single minimum value, our approach provided a range for canopy cover and distance to motorized routes based on the cumulative elk use and our definitions of security and preferred security areas (i.e., 75% and 50% area under the curve, respectively). Our results suggest that minimum canopy cover values between 23% and 60% and distances to motorized routes between 1,846 m and 3,679 m characterize most elk use within our Elkhorn Mountain study area for both sexes in both the archery and rifle hunting seasons. Rather than provide a single minimum recommendation, the range of covariate values and their relationship with elk use provides an alternative approach to providing management recommendations that better accommodates the varied shapes of our predictive curves and meaningfully translates to elk resource use.

In addition to different methodologies for defining threshold values, there were important differences between the study areas and study designs that may have contributed to the more conservative recommendations regarding canopy cover and security from this study compared to Ranglack et al. (2017) and Hillis et al. (1991). First, Ranglack et al. (2017) included female elk on public and private lands when defining elk security. In contrast, our recommendations were developed for male and female elk that predominately occupied lands open to unrestricted public hunting and censored elk on private lands with restricted hunter access where elk may be less reliant on areas of dense canopy cover far from motorized routes because hunting pressure is relatively lower. Second, although we were unable to estimate hunter densities specifically for lands open to unrestricted public hunting within our study area or across Montana, the Elkhorn Mountains hunting district (HD 380) is subject to relatively high hunting pressure with record highs of 3,936 hunters and 31,786 hunter days in 2015. Moreover, on average the Elkhorn Mountains received 5.4 times (range = 1.4–35) more hunters and 5.7 times (range = 1.4–28) more hunter days than the populations from Ranglack et al. (2017). Lastly, in contrast to Hillis et al. (1991), which was conducted in densely forested systems in western Montana, our study was located in southwest Montana where cover is patchily distributed. Given our focus on public lands in the most heavily hunted district in Montana with patchily distributed forest cover, one would expect elk to have stronger selection for forest cover and security areas than when pooling elk across areas with restricted and unrestricted hunting access (Ranglack et al. 2017) or when working in systems that are largely forested (Hillis et al. 1991).

Although our results indicated elk continued to use areas affected by pine beetles post-infestation during the hunting season, regionally there have been mixed results regarding elk use of pine beetle-infested forests. In south-central Wyoming, USA, Lamont et al. (2019) reported that during summer female elk avoided pine beetle-infested forests during nearly all parts of the day and selected for intact coniferous forests during daytime. The selection for intact forests during daytime in summer highlights the need for thermal refuge, which may be compromised in pine beetle-infested forests (Lamont et al. 2019). In contrast, Ivan et al. (2018) documented an increase in the use of pine beetle-infested forests by elk, mule deer, and moose during summer months across Colorado, USA, presumably influenced by increases in understory forage associated with the decrease in canopy cover post-defoliation. Given the host of dynamic factors that can influence wildlife response to pine beetle infestations, many of which are spatially and temporally variable and difficult to quantify over broad spatial scales (i.e., number of downed trees), regional and taxonomic generalizations may prove difficult given the limited number of studies examining wildlife responses to pine beetle infestations that have been completed currently (Saab et al. 2014).

An additional consideration in interpretations and applications of our work was the successional stage of pine beetle-infested forests in our study area. An increase in fallen trees is presumed to influence wildlife mobility in pine beetle-infested forests with a hypothesized reduction in use associated with increased costs in mobility (Saab et al. 2014, Lamont et al. 2019). Our study occurred between 2015 and 2017, 7–10 years after the peak of the pine beetle infestation in 2008, yet most of the infested trees were standing dead. Given the lack of downed trees during our study period, our results are specific to the period post-defoliation but before tree blowdown. Although Lamont et al. (2019) did document an increase in downed trees in pine beetle-infested forests, their study also occurred when trees were just beginning to fall, 3–7 years following the peak outbreak. Although downfall in infested areas was not widespread, the avoidance of pine beetle-infested areas during all parts of the day in their study area was suggestive of a down-tree effect and lends correlative support to elk avoiding pine beetle-infested areas and the associated increase in locomotion costs (Lamont et al. 2019). In Colorado, Ivan et al. (2018) did not describe the degree of downfall across their broad study region but did see increases in ungulate us of pine beetle-infested forest across 3 different severities (10%, 50%, and 90% mortality) spanning 0–11 years after initial outbreak.

As pine beetle-infested forests transition from standing dead to blowdown, understanding the effect on elk habitat and security is an important consideration. Given the large degree of canopy cover provided by lodgepole forests on public lands and the large percentages that have been affected throughout the western United States, relatively minor avoidance of these areas could result in a notable net loss of habitat that provides demographic benefits through increased security and thermal cover. As pine beetle-infested forests continue to mature with increasing proportions of fallen trees, assessing the degree to which these forests are selected or avoided by elk and the effect on demography will be a management priority. Additionally, the management of adjacent or nearby intact forests may become increasingly important in providing elk security as infested stands mature and potentially become inadequate for providing elk security.

MANAGEMENT IMPLICATIONS

We recommend that managers include measures of canopy cover and distance to motorized routes in definitions of elk security and encourage managers to implement seasonal or permanent road closures in areas with forest canopy cover to provide elk security during hunting seasons. In the Elkhorn Mountains and other landscapes with similar forest characteristics and hunting pressures, we recommend managing for security with canopy cover values of 23–60% and distance from motorized routes of 1,846–3,679 m based on our definitions of security and preferred security areas, respectively. Where possible, the implementation of more stringent objectives at the upper end of the canopy cover and distance to road thresholds will more strongly reflect preferred security areas. In addition, locating security areas in relatively rugged terrain and moderate slope angles (i.e., 18°) can further enhance elk security. Given the relatively small difference in selection patterns between males and females and the archery and rifle hunting seasons, these recommendations are appropriate for both sexes and both the archery and rifle hunting seasons. Managing for these habitat security attributes on public lands may help to reduce the redistribution of elk to private lands that restrict hunting during the archery and rifle seasons. In the presence of pine beetle infestations, we also recommend prioritizing intact forests with high canopy cover to help retain security areas as infestations mature and blowdown, potentially further affecting habitat conditions within infested forests. Because of the relatively narrow temporal window of our study, approximately 7–10 years after the beetle infestation, we recommend future work that investigates the relationships with pine beetle-infested areas post-blowdown as changes in ground structure may provide the biggest effect to habitat and security for elk and other ungulates.

ACKNOWLEDGMENTS

We thank the numerous private landowners for facilitating field sampling efforts, B. R. Malo, M. S. Scott, and N. S. Cadwell for their work in capturing animals, D. M. Pengeroth and A. J. Grove for important logistical support, information and advice throughout the project, B. M. Cascaddan for field logistics, J. T. Paterson for technical support and helping to develop our threshold estimation methods, and members of the Elkhorn Working Group for their collaborative support and input. J. A. Gude, A. J. Grove, and D. M. Pengeroth provided valuable insight through reviews of this manuscript. Funding was provided by the Elkhorn Working Group, United States Forest Service, Montana Department of Military Affairs, Rocky Mountain Elk Foundation, and Cinnabar Foundation. Additional funding was provided by revenues from the sale of Montana hunting and fishing licenses and matching Federal Aid in Wildlife Restoration grants to Montana Fish, Wildlife, and Parks.

APPENDIX A: TOP-RANKED HABITAT SECURITY MODEL COEFFICIENT ESTIMATES

Table A1. Coefficient estimates for the top-ranked habitat security models for male and female elk during the archery and rifle hunting seasons, Elkhorn Mountains, southwest Montana, USA, 2015–2018

Sex-season	Covariatea	Beta	SE
Female-archery	Intercept	−2.629	0.047
	Canopy cover	0.294	0.040
	DMOT	0.196	0.039
	Slope	0.222	0.048
	Slope²	−0.188	0.035
	Ruggedness	0.032	0.036
Female-rifle	Intercept	−2.811	0.063
	Canopy cover	0.595	0.056
	DMOT	0.343	0.051
	Slope	0.285	0.060
	Slope²	−0.169	0.041
	Ruggedness	−0.094	0.045
Male-archery	Intercept	−2.471	0.075
	Canopy cover	0.009	0.050
	DMOT	0.479	0.063
	Slope	0.063	0.064
	Slope²	−0.424	0.059
	Ruggedness	0.225	0.043
Male-rifle	Intercept	−2.466	0.101
	Canopy cover	0.207	0.079
	DMOT	0.289	0.077
	Slope	0.491	0.099
	Slope²	−0.533	0.087
	Ruggedness	0.189	0.059

^a DMOT= distance to motorized route.

LITERATURE CITED

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67: 1–48.
10.18637/jss.v067.i01
Web of Science®Google Scholar
Boyce, M. S., P. R. Vernier, S. E. Nielsen, and F. K. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157: 281–300.
10.1016/S0304-3800(02)00200-4
PubMedWeb of Science®Google Scholar
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer Science & Business Media, New York, New York, USA.
CASGoogle Scholar
Chan-McLeod, A. C. A. 2006. A review and synthesis of the effects of unsalvaged mountain-pine-beetle-attacked stands on wildlife and implications for forest management. Journal of Ecosystems and Management 7.
Google Scholar
Christensen, A., L. Lyon, and J. Unsworth. 1993. Elk management in the Northern Region: considerations in forest plan updates or revisions. U.S. Department of Agriculture Forest Service, Intermountain Research Station, Ogden, Utah, USA.
Google Scholar
D'eon, R. G., and D. Delparte. 2005. Effects of radio-collar position and orientation on GPS radio-collar performance, and the implications of PDOP in data screening. Journal of Applied Ecology 42: 383–388.
10.1111/j.1365-2664.2005.01010.x
Web of Science®Google Scholar
DeSimone, R. M., and J. M. Vore. 1992. Elkhorn elk monitoring program: 1991 annual report. Montana Fish, Wildlife and Parks, Helena, USA.
Google Scholar
DeVoe, J. D., K. M. Proffitt, M. S. Mitchell, C. S. Jourdonnais, and K. J. Barker. 2019. Elk forage and risk tradeoffs during the fall archery season. Journal of Wildlife Management 83: 801–816.
10.1002/jwmg.21638
Web of Science®Google Scholar
Duffield, J., and J. Holliman. 1988. The net economic value of elk hunting in Montana. Montana Fish, Wildlife and Parks. <http://archive.org/details/B55D9C76-4A8A-4A30-8711-38EDB312FA55>. Accessed 14 Feb 2019.
Google Scholar
Frank, D. A., and R. D. Evans. 1997. Effects of native grazers on grassland N cycling in Yellowstone National Park. Ecology 78: 2238–2248.
10.1890/0012-9658(1997)078[2238:EONGOG]2.0.CO;2
Web of Science®Google Scholar
Gillies, C. S., M. Hebblewhite, S. E. Nielsen, M. A. Krawchuk, C. L. Aldridge, J. L. Frair, D. J. Saher, C. E. Stevens, and C. L. Jerde. 2006. Application of random effects to the study of resource selection by animals: random effects in resource selection. Journal of Animal Ecology 75: 887–898.
10.1111/j.1365-2656.2006.01106.x
PubMedWeb of Science®Google Scholar
Haggerty, J. H., and W. R. Travis. 2006. Out of administrative control: absentee owners, resident elk and the shifting nature of wildlife management in southwestern Montana. Geoforum 37: 816–830.
10.1016/j.geoforum.2005.12.004
Web of Science®Google Scholar
Hijmans, R. J. 2017. raster: geographic data analysis and modeling. R package version 2.6-7. https://CRAN.R-project.org/package=raster. Accessed 14 Feb 2019.
Google Scholar
Hillis, J. M., M. J. Thompson, J. E. Canfield, L. J. Lyon, C. L. Marcum, P. M. Dolan, and D. W. McCleerey. 1991. Defining elk security: the Hillis paradigm. Proceedings of the elk vulnerability symposium. Montana State University, Bozeman, USA.
Google Scholar
Hobbs, N. T. 2003. Challenges and opportunities in integrating ecological knowledge across scales. Forest Ecology and Management 181: 223–238.
10.1016/S0378-1127(03)00135-X
Web of Science®Google Scholar
Ivan, J. S., A. E. Seglund, R. L. Truex, and E. S. Newkirk. 2018. Mammalian responses to changed forest conditions resulting from bark beetle outbreaks in the southern Rocky Mountains. Ecosphere 9(8):e02369.
10.1002/ecs2.2369
Web of Science®Google Scholar
Jenkins, M. J., E. Hebertson, W. Page, and C. A. Jorgensen. 2008. Bark beetles, fuels, fires and implications for forest management in the Intermountain West. Forest Ecology and Management 254: 16–34.
10.1016/j.foreco.2007.09.045
Web of Science®Google Scholar
Johnson, C. J., and M. P. Gillingham. 2008. Sensitivity of species-distribution models to error, bias, and model design: an application to resource selection functions for woodland caribou. Ecological Modelling 213: 143–155.
10.1016/j.ecolmodel.2007.11.013
Web of Science®Google Scholar
Jones, M. O., B. W. Allred, D. E. Naugle, J. D. Maestas, P. Donnelly, L. J. Metz, J. Karl, R. Smith, B. Bestelmeyer, C. Boyd, J. D. Kerby, and J. D. McIver. 2018. Innovation in rangeland monitoring: annual, 30 m, plant functional type percent cover maps for U.S. rangelands, 1984–2017. Ecosphere 9:e02430.
10.1002/ecs2.2430
Web of Science®Google Scholar
Klenner, W., and A. Arsenault. 2009. Ponderosa pine mortality during a severe bark beetle (Coleoptera: Curculionidae, Scolytinae) outbreak in southern British Columbia and implications for wildlife habitat management. Forest Ecology and Management 258: S5–S14.
10.1016/j.foreco.2009.08.035
Web of Science®Google Scholar
Lamont, B. G., K. L. Monteith, J. A. Merkle, T. W. Mong, S. E. Albeke, M. M. Hayes, and M. J. Kauffman. 2019. Multi-scale habitat selection of elk in response to beetle-killed forest. Journal of Wildlife Management 83: 679–693.
10.1002/jwmg.21631
Web of Science®Google Scholar
Lyon, L. J., and J. E. Canfield. 1991. Habitat selections by Rocky Mountain elk under hunting season stress. Proceedings of the elk vulnerability symposium. Montana State University, Bozeman, USA.
Google Scholar
Manly, B. F., L. McDonald, D. Thomas, T. L. McDonald, and W. P. Erickson. 2002. Resource selection by animals: statistical design and analysis for field studies. Springer Science & Business Media, Berlin, Germany.
Google Scholar
Martin, K., A. Norris, and M. Drever. 2006. Effects of bark beetle outbreaks on avian biodiversity in the British Columbia interior: implications for critical habitat management. Journal of Ecosystems and Management 7: 1–16.
Google Scholar
Meddens, A. J. H., J. A. Hicke, and C. A. Ferguson. 2012. Spatiotemporal patterns of observed bark beetle-caused tree mortality in British Columbia and the western United States. Ecological Applications 22: 1876–1891.
10.1890/11-1785.1
PubMedWeb of Science®Google Scholar
Montana Fish Wildlife & Parks. 2018. Montana harvest reports. https://myfwp.mt.gov/fwpPub/harvestReports. Accessed 14 Feb 2019.
Google Scholar
Northrup, J. M., M. B. Hooten, C. R. Anderson, and G. Wittemyer. 2013. Practical guidance on characterizing availability in resource selection functions under a use–availability design. Ecology 94: 1456–1463.
10.1890/12-1688.1
PubMedWeb of Science®Google Scholar
Pebesma, E. 2018. sf: simple features for R. R package version 0.6-3. https://CRAN.R-project.org/package=sf. Accessed 14 Feb 2019.
Google Scholar
Pfeifer, E. M., J. A. Hicke, and A. J. H. Meddens. 2011. Observations and modeling of aboveground tree carbon stocks and fluxes following a bark beetle outbreak in the western United States: carbon cycling following bark beetle outbreak. Global Change Biology 17: 339–350.
10.1111/j.1365-2486.2010.02226.x
Web of Science®Google Scholar
PRISM Climate Group. 2016. Oregon State University. <http://prism.oregonstate.edu>. Accessed 27 Jul 2018.
Google Scholar
Proffitt, K. M., J. L. Grigg, R. A. Garrott, K. L. Hamlin, J. Cunningham, J. A. Gude, and C. Jourdonnais. 2010. Changes in elk resource selection and distributions associated with a late-season elk hunt. Journal of Wildlife Management 74: 210–218.
10.2193/2008-593
Web of Science®Google Scholar
Proffitt, K. M., J. A. Gude, K. L. Hamlin, and M. A. Messer. 2013. Effects of hunter access and habitat security on elk habitat selection in landscapes with a public and private land matrix. Journal of Wildlife Management 77: 514–524.
10.1002/jwmg.491
Web of Science®Google Scholar
R Core Team. 2018. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
Google Scholar
Ranglack, D. H., K. M. Proffitt, J. E. Canfield, J. A. Gude, J. Rotella, and R. A. Garrott. 2017. Security areas for elk during archery and rifle hunting seasons. Journal of Wildlife Management 81: 778–791.
10.1002/jwmg.21258
Web of Science®Google Scholar
Saab, V. A., Q. S. Latif, M. M. Rowland, T. N. Johnson, A. D. Chalfoun, S. W. Buskirk, J. E. Heyward, and M. A. Dresser. 2014. Ecological consequences of mountain pine beetle outbreaks for wildlife in western North American forests. Forest Science 60: 539–559.
10.5849/forsci.13-022
Web of Science®Google Scholar
Sappington, J. M., K. M. Longshore, and D. B. Thompson. 2007. Quantifying landscape ruggedness for animal habitat analysis: a case study using bighorn sheep in the Mojave Desert. Journal of Wildlife Management 71: 1419–1426.
10.2193/2005-723
Web of Science®Google Scholar
Sawyer, H., R. M. Nielson, F. G. Lindzey, L. Keith, J. H. Powell, and A. A. Abraham. 2007. Habitat selection of Rocky Mountain elk in a nonforested environment. Journal of Wildlife Management 71: 868–874.
10.2193/2006-131
Web of Science®Google Scholar
Simard, M., W. H. Romme, J. M. Griffin, and M. G. Turner. 2011. Do mountain pine beetle outbreaks change the probability of active crown fire in lodgepole pine forests? Ecological Monographs 81: 3–24.
10.1890/10-1176.1
Web of Science®Google Scholar
Unsworth, J. W., L. Kuck, M. D. Scott, and E. O. Garton. 1993. Elk mortality in the Clearwater Drainage of northcentral Idaho. Journal of Wildlife Management 57: 495–502.
10.2307/3809273
Web of Science®Google Scholar
U.S. Forest Service [USFS]. 2014. Region one vegetation classification, mapping, inventory and analysis report. Helena-Lewis & Clark National Forest–VMap 2014 tree dominance type (DOM40), tree canopy cover, tree size class, and lifeform accuracy assessment. Numbered Report NRGG14-01. https://www.fs.usda.gov/detailfull/r1/landmanagement/gis/?cid=stelprdb5331054&width=full. Accessed 14 Feb 2019.
Google Scholar
U.S. Forest Service [USFS]. 2017. Aerial survey detection GIS data. https://www.fs.usda.gov/detail/r4/forest-grasslandhealth/?cid=stelprdb5366459. Accessed 14 Feb 2019.
Google Scholar
U.S. Forest Service [USFS]. 2018. Insect and disease detection survey data explorer. https://foresthealth.fs.usda.gov/portal/Flex/IDS. Accessed 14 Feb 2019.
Google Scholar
U.S. Forest Service [USFS], Bureau of Land Management, and Montana Fish Wildlife & Parks. 1993. Elkhorns landsape analysis documentation: Elkhorn Cooperative Management Area. http://archive.org/details/elkhornslandscap00unit. Accessed 14 Feb 2019.
Google Scholar
Wickham, H. 2017. tidyverse: easily install and load the ‘Tidyverse’. R package version 1.2.1. https://CRAN.R-project.org/package=tidyverse. Accessed 14 Feb 2019.
Google Scholar

Associate Editor: Kathryn Schoenecker.

Citing Literature

Volume84, Issue1

January 2020

Pages 138-149

Hiding Without Cover? Defining Elk Security in a Beetle-Killed Forest

ABSTRACT

STUDY AREA