Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Caribou in the cross-fire? Considering terrestrial lichen forage in the face of mountain pine beetle (Dendroctonus ponderosae) expansion

  • Barry R. Nobert,

    Roles Data curation, Formal analysis, Software, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Caribou Program, fRI Research, Hinton, Alberta, Canada, Alberta Environment and Parks, Grande Prairie, Alberta, Canada

  • Terrence A. Larsen,

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Writing – review & editing

    Affiliation Grizzly Bear Program, fRI Research, Hinton, Alberta, Canada

  • Karine E. Pigeon,

    Roles Conceptualization, Data curation, Investigation, Methodology, Writing – review & editing

    Affiliations Caribou Program, fRI Research, Hinton, Alberta, Canada, Geomatics and Landscape Ecology Research Lab, Carleton University, Ottawa, Ontario, Canada

  • Laura Finnegan

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing

    lfinnegan@friresearch.ca

    Affiliation Caribou Program, fRI Research, Hinton, Alberta, Canada

Abstract

Mountain pine beetle (MPB) has become an invasive forest pest of mature pine in western North America as it spreads beyond its former endemic range. Management actions such as timber harvest can reduce the spread of MPB but may affect species of conservation concern like woodland caribou. Our goal was to inform MPB management within caribou ranges by exploring the impacts of MPB on caribou habitat–focusing on terrestrial lichens, an important winter food for caribou. We evaluated differences in lichen cover among four MPB management actions: timber harvest, wildfires, leaving MPB killed trees as-is, and single-tree cut-and-burn control. We found little evidence that leaving MPB killed trees as-is or controlling MPB using single-tree cut-and-burn impacted lichen cover. However, we found that lichen cover was lower in timber harvested and burned areas compared to intact undisturbed forest but only 10 to 20 years post-disturbance, respectively. Our results suggest that despite short-term reductions in lichen cover, using timber harvesting and prescribed burns to control MPB may balance management needs for MPB while maintaining lichen cover over time. However, using timber harvesting and prescribed burns to manage MPB is likely to have detrimental population-level effects on caribou by increasing the proportion of disturbed habitat and thus predators within caribou ranges. Among the four management actions that we evaluated, the cut-and-burn control program balances the need to limit the spread of MPB while also limiting negative impacts on caribou food. Our work addresses some of the challenges of managing competing forest and ecosystem values by evaluating the consequence of forest pest management actions on an important food resource for a species-at-risk.

Introduction

Invasive species are a major source of ecological and economic loss [13]. In an effort to mitigate negative impacts of invasive species, land managers typically employ aggressive eradication programs [4,5]. However, management actions for species eradication can have unintended and detrimental ecological consequences on non-target organisms [6,7]. Because of this risk, managers should evaluate the potential for unintended outcomes prior to any intervention, especially in areas where management actions for invasive species could negatively impact species of conservation concern [8,9]. In such areas, evaluating the consequences of management actions on non-target species under alternate management scenarios could allow for proactive and informed invasive species mitigation [10,11].

Since the early 1990’s, land managers in western North America have attempted to control the spread of mountain pine beetle (Dendroctonus ponderosae; hereafter MPB); an endemic forest pest of mature pine west of the Rocky Mountains [12]. MPB breached the Rocky Mountains in the mid-2000s, spreading north and east at approximately 80km/year [13]. MPB has also spread from lodgepole (Pinus contorta) into jack pine (Pinus banksiana) [14,15] and whitebark pine (Pinus albicaulis) [16]. The spread of MPB into new regions, including western Alberta, is thought to be a result of decades of fire suppression associated with forest management coupled with increasingly warm winters and summers [15]. MPB is responsible for the death of millions of hectares of forest in Canada and the United States [17,18], with cascading and significant impacts on ecosystem function [1921]. In an effort to eradicate and slow the spread of MPB, and to mitigate the economic impact on the forest industry, MPB control has focused on accelerated harvest of mature pine [22], single-tree cut-and-burn control programs of infested trees [23,24], and prescribed burning in protected areas [25]. However, MPB infestation and associated management actions can affect non-target organisms; a process that has already been observed in the boreal forest [26].

MPB and MPB management alter the forest structure by decreasing canopy cover and creating canopy gaps [27], which impacts understory vegetation by increasing light penetration and reducing snow interception [26,28,29]. For some boreal species, these impacts may be beneficial since species that prefer early seral habitats may benefit from changes in understory vegetation resulting in more food resources, and cavity-nesting birds may benefit from an increase in standing dead trees [26,29,30]. However, not all species benefit from MPB. For instance, elk (Cervus canadensis) avoid beetle-killed stands despite having abundant early seral forage, likely because of the increased costs of locomotion necessary to move through MPB-killed stands with dead and downed trees [31]. For species that prefer mature forest, MPB and MPB management may have significant and detrimental impacts on the availability of habitat and food resources [26,32]. For such species, specifically species that are of conservation concern, there is a need to evaluate whether MPB infestations could be less detrimental for species-at-risk compared to a range of MPB management actions.

Woodland caribou (Rangifer tarandus; hereafter ‘caribou’) are a threatened species [3335] that prefer mature forest stands [36,37], which are habitats susceptible to MPB (e.g., mature pine). This preference for mature forest is driven by multiple factors but in part mature, open conifer stands provide abundant terrestrial and arboreal lichen that are important food for caribou during winter [3840]. In addition, mature forest with sparse understory vegetation supports low densities of other ungulates and consequently, predators occur at low densities–effectively making contiguous mature forest stands predator refugia for caribou [4143]. Loss and fragmentation of mature forest caused by habitat disturbances resulting in unstainable high predation rates is the main driver of caribou declines [4446]. Because of this, federal and provincial caribou recovery strategies aim to reduce habitat disturbances such as timber harvest and wildfires within caribou ranges [34,35,47]. Because caribou conservation plans resolve to protect mature forest, they directly contradict the management actions for MPB eradication.

Management actions used to eradicate and slow the spread of MPB mainly accelerate harvest of mature pine in combination with single-tree cut-and-burn of infested trees at the leading edge of MPB spread [22,23,48]. In addition, stands that have already been killed by MPB may be salvage logged or burned [49] with prescribed burns being the main MPB management approach used in protected areas like National Parks [25]. These management actions are likely detrimental to caribou, but allowing MPB infestations to linger in caribou range may not guarantee the protection of caribou habitat either. For example, in British Columbia, Cichowski and Haeussler [50] reported a 9% decrease in percent cover of terrestrial lichens a decade after MPB infestation. These opposing management actions operating in the same region create a need to understand their potential impact to caribou habitat.

The goal of our study was to determine how MPB and alternative MPB management actions affect the distribution and abundance of terrestrial lichen in western Alberta. We focused on the impacts of MPB and MPB management on terrestrial lichen (hereafter “lichen”) because adequate food resources and nutrition are necessary to maintain sustainable caribou populations [51,52], and because caribou habitat use can be closely linked to the availability of forage [53,54]. First, we determined how MPB and actions to manage MPB affect lichen cover by constructing spatiotemporal lichen cover models. Specifically, we 1) modeled lichen cover in a) timber harvested stands, b) stands burned by wildfire, c) MPB single-tree cut-and-burn control stands, d) MPB infested stands, and e) intact stands. We then 2) simulated future lichen cover under different management actions. Second, we used resource selection functions (RSF; [55]) to evaluate the predictive ability of the lichen cover models based on the presumption that caribou should select habitats with higher predicted lichen cover. The results of this research are intended to help guide MPB management actions in support of caribou recovery in the boreal forest.

Material and methods

Ethical statement

Weyerhaeuser Company provided caribou GPS collar data with animal care protocols completed by Alberta Environment and Parks (AEP). AEP adhered to capture and handling guidelines under the Canadian Council on Animal Care [56] and the Government of Alberta’s Animal Care Protocol No. 008 [57]. Lichen data collection occurred on public lands and in provincial parks, and permission for field sampling and helicopter access was granted under the authority of the Government of Alberta (permit #14–109).

Study area

The study area was approximately 33,000 km2, encompassed nine natural sub-regions [58], and included caribou ranges in west-central (53.857, -119.109) and north-western Alberta, Canada (57.675, -119.037). In west-central, forests are a mosaic of lodgepole pine, white spruce, and aspen, with black spruce, larch, and muskeg in low-lying areas [5860]. Higher elevations have Engelmann spruce and subalpine fir below tree-line and graminoid, sedge, and herbaceous ground cover or exposed rock above tree-line. In the north-west, forests are white spruce, trembling aspen, and balsam poplar with black spruce, larch, and muskeg and fen in low-lying areas [58,61]. The study area included 2,032 km2 of federally protected land, 5,410 km2 of provincially protected land, and 25,761 km2 of provincial land-base. Hunting and other recreational activities occurred within protected lands but mostly on the provincial land-base. Industrial activities associated with the energy (mining, oil, and natural gas) and forest industry occur exclusively within the provincial land-base.

Lichen absence and abundance

Field data collection.

We used a geographic information system (GIS) and a random number generator to identify transects within forests stratified into five categories: timber harvest (Cut), wildfire (Fire), MPB kill (MPB), single-tree cut-and-burn control program (SingleTree), and intact undisturbed forest (Forest)–see S1 Appendix for details. For ease of access, we constrained 80% of transects to within 1 km of roads or pipelines and accessed the remainder via helicopter. We did not survey SingleTree within the north-western study area because at the time of data collection (2014 and 2015), there was no single-tree cut-and-burn management in the area. We collected data from 776 transects between June and October of 2014 and 2015 (S1 Appendix, Table 1).

Field surveys and field-derived explanatory variables.

We focused field surveys on four terrestrial lichen genera (Cetraria spp., Cladina spp., Cladonia spp., and Flavocetraria spp.) preferred by caribou [62]. At each transect, we visually estimated percent cover of lichens within six subplots placed at 5-m increments along a 25-m transect line. Because forest canopy cover and over-story species are known to influence the distribution and percent cover of lichen [63,64], we also estimated percent canopy cover at each subplot and recorded characteristics of the three nearest trees (species, MPB killed, and single-tree cut-and-burn control). Field data collection described in detail in S1 Appendix and field variables are in S2 Appendix.

GIS-derived explanatory variables.

We linked transects to GIS-derived variables previously reported to influence the distribution and percent cover of lichen (Table B in S2 Appendix). For forest stand age, we used forest inventory data provided by forest companies to calculate years since timber harvest, or we used provincial wildfire data to calculate years since wildfire. For climate, we used data from western Canada adjusted for elevation [65] to interpolate climate normals (circa 1961–1991) across our study area. Climatic growing condition data included mean annual precipitation (cm), mean summer precipitation, number of consecutive frost free days, degree-days > 5 °C, and summer heat-moisture index. For forest canopy, in west-central we used a percent canopy cover and height layer derived from LiDAR data [66]. For north-western, we used field-derived visual estimate of canopy cover because LiDAR-derived canopy cover data were not available (S1 Appendix).

For terrain, we used a LiDAR-derived depth to water estimation, a metric of soil wetness based on local topography and modeled hydrologic flow [67,68]. To represent the diminishing effect of the depth to water on vegetation growth, we transformed the variable using an exponential decay function 1 –e−1.55×Depth2Wat(m) [69]. This decay function caused depth to water to rapidly decrease at depths greater than 2 m and to become constant at depths greater than 3 m, reflecting the root depth of boreal forest vegetation [70]. We also used the Canadian Digital Elevation Model [71] to extract values of elevation, terrain wetness (compound topographic index, CTI; [72]), and solar radiation based on latitude, topographic position, and terrain shadowing intersecting each transect [73,74]. We used CTI rather than depth to water for north-western because LiDAR-derived depth-to-water data were not available for all transects surveyed in that region. We used ArcGIS 10.3 [75] to extract GIS-derived variables intersecting each transect.

Data analysis

We carried out statistical analysis using R [76] within R-studio [77]–package names are indicated with quotations. To assess differences in mean percent lichen cover among the five sampling strata in each region, we used a Kruskal–Wallis test (‘stat’ [78]) and post hoc pairwise Nemenyi-tests [79] in ‘PMCMR’ [80].

Modelling lichen occurrence and abundance.

Before analyzing lichen occurrence and abundance, we screened explanatory variables following Zuur et al. [81]. We did not include variables in the same model if they were strongly correlated (|rp| > 0.60); using univariate models and Deviance Information Criterion (DIC, [82]) to identify which of any two correlated variables to include in downstream analyses. We also excluded variables from models with variance inflation factor (VIF) >2 (‘usdm’; [83]). We standardized continuous variables before fitting models (Table C in S2 Appendix).

We used zero-inflated beta regression [84] to model presence-absence and abundance of lichen along transects within each region. Beta regression is appropriate for analysis of proportional data [8587], and has previously been used to model terrestrial lichen [88]. We fit zero-inflated beta regression models with ‘zoib’ [89,90], which derives inference for model parameters using a Bayesian framework via the Markov Chain Monte Carlo (MCMC) approach implemented in JAGS [91]. We chose a Bayesian framework over a likelihood-based approach because a Bayesian framework helped avoid issues of non-convergence and biased parameters. We surveyed lichen within a 1 m2 subplot during the first year of data collection, but increased the subplot size to 10 m2 during the second survey year to capture more of the variability present in lichen distribution. We accounted for potential bias in occurrence or percent cover of lichen caused by combining data from 1 m2 subplots in the first survey year with 10 m2 subplots in the second survey year by including a fixed effect ‘scale’ variable in west-central models (we only surveyed lichen in north-western during 2014). We accounted for the clustered nature of the dataset (i.e., six subplots along a 25-m transect) by treating transect as the sample unit with subplots nested within.

Before fitting models, we combined percent cover of the four lichen genera because combining information from similar rare species improves model predictability relative to individual species models [92]. We built separate models for west-central and north-western, and separate models for Fire and Cut. We combined Forest, MPB, and SingleTree within a single model. Combining multiple strata into a single model allowed us to explore MPB and single-tree cut-and-burn control effects relative to intact forest by including covariates within the combined model, specifically percent of MPB killed trees and the presence/absence of MPB control along the transect. We expected explanatory variables for occurrence and percent cover to differ, and therefore performed model selection on each part of the zero-inflated equation separately while holding the other side of the equation at the intercept. We evaluated competing models using DIC, and if any two models were within ≤4 ΔDIC of one another, we chose the model with fewer parameters. We considered non-linear effects for the stand age and canopy structure variables by including squared terms (Table B in S2 Appendix). Additional variable details within each strata and region are provided in Table B in S2 Appendix.

We carried out model selection using an iterative process. We started with a global model that included all of the variables of interest for that sampling strata/region (Table B in S2 Appendix), and following the principle of parsimony, we removed uninformative variables [93] from the global model for each sampling strata using a “drop one” approach. For the “drop one” approach, we used DIC to compare alternative models with each variable removed in turn, and removed uninformative variables in an iterative manner from the downstream analysis until removing a variable did not further reduce model DIC. We reported final model results as mean beta (β) coefficients with 2.5% and 97.5% posterior predictive values, or as the relative probability of occurrence (Eq 1) or abundance (Eq 2). (1) (2) Because we modelled presence-absence using a zero-inflated model, positive β coefficients indicate a negative relationship between lichen occurrence and a variable (i.e. probability of lichen being absent), and a positive relationship between percent lichen cover and a variable. We also reported final model results as spatial maps of the predicted mean percent lichen cover given occurrence (Eq 1 * Eq 2) for landscape conditions in 2017. We evaluated the predictive ability of final models using mean absolute error (MAE) and root mean square error (RMSE) calculated from model residuals [94]. MAE is the mean difference between the observed and predicted percent cover in absolute terms. RMSE can be interpreted as the standard error in a model’s unexplained variance. Lower values of MAE and RMSE indicate a better predictive model.

Simulating lichen abundance across MPB management actions.

To evaluate how percent lichen cover may change in the future under different MPB management approaches, we used the final zero-inflated lichen models to simulate changes. Because the final models for MPB and Control did not include age since disturbance (see Results), we focused on Cut and Fire; simulating changes in lichen cover over a forty year period, the maximum age of the strata sample. We used the final lichen model for each strata and region as our baseline landscape condition and evaluated the potential effects of timber harvesting and wildfires on lichen cover into the future by increasing age of the disturbance within models from 0 to 40 years while holding all other variables at their mean. For Forest, we added 0 to 40 years to the mean stand age within each region (mean stand age west-central = 119 years, north-western = 97 years). When simulating changes in percent cover of lichen over time, we held all other spatial variables constant at their respective means for each region.

Model evaluation and caribou habitat selection

To help evaluate the predictive ability of the lichen cover models, we assessed whether caribou in west-central Alberta selected for areas predicted to have higher lichen cover. We focused our analysis on the early and late winter seasons (30 November– 5 February; 6 February– 9 May respectively; see MacNearney et al. [95]) because lichens are an important food resource during winter [39,40]. We used GPS data collected from 100 caribou collared between 1998 and 2016 in the Redrock Prairie Creek population. We only used GPS collar data with a dilution of precision (DOP) ≤ 12 for analysis. We rarefied GPS location data to 2-hr intervals before building models to account for variable fix rates. We then used mixed effects logistic regression within ‘lme4’ [96] to build Resource Selection Function (RSF) models. We constructed RSFs at the home range scale (i.e., 3rd order, [97]); generating 20 random available locations for each GPS location within seasonal caribou home ranges defined by a minimum convex polygon (MCP).

To evaluate the link between caribou habitat selection and lichen, we used a two-step process. First, we used Akaike’s information criterion (AIC) [93,98] and a “drop one” approach to identify a suite of variables related to topography and habitat disturbance to include within a baseline model explaining caribou habitat selection (Table D in S5 Appendix). These variables are known to be important predictors of caribou habitat selection in our study area [95,99,100]. Second, we added the predicted percent lichen cover derived from our zero-inflated lichen models to the baseline model and used AIC to evaluate performance of the baseline model with and without percent lichen cover. We present RSF results as Relative Selection Strength (RSS) and lower and upper 95% confidence intervals (LCL, UCL) of the predictor variables [101]. RSS greater than one indicated a positive relationship between caribou habitat selection and a variable, whereas RSS less than one indicated a negative relationship between habitat selection and a variable.

We evaluated the ability of the final RSF models to predict caribou habitat selection with k-fold cross validation where 20% of the data were withheld for testing [102]. We followed the approach of Boyce et al. [103] and calculated the spearman correlation (rs) between RSF ranked values and the frequency of used points within ten equal area bins across 100 iterations. For k-fold cross validation, rs values closer to 1 indicate a model with better predictive ability. We also calculated the area under the receiver operator curve (AUC) [104] with ‘caret’ [105]; a measure of model performance [106]. AUC values between 0.7 and 0.8 are considered acceptable discrimination, 0.8 to 0.9 are considered excellent, and above 0.9 is considered outstanding [106].

Results

Mean differences in lichen among sampling strata

Mean percent lichen cover differed across strata (west-central χ2 = 42.3, df = 4, P < 0.001; north-western χ2 = 33.0, df = 4, P < 0.001; Fig B in S3 Appendix). In west-central, SingleTree and MPB had lower lichen cover compared to Cut (SingleTree P = 0.008; MPB P < 0.001) and Forest (SingleTree P = 0.001; MPB P < 0.001). In north-western, Cut and MPB had lower lichen cover relative to Fire (Cut P < 0.001; MPB P < 0.001) and Forest (Cut P < 0.001; MPB P < 0.001).

Lichen occurrence and abundance

Final zero-inflated model coefficients are presented in Tables 2 and 3. In west-central, the model for Cut indicated that the probability of lichen occurrence increased in conifer forest and was highest at intermediate cutblock age (~25 years; Fig C in S3 Appendix). The Cut model indicated that lichen abundance increased with cutblock age. The model for Fire indicated that the probability of lichen occurrence increased with wildfire age, but that there was no relationship between wildfire age and lichen abundance. The model for Forest, MPB, and SingleTree suggested that the probability of lichen occurrence was higher in conifer forest, decreased with decreasing stand age, and increased with decreasing summer precipitation. This model also showed that the probability of lichen occurrence was highest when canopy height was ~9 m (Fig C in S3 Appendix), and that percent lichen cover increased linearly with increasing stand canopy height. The model for Forest, MPB, and SingleTree also indicated that the probability of lichen occurrence decreased with increasing percent of mountain pine beetle killed trees.

thumbnail
Table 2. Lichen occurrence and abundance model coefficients for the west-central study area.

https://doi.org/10.1371/journal.pone.0232248.t002

thumbnail
Table 3. Lichen occurrence and abundance model coefficients for the north-western study area.

https://doi.org/10.1371/journal.pone.0232248.t003

In north-western, the model for Cut indicated that the probability of lichen occurrence increased at dry sites (i.e., high CTI) within cutblocks and at high elevation. This model also showed that lichen occurrence and percent cover increased with cutblock age. The model for Fire suggested that the probability of lichen occurrence increased with wildfire age. The model for Forest, MPB, and SingleTree indicated that the probability of lichen occurrence decreased with increasing percent of MPB-killed trees.

In west-central, MAE and RSME pointed to better model fit for the Forest, MPB, and SingleTree model (MAE 1.7%, RSME 1.8%) when compared to Cut (MAE 2.7%, RSME 2.9%) and Fire models (MAE 1.8%, RSME 1.9%). In north-western, MAE and RSME indicated better model fit for Cut (MAE 0.8%, RSME 0.9%), relative to Forest, MPB, and SingleTree (MAE 5.1%, RSME 5.8%) and Fire (MAE 9.2%, RSME 10.4%) models. Based on landscape conditions in 2017, we found that in west-central, predicted percent lichen cover tended to be higher in the mountains when compared to the lower elevation foothills (Fig 1; Fig D in S4 Appendix). Higher elevation was associated with older forest stands with lower canopy heights, and with areas with higher mean summer precipitation (Fig E in S4 Appendix).

thumbnail
Fig 1. Map of predicted terrestrial lichen cover.

Predicted terrestrial lichen cover (Cetraria spp., Cladina spp., Cladonia spp. and Flavocetraria spp.) in west-central and north-western Alberta, Canada, mapped using landscape conditions in 2017. Blank areas within caribou range in west-central denote rock and ice covered mountain tops.

https://doi.org/10.1371/journal.pone.0232248.g001

Simulating changes in lichen abundance in timber harvested and burned areas

In west-central, although Cut and Fire had the lowest initial percent lichen cover, our models predicted that percent lichen cover in Cut and Fire would exceed percent lichen cover in Forest within forty years (Fig 2). In north-western, Cut had the lowest initial percent lichen cover and Fire had the highest initial percent lichen cover, and our models predicted that percent lichen cover would remain relatively stable over time (Fig 2).

thumbnail
Fig 2. Simulated terrestrial lichen cover under different management actions.

Lichen cover simulated over forty years using the zero-inflated lichen models in west-central and north-western Alberta, Canada. Cut and Fire simulations assumed that the disturbances were created at year zero (2017), whereas for the Forest strata simulations the stand age at year zero was the mean stand age in each study area (west-central = 119 years, north-western = 97 years).

https://doi.org/10.1371/journal.pone.0232248.g002

Caribou habitat selection and lichen abundance

During early winter, the baseline RSF included elevation, cutblocks, seismic lines, and winter roads, while during late winter, the baseline RSF included cutblocks, seismic lines, and roads. During early and late winter, adding percent lichen cover to the baseline model improved model AIC, fit, and predictive ability (Table 4). The combined base and lichen model indicated that caribou selected areas with higher predicted lichen cover during early (RSS 1.58, LCL 1.57, UCL 1.60) and late winter (RSS 1.63, LCL 1.62, UCL 1.64]. Complete model parameters are in Table E in S5 Appendix.

thumbnail
Table 4. RSF model comparison between the baseline and lichen models.

https://doi.org/10.1371/journal.pone.0232248.t004

Discussion

Landscape management that aims to balance resource extraction and species conservation is complex. Our study evaluated the potential impact of MPB and MPB management on an important winter food resource of threatened caribou. Overall, we found little effect of single-tree MPB cut-and-burn control or standing dead MPB killed trees on terrestrial lichen cover in western Alberta, at least in the eight years after infestation. We did find that timber harvested areas and wildfires had lower lichen cover compared to forests that have not been disturbed, but lichen cover increased with age of timber harvest and wildfire. Our results suggest that in areas where caribou and MPB overlap, single tree cut-and-burn control or leaving infested forest stands as-is may be the preferred management approach. By evaluating the impacts of management actions, the results of this study help to mitigate the unintended consequences of MPB management on caribou where MPB and caribou co-occur.

Our study linking lichen cover to MPB kill and MPB control showed that although lichen abundance was not affected by MPB-killed trees, age of MPB, and MPB control, lichens were less likely to occur in areas with more MPB-killed trees. It is possible that we did not detect an effect of MPB control on lichen cover because cut-and-burn crews operate on foot and during winter, therefore limiting ground disturbances that would damage terrestrial lichen. Even forests timber harvested by mechanized equipment in winter retain high cover of terrestrial lichen, at least in the short term [107]. However, our findings may also be an artifact of the age of the MPB infestation in Alberta because eight years may not be sufficient to detect any appreciable change in slow-growing species such as lichen. Similar research in British Columbia only detected a decrease in lichen cover 10 to 15 years after the initial MPB infestation [108]. It is also possible that the impacts of MPB and MPB control may be more apparent with faster growing understory species like shrubs and forbs [109]. Continuing to assess the availability of lichen in MPB infested and controlled stands at later stages of infestation (> 8 years) would provide additional information to guide forest management decisions.

If we consider alternate MPB management actions, which are timber harvesting and wildfire, our research showed that there was less lichen cover initially in timber harvested and burned areas, but that lichen increased with age. These findings are in line with previous research [110,111], because timber harvested areas generally have more terrestrial lichen cover than fire-origin stands of the same age [112114]–with low lichen cover after timber harvesting [107]. Lichen cover can remain low in timber harvested stands until they reach 30 years old, with the highest lichen abundance occurring in stands between 50 and 100 years old [113]. Our results and simulations for the north-western region support this pattern of higher lichen cover in burned areas and a slight increase in lichen cover over time. However, in the west-central region, lichen cover in timber harvested areas was slightly higher than lichen cover in burned areas. These regional differences in lichen cover may be driven by differences in local environmental conditions with the north-western region having flat topography and wet conditions relative to the west-central region.

Indeed, even within regions, we found that lichen cover was higher within older, higher elevation, drier mature forests with low-to intermediate canopy heights. The association between lichens, stand age, and stand height was expected because lichens are slow to establish and grow, and they typically reach a peak in abundance within forests with moderate canopy cover and age [115117]. Forests with more enclosed canopy, humidity, and reduced light transmission to the forest floor are often dominated by mosses [108,113]. We found that lichen cover increased with elevation in the west-central region, consistent with the transition from foothills forests to subalpine areas [58]. In the subalpine, the long 110 to 162 year fire interval [118,119], and harsh climatic conditions [58], likely promote high percent lichen cover by allowing for very old forest stands (i.e., > 300 years), while still maintaining the short-open canopy that lichen thrive in [120]. Our study showed that quantifying relationships between forest attributes and percent lichen cover could help identify forest stands with more abundant winter forage for caribou.

Our habitat selection analysis helped to support the lichen cover models because caribou were more likely to select areas predicted to have greater lichen cover. Other studies that have considered caribou food availability with broad scale habitat characteristics within models have found similar links between caribou and lichen [121123]. The purpose of the habitat selection analysis herein was to simply evaluate the predictive ability of the lichen cover models. Concluding that lichen cover is the main driver of caribou distribution in west-central Alberta would require a comprehensive comparison of caribou habitat selection relative to lichen cover and competing habitat variables such as terrain, land cover, and predation risk [59,100].

If the ultimate goal of forest management associated with MPB were to retain caribou food supply, then our results would suggest that despite short-term reductions in caribou forage, using timber harvesting and prescribed fire to control MPB could balance management needs and caribou food supply over time. However, we would caution against such an approach because timber harvesting and prescribed fires could have long-term population-level effects on caribou by reducing available caribou habitat [59,123,124], increasing predation risk [45,125,126], and contributing to population declines [36,127]. To address this uncertainty, future studies should expand upon our examination of MPB management actions and caribou food supply by exploring how different MPB management actions change caribou predation risk, especially because unsustainably high predation rates is the primary cause of caribou population declines [128130].

Conclusions

We evaluated the potential impacts of managing an invasive forest pest on a species-at-risk. Overall, we found limited evidence that MPB killed trees impact lichen cover. However, our study was restricted to eight years after infestation and management, and further impacts may emerge over time. Leaving MPB killed forest as-is could benefit caribou conservation but this would need to be evaluated against the potential for increased wildfire risk and need to be balanced with socio-economic considerations [131]. Of the four MPB management actions that we evaluated, the MPB cut-and-burn control program appears to balance the need to limit the spread of mountain pine beetle and negative impacts of MPB and MPB management on caribou food. Our work helps address the challenge of managing forests under competing ecological values, specifically species-at-risk conservation versus invasive species control. When developing management strategies across the boreal forest, understanding potential unintended consequences of management actions on non-target species can improve conservation planning in a changing landscape.

Supporting information

S1 Appendix. Details of field data collection among sampling transects.

https://doi.org/10.1371/journal.pone.0232248.s001

(PDF)

S2 Appendix. Explanatory variables used to model lichen occurrence and percent cover.

https://doi.org/10.1371/journal.pone.0232248.s002

(PDF)

S3 Appendix. Relationship between lichen cover and linear variables.

https://doi.org/10.1371/journal.pone.0232248.s003

(PDF)

S4 Appendix. Relationship between predicted lichen cover and linear variables.

https://doi.org/10.1371/journal.pone.0232248.s004

(PDF)

S5 Appendix. Caribou RSF explanatory variables and model parameters.

https://doi.org/10.1371/journal.pone.0232248.s005

(PDF)

Acknowledgments

In-kind support for this project was provided by Manning Diversified Forest Products [now West Fraser Timber Co. Ltd.], Daishowa-Marubeni International (DMI) [Mercer International Inc.], Canfor Corporation, Weyerhaeuser Co. Ltd., Alberta Newsprint Company, Tolko Industries Ltd., and West Fraser Timber Co. Ltd. S Blanton, J Witiw, G Whitmore, Emend, L Fullerton, D Walty, B White, J Stadt, Alberta Environment and Parks, Alberta Parks, and the Hinton Training Centre helped with field logistics and accommodation. Assistance with GIS and mapping was provided by J Crough, J Duval, and D Wismer. D MacNearney trained field technicians and coordinated field logistics, K Ridley entered field data. Pilots at Peregrine Helicopters and Valley B Aviation kept our crews safe during aerial sampling. This project would have been impossible without our dedicated field technicians–M Anderson, A Barre, G Degre-Timmons, J Dillon, S Fassina, J Halbert, J Hayden, M Hull, A MacDonald, S Murray, K Ridley, A Sprott, and K Trepanier.

References

  1. 1. Clavero M, García-Berthou E. Invasive species are a leading cause of animal extinctions. Trends Ecol Evol. 2005;20: 110. pmid:16701353
  2. 2. Lodge DM, Williams S, MacIsaac HJ, Hayes KR, Leung B, Reichard S, et al. Biological invasions: Recommendations for U.S. policy and management. Ecol Appl. 2006;16: 2035–2054. pmid:17205888
  3. 3. Pimentel D, Zuniga R, Morrison D. Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecol Econ. 2005;52: 273–288.
  4. 4. Myers JH, Simberloff D, Kuris A, Carey JR. Eradication Revisited: Dealing with Exotic Species. Trends Ecol Evol. 2000;15: 316–320. pmid:10884695
  5. 5. Rejmanek M, Pitcairn MJ. When is eradication of exotic pest plants a realistic goal? In: Veitch CR, Clout MN, editors. Turning the Tide: The Eradication of Invasive Species. Auckland, New Zealand: Hollands Printing Ltd; 2002. pp. 249–253.
  6. 6. S.Zavaleta E, J.Hobbs R, A.Mooney H. Viewing invasive species removal in a whole-ecosystem context. Trends Ecol Evol. 2001;16: 454–459.
  7. 7. Bergstrom DM, Lucieer A, Kiefer K, Wasley J, Belbin L, Pedersen TK, et al. Indirect effects of invasive species removal devastate World Heritage Island. J Appl Ecol. 2009;46: 73–81.
  8. 8. Strong DR, Pemberton RW. Biological control of invading species—Risk and reform. Science (80-). 2000;288: 1969–1970. pmid:10877714
  9. 9. Kopf RK, Nimmo DG, Humphries P, Baumgartner LJ, Bode M, Bond NR, et al. Confronting the risks of largescale invasive species control. Nat Ecol Evol. Macmillan Publishers Limited; 2017;1: 1–4. pmid:28812629
  10. 10. Schindler DE, Hilborn R. Prediction, precaution, and policy under global change. Science (80-). 2015;347: 953–954. pmid:25722401
  11. 11. A DP, M PM, E K, B PWJ, B AE. Inferential and forward projection modeling to evaluate options for controlling invasive mammals on islands. 2016;26: 2548–2559. pmid:27880019
  12. 12. Safranyik L, Carroll A. The biology and epidemiology of the mountain pine beetle in lodgepole pine forests. Safranyik L, Wilson WR, editors. The Mountain Pine Beetle: A Synthesis of Its Biology, Management and Impacts on Lodgepole Pine. Victoria, British Columbia: Natural Resources Canada, Canadian Forest Service; 2006.
  13. 13. Bleiker KP, Boisvenue C, Campbell EM, Cooke BJ, Erbilgin N, Friberg RF, et al. Risk assessment of the threat of mountain pine beetle to Canada’s boreal and eastern pine forests. 2019.
  14. 14. Cullingham CI, Cooke JEK, Dang S, Davis CS, Cooke BJ, Coltman DW. Mountain pine beetle host-range expansion threatens the boreal forest. Mol Ecol. 2011;20: 2157–2171. pmid:21457381
  15. 15. Cook BJ, Carroll AL. Predicting the risk of mountain pine beetle spread to eastern pine forests: Considering uncertainty in uncertain times. For Ecol Manage. 2017;396: 11–25.
  16. 16. Logan JA, Macfarlane WW, Wilcox L. Whitebark pine vulnerability to climate‐driven mountain pine beetle disturbance in the Greater Yellowstone Ecosystem. Ecol Appl. 2010;20: 895–1190. pmid:20597278
  17. 17. Nealis VG, Cooke BJ. Risk assessment of the threat of mountain pine beetle to Canada’s boreal and eastern pine forests. Canadian Council of Forest Ministers Report Fo79-14/2014E-PDF. Victoria, British Columbia; 2014.
  18. 18. Hicke JA, Logan JA, Powell J, Ojima DS. Changing temperatures influence suitability for modeled mountain pine beetle (Dendroctonus ponderosae) outbreaks in the western United States. J Geophys Res Biogeosciences. 2006;111.
  19. 19. Dhar A, Parrott L, Heckbert S. Consequences of mountain pine beetle outbreak on forest ecosystem services in western Canada. Can J For Res. 2016;46: 987–999.
  20. 20. Mikkelson KM, Maxwell RM, Ferguson I, Stednick JD, McCray JE, Sharp JO. Mountain pine beetle infestation impacts: modeling water and energy budgets at the hill-slope scale. Ecohydology. 2013;6: 64–72.
  21. 21. Kurz WA, Ebata T, Neilson ET, Safranyik L, Rampley GJ, Dymond CC, et al. Mountain pine beetle and forest carbon feedback to climate change. Nature. 2008;452: 987–990. pmid:18432244
  22. 22. Amoroso MM, David Coates K, Astrup R. Stand recovery and self-organization following large-scale mountain pine beetle induced canopy mortality in northern forests. For Ecol Manage. Elsevier B.V.; 2013;310: 300–311.
  23. 23. Government of Alberta. Mountain pine beetle action plan Alberta. 2007.
  24. 24. ASRD. Mountain Pine Beetle Management Strategy. Edmonton, AB, Canada; 2007.
  25. 25. McFarlane BL, Stumpf-Allen RCG, Watson DO. Public perceptions of natural disturbance in Canada’s national parks: The case of the mountain pine beetle (Dendroctonus ponderosae Hopkins). Biol Conserv. 2006;130: 340–348.
  26. 26. Saab VA, Latif QS, Rowland MM, Johnson TN, Chalfoun AD, Buskirk SW, et al. Ecological consequences of mountain pine beetle outbreaks for wildlife in Western North American forests. For Sci. 2014;60: 539–559.
  27. 27. Dhar A, Hawkins CDB. Regeneration and growth following mountain pine beetle attack: a synthesis of knowledge. BC J Ecosyst Manag. 2011;12: 1–16.
  28. 28. Fornwalt PJ, Rhoades CC, Hubbard RM, Harris RL, Faist AM, Bowman WD. Short-term understory plant community responses to salvage logging in beetle-affected lodgepole pine forests. For Ecol Manage. Elsevier; 2018;409: 84–93.
  29. 29. Chan-Mcleod ACA. A review and synthesis of the effects of unsalvaged mountain-pine-beetle- attacked stands on wildlife and implications for forest management. BC J Ecosyst Manag. 2006;7: 119–132.
  30. 30. Larsen TA. The potential influence of mountain pine beetle (Dendroctonus ponderosae) control harvesting on grizzly bear (Ursus arctos) food supple and habitat conditions in Alberta. M Sc Thesis, Dep Biol Sci Univ Alberta. 2012; 133 pp.
  31. 31. Lamont BG, Monteith KL, Merkle JA, Mong TW, Albeke SE, Hayes MM, et al. Multi-scale habitat selection of elk in response to beetle-killed forest. J Wildl Manage. 2019;83: 679–693.
  32. 32. Matsuoka SM, Handel CM, Ruthrauff DR. Densities of breeding birds and changes in vegetation in an Alaskan boreal forest following a massive disturbance by spruce beetles. Can J Zool. 2001;79: 1678–1690.
  33. 33. Alberta Sustainable Resource Development, Alberta Conservation Association. Status of the Woodland Caribou in Alberta: Alberta Wildlife Status Report No. 30 (Update 2010). Alberta Sustainable Resource Development. Edmonton, Alberta, Canada; 2010.
  34. 34. Environment Canada. Recovery Strategy for the Woodland Caribou, Southern Mountain population (Rangifer tarandus caribou) in Canada [Proposed]. Ottawa: Species at Risk Act Recovery Strategy Series. Environment Canada, Ottawa.; 2014. p. viii + 68 pp.
  35. 35. Environment Canada. Recovery Strategy for the Woodland Caribou (Rangifer Tarandus Caribou), Boreal Population, in Canada. Ottawa, Ontario, Canada; 2012.
  36. 36. Environment Canada. Scientific assessment to inform the identification of critical habitat for woodland caribou (Rangifer tarandus caribou), boreal population, in Canada. Update. 2011.
  37. 37. Dzus EH. Status of the Woodland Caribou (Rangifer tarandus caribou) in Alberta. Environment. 2001.
  38. 38. Thompson I, Wiebe P, Mallon E, Rodgers A, Fryxell J, Baker J, et al. Factors influencing the seasonal diet selection by woodland caribou (Rangifer tarandus tarandus) in boreal forests in Ontario. Can J Zool. 2015;93: 87–98.
  39. 39. Thomas DC, Edmonds EJ, Brown WK. The diet of woodland caribou populations in west-central Alberta the. Methods. 1996;9: 1–4. Available: http://scholar.google.ca/scholar?hl=en&q=edmonds+caribou&btnG=Search&as_sdt=2000&as_ylo=&as_vis=0#8.
  40. 40. Johnson C, Parker K, Heard D. Feeding site selection by woodland caribou in north-central British Columbia. Rangifer Spec Issue. 2000;12: 159–172.
  41. 41. DeCesare NJ, Hebblewhite M, Robinson HS, Musiani M. Endangered, apparently: The role of apparent competition in endangered species conservation. Anim Conserv. 2010;13: 353–362.
  42. 42. Rettie WJ, Messier F. Hierarchical Habitat Selection by Woodland Caribou: Its Relationship to Limiting Factors. Ecography (Cop). 2000;23: 466–478.
  43. 43. Sorensen T, McLoughlin PD, Hervieux D, Dzus E, Nolan J, Wynes B, et al. Determining sustainable levels of cumulative effects for boreal caribou. J Wildl Manage. 2008;72: 900–905.
  44. 44. Vors LS, Schaefer JA, Pond BA, Rodgers AR, Patterson BR. Woodland Caribou Extirpation and Anthropogenic Landscape Disturbance in Ontario. J Wildl Manage. 2007;71: 1249–1256.
  45. 45. Hervieux D, Hebblewhite M, DeCesare NJ, Russell M, Smith K, Robertson S, et al. Widespread declines in woodland caribou (Rangifer tarandus caribou) continue in Alberta. Can J Zool. 2013;91: 872–882.
  46. 46. Vors LS, Boyce MS. Global declines of caribou and reindeer. Glob Chang Biol. 2009;15: 2626–2633.
  47. 47. Government of Alberta. Draft Provincial Woodland Caribou Range Plan. Edmonton, AB; 2017.
  48. 48. Sasketchewan Ministry of the Environment. State of the Environment 2019: Focus on forests. 2019.
  49. 49. Safranyik L, Linton DA, Shore TL, Hawkes BC. The effects of prescribed burning on Mountain Pine Beetle in lodgepole pine. Information Report BC-X-391. Victoria, BC; 2001.
  50. 50. Cichowski D, Haeussler S. The response of caribou terrestrial forage lichens to forest harvesting and mountain pine beetle in the East Ootsa and Entiako areas [Internet]. Smithers, BC; 2013. https://www.for.gov.bc.ca/hfd/library/fia/html/fia2008mr189.htm
  51. 51. Parker KL, Barboza PS, Gillingham MP. Nutrition integrates environmental responses of ungulates. Funct Ecol. 2009;23: 57–69.
  52. 52. Brown GS, Landriault L, Sleep DJH, Mallory FF. Comment arising from a paper by Wittmer et al.: Hypothesis testing for top-down and bottom-up effects in woodland caribou population dynamics. Oecologia. 2007;154: 485–492. pmid:17891419
  53. 53. Joly K, Stuart Chapin F, Klein DR. Winter Habitat Selection by Caribou in Relation to Lichen Abundance, Wildfires, Grazing, and Landscape Characteristics in Northwest Alaska. Ecoscience. 2010;17: 321–333.
  54. 54. Avgar T, Mosser A, Brown GS, Fryxell JM. Environmental and individual drivers of animal movement patterns across a wide geographical gradient. 2013; 96–106. pmid:23020517
  55. 55. Johnson CJ, Nielsen SE, Merrill EH, McDonald TL, Boyce MS. Resource selection functions based on use-availability data: Theoretical motivation and evaluation methods. J Wildl Manage. BETHESDA; 5410 GROSVENOR LANE, BETHESDA, MD 20814–2197 USA: WILDLIFE SOC; 2006;70: 347–357.
  56. 56. Canadian Council on Animal Care. Guide to the Care and Use of Experimental Animals [Internet]. Ottawa, Ontario; 2017. https://www.ccac.ca/Documents/Standards/Guidelines/Experimental_Animals_Vol1.pdf
  57. 57. Hervieux D, Hebblewhite M, Decesare NJ, Russell M, Smith K, Robertson S, et al. Widespread declines in woodland caribou (Rangifer tarandus caribou) continue in Alberta. 2013;882: 872–882.
  58. 58. Natural Regions Committee. Natural Regions and Subregions of Alberta. Edmonton, AB; 2006.
  59. 59. Smith KG, Ficht EJ, Hobson D, Sorensen TC, Hervieux D. Winter distribution of woodland caribou in relation to clear-cut logging in west-central Alberta. Can J Zool. 2000;78: 1433–1440.
  60. 60. Saher DJ, Schmiegelow FK. Movement pathways and habitat selection by woodland caribou during spring migration. Rangifer. 2005;16: 143–154.
  61. 61. Tigner J, Bayne EM, Boutin S. Black bear use of seismic lines in Northern Canada. J Wildl Manage. 2014;78: 282–292.
  62. 62. Denryter KA, Cook RC, Cook JG, Parker KL. Straight from the caribou’s (Rangifer tarandus) mouth: detailed observations of tame caribou reveal new insights into summer–autumn diets. Can J Zool. 2017;95: 81–94.
  63. 63. Lance AN, Eastland WG. A Guide to Evaluating Forest Stands as Terrestrial Lichen Forage Habitat for Caribou. Prince George, BC; 2000.
  64. 64. Lance AN, Mills B. Attributes of woodland caribou migration habitat in west-central British Columbia. Rangifer. 1996;9: 355–364.
  65. 65. Mbogga M, Hamann A, Wang T. Historical and projected climate data for natural resource management in western Canada. Agric For Meteorol. 2009;149: 881–890.
  66. 66. Nijland W, Nielsen SE, Coops NC, Wulder MA., Stenhouse GB. Fine-spatial scale predictions of understory species using climate- and LiDAR-derived terrain and canopy metrics. J Appl Remote Sens. 2014;8: 83572.
  67. 67. White B, Ogilvie J, Campbell DMHH, Hiltz D, Gauthier B, Chisholm HK, et al. Using the Cartographic Depth-to-Water Index to Locate Small Streams and Associated Wet Areas across Landscapes. Can Water Resour J. 2012;37: 333–347.
  68. 68. Alberta Environment and Parks. Resource Data Product Catalog: Hyrdrological [Internet]. 2017 [cited 5 Jan 2017]. http://aep.alberta.ca/forms-maps-services/maps/resource-data-product-catalogue/hydrological.aspx
  69. 69. Finnegan L, Pigeon KE, Cranston J, Hebblewhite M, Musiani M, Neufeld L, et al. Natural regeneration on seismic lines influences movement behaviour of wolves and grizzly bears. 2018.
  70. 70. Canadell J, Jackson R, Ehleringer J, Mooney HA, Sala OE, Schulze E-D. Maximum rooting depth of vegetation types at the global scale. Oecologia. 1996;108: 583–595. pmid:28307789
  71. 71. Natural Resources Canada. Canadian Digital Elevation Model [Internet]. 2015 [cited 10 Jan 2017]. http://open.canada.ca/data/en/dataset/7f245e4d-76c2-4caa-951a-45d1d2051333.
  72. 72. Gessler P, Chadwick O, Chamran F, Althouse L, Holmes K. Modeling soil-landscape and ecosystem properties using terrain attributes. Soil Sci Soc Am J. 2000;64: 2046–2056.
  73. 73. Fu P, Rich PM. A Geometric Solar Radiation Model with Applications in Agriculture and Forestry. Comput Electron Agric. 2002;37: 25–35.
  74. 74. Rich PM, Hetrick WA, Saving SC, Dubayah RO. Using Viewshed Models to Calculate Intercepted Solar Radiation: Applications in Ecology. Am Soc Photogramm Remote Sens Tech Pap. 1994; 524–529.
  75. 75. ESRI. ArcGIS Desktop [Internet]. Relase 10. Redlands, CA: Environmental Systems Research Institute; 2011. https://www.esri.com
  76. 76. R Development Core Team. R: A Language and Environment for Statistical Computing [Internet]. R Foundation for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2017.
  77. 77. RStudio Team. RStudio: Integrated Development for R [Internet]. Boston, MA: RStudio, Inc.; 2016. http://www.rstudio.com/
  78. 78. McDonald JH. Handbook of Biological Statistics. 3rd ed. Baltimore, Maryland: Sparky House Publishing; 2014.
  79. 79. Sachs L. Angewandte Statistik. Berlin, Germany: Springer; 1997.
  80. 80. Pohlert T. The Pairwise Multiple Comparison of Mean Ranks Package (PMCMR) [Internet]. 2014. https://cran.r-project.org/package=PMCMR
  81. 81. Zuur AF, Ieno EN, Elphick CS. A protocol for data exploration to avoid common statistical problems. Methods Ecol Evol. 2010;1: 3–14.
  82. 82. Spiegelhalter DJ, Best NG, Carlin BP, Van Der Linde A. Bayesian measures of model complexity and fit. J R Stat Soc Ser B Stat Methodol. 2002;64: 583–616.
  83. 83. Naimi B. usdm: Uncertainty Analysis for Species Distribution Models [Internet]. R Foundation for Statistical Computing; 2015. https://cran.r-project.org/package=usdm
  84. 84. Ferrari S, Cribari-Neto F. Beta Regression for Modelling Rates and Proportions. J Appl Stat. 2004;31: 799–815.
  85. 85. Eskelson BNI, Madsen L, Hagar JC, Temesgen H. Estimating riparian understory vegetation cover with beta regression and copula models. For Sci. 2011;57: 212–221.
  86. 86. Finnegan L, MacNearney D, Pigeon KE. Divergent patterns of understory forage growth after seismic line exploration: Implications for caribou habitat restoration. For Ecol Manage. Elsevier; 2018;409: 634–652.
  87. 87. Latifi H, Hill S, Schumann B, Heurich M, Dech S. Multi-model estimation of understorey shrub, herb and moss cover in temperate forest stands by laser scanner data. Forestry. 2017;90: 496–514.
  88. 88. Keim JL, DeWitt PD, Fitzpatrick JJ, Jenni NS. Estimating plant abundance using inflated beta distributions: Applied learnings from a lichen–caribou ecosystem. Ecol Evol. 2017;7: 486–493. pmid:28116045
  89. 89. Liu F, Kong Y. zoib: Bayesian Inference for Beta Regression and Zero-or-One Inflated Beta Regression [Internet]. 2016. https://cran.r-project.org/package=zoib
  90. 90. Liu F, Kong Y. zoib: An R Package for Bayesian Inference for Beta Regression and Zero/One Inflated Beta Regression. R J. 2015;7: 34–51.
  91. 91. Plummer M. JAGS: Just Another Gibbs Sampler [Internet]. 2017. p. 1. https://sourceforge.net/projects/mcmc-jags/
  92. 92. Hui FKC, Warton DI, Foster SD, Dunstan PK. To mix or not to mix: Comparing the predictive performance of mixture models vs. separate species distribution models. Ecology. 2013;94: 1913–1919. pmid:24279262
  93. 93. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach [Internet]. New York: Springer; 2002. Available: Grande Prairie Regional College Internet Access; University of Alberta Access; Red Deer College Access; Concordia University College Internet Access
  94. 94. Willmott CJ, Matsuura K. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Clim Res. 2005;30: 79–82.
  95. 95. MacNearney D, Pigeon K, Stenhouse G, Nijland W, Coops NC, Finnegan L. Heading to the hills? Evaluating spatial distribution of woodland caribou in response to a growing anthroprogenic disturbance footprint. Ecol Evol. 2016;6: 6484–6509. pmid:27777724
  96. 96. Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed-effects models using S4 classes. 2014.
  97. 97. Johnson CJ, Nielsen SE, Merrill EH, Trent L, Boyce MS, Science E, et al. Resource Selection Functions Based on Use–Availability Data: Theoretical Motivation and Evaluation Methods. 2006;70: 347–357.
  98. 98. Burnham KP, Anderson DR, Huyvaert KP. AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behav Ecol Sociobiol. 2011;65: 23–35.
  99. 99. Nobert BR, Milligan S, Stenhouse GB, Finnegan L. Seeking Sanctuary–the Neonatal Calving Period among Central Mountain Caribou (Rangifer tarandus caribou). Can J Zool. 2016.
  100. 100. DeCesare NJ, Hebblewhite M, Schmiegelow F, Hervieux D, McDermid GJ, Neufeld L, et al. Transcending scale dependence in identifying habitat with resource selection functions. Ecol Appl. 2012;22: 1068–1083. pmid:22827119
  101. 101. Avgar T, Lele SR, Keim JL, Boyce MS. Relative Selection Strength: Quantifying effect size in selection inference. 2017; 1–9. pmid:28770070
  102. 102. Kohavi R. A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection. 1995;
  103. 103. Boyce MS, Vernier PR, Nielsen SE, Schmiegelow FKA. Evaluating resource selection functions. Ecol Modell. 2002;157: 281–300.
  104. 104. Fielding AH, Bell JF. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ Conserv. 1997;24: 38–49.
  105. 105. Kuhn M. caret: Classification and Regression Training. 2014.
  106. 106. Hosmer DW, Lemeshow S, May S. Applied survival analysis: regression modeling of time-to-event data [Internet]. Wiley series in probability and statistics. Hoboken, N.J: Wiley-Interscience; 2008. Available: Contributor biographical information; Publisher description; Table of contents only
  107. 107. Waterhouse MJ, Armleder HM, Nemec A. Terrestrial lichen response to partial cutting in lodgepole pine forests on caribou winter range in west—central British Columbia. Rangifer Spec Issue. 2011;19: 119–134.
  108. 108. Cichowski DB, Haeussler S. The reponse of caribou terrestrial forage lichens to mountain pine beetles and forest harvesting in the East Ootsa and Entiako areas. Annual report—2012/13—Year 11. Smithers, BC; 2013.
  109. 109. Pec GJ, Karst J, Sywenky AN, Cigan PW, Erbilgin N, Simard SW, et al. Rapid Increases in Forest Understory Diversity and Productivity following a Mountain Pine Beetle (Dendroctonus ponderosae) Outbreak in Pine Forests. PLoS One. 2015;10: e0124691. pmid:25859663
  110. 110. Woodard PM. The effects of harvesting on lichen regeneration rates in west-central Alberta. Edmonton, AB; 1995.
  111. 111. Snyder J, Woodard PM. Lichen regeneration rates in Alberta following various types of logging and wildfire disturbances. Edmonton, AB; 1992.
  112. 112. Webb ET. Survival, persistence, and regeneraton of reindeer lichens, Cladian stellaris, C. rangiferina, and C. mitis following clearcut logging and forest fire in northwestern Ontario. Rangifer Spec Issue. 1998;10: 41–47.
  113. 113. Coxson DS, Marsh J. Lichen chronosequences (postfire and postharvest) in lodgepole pine (Pinus contorta) forests of northern interior British Columbia. Can J Bot. 2001;1464: 1449–1464.
  114. 114. Lafleur B, Zouaoui S, Fenton NJ, Drapeau P, Bergeron Y. Short-term response of Cladonia lichen communities to logging and fire in boreal forests. For Ecol Manage. 2016;372: 44–52.
  115. 115. Benedict JB, Nash TH. Radial growth and habitat selection by morphologically similar chemotypes of Xanthoparmelia. Bryologist. 1990;93: 319–327.
  116. 116. Dettki H, Klintberg P, Esseen P-A. Are epiphytic lichens in young forests limited by local dispersal? Ecoscience. 2000;7: 317–325.
  117. 117. Sancho LG, Green TGA, Pintado A. Slowest to fastest: Extreme range in lichen growth rates supports their use as an indicator of climate change in Antarctica. Flora. 2007;202: 667–673.
  118. 118. Graham L, Quintilio K. Willmore Wilderness Park fire management plan [Internet]. Edmonton, AB; 2006. https://open.alberta.ca/publications/3844388#detailed
  119. 119. Andison DW. Landscape-Level Fire Activity on Foothills and Mountain Landscapes of Alberta [Internet]. Belcarra, British Columbia; 2000. https://friresearch.ca/sites/default/files/null/HLP_2000_07_Rpt_LandscapeLevelFireActivityonFoothillsandMountainLandscapesofAlbertaNDreport2.pdf
  120. 120. Silva JA, Nielsen SE, Lamb CT, Hague C, Boutin S. Modelling Lichen Abundance for Woodland Caribou in a Fire-Driven Boreal Landscape. 2019;
  121. 121. Bradshaw C, Hebert DM, Rippin AB, Boutin S. Winter peatland habitat selection by woodland caribou in northeastern Alberta. Can J Zool. 1995;73: 1567–1574.
  122. 122. Johnson CJ, Parker KL, Heard DC. Foraging across a variable landscape: Behavioral decisions made by woodland caribou at multiple spatial scales. Oecologia. 2001;127: 590–602. pmid:28547497
  123. 123. Hébert I, Weladji RB. The use of coniferous forests and cutovers by Newfoundland woodland caribou. For Ecol Manage. 2013;291: 318–325.
  124. 124. DeCesare NJN. Resource selection, predation risk, and population dynamics of Woodland caribou. PhD. 2012;
  125. 125. Serrouya R, McLellan B, van Oort H, Mowat G, Boutin S. Experimental moose reduction lowers wolf density and stops decline of endangered caribou. PeerJ. 2017;5: e3736. pmid:28875080
  126. 126. Avgar T, Baker JA, Brown GS, Hagens JS, Kittle AM, Mallon EE, et al. Space-use behaviour of woodland caribou based on a cognitive movement model. J Anim Ecol. 2015;84: 1059–70. pmid:25714592
  127. 127. Wittmer HU, McLellan BN, Serrouya R, Apps CD. Changes in landscape composition influence the decline of a threatened woodland caribou population. J Anim Ecol. 2007;76: 568–579. pmid:17439473
  128. 128. Wittmer HHU, Sinclair AEAARE, McLellan BNB, Sinclair ÆARE, McLellan BNB. The role of predation in the decline and extirpation of woodland caribou. Oecologia. Springer-Verlag; 2005;144: 257–267. pmid:15891849
  129. 129. Seip DR. Factors limiting woodland caribou populations and their interrelationships with wolves and moose in southeastern British Columbia. Can J Zool. 1992;70: 1494–1503.
  130. 130. Serrouya R, Seip DR, Hervieux D, McLellan BN, McNay RS, Steenweg R, et al. Saving endangered species using adaptive management. Proc Natl Acad Sci U S A. 2019;116: 6181–6186. pmid:30858314
  131. 131. Dhar A, Parrott L, Hawkins CDB. Aftermath of mountain pine beetle outbreak in british columbia: Stand dynamics, management response and ecosystem resilience. Forests. 2016;7: 1–19.