Are forestry clearcuts ecological traps for caribou?

Abstract

Context

Woodland caribou (Rangifer tarandus caribou) declines are linked to increased predation facilitated by industrial land-use changes, including logging, through mechanisms such as disturbance-mediated apparent competition.

Objectives

We hypothesized clearcuts could exacerbate predation risk by functioning as ecological traps in low-productivity landscapes, drawing caribou during summer when nutritional demands are high. We predicted caribou would select clearcuts as or more strongly than other landcovers, with clearcuts providing comparable or greater forage but imposing fitness costs.

Methods

Using step-selection functions, we analyzed habitat selection for 71 GPS-collared caribou in west-central British Columbia, Canada. We assessed forage availability through vegetation surveys at 56 paired clearcut and uncut plots, and inferred predation risk from wolf (Canis lupus), coyote (C. latrans), grizzly (Ursus arctos), and black bear (U. americanus) detections from 149 camera traps.

Results

Caribou selected regenerating (≤ 25 years old) clearcuts as or more strongly than other landcovers, whereas older clearcuts (26–39 years old) were used neutrally or avoided. Compared to paired uncut areas, forage was greater in regenerating clearcuts, but not in older clearcuts. Predation risk was elevated in clearcuts and near roads, with higher activity of wolves (clearcuts), coyotes (roads), and black bears (both).

Conclusions

We provided partial evidence that regenerating clearcuts may function as ecological traps during summer in low‐productivity landscapes. Elevated predation risk does not necessarily equate to reduced fitness, and alternative mechanisms (e.g., risk-sensitive foraging) require further investigation. Regardless, selection of high-risk clearcuts by caribou represents an understudied pathway potentially contributing to population declines. We recommend limiting new clearcutting and improving protection of caribou summer ranges.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s10980-026-02411-1.

Introduction

Wildlife species inhabit heterogeneous landscapes composed of mosaics of habitat patches associated with varying fitness outcomes. Over evolutionary time, species have developed cue-response systems that guide individuals toward adaptive habitat selection, enabling them to recognize and avoid suboptimal habitats and thereby maximize fitness (Sih et al. 2011; Northrup et al. 2022). However, human-induced rapid environmental change (HIREC) can disrupt these evolved relationships by decoupling the environmental cues animals use from the actual quality of habitats. This may prevent individuals from detecting risks and could result in selection of habitats that lead to lower fitness than other available options, i.e., ecological traps (Robertson and Hutto 2006; Robertson et al. 2013). Such traps can often accelerate population declines and threaten the long-term persistence of species (Hawlena et al. 2010; Penteriani et al. 2018; Johnson H.E. et al. 2020; Singh et al. 2024).

Over the past century, HIREC has contributed to severe population declines in woodland caribou (Rangifer tarandus caribou; hereafter caribou) through widespread habitat loss (Festa-Bianchet et al. 2011; Latham et al. 2011; Johnson C.A. et al. 2020; Stewart et al. 2020; Lessard et al. 2025). As a result, the three recognized populations of woodland caribou in western Canada are listed under Canada’s Species at Risk Act as either Threatened (boreal, southern mountain) or of Special Concern (northern mountain) (Environment Canada 2012, 2014); Environment and Climate Change Canada 2020). The rapid expansion of forestry and energy development has degraded habitat and triggered unsustainable mortality for caribou from predators such as gray wolves (Canis lupus), and cougars (Puma concolor), primarily through two mechanisms. First, disturbance-mediated apparent competition (DMAC; Serrouya et al. 2021) arises when early-seral vegetation created by logging, wildfire, or energy development, increases forage availability for prey species such as moose (Alces alces), deer (Odocoileus spp.), and elk (Cervus canadensis). This leads to higher predator densities, and thus increasing caribou mortality due to opportunistic predation (Kinley and Apps 2001; Hebblewhite et al. 2007; Latham et al. 2011; Whittington et al. 2011; Peters et al. 2013). Second, linear features such as forestry roads and seismic lines enable wolves to move faster and farther into caribou habitats, potentially increasing their hunting efficiency (McKenzie et al. 2012; Dickie, et al., 2017, 2020; DeMars and Boutin 2018).

Caribou must try to avoid predators while meeting nutritional demands, which is especially difficult for mountain caribou during calving and summer seasons. During calving season (May–June), parturient females primarily adopt an anti-predator strategy by selecting high-elevation alpine and rugged terrain, distancing themselves from predators and other prey more common in valley bottoms (Bergerud et al. 1984; Bergerud and Page 1987; Cichowski 1989; Gustine et al. 2006; Apps et al. 2013). These high-elevation areas, however, offer limited foraging. By late June, nutritional demands peak due to lactation, depleted energy reserves from gestation, and protein deficits from a lichen-based winter-spring diet (Robbins 1993; Gerhart et al. 1996; Barboza and Parker 2008; Parker et al. 2009). To meet their nutritional needs, these caribou migrate to low-elevation areas where nutritious forage is more abundant (Bergerud et al. 1984; Cichowski 1989; Gustine et al. 2006). Use of these low-elevation areas increases exposure to predators and coincides with their highest seasonal mortality (Seip and Cichowski 1996; Wittmer et al. 2005; Whittington et al. 2011; Apps et al. 2013).

Elevated predation risk, driven by changes in predator distribution, abundance, or behaviour, is a common feature of ecological traps and is often coupled with increased food availability (Weldon and Haddad 2005; Robertson and Hutto 2007; Leighton et al. 2008; Hawlena et al. 2010; Robertson et al. 2013). Protein-rich plants preferred by caribou during summer are typically more abundant in early seral forests, such as those created by logging (hereafter clearcuts), than in mature stands (Denryter et al. 2017, 2022). Yet whether clearcuts can function as ecological traps, by drawing caribou toward attractive forage and away from areas where they could potentially segregate more effectively from predators and apparent prey (Rettie & Messier 2000; Tjaden-McClement et al. 2025), remains unclear.

While fine-scale selection of clearcuts during the summer has been documented in a few cases in boreal caribou (Dussault et al. 2012; Leclerc et al. 2014), caribou are generally understood to avoid anthropogenically disturbed habitats at both broad and fine scales (Johnson et al. 2015; Apps and Dodd 2017; Palm 2021; but see Apps 2020). However, vegetation communities with high nutritional value are low in availability during summer for mountain and boreal caribou in some areas such as northeastern British Columbia (BC). Consequently, nutritional inadequacies for lactating caribou are estimated to be widespread in these areas (i.e., inability to meet protein and energy requirements, even without replenishment of body reserves) (Denryter et al. 2022). Prevalence of poor body condition during summer has increased in these subpopulations, indicative of starvation and elevated vulnerability to other mortality sources (Cook R.C. et al. 2021). Coupled with evidence that summer forage strongly influences caribou fitness (Parker et al. 2009; Cook J.G. et al. 2021), these patterns highlight the need to investigate the short-term but potentially magnified risks associated with using clearcuts during summer. Examining the trade-off between forage acquisition and predation risk is particularly important in low-productivity landscapes, where low availability of nutritious forage may draw foraging caribou to clearcuts.

In this study, we aimed to examine whether clearcuts could function as ecological traps for mountain caribou in low-productivity landscapes during summer. We developed the three following predictions grounded in the key criteria defining ecological traps (Robertson et al. 2013):

1. Caribou select clearcuts over, or equally to, uncut landcovers. We analysed males and females separately to account for potential sex-specific behaviours (Harestad and Bunnel 1979; Ruckstuhl and Neuhaus 2002; Ofstad et al. 2016).

   1. We additionally investigated whether females with calves (i.e., lactating individuals with elevated nutritional demands) responded differently to clearcuts than females without calves. Females often adjust their habitat-use to reduce calf vulnerability during the neonatal phase (Viejou et al. 2018). However, as the season progresses into summer, the cumulative energetic costs of gestation and lactation, combined with the need to replenish depleted body reserves, can constrain continued anti-predator behaviour, particularly if clearcuts are not perceived to impose immediate fitness costs. As a result, we expected females with calves to select clearcuts as or more strongly than those without calves.

2. Clearcuts provide forage in similar or greater abundance compared to uncut areas.

3. Clearcuts are associated with higher fitness costs for caribou compared to uncut areas.

We used a combination of satellite telemetry tracking, vegetation surveys, and camera trap sampling to test our predictions. Additionally, as a supplementary post-hoc analysis, we further investigated whether caribou have shifted their summer home ranges toward clearcuts over time by comparing recent range use to an earlier baseline period when clearcuts were largely absent (Supplementary Information 1). We interpret our findings within the broader context of caribou–predator dynamics and potential alternative hypotheses, emphasizing the need for research that addresses the unique vulnerabilities of caribou subpopulations inhabiting low-productivity landscapes amid ongoing conservation challenges driven by increasing anthropogenic disturbances.

Methods

Study area

Our study area was within the Chilcotin Plateau of west-central BC, a region characterized by a cold, dry climate (average annual temperature = 4 °C; average annual precipitation = 400–500 mm) due to its position in the rain shadow of the Coast Mountains, resulting in relatively low net primary productivity and slow vegetation succession. Tjaden-McClement et al. (2025) used the remotely sensed enhanced vegetation index to characterize trends in vegetation productivity following disturbances in the Itcha-Ilgachuz caribou range, and found it to have significantly lower productivity than several other woodland caribou ranges in western Canada. The study area covered the 3540 km² summer range of the Itcha-Ilgachuz mountain caribou. During the calving season (mid-May to mid-June), female caribou use the high-elevation alpine habitats across the Itcha and Ilgachuz mountain ranges, while most males and subadults remain below the treeline, with some individuals moving to alpine areas from late June to early July (Cichowski 1989). The high-elevation winter-summer range (Fig. 1a; hereafter core summer range; mean elevation [range]: 1681 m [1277–2407]) includes sparsely vegetated alpine slopes, scrub birch (Betula glandulosa) or Altai fescue (Festuca altaica) in mesic sites, and subalpine parkland dominated by krummholz and meadow patches (Coupé et al. 1991; Himmer and Power 1999), and adjacent lower-elevation forested areas surrounding the Itcha and Ilgachuz mountains (Fig. 1a). The low-elevation summer range (Fig. 1a; hereafter peripheral summer range; mean elevation [range]: 1440 m [1159–1923]) refers to areas farther from the mountains to the east and south with predominantly forested habitat. We use the terms core and peripheral to better represent how these two ranges are spatially positioned relative to one another (Fig. 1a), rather than to imply core-matrix habitat definitions (Environment Canada 2014; Caribou Recovery Program 2025). Transitioning into summer (July–August), caribou increase their use of lower-elevation areas either within the core summer range (Cichowski 1989; Youds et al. 2002), or further into the peripheral summer range. The low-elevation areas are dominated by open even-aged lodgepole pine (Pinus contorta) forests, with hybrid spruce (Picea glauca × engelmannii) and trembling aspen (Populus tremuloides) occasionally found in wetland patches. Wetlands, broadleaf, and mixed-wood forests account for < 3% of the combined core and peripheral summer ranges. Common understory shrubs include common juniper (Juniperus communis), soopalallie (Shepherdia canadensis), and kinnikinnick (Arctostaphylos uva-ursi).

Fig. 1.

Study area in west-central BC, showing the distribution of camera traps deployed across the high-elevation summer range (i.e., core summer range) and low-elevation summer range (i.e., peripheral summer range) of Itcha-Ilgachuz caribou (a), and vegetation sampling plots in regenerating and older clearcuts and paired uncut sites within the peripheral summer range (b). The Itcha-Ilgachuz caribou range includes the high-elevation winter-summer range (HEWSR), low-elevation summer range (LESR), and low-elevation winter range (LEWR)

The landscape is highly affected by logging and wildfires (Nagy-Reis et al. 2021; Maltman et al. 2024). As of 2021, approximately 10% of the peripheral summer range had been harvested. In contrast, harvesting within the core range has been minimal (~ 1%), largely because most of the unforested alpine area lies within Itcha-Ilgachuz Provincial Park (Fig. 1, Fig. S3.1). Over the past century, wildfires have burned ~ 11% of the core, and ~ 27% of peripheral summer ranges. However, prior to the large megafires of 2017–2018, only about 7% of the peripheral summer range had burned (Tjaden-McClement et al. 2025). Extreme outbreaks of mountain pine beetle (Dendroctonus ponderosae) in the Chilcotin have also led to intense salvage logging and increased wildfire fuel loads (Alfaro et al. 2015; Sagar and Waterhouse 2015; Maltman et al. 2024).

In response to the decline in Itcha-Ilgachuz caribou (from more than 2800 individuals in 2003 to 385 in 2019; Young & Freeman 2003; Shores 2019; Lamb et al. 2024), a wolf reduction program was initiated by the provincial government in February 2020 and has been implemented each February since. As a result, wolf densities declined from an estimated 6.4 wolves per 1,000 km² in 2019 to 1.4 in 2020, 1.3 in 2021, and 0.8 in 2023 and 2024 (FLNRORD 2018b; Caribou Recovery Program 2021, 2022, 2023, 2024).

Caribou telemetry and reproductive status

We used geographic positioning system (GPS) data from 101 adult female and 32 male collared caribou, collected by the provincial government between 2011–2015 and 2018–2021. Caribou were captured via helicopter net-gunning and fitted with Vectronic Vertex Lite 2D (Berlin, Germany) or Lotek Lifecycle Pro 360 (Newmarket, Ontario, Canada) collars. Fix intervals were every 4 h in 2011–2015. In 2018–2021, collars recorded every 4 h from May to mid-July and every 12 h for the remainder of the year. Five individuals in 2021 had 2-h fixes from February to July and 4-h fixes in August before ceasing data collection. We excluded low-accuracy fixes (i.e., fixes made with < 4 satellites or < 3D for 2012–2014, and fixes with dilution of precision values ≥ 6 for 2018–2021 data), prior to further analysis (Table S2.1).

We used a modified version of the individual-based method (IBM) (DeMars et al. 2013; Gharajehdaghipour et al. 2025) to infer the reproductive status of collared females in the absence of direct parturition and calf survival data. The IBM classifies females into three groups based on shifts in their mean step length: (1) did not calve, (2) calved and the calf survived up to 4 weeks, (3) calved and the calf died within 4 weeks. The IBM only tracks calf survival up to 4 weeks, as the movement rates of females and calves tend to become indistinguishable after this period (DeMars et al. 2013). Most females gave birth in late May to early June (Gharajehdaghipour et al. 2025). Thus, IBM can monitor calf mortality only up to the end of June, before the start of summer season (15 July–15 Sep). However, given IBM’s high accuracy (DeMars et al. 2013; Gharajehdaghipour et al. 2025), we assumed that females who did not give birth or lost their calf experienced lower metabolic demands (due to absence or cessation of lactation) compared to females that retained their calf during the neonatal period. This assumption holds even if some females in the latter group might have lost their calf after IBM’s tracking period and before summer, as they would have experienced a longer lactation period compared to females that were nonparturient or lost their calf within the neonatal period. Our sample size after IBM application was 78 individual-years (Table S2.1).

Landcover layers

We used National Terrestrial Ecosystem Monitoring System geospatial layers corresponding to clearcuts, fires and landcover categories developed based on data from the Landsat series of satellites (Hermosilla et al. 2016, 2018, 2022; Additional details on landcover designation is provided in Supplementary Information 3). We excluded a small proportion (< 5%) of harvested areas cut according to an irregular group shelterwood system (Waterhouse et al. 2011) as our analysis focused solely on conventional clearcuts. We used conifer, shrub, and herbaceous landcover classes as natural (i.e., non-anthropogenic) landcovers, excluding pixels with disturbance.

Habitat selection analysis

To assess caribou response to clearcuts at fine spatial scales (i.e., third-order selection within home ranges; Johnson 1980), we used step selection functions (SSFs; Fortin et al. 2005; Thurfjell et al. 2014; Muff et al. 2020). We defined summer as 15 July–15 September to exclude early summer movements related to the calving period and late summer/fall movements associated with returns to core summer range at the start of the rut season in late September. We limited the telemetry data to this period. As a result, we excluded 41 individual-years due to insufficient fixes (i.e., < 50% fix rate, Table S2.1). Approximately 30% (8 of 27) of females with calves, 12% (5 of 42) of females without calves, and 12% (2 of 17) of males remained entirely in the core summer range with no clearcut exposure, resulting in no within-individual variation in this covariate. We excluded these individual-years (17%; 15 of 86) from analysis, as their inclusion would lead to shrinkage toward the population mean and potentially biased inference about clearcut selection. Our final sample size for SSF analysis included 19, 37, and 15 individual-years for females with calves, females without calves, and males, respectively (total of 71 individual-years; Table S2.1).

To standardize across individual-years with different fix rates, we resampled GPS tracks to 12 h. For each movement step of an individual-year, we characterized available habitat by generating 10 random steps from the same starting point by resampling from a gamma distribution of observed step lengths and a von Mises distribution of observed turning angles, based on all telemetry points from that individual-year (Duchesne et al. 2015; refer to Supplementary Information 2 for a sensitivity analysis).

We fit a separate conditional Poisson generalized linear mixed effects model (GLMM) for each caribou sex-reproductive class (hereafter class) using the glmmTMB package (Brooks et al. 2017; Muff et al. 2020). To do so, we included a random intercept by step ID and random slopes for all fixed effects by individual-year, and used a large (10⁶) fixed intercept variance (Muff et al. 2020). For fixed effects, we extracted the proportion of area covered by clearcuts within a 500 m radius around the end of each used and available step in two age classes: regenerating (≤ 25 years old) and older (26–39 years old), along with proportions of conifers, shrubs, and herbs (Supplementary Information 2.1). The 25-year threshold reflects the provincial definition of early seral habitat for clearcuts (FLNRO 2017), and was used by Mumma et al. (2021) to record moose response to clearcuts based on vegetation regeneration data. Clearcuts ≥ 40 years old were scarce across the Itcha-Ilgachuz caribou range (Lamb et al. 2025; Supplementary Information 3). Although caribou response to fire was not a primary focus of our study, fire is a well-established driver of caribou habitat selection (Palm et al. 2022; Stevenson et al. 2024), and thus proportion of regenerating burns (≤ 25 years old) was also included as an additional landcover variable. We excluded other landcover classes including older burns, and ones with multiple legacies of disturbance in our analysis as < 1% of all available steps fell within these landcovers. Additionally, we included the distance from each step’s end point to the nearest road (MOF 2025), and transformed this variable using an exponential decay function (1 − e^{−α × distance}, α = 0.5) to account for the decreasing influence of roads on caribou habitat selection as the distance between a road and a caribou location increases (Nielsen et al. 2009; MacNearney et al. 2021). The exponentially decayed distance to roads becomes increasingly close to 1 as distance increases beyond 5 km.

We also included elevation (GeoBC 2014), slope (Horn 1981; raster package [Hijmans 2023]), terrain ruggedness (Sappington et al. 2007; MultiscaleDTM package [Ilich et al. 2023]), and distance to water (Natural Resources Canada 2023) at the end of each used and available step in our models as these variables are often associated with caribou habitat selection (e.g., Mumma et al. 2018; Russell et al. 2025). We exponentially decayed distance to water (as described above). We additionally included the step-length between two consecutive steps as an additional covariate in our models to reduce bias in parameter estimations (Forester et al. 2009).

Prior to fitting models, we tested for collinearity between covariates using variance inflation factor (VIF) with threshold value of 3 (Zuur et al. 2009). Elevation VIF exceeded 3 for all caribou classes and thus was omitted from models. The VIF for proportion of conifers was slightly above 3 for all classes (< 3.08), but we retained this covariate due to the almost ubiquitous cover of lodgepole pines in the peripheral summer range. We instead omitted the proportion of herbaceous cover as this landcover was fairly uncommon (Fig. S3.1; ~ 3% of all available steps fell within this landcover). This omission decreased the VIF of proportions of conifers below 3. We evaluated the predictive strength of our model using k-fold cross-validation (k = 5) blocked by random individual-years (Boyce et al. 2002; Roberts et al. 2017). We then used Spearman’s rho to measure correlation between the mean area-adjusted frequencies and 10 equal-area bins as well as frequencies by bin for each withheld set (Boyce et al. 2002).

Functional responses in habitat selection occur when individuals modify their selection based on habitat availability (Mysterud and Ims 1998; Hebblewhite and Merrill 2008; Aarts et al. 2013; Griffin et al. 2025). Investigating functional responses provides insight into how individuals respond to a range of conditions, including evolutionarily novel habitats such as clearcuts. For example, habitat selection behaviours can be informed by recent and long-term memory (i.e., learning; Rheault et al. 2021). Thus, we expected individuals with higher potential exposure to clearcuts to exhibit stronger selection for these areas, if clearcuts are not perceived to reduce fitness. We tested for evidence of functional responses to regenerating clearcuts in each caribou class by running a linear model with the selection coefficient for regenerating clearcuts as the response variable and the mean availability of regenerating clearcuts across movement-defined neighborhoods as the predictor variable (Gigliotti et al. 2023; details in Supplementary Information 2).

As a post-hoc analysis, we further investigated whether caribou have shifted their summer home ranges toward clearcuts over time by comparing recent range use to an earlier baseline period when clearcuts were largely absent (Supplementary Information 1).

For this and all following analysis, we centred and scaled (mean = 0, SD = 1) continuous explanatory variables before fitting models, and conducted all analyses in R Statistical Software (R Core Team 2025). We report results using the language of evidence, which is based on gradual interpretation of P-values rather than a dichotomous classification of significance (Muff et al. 2022).

Vegetation sampling and estimating caribou forage availability

We used vegetation sampling to compare caribou forage availability between clearcuts and uncut areas within the peripheral summer range. From 6 July to 6 August 2021, we surveyed 28 paired sites, each consisting of one clearcut and one uncut (control) plot (n = 56; Fig. 1b). Clearcuts spanned seven age classes in 5-year intervals (4 blocks per interval) from 1983 (age class 8) to 2019 (age class 1), with no clearcuts occurring between 2006 and 2010. For each interval, clearcuts were chosen randomly. Uncut plots were also selected randomly in conifer landcover within 2 km of their paired clearcut and at least 100 m from the nearest edge to reduce edge effects. At each plot, we established a 3.99 m radius (50 m²) circular sampling area and recorded the dripline cover proportion of plant species using the species-by-layer method (FLNRO 2016). Plants were identified to species or genus when species-level identification was not possible. We defined acceptable forage (hereafter forage) as species not avoided by caribou (lists given in Denryter et al. 2017; Cook et al. 2023). We grouped plots into three categories: regenerating clearcuts (n = 16), older clearcuts (n = 12), and controls (n = 28, 16 paired with regenerating, and 12 with older clearcuts, respectively) and compared forage proportion across these categories using a beta-distributed GLMM, with a random intercept for pair ID.

Camera trap deployments and estimating predation risk

We used predator counts from camera traps as a proxy for predation risk to caribou, as predation-caused mortality data for collared caribou prior to onset of wolf reduction (i.e., 2020) that could be linked with caribou habitat selection behaviour was scarce (n = 5). This limitation was primarily due to lack of georeferenced mortality data in 2012–2014, and collar malfunctions resulting in low fix rates for all individuals in 2018 and 35% of individuals in 2019 (Table S2.1).

We deployed 149 camera traps (Reconyx Hyperfire 2, Holmen, WI, USA) in a systematic sampling design across 5 grids of 30 cameras each (except one grid with 29 cameras) across the core (n = 3) and peripheral (n = 2) summer range of caribou in September 2020 (Fig. 1a). Grids were rectangular and covered ~ 200 km² each, with most cameras evenly spaced about 2.6 km apart (minimum of 1.4 km). Cameras were revisited between June–Aug of 2021–2024 to retrieve images. Supplementary Information 5 provides further information on camera deployments and image processing. To reduce the influence of repeat images of the same individuals during a short period, we organized predator detections into independent detection events, defined as consecutive images of the same species at the same camera, separated by the conventional minimum of 30 min (Burton et al. 2015). For each event, we counted the number of individuals present and summed these counts across all events at each camera during the 2021–2024 summer periods (15 July–15 Sep). This total count served as our response variable, which represented predator habitat-use intensity across cameras (Burton et al. 2015). Given that cameras and caribou telemetry locations overlapped (Fig. S5.1), we assumed that predator habitat-use provided a proxy for predation risk to caribou in the area surrounding each camera over the sampling period (i.e., risky places; sensu Dröge et al. 2017) (Tattersall et al. 2020; Darlington et al. 2022; Tjaden-McClement et al. 2025). Cameras without detections were assigned a total count of 0. Although not all wildlife use is captured by cameras (Burton et al. 2015), we standardized deployment features (e.g., camera model, height, distance to target; Supplementary Information 5) to minimize variability in imperfect detections among cameras, and opted not to model this variation as sampling error, as is done in occupancy frameworks (Burton et al. 2015).

We focused our analysis on wolves, grizzly bears (Ursus arctos), black bears (U. americanus) and coyotes (C. latrans). Wolves and bears are major predators of both adult and calf mountain caribou (Seip and Cichowski 1996). Prior to the onset of wolf reduction, wolf predation accounted for 60% (3 of 5) of collared caribou mortalities with known causes during summer 2018–2019, while bear predation accounted for 20% (1 of 5). The remaining mortality was attributed to health-related causes. We included coyotes in our analysis because, in the absence of wolves, they contribute significantly to caribou calf mortality in eastern Canada (Mahoney et al. 2016; Mumma et al. 2019a). Wolverines (Gulo gulo) are also key predators of caribou calves during the neonatal period, before calves gain sufficient mobility (Gustine et al. 2006). However, since our study was outside this period, we did not include wolverines in our analysis. We also excluded cougars, as they were detected at only 3 cameras.

We used negative binomial GLMMs to model total counts of each predator at each camera, using the same predictors as in the caribou models. Landcover was summarized within 500 m buffers around each camera station. Distance metrics were transformed as described previously. We included a random intercept for grid ID to account for spatial clustering among stations,and the log-transformed sampling effort (i.e., the number of active camera days per station) as an offset.

We omitted elevation from our models due to its VIF exceeding the threshold value (VIF = 4.35). We estimated the explanatory power of the model by calculating both the marginal and conditional R² values (Nakagawa et al. 2017; performance package [Lüdecke et al. 2021]), and assessed model assumptions and fit using residual diagnostic plots in the DHARMa package (Hartig 2022).

While responses of potential apparent competitors such as moose and mule deer (Odocoileus hemionus) were not a primary focus of this study, we conducted a supplementary analysis to examine habitat associations of these species during our focal period using camera trap detections. This analysis was intended to provide insight into whether caribou spatially segregate from potential apparent competitors (Rettie & Messier 2000; Supplementary Information 5).

Results

Caribou habitat selection

All caribou classes selected areas with a higher proportion of regenerating clearcuts. The evidence for this selection was strong for females and weak for males (Figs. 2, Fig. S2.1, Fig. S2.2; full model estimates are reported in Table S2.2). Additionally, we found evidence of a functional response in females, where selection for regenerating clearcuts increased with their mean availability (Supplementary Information 22; Table S2.4). Females with and without calves showed comparable selection for regenerating clearcuts (i.e., overlapping 95% confidence intervals; Fig. 2). However, females with calves generally exhibited higher selection coefficients for regenerating clearcuts than females without calves and males (Figs. S2.1, Fig. S2.2). Females with calves and males avoided older clearcuts. However, this variable was not associated with habitat selection by females without calves. The models indicated avoidance of roads for all classes. The evidence for this avoidance was strong for females, and weak for males. Caribou response to natural landcovers was similar across all classes. Models showed strong evidence for selection of shrubs, and avoidance of conifer landcovers. Caribou response to regenerating burns varied (males: negative, females with calves: neutral, females without calves: positive). Results from blocked cross-validation indicated that models for all three caribou classes had strong predictive power (females with calves: r_s = 0.99, P < 0.0001; females without calves: r_s = 0.99, P < 0.0001; males: r_s = 0.96, P < 0.0001; Spearman test results for each fold separately are provided in Table S2.3).

Fig. 2.

Coefficient estimates (white points) with 95% confidence intervals (color fades) from habitat selection models for 71 individual-years across 3 classes of caribou in west-central BC during summer (15 July–15 September, 2012–2014 and 2019–2021), as a function of disturbed and natural landcovers and control variables. Numbers to the right of the graph indicate P-values

Forage availability

The most commonly detected species were common juniper, twin flower (Linnaea borealis), and reindeer lichens (Cladina spp.). Acceptable forage with the highest mean cover was grouseberry (Vaccinium scoparium) in control areas, fireweed (Chamerion angustifolium) and soopolallie in regenerating clearcuts, and soopolallie in older clearcuts. Regenerating clearcuts had a higher cover of herbaceous forage compared to uncut plots (Fig. 3; full model estimates are reported in Table S4.1). There was weak evidence that deciduous shrub cover was greater in regenerating clearcuts than uncut plots. Lichen coverage was lower across regenerating clearcuts. Older clearcuts also had greater coverage of herbaceous forage compared to uncut plots (Fig. 3; Table S4.1). Deciduous shrub cover was similar in older clearcuts and uncut areas. Lichen coverage was lower across older clearcuts compared to uncut areas. When all acceptable forage types (lichens, herbs, and deciduous shrubs) were combined, regenerating clearcuts had 1.7-fold greater coverage than control sites. In contrast, we found no evidence of a difference in combined forage cover between older clearcuts and uncut areas.

Fig. 3.

Estimated mean forage cover and SE from models comparing caribou summer forage cover in regenerating and older clearcuts to paired uncut areas. Forage cover was measured at 56 vegetation plots surveyed in July–August 2021 across the peripheral summer range of Itcha-Ilgachuz caribou in west-central BC. Numbers above bars indicate P-values

Predation risk

Cameras recorded 98 wolf, 231 coyote, 44 grizzly bear, and 365 black bear independent detections across 32,685 camera days during the focal period. Wolves, coyotes, grizzly, and black bears were detected across 44, 45, 27, and 85 cameras respectively. The majority of detections were of solitary animals (wolves: 67%; coyotes: 89%; grizzly bears: 95%; black bears: 98%), with the largest group sizes being 11 for wolves, 3 for coyotes, 2 for grizzly bears and 3 for black bears.

Wolves were associated with areas with higher coverage of regenerating and older clearcuts (Fig. 4; full model output in Table S5.1). Wolves' response to roads was negative, however the evidence for this was weak. Wolves' response to conifer landcover was neutral. Coyotes were positively associated with roads, and responded negatively to conifer landcover. We found no evidence for the impact of regenerating or older clearcuts on coyotes. Grizzly bears were not associated with clearcuts, or roads, and were negatively associated with conifer landcover. Conversely, black bears were associated with regenerating clearcuts and roads, and did not respond to older clearcuts. Response by all species to shrub landcover was neutral. The models explained 19, 53, 21, and 19% of variability in total counts of wolves, coyotes, grizzly bears, and black bears, respectively.

Fig. 4.

Regression coefficient estimates (white points) with 95% confidence intervals (color fades) from models describing summer (15 July–15 September, 2021–2024) predator camera trap total counts, based on disturbed and natural landcovers, and control variables. Data were collected from 149 camera traps deployed across five grids spanning the core and peripheral summer ranges of Itcha-Ilgachuz caribou in west-central BC. Numbers to the right of the graph indicate P-values

Discussion

Wildlife face a range of direct and indirect anthropogenic stressors, leading to declines in many species globally (Díaz et al. 2019). Logging is a widespread driver of landcover change that can have negative effects on many mammal species (Bowman et al. 2010; Venier et al. 2014; Pfeifer et al. 2017). For caribou, clearcuts can increase predation risk through mechanisms such as DMAC and the expansion of road networks (Dickie et al. 2017; Mumma et al. 2018; Lochhead et al. 2022). We found evidence consistent with predictions from the ecological trap hypothesis, representing a potential additional pathway through which logging may contribute to caribou declines. Caribou that moved into the peripheral summer range during summer selected regenerating clearcuts at fine scales (i.e., third-order selection; Johnson 1980). This selection was stronger than for undisturbed coniferous forests in all three caribou classes and regenerating burns in males and females with calves (as indicated by non-overlapping 95% CIs), and comparable to that for shrub-dominated landcover in all classes. In our study area, shrub-dominated landcover was largely confined to core summer range, whereas clearcuts were located in peripheral summer range and primarily surrounded by coniferous forest (Fig. S3.1).

The higher forage availability observed in regenerating clearcuts relative to uncut areas is also consistent with our hypothesis that these clearcuts are attractive to caribou during summer. Regenerating clearcuts were also associated with greater use by wolves and black bears, and thus likely elevated predation risk, consistent with previous literature (Muhly et al. 2019; Mumma et al. 2019b; Dickie et al. 2020; Tjaden-McClement et al. 2025). The increased use by black bears is likely linked to the higher forage availability in regenerating clearcuts (Best et al. 2024; Colton et al. 2021). The positive response by wolves was consistent with our finding of increased moose use of these habitats (Table S5.2) and with previous studies (McKay and Finnegan 2023; Best et al. 2024).

Caribou did not select older clearcuts and thus older clearcuts will likely not function as ecological traps. This behavioral pattern was consistent with our finding that older clearcuts offered no forage advantage over uncut areas, providing support for a forage-driven mechanism for clearcut selection and their potential as ecological traps.

Females with calves selected regenerating clearcuts as strongly as those without calves. While compared to other classes, a greater proportion of females with calves avoided clearcuts during summer by remaining within the core summer range and the immediate adjacent areas, many of those that did move to the peripheral summer range showed a stronger selection for clearcuts than individuals in other classes. Caribou face elevated nutritional demands between late spring and early fall (Cook J.G. et al. 2021). For instance, lactation can increase energy demands by 65–215% (Oftedal 1985; Robbins 1993). The greater forage availability in regenerating clearcuts is therefore consistent with lactating individuals exhibiting the strongest selection to meet their higher energetic demands. These findings also highlight the importance of individual-level inference in habitat selection studies (Jones et al. 2020; Muff et al. 2020), as ecological traps result from maladaptive habitat-use decisions and their fitness consequences at the individual level, which can scale up to population-level patterns such as attractive sinks (Delibes et al. 2001).

Consistent with our main results, our post-hoc analysis of historical home ranges revealed that Itcha-Ilgachuz caribou shifted their summer home ranges toward areas with regenerating clearcuts in the 1990s, 2010s, and 2020s compared to the 1980s, when clearcuts were largely absent (Supplementary Information 1). In the 1980s, caribou made short migrations from high-elevation calving areas to nearby low-elevation areas within the core summer range. Conversely, in later decades, many caribou home ranges overlapped with clearcuts, which were primarily harvested in the peripheral summer range, at least ~ 30 km from the low-elevation areas within the core summer range used exclusively in the 1980s. Our results are consistent with Lamb et al. (2025), who found that Itcha-Ilgachuz caribou increased migration distances southward from their summer range to winter range over time in response to disturbance. Similarly, our results suggest an additional long-distance shift between calving and rut seasons (i.e., summer), likely driven by clearcut emergence. In both cases, these shifts appear to expose caribou to riskier habitats: the shift in winter range has brought caribou into lower elevation areas with higher densities of predators like cougars and potential apparent competitors such as white-tailed deer (Odocoileus virginianus) and feral horses (Equus caballus) (White et al. 2020; Tjaden-McClement et al. 2025). Likewise, our study shows that the summer home range shift toward regenerating clearcuts likely increases exposure to predators such as wolves, coyotes, and black bears.

Logging roads can contribute to caribou population declines independent of post-disturbance increases in apparent competitors (Mumma et al. 2018; Lochhead et al. 2022). In our study area, coyotes and black bears were associated with roads. Use of roads and other linear features (e.g., seismic lines) by these species has been commonly reported (Latham et al. 2013; Tattersall et al. 2020; Tjaden-McClement et al. 2025). We detected a neutral to weak negative association between wolves and roads. Although wolves are typically positively associated with anthropogenic linear features (Dickie et al. 2020), avoidance of roads has been reported in areas where wolf populations are reduced (Baillie-David et al. 2024). This behavior may reflect avoidance of humans, as wolves face high mortality risk from humans (e.g., aerial shooting along roads) (Cristescu et al. 2013; Llaneza et al. 2016; Baillie-David et al. 2024). Additionally, our camera survey began after the onset of wolf reduction program, and given ongoing annual reductions, the population likely remains in a near-constant recolonization state. Consequently, detections could reflect the movements of transient or recolonizing individuals which may exhibit different habitat-use patterns compared to established packs (Roffler and Gregovich 2018). Caribou also avoided roads, a behaviour documented in other herds (Beauchesne et al. 2013, 2014; Johnson et al. 2015). However, we expect that caribou use of clearcuts, combined with the general proximity of many road segments to clearcuts at fine scales, could increase the risk of encounters with multiple predators and thus predation risk.

Alternative hypotheses and research gaps

Although our findings are consistent with several predictions of the ecological trap framework, they do not fully satisfy all criteria required to demonstrate an ecological trap. This motivates consideration of alternative processes that could generate similar patterns of habitat selection. One such alternative is density‐dependent habitat selection (Rosenzweig 1981; Morris 2003; Webber et al. 2024). Under this framework, selection for regenerating clearcuts could arise if these habitats offer higher forage availability (i.e., reduced competition) relative to core summer ranges, even when associated with elevated predation risk. The small proportion of individuals remaining in core summer ranges and thus avoiding clearcuts at the second-order of selection may experience lower predation risk but also lower forage availability (i.e., higher competition), such that the net fitness consequences for individuals selecting clearcuts are not necessarily lower relative to those remaining in core summer ranges. However, our post-hoc analysis of subpopulation-level home ranges suggests that density dependence may not fully explain patterns of space use over multiple decades. The subpopulation roughly doubled from ~ 1400 individuals in the 1980s to ~ 2800 in 2003 (FLNRORD 2018a), and the higher density may have driven the expansion of individual summer home ranges from the core summer range into peripheral summer range with clearcuts. Nevertheless, subpopulation sizes in the 1990s (~ 1550), 2010s (~ 1300), and 2020s (~ 450) were similar to or smaller than those in the 1980s, yet caribou did not revert back to the core summer range used during that period of similar density. Rather, caribou continued to use areas near clearcuts, which seems inconsistent with density-dependent habitat selection.

Another plausible alternative is the risk-sensitive foraging hypothesis (Lima and Dill 1990), which predicts that during energetically demanding periods individuals may select habitats with higher predation risk in exchange for energetic benefits, without implying maladaptive decision-making. In ungulates, tolerance of predation risk varies with nutritional state, with food-limited individuals more likely to prioritize foraging opportunities over risk avoidance (Oates et al. 2019). Under this framework, selection for regenerating clearcuts may reflect adaptive trade-offs between energy acquisition and safety, particularly if landscape-level changes have altered the spatial distribution of forage and risk.

Both density-dependent habitat selection and risk-sensitive foraging can therefore produce patterns of selection for higher-risk habitats similar to those observed in this study, without requiring a mismatch between habitat cues and fitness outcomes. By contrast, ecological traps arise specifically from maladaptive responses that reduce fitness. Distinguishing among these competing mechanisms requires direct evidence that selection for regenerating clearcuts results in disproportionate fitness costs, such as reduced survival or reproductive success for individuals selecting clearcuts relative to those selecting alternative habitats. Accordingly, future work that explicitly links habitat selection to individual fitness outcomes will be essential for determining whether clearcuts function as ecological traps or instead reflect adaptive, context-dependent responses to spatial variation in resources and predation risk. Relatedly, we recommend that future research examine how cumulative habitat disturbances, including logging and climate-driven events such as large-scale mountain pine beetle outbreaks and wildfires (Griffin et al. 2025; Stevenson et al. 2025) have reshaped the spatiotemporal distribution of risks and rewards across landscapes, potentially altering the energetic trade-offs influencing caribou space use. As part of this, expanding vegetation surveys to include shrub- and herb-dominated landcovers, particularly in the core summer range, would improve understanding of habitat quality across the broader landscape. We additionally acknowledge that our telemetry resolution did not allow direct inference about fine-scale foraging decisions, and vegetation cover does not necessarily reflect realized diet. Thus, we encourage future work to explicitly link foraging behaviour, forage availability, nutritional quality, and diet composition in caribou inhabiting low-productivity ranges (Denryter et al. 2017, 2022). Finally, camera trap counts of predators may provide an incomplete representation of predation risk, given that predation is a multi-stage hierarchical process involving spatial overlap, encounter, attack and kill probabilities. Thus, future research examining the entire predation sequence would improve our inference on predation risk for caribou in heterogeneous landscapes (Suraci et al. 2022).

Landscape management implications

While caribou are generally thought to reduce predation risk by spatially segregating from predators and their primary prey, our results show that caribou in a low-productivity landscape selected regenerating clearcuts, areas that offered higher forage while exposing caribou to elevated predation risk from wolves, black bears, and coyotes. In contrast, older clearcuts may not pose the same risk, as they were not associated with increased forage availability, and caribou did not select them. These findings suggest that recovery of logged habitats in the summer range, such that they no longer attract caribou, may require up to 25 years. During this period, we recommend against harvesting of new clearcuts. Such additional disturbance risks exacerbating the problem by further increasing female caribou selection for these high-risk areas, as indicated by the functional response we observed to clearcut availability. While adopting silvicultural practices that reduce herbaceous and deciduous forage in clearcuts may help decrease the attractiveness of these areas to caribou, such management actions could negatively affect other ungulates, such as moose, which are an important contributor to food security and food sovereignty for many Indigenous communities (Popp et al. 2019; Alderhill Planning Inc. 2022; Werner et al. 2022; McKay and Finnegan 2023). More broadly, our study adds to the growing calls for stronger conservation and recovery actions for caribou habitats (Palm et al. 2020; Nagy-Reis et al. 2021; Maltman et al. 2024). Greater overlap of federally designated critical habitat, and provincially designated protected areas and no-harvest zones with caribou low-elevation summer ranges (FLNRO 2011); Environment Canada 2014), particularly undisturbed patches with high nutritional value, is needed.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CRediT statement—T.G.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing (original draft), Writing (reviewing and editing). K.T.-M.: Data curation, Investigation, Methodology, Project administration, Supervision, Writing (reviewing and editing). T.M.: Methodology, Supervision, Writing (reviewing and editing). J.K.: Supervision, Writing (reviewing and editing). Z.K.: Data curation, Investigation, Writing (reviewing and editing). A.C.B.: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing (reviewing and editing).

Funding

This research was supported by the British Columbia Caribou Recovery Program under the Ministry of Water, Land and Resource Stewardship, University of British Columbia, Natural Sciences and Engineering Research Council of Canada Discovery Grant (grant number RGPIN-2018–03958) and Canada Research Chairs program, and the Audain Foundation.

Data availability

Caribou telemetry and mortality investigation data are subject to the Government of British Columbia restrictions and can be requested directly from them. Camera trap and vegetation survey data are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.