Habitat loss on seasonal migratory range imperils an endangered ungulate

1. Endangered species policies and their associated recovery documents and manage-mentactions do notalways sufficientlyaddress theimportance ofmigratory behaviour and seasonal ranges for imperilled populations. 2. Using a telemetry location dataset spanning 1981–2018, we tested for changes in prevalence of migratory tactics (resident, migrant) over time, switching between tactics, shifts in seasonal space use including migration corridors, and survival consequences of migrant and resident tactics for 237 adult female endangered woodland mountain caribou in one population in western Canada. 3.Overmorethanthreedecades,theproportionofindividualsdisplayingannualmigra-tion habitats needed for the entire life history of the species, including all seasonal ranges and migratory habitat.

habitats needed for the entire life history of the species, including all seasonal ranges and migratory habitat.

K E Y W O R D S
Ecological trap, habitat loss, migration, Rangier tarandus caribou, survival, woodland caribou INTRODUCTION Migration is a crucial ecological phenomenon that contributes to maintaining biodiversity, supports populations of many species, and connects ecosystems across spatial scales (Bauer & Hoye, 2014;Dingle, 1996;Milner-Gulland, Fryxell, & Sinclair, 2011). However, across the globe, species that migrate are disproportionately vulnerable to extinction, threatening the persistence of this behaviour and its ecological functions (Berger, 2004;Bolger, Newmark, Morrison, & Doak, 2008).
There are notable examples of government policies and laws to accommodate migratory species, such as the North American Migratory Bird Convention and associated legislation. Yet, for many migratory species, important migration areas and habitats are not conserved or adequately managed. This can occur due to a lack of information on migratory behaviour and its demographic benefits, and due to challenges in implementing conservation strategies across jurisdictional boundaries (Bolger et al., 2008;Milner-Gulland et al., 2011;Pierce, Bleich, Wehausen, & Bowyer, 1999).
Migratory terrestrial species could benefit from approaches that identify and conserve seasonal ranges and migration corridors (Berger, 2004;Bolger et al., 2008;Seidler, Green, & Beckmann, 2018). For example, 20 years ago researchers in Wyoming identified a 160km pronghorn (Antilocapra americana) migration route extending from Grand Teton National Park into unprotected lands threatened by energy and housing development (Sawyer & Lindzey, 2000). Berger and Cain (2014) highlighted the conservation difficulties associated with protecting this population's well defined, but narrow, migratory route and seasonal ranges, which together, span multiple land ownership and management types. Identification of this migration route and emphasis of its importance has led to numerous protections ), yet the migration route and seasonal ranges are still not yet protected in entirety. Summer range was protected because of its location within a national park, and the migratory route itself benefited from a suite of conservation measures; yet habitat degradation on winter range continued (Middleton et al., 2020;Sawyer, Beckmann, Seidler, & Berger, 2019). This case study highlights the need to consider different threats and solutions on the entire seasonal migratory cycle for such species.
Other terrestrial mammal migrations have fewer protective measures than the pronghorn example, and future conservation will be unsuccessful unless key conditions are met. First, conservation of all seasonal ranges needs to be a core element of a migratory conservation strategy (Johnstone et al., 2020;Kauffman et al., 2018;Peters et al., 2019). In the pronghorn case, one seasonal range and the migration corridor benefitted from various protections, although erosion of the winter range continued, eventually resulting in altered pronghorn behaviour and possible demographic consequences . For other species, protections focus on one or more seasonal ranges, but the land used during migration itself is overlooked (Runge et al., 2015). Second, conservation of static migration corridors and stopovers will only be effective if animals consistently use defined migration corridors and stopovers. Again in the pronghorn example, individuals in this population used a narrow corridor with high fidelity (Middleton et al., 2020), but this is not the case for other species and populations. Studies show that some terrestrial mammals demonstrate variation in migratory tactics (e.g. timing of migration and route an individual uses) within populations and among years (Berg, Hebblewhite, St. Clair, & Merrill, 2019). For ungulates in particular, flexibility in migratory behaviour frequently results in uncertainty in the extent to which migration corridors are predictable; individuals within a population use different routes, and the route any individual uses may vary over time (Bolger et al., 2008;Cagnacci et al., 2016). Thus, while terrestrial mammal migratory populations are globally threatened, usually by anthropogenic land use (Berger, 2004;Dobson et al., 2010;Harris, Thirgood, Hopcraft, Cromsigt, & Berger, 2009), the inherent variability of migratory behaviour can often make meeting these key conditions challenging.
Caribou (Rangifer tarandus) populations demonstrate some of the most pronounced migratory behaviours among ungulates (Joly et al., 2019), and the species is threatened and declining across North America (Festa-Bianchet, Ray, Boutin, Côté, & Gunn, 2011;Hervieux et al., 2013;C. J. Johnson, Ehlers, & Seip, 2015;Wittmer, McLellan, Serrouya, & Apps, 2007). Yet, within and across caribou populations, there is wide variability (Gurarie et al., 2019). Migratory behaviour varies between barren-ground caribou, where individuals migrate 1,000′s of kilometres; mountain woodland caribou, which migrate between alpine habitats and low elevation forests; and sedentary boreal woodland caribou (Joly et al., 2019). Boreal and mountain caribou are increasingly threatened by anthropogenic land use change (e.g. C. J. Johnson et al., 2015;Palm, Jacob, Fluker, Nesbitt, & Hebblewhite, 2020;Wittmer et al., 2007), and the majority of woodland caribou populations are declining (Festa-Bianchet et al., 2011) with many notable recent extirpations. The causes of these declines are anthropogenic, resulting from direct and indirect habitat loss, increased efficiency of predators, and altered food-webs, which negatively affect caribou through apparent competition (Holt, 1977;Serrouya et al., 2019;Wittmer et al., 2007). Conservation of woodland caribou is one of North America's most pressing management issues, affecting boreal biodiversity conservation and carbon sequestration through the role of caribou as an umbrella species for both (Bichet, Bradshaw, Warkentin, & Sodhi, 2009;Yona, Cashore, & Schmitz, 2019). Their conservation is complex in consideration of the large economic benefits available from the exploitation of natural resources in and adjacent to woodland caribou ranges (Hebblewhite, 2017).
Woodland caribou conservation and recovery is mandated nationally in Canada through the federal Species At Risk Act (SARA; Government of Canada). Like many endangered species policies globally, associated recovery strategies contain the central elements of identification and conservation of critical habitat (defined as the habitat necessary for the survival or recovery of a listed species) for woodland caribou. For the Southern Mountain woodland caribou, critical habitat is identified as a specified minimum level of undisturbed habitat measured at landscape levels across low elevation winter ranges, high elevation winter and/or summer range, biophysical habitat attributes, and a concept of matrix range that includes low use, possible migration ranges, and specific predator-prey conditions (Environment Canada, 2014).
Almost all mountain woodland caribou were historically migratory, using mountainous high elevation areas during the summer, and lower elevation forested foothills during the winter (McDevitt et al., 2009). Anthropogenic-caused habitat loss, particularly clustered in low elevation areas, has reduced caribou distribution, altered resource selection, and reduced overall survival (e.g. Decesare et al., 2012;MacNearney et al., 2016). Anthropogenic disturbance occurs as both polygonal (e.g. cut blocks, fires) and linear (e.g. seismic lines, roads) features and impact caribou in varying ways. Removal of large sections of mature and old forest directly removes (direct habitat loss) forage critical for overwinter survival (Decesare et al., 2012;Wittmer et al., 2007). Moreover, early seral forage promoted by cutblocks enhances forage availability for moose and deer, leading to increases in primary prey abundance and subsequently, predator population growth (Seip, 1992;Serrouya, McLellan, Boutin, Seip, & Nielsen, 2011). Simultaneously, linear disturbances increase predator efficiency at searching for, encountering, and killing caribou (Spangenberg et al., 2019;Whittington et al., 2011). In response, caribou attempt to avoid anthropogenic disturbances, leading to indirect habitat loss, and yet are still often unsuccessful at predator avoidance. As a result, woodland caribou populations experience lower adult and juvenile survival (Decesare et al., 2014), leading to population declines across North America (C. A. Serrouya et al., 2019). For mountain woodland caribou, lower elevation ranges are frequently subjected to higher human disturbance, and avoidance of disturbance at low elevation winter range may lead to residency in lower quality, higher elevation summer ranges year-round (Edmonds, 1988). Such high elevation ranges are inherently lower quality (Decesare et al., 2014) because available forage is often covered by unfavourable snow conditions, increased wind and colder temperatures, and risk of fatal avalanches in mountainous regions (Hebblewhite, White, & Musiani, 2010). Yet, few studies have explicitly tested how differential anthropogenic development of seasonal ranges of migratory caribou affects behaviour and demography.
Here, we used a long-term dataset spanning more than three decades, 1981 -2018, to understand consequences of increasing anthropogenic disturbance on migratory behaviours and adult female survival rates within a declining migratory mountain woodland caribou population in Alberta and British Columbia, Canada ( Figure 1). First, we tested whether there were changes in migratory propensity for the population and whether individual caribou changed their migratory behaviour. Second, we tested for changes in the seasonal spatial distribution of caribou over three roughly decadal time periods and examined whether increasing land use change from natural resource development over time in these three periods were correlated with changes in seasonal distributions. We focused primarily on the winter range area as disturbance due to anthropogenic development is less of a concern in the largely protected summer range. Next, we tested whether caribou migrations adhered to discrete identifiable migration corridors that could be used in conservation and recovery planning (sensu Berger & Cain, 2014). Finally, we tested whether any identifiable changes in migratory behaviour and space use had impacts on adult female caribou survival.

Study area
The land use, particularly forestry and oil and gas developments, which negatively affect caribou (e.g. Decesare et al., 2012Decesare et al., , 2014Smith, Ficht, Hobson, Sorensen, & Hervieux, 2000). While fire has been shown to have negative effects on caribou behaviour, habitat, and demography through the boreal forest (e.g. , it was rare in our study area < 2%; see Results). Therefore, we focus on anthropogenic disturbance. For more study area details, see Decesare et al. (2014). a key winter forage, Thomas, Edmonds and Brown 1996), but higher anthropogenic landscape disturbance (Decesare et al., 2014). High elevation resident individuals have reduced access to high-quality for-age (Thomas, Edmonds, & Brown, 1996) and are exposed to the more extreme and hazardous conditions (e.g., avalanches) found in mountainous environments in winter (Hebblewhite et al. 2010;MacNearney et al., 2016; Alberta Environment and Parks, unpublished data).

Classification of migratory tactics
VHF and GPS locations were used to assess the occurrence and timing of migration to and from winter (January-March) and summer (June-August) ranges and to classify individuals as employing either 'migrant' (moving between high elevation summer range and low elevation winter range) or 'resident' (remaining in high elevation range throughout the year) tactics. Migratory tactic was assessed over the entire migration year (January to December) for each year an animal was monitored.
We classified migratory tactic using net-squared displacement (NSD) models in the 'migrateR' package in the program R (

Migration tactic trends
We fit Bayesian logistic regression models to estimate the probability of an individual using a migrant tactic as a function of time and to assess the probability of an individual switching migratory tactic from the previous year as a function of time and the animal's tactic during the current year (sensu Eggeman et al., 2016). We included a random effect for individual to account for non-independence of repeated records from the same individual in both models. We fit models using the program JAGS (Plummer, 2003) and the R package 'jagsUI' (Kellner, 2019

Space-use across seasons and time
We estimated space use during seasonal periods (winter and summer, We grouped the UDs into three time periods: early-historical (1981-1998), mid-historical (1999-2008), and current (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) Finally, we assessed the extent to which winter range area was impacted by fire and anthropogenic activities (energy and forestry) across the three time periods of the study using available spatial data.
Publicly available data for well sites, mines, pipelines seismic lines, transmission lines, roads, and cut blocks (Alberta Biodiversity Monitoring Institute and Alberta Human Footprint Monitoring Program, 2019) were combined with previously compiled date-time stamped seismic line and road data from previous studies (Decesare et al., 2014). Attribute data were identified for each feature wherever possible to assign a year and/or time period of the initiation of the feature. Each feature was also buffered by 500 m following previous studies that demonstrated indirect habitat loss affected demography of caribou . We calculated the total area of impact by each category during each time period, individually and cumulatively, within the early-historical winter range boundary. Additional details on spatial data processing steps can be found in the Supporting Information.

Spatial analysis of seasonal migration corridors
We estimated spring and fall migration corridors using locations only from migrant caribou individual-season-years that had an NSD model successfully converge and had more than two observations during the migratory movement (unique individuals n = 77, total migration events n = 155). We used only locations that occurred within the NSD-estimated start and end date to delineate migratory movements and estimate UDs. We followed the approach of Sawyer et al. (2019) where individual-season-year migration UDs were combined across years and seasons per individual, so that each individual had a single UD surface showing the probability of migratory movement for that individual across the study area (all cells scaled to sum to one). We overlaid all processed migratory UDs to obtain a cumulative probability surface of use during spring and fall migrations over the course of the study period and over all individuals to determine if highly used migration corridors existed. We log-transformed the total and used a threshold of the top 90% of the probability surface to aid in visualization.

Survival consequences of migration
We estimated Kaplan-Meier survival for known migration-tactic caribou using the R 'survival' package (Therneau, 2020;Therneau & Grambsch, 2000). Individuals were assigned a fate of alive, dead, or censored for each year that they were monitored following capture.
An individual that was not detected for a period greater than 200 days was right-censored and removed from the at-risk pool of caribou after its last known alive location. Previous analyses demonstrated minimal bias in survival estimates with this sampling design so long as fate was known, which functioning radiocollars ensured (Decesare et al., 2016).
We tested for effects of migration tactic and time period on survival, corresponding with the time period of differing anthropogenic disturbance intensity (Table 1). We selected the best survival model using  (Therneau & Grambsch, 2000). Finally, we tested the proportional-hazards assumption of the Cox model using Schoenfeld residuals (Grambsch & Therneau, 1994).

Monitoring movement
We monitored

Classification of migration tactic and trends
We

Space-use across seasons
We Estimated winter ranges showed change over the study, in terms of both location and size (Figures 3 and S1). In the early-historical period The area delineated by the early-historical winter range boundary impacted by fire was small across the entire study period (Figure 1;

Seasonal migration
Migration routes varied greatly across individuals and covered large portions of the winter and summer ranges, as well as intervening areas.
The 90% contour of the cumulative probability of all migratory UDs covered 280,266 ha and 53% of the migration area occurred within the interface of summer and winter ranges ( Figure 5). These results indicate no discretely identifiable migration corridors were used by migrating individuals, instead pointing to widespread overlap between migration corridors and seasonal caribou ranges.

Survival
We recorded fates for 231 known-migration tactic individuals, over 562 individual-years of data.  Time period also had an influence on survival. The model including both migration tactic, and time period was the second most-supported model with an AIC < 1 higher than the most supported model ( Figure S3; Table 2). For the three time periods, only the current time period was significantly different at the 90% confidence level indicating declining survival over time ( Figure S3).

DISCUSSION
Our results provide another example of disappearing migratory behaviour in the world's large ungulates associated with rapid land use change (Berger, 2004;Harris et al., 2009;Wilcove & Wikelski, 2008).
We tested for changes in adult female seasonal range use, migration and survival over time in an endangered woodland caribou population.
We found that the probability of individuals migrating decreased over time and that the probability of an individual switching between tactics Canada and the United States (Hervieux et al., 2013;Johnson et al., 2015). Timber harvesting, oil and gas development and associated road and pipeline infrastructure increase landscape-level habitat disturbance and remove essential biophysical elements of caribou habitat, including mature and old forest stands containing lichens, a crucial winter forage for woodland caribou (Shepherd, 2006;Thomas et al., 1996). Changes to forest age-class structure can negatively affect caribou through apparent competition (Holt, 1977;Serrouya et al., 2019;Wittmer et al., 2007)  Our results indicate that survival is poorest for caribou employing the high elevation residency tactic, despite that area being largely undisturbed and protected. High elevation dwelling resident mountain caribou are exposed to lower forage availability (Thomas et al., 1996), harsher weather conditions, and additional mortality from avalanches (Alberta Environment and Parks, unpublished data). Indeed, another nearby high elevation resident caribou population without the ability to migrate to lower elevation range outside a national park (Banff) was extirpated in 2009 by an avalanche (Hebblewhite et al., 2010). The increased mortality risk for high elevation resident caribou, in combination with the growing tendency for animals to switch to a resident tactic, suggest that RRPC caribou are displaying maladaptive selection and therefore falling into an ecological trap (Robertson & Hutto, 2006). to year increased and the trend was towards more switching from low elevation migration to high elevation residency. This asymmetry in switching behaviour between different tactics was similarly reported for elk by Eggeman et al. (2016), who found migrants more likely to switch than residents because of predation risk refugia on the resident range. The strong directional trend in switching that we documented suggests that weather variation did not drive the loss of this behaviour, as in wildebeest. Instead, we conclude that cumulative land use change on the low elevation seasonal range was responsible for loss of the migratory tactic as individuals sought refuge from increased risk and access to forage (MacNearney et al., 2016). While seasonal range abandonment of this nature has been reported across species and taxa (Dinkins et al., 2017) (Dinkins et al., 2017). If disturbance is measured collectively across all seasonally used areas, it may appear that the overall impact level is low. A recovery plan for an endangered species might then assume that the population should be resilient, while in reality, a high level of disturbance in one essential seasonal habitat area could threaten population viability. Our results demonstrate this challenge.
During the early years of our study, caribou made wide-ranging use of low elevation areas during winter (Edmonds, 1988 (Palm et al., 2020). Accordingly, most caribou populations throughout the Southern Mountain distribution have very low population viability (Wittmer, Ahrens, & McLellan, 2010). Failure to protect all seasonally critical habitats may render protection of other seasonal habitats ineffective for species conservation.
The unique genetics and associated migratory behaviour of these mountain caribou has led to their identification as their own Designatable Unit 8 by COSWEIC, signifying the value of protecting this distinctive population (COSEWIC, 2014). Yet, the very behaviour contributing to the identification of Designatable Unit 8 is being lost due to habitat change on low elevation winter range. As we show here, however, the shift in migratory behaviour to adopt a high elevation resident tactic resulted in decreased survival. Thus, while the change in tactic from migratory to resident behaviour is becoming increasingly common, coincidental with and likely as a consequence of growing anthropogenic disturbance, the lower survival represents an insidious ecological trap (Decesare et al., 2014;Robertson & Hutto, 2006).

CONCLUSIONS
In our study, habitat change and degradation was concentrated on the low elevation winter seasonal range, resulting in declining survival and population size of the entire population over more than 30 years.
It is important, therefore, that effective critical habitat targets be

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
Processed datasets used in probability of migration, switching, and survival analyses are available from the Dryad Digital Repository: https:// doi.org/10.5061/dryad.cc2fqz652 (Williams et al., 2021). Spatial location data itself (X,Y locations) are prevented from being shared due to their status as threatened /endangered species in Alberta/Canada and legal restrictions on sharing raw location data, and associated thirdparty data-sharing agreements.