Our study was novel in that we combined hidden Markov models with SSFs to assess the influence of current and future anthropogenic development on multiple movement processes including state-dependent movement behaviour, habitat use, and connectivity. We found that the cumulative effects of anthropogenic development caused more extensive habitat degradation for slow movements than for fast movements. Overall, grizzly bears and especially wolves increased their speed of travel near anthropogenic development to minimize encounters with people and avoided anthropogenic development for slow movements, especially during the day when people were more active. This habitat degradation reduced the amount of high quality habitat available for foraging and resting, reduced the ability of carnivores to regulate prey populations that seek human settlements as prey refuges (30), and reduced the functionality of wildlife corridors (5). We found that that towns, roads, and trails together reduced connectivity around mountain towns by > 80%. Our study supports the growing body of research showing the negative effects of anthropogenic development on wildlife movements (e.g., 46, 47, 48).
Anthropogenic development in our study increased transitions rates from slow to fast movements for both grizzly bears and wolves in all seasons except for grizzly bears in summer when they spent most of their time in fast states of movement and for wolves in spring when some packs denned near trails and roads. Globally, human activity has variable effects on animal movement, including movements of large carnivores (47). In many cases, human activity has reduced movement rates of animals through barrier effects or by providing resource rich environments for concentrated foraging (46). For example, puma in California exhibited slower movement rates near anthropogenic developments, perhaps because they were forced to travel in rugged terrain that slowed movements (49). Similarly, wild dogs in Africa decreased movement rates near human settlements but increased rates of travel outside of protected areas, perhaps because of lower prey availability (17). Conversely, African lions (Panthera leo) increased their speed of travel near bomas (livestock enclosures), perhaps to reduce their risk of encountering and being detected by people (49). The combined movement models and step selection functions from our study suggest that grizzly bears and wolves sped up movement rates near anthropogenic developments due to a combination of factors including increased encounter rates with people, reduction in secure habitat for foraging and resting near towns and areas with high densities of trails and roads. These models also suggest that grizzly bears and wolves use linear features to increase travel efficiency, but this can often subject them to increased mortality risk due to vehicle collisions, human hunting and management actions (31, 50).
Grizzly bear and wolf resource selection responses to anthropogenic development depended both on behavioural state and time of day, with larger effect sizes for behavioural state. Our results were consistent with the few studies to assess the effects of movement state and time of day on resource selection (13, 51). Like grizzly bears and wolves in our study, African lions avoided human activity during the day when foraging and resting, yet had higher tolerance for human activity when travelling and at night (13). Our results are consistent with other research showing that wildlife are more likely to use habitat and travel through areas with people at night than during the day (51, 52). For example, in a meta-analysis Gaynor, Hojnowski (52) found that many taxa adapted to human disturbance by increasing their activity at night by an average factor of 1.36. This nocturnal temporal shift in movement and foraging behaviour allows animals to access habitat required to maintain fitness while minimizing encounters with people.
Grizzly bears and wolves in our study avoided areas near towns when in slow movement states even though towns contained attractive natural and anthropogenic food sources (31, 53). For instance, elk, which are an important prey species for grizzly bears and wolves, congregated and calved near towns to reduce predation risk (30). Even with this attractive food source, grizzly bears avoided areas 200 to 300 m from towns while wolves avoided areas 400 to 500 m from towns. Avoidance of towns tapered at night and was negligible for wolves in their fast state of movement. Together, this suggests towns had stronger effects on habitat required for foraging and resting compared to connectivity habitat required for travel.
Grizzly bears and wolves both avoided areas with high trail and road density when in slow states of movement, likely to reduce encounter rates with people. The exception occurred in the fall, when grizzly bears selected areas with high trail densities. Trails at this time of year had low levels of visitation and grizzly bears likely selected for seasonal foods associated with high trail density. For example, Sheperdia canadensis berries and Hedysarum spp. roots are important food sources for grizzly bears in the summer and early fall and can be found along forest edges and in open canopy forests that receive higher levels of solar radiation (54). While grizzly bears selected areas with higher trail density in the fall, their selection for these features likely depends more on the combination of available foods and levels of human activity than on the trails themselves.
Our study could be improved with better estimates of recreational activity on trail networks (37). While we used trail and road density as a surrogate for intensity of human use, carnivores typically avoid encounters with people rather than the physical density of linear features (35). Recent studies show promising approaches for predicting recreational activity by directly tracking recreationists’ movements (44), inferring activity from mobile device and crowdsourced data (55, 56), or modelling spatial and temporal trends in trail use (57). Stronger links between recreational activity and wildlife movements would improve our understanding of recreational thresholds for wildlife and our ability to manage human-wildlife coexistence (58, 59).
Numerous studies have found that grizzly bears (60, 61) and wolves (33, 59, 62) avoid human activity, which can contribute to the fragmentation of populations (63, 64). However, few studies have compared the behaviour of the two species. Wolves in our study exhibited stronger avoidance of towns and areas of high trail-road density relative to grizzly bears. The muted response of grizzly bears was likely influenced by high individual variability in responses to anthropogenic development (57, 65). Model fit from the predicted habitat use had high variability among individual grizzly bears (but not wolves), which reflects high individual variability in resource selection. Thus, our results for grizzly bears likely averaged results from both wary and habituated individuals. This could lead to an underestimate of the effect of human activity on surviving bears because habituated bears have dismal survival prospects in busy landscapes such as the Bow Valley (31). Simulating movements from random coefficients could highlight estimates of connectivity for both wary and habituated animals and could help identify areas likely to have high levels of human wildlife conflict (31, 66). Finally, grizzly bears are highly motivated to find food, including human and natural foods in and around residential areas, which can lead to increased human-wildlife conflict (53). As such, grizzly bears that use areas near people face high risk of mortality, which can lead to population level source-sink dynamics (31, 67). Pairing demographic outcomes such as survival and reproduction with individual-level behavioural responses to human activity could help bridge the gap to what Fahrig, Arroyo-Rodríguez (7) and identified as one of the missing links in connectivity science, population-level connectivity.
Integrated step selection analyses that interact movement parameters with anthropogenic features could also be used to estimate the effects of human activity on habitat use and connectivity without the added step of developing hidden Markov movement models (Avgar et al. 2016, Signer et al. 2017). However, we found that classifying movements into discrete behavioural states simplified our interpretation about how human activity affected movement processes. Moreover, animal motivations to move include accessing habitat required for fitness enhancing behaviours such as foraging, resting, and reproduction (68). Understanding and conserving habitat for slow-state behaviours could affect realized movement rates and could have important consequences for fitness and population level-connectivity.