Coyote family activity in a landscape of fear

Global urban expansion has enmeshed many wildlife species in human landscapes. Examples include coyotes in Calgary¹, wild boar (Sus scrofa) in Berlin² and caracals (Caracal caracal) in Cape Town³. It has been argued that several wildlife species now complete their entire lifecycles in cities⁴, leading to increased reports of human-wildlife conflict⁵. Coyotes are at the forefront of this conflict in North America.

Evidence shows coyotes have a plethora of adaptive responses that enable urban living. For instance, coyotes can alter their population size, prey selection, and temporal use of habitat^6,7,8. This resilience however can keep coyotes closer to humans, increasing the risk of negative encounters⁹ (e.g., a coyote bites a pet), which can cause lasting financial and emotional damage for humans and their pets.

Coyote behavior throughout their lifecycle is relatively stable and predictable¹⁰, which should make coexistence straightforward. However, coexistence remains challenged by mixed human perceptions and incomplete knowledge about coyote behavior and adaptive responses in cities¹¹. In tandem, while lethal removal of coyotes remains a core response to conflict¹², there is a growing societal demand for non-lethal coexistence strategies¹³. This push for non-lethal approaches stems from a widespread acceptance that lethal approaches are not an effective long-term solution to human-coyote conflict, may be ecologically damaging, and are inhumane when compared to alternative non-lethal methods (e.g., aversion conditioning)^14,15. Moreover, without addressing the root cause of the conflict, which is most often human behavior, there is compelling evidence that the conflict will repeat with another coyote or species^15,16.

In the case of urban coyotes, the pup-rearing season (May to August) is a time when conflict with humans typically spikes¹⁰ as coyotes are defensive of their pups, especially toward humans with dogs (Canis lupus familiaris)¹⁷. Previous research related to the pup-rearing stage has explored behaviors such as boldness, exploration, and aggression in both captive¹⁸ and wild coyotes¹⁹. Other work has developed a generic ethogram to measure coyote behaviors²⁰. To our knowledge, no research has compared activity patterns and behaviors specific to pup-rearing in urban coyote families to those in less disturbed settings. Yet, changes in adaptive pup-rearing behaviors that result from urban pressure could impact conflict. For instance, a reduction in play activity amongst pups, or between pups and adults due to a heightened need for adults to be vigilant could result in less socialization of pups. In turn, lower socialization could affect pup survival after dispersal²¹.

Consequently, we believe there exists an opportunity to develop a baseline understanding of how urbanization affects activity and behavior of coyote families. Such understanding is important in and of itself but also may foster best practices in non-lethal conflict management practice. Presently, non-lethal coexistence programs may integrate aversion conditioning (AC) to respond to conflict. Some AC approaches used on carnivores (coyotes especially) are predicated on an assumed need to re-instil or heighten an animal’s fear response to humans^12,22. In the case of coyotes, some managers have used these high-intensity AC approaches when coyote behavior is labelled as ‘bold,’ ‘aggressive,’ or a risk to human safety²³. Yet, no studies appear to examine fear behavior in coyotes, nor any change in behaviors because of living under urbanization pressure. If such a behavioral study demonstrated that urbanization relates to greater vigilance and fear in adult coyotes, that could be a reason to re-evaluate the use of using fear-evoking AC.

To understand whether fear in coyotes is affected by urbanization we must first address the question “What is natural fear in wild animals?” Here, we can borrow from the “landscape of fear” concept, in which prey species will adjust their behavior in response to the threat of predation²⁴. This concept derives from fear ecology work in which the behavioral response of prey was compared between fear-driven systems and direct predation-driven systems²⁵. The concept explains how the knowledge of predation risk impacts prey choices in space use, foraging, and vigilance behavior²⁶. For example, black-tailed jackrabbits (Lepus californicus) and desert cottontails (Sylvilagus audobonii) adjust their movements and behaviors in response to fear of predators²⁷. The landscape of fear concept also has been invoked in situations where humans play the role of the “super-predator,” such as with elk (Cervus canadensis)²⁸ and marsupials²⁹. Importantly though, while fear can be adaptive for reducing predation risk and increasing life expectancy in prey species³⁰, it can come at a cost of foraging³¹. In turn, this can add food stress, which can increase conflict within the prey populations. Similarly, human activity may drive changes in top predator behavior, impacting the animals’ ability to regulate prey populations. In the case of coyotes – the top predator in many cities – additive fear in a landscape could lead to unchecked populations of small mammal species³² and larger mammals like deer (Odocoileus spp)³². This may lead to spill-over effects for humans, such as rodent infestations or greater deer-vehicle impacts³³. Most critically, if coyotes spend more time in behaviors that arise out of fear (e.g., vigilance) at the expense of foraging, this could lead to opportunistic feeding on anthropogenic food sources, which may increase stress and conflict in coyotes³⁴. Fear in urban-adapted wildlife has been previously studied in the foraging behavior of smaller, prey mammals, with lower levels of vigilance in treatments closer to urban areas but higher responses to fear stimuli in treatments in a peri-urban environment³⁵. However, fear in non-foraging coyotes across an urban to rural gradient is not well understood.

To explore whether urbanization affected adult coyote activity and behavior during the pup-rearing season, and specifically whether fear was higher in urban coyotes, we narrowed our analysis to the following general questions: (1) Were there changes in adult presence and fear-related behavior (i.e., den-guarding) around pups across rural to urban sites?; (2) Did the percent of time spent by adult coyotes on high alert during captured activity sequences increase across the rural to urban gradient?

We coded 81 442 images from camera traps (CTs), of which 62 822 captures showed coyotes (including adults and pups), across 923 total trap nights (402 from Campus [the urban site], 188 from Spyhill [the peri-urban site], and 333 from WA Ranches [the rural site]). We used a selection of the coded CT photos that were isolated to the pup-rearing season in 2022, totalling 15 000 captures from Campus, 15 108 captures from Spyhill, and 14 808 captures from WA Ranches. Amongst the latter subset, 21 386 contained adults, 31 663 contained pups, and 8 513 had both pups and adults present. We converted the CT photos into 4 556 sequences of activity across the entire sample area, 2 021 of which were used for proportional analysis of behavior Fig 1.

Adult Coyote presence and behavior around pups

In our first set of comparisons, we examined adult behavior and presence around pups and pup play (Fig. 2). We observed adult attendance to pups in behaviors such as interacting, nursing, and guarding (Table 1). While pups were attended by adults in ~ 27% of photos across all three sites, we found significant differences between adult presence around pups by site (X² = 37.717, P = 6.454e-09, df = 2), adult presence around pups playing by site (X² = 137.63, P < 2.2e-16, df = 2), and den guarding around pups by site (X² = 352.2, P < 2.2e-16, df = 2). Using Pearson residuals, we investigated which frequencies deviated the most from what would be expected if there was no difference between sites. We found significantly more adult presence around pups at WA Ranches (i.e., rural). We observed significantly greater adult presence around pups that were playing and adult guarding at Campus (i.e., urban).

Adult Coyote alertness by site and conditional variables

We compared the percent of activity time spent by adults on high alert by site including other conditional variables (e.g., pup presence, time of day, novel object presence). The mean proportion of images showing high alert behavior relative to not high alert behavior was 12.8% across all three sites during the spring/summer sample period. At Campus, the mean proportion of high alert behavior was 16.7%, at Spyhill it was 14.3%, and at WA Ranches it was 4.8%. The results of our zero-inflated binomial mixed-effects model of the percentage of high alert behavior per activity sequence, including all independent variables with significant interactions with each other, are presented in Table 2. Note that for novel object presence, only six out of 2 021 image sequences captured a novel object.

Due to the high number of pairwise interactions between independent variables in the model, the post-hoc Tukey analysis was Sidak-adjusted for the comparison of means³⁶. The variation in these means of the proportions of high alert behavior per sequence by site, time of day, and pup presence is presented in Fig. 3. The highest estimated marginal means for the proportion of high alert behavior per sequence occurred at the Campus site in the daytime with pups present, while the lowest occurrence of high alert behavior occurred at the WA Ranches site at twilight with pups present. Significant differences included higher marginal means for alertness at Campus than WA Ranches at twilight without pups present (Z ratio = 2.638, P = 0.0227, df = inf) and at any time of day with pups present (Day: Z ratio = 4.849, P < 0.0001, df = inf; Night: Z ratio = 3.036, P = 0.0068, df = inf; Twilight: Z ratio = 4.499, P < 0.0001, df = inf), and higher marginal means for alertness at Spyhill than WA Ranches in the daytime with pups present (Z ratio = 2.473, P < 0.0357, df = inf) and at twilight with pups present (Z ratio = 3.182, P < 0.0042, df = inf). Degrees of freedom are labeled as infinite using the emmeans package³⁶ as estimates were compared against the standard normal distribution.

Coyotes have a high investment in their pups, as seen by the frequent attendance of them by both parents³⁷ and the contribution of non-breeding helpers to pup-rearing³⁸. At the WA Ranches (i.e., rural) site, we observed slightly more adult presence with the pups, which could be a positive indicator for pup survival, as reported previously by Bekoff and Wells³⁸. However, when engaged in play activities, we observed that the pups were left unsupervised significantly more at WA Ranches and Spyhill (i.e., peri-urban) when compared to Campus (i.e., urban). We also observed that guarding behavior by adults was significantly more common amongst the Campus coyotes. One explanation is that the Campus coyote family perceived this urban homesite to be riskier for pups due to being embedded in high density urban matrix. When pups engaged in play, we observed routinely that they spread quickly outside the visual range of adult coyotes. In an urban site, the higher incidence of potentially dangerous novel objects may demand higher levels of pup supervision and guarding by adults.

The significantly higher incidence of guarding pups at Campus could indicate that urban coyotes may experience more concern and fear for themselves and their offspring. To mitigate the risks to pups, one might expect that adult coyotes may have to adjust their activity budgets. Particularly concerning would be if increased time spent in pup-supervision leads to adult coyotes making trade-offs with essential behaviors like hunting. The latter was observed with African wild dogs (Lycaon pictus), where an increase in pup-guarding, though effective in decreasing pup mortality, negatively impacted hunting as the wild dog groups were constrained by the absence of pup-guarders to assist with hunts³⁹. Arguably, we are seeing evidence that suggests urban coyotes (at least in the Campus focal family) may be redirecting time to protect pups.

Another potential indicator of a heightened fear response is seen in vigilance behavior. We measured vigilance as high alert behavior and den-guarding. Vigilance has been associated with fear hormonally as it can be paired with the release of fear-response hormones⁴⁰. Vigilance has been observed to increase in other species around human presence^28,32,41, but it has not been studied before in this manner in wild coyotes. In previous studies, vigilance has been measured in behavioral experiments of urban coyotes¹⁹ and in understanding the response of captive coyotes to human activity⁴². We found that adult coyotes spent significantly more time displaying vigilance behaviors (high alertness, den-guarding) at the urban site compared to the peri-urban and rural sites. For instance, in the rural site (WA Ranches), where coyotes share the landscape with larger predators caught on camera like grizzly bears (Ursus arctos) and cougars (Puma concolor), rural coyotes only spent ~ 5% of their time on high alert, while the urban coyotes spent ~ 17% of their time on high alert (as captured in the data). Given the potential for a trade-off to occur between fear and foraging³¹, the significant increase in time spent in vigilance (assuming it is fear related) by urban coyotes may drive them to procure opportunistic easier anthropogenic food sources like garbage, which can increase human-coyote conflict^34,43. This behavioral shift may in turn be paired with an avoidance of certain high-human use areas, potentially resulting in an increase in prey animal presence, as has been suggested in the “human-shield hypothesis”⁴⁵.

In grouping all three coyote families, we also found that alertness was higher overall in the presence of pups. This would be expected, given demonstrated parental investment by coyote parents in their young³⁷. However, examining WA Ranches on its own, adult alertness was significantly lower in the presence of pups than the absence of pups. This may suggest that these rural coyotes perceive less risk and therefore experience a lower baseline level of fear than their urban cousins. Certainly, the rural site is protected from hunting, trapping, and poisoning (within the ranch) and very few people enter the site, which would create a sense of security from humans.

When examining the relationship between time of day and high alertness, we found a significant increase in alertness during the twilight period. We observed more high alertness overall in the twilight period with no significant variation between sites but less high alertness in the presence of pups in the twilight and night periods, suggesting there is heightened concern for pups in the daylight. Previous urban coyote studies have shown a decrease in coyote activity in the twilight period corresponding with higher human activity, suggesting coyotes use an avoidance technique at this time of day⁴⁵. As the coyotes in this study were viewed at homesites where they can avoid humans, the heightened alertness may be another response to an increase in human activity.

Examining the relationship between the length of an activity sequence (as determined by the number of images of coyotes in the sequence) and high alertness, there was a slight but significant negative effect of more photos. This may relate to longer sequences having more behaviors visible, which could offset the overall presence of high alertness; alternatively, a short sequence of photos could have three photos all of which display high alertness making for a much higher proportion.

Exploring the relationship between the maximum number of adults in a sequence and high alertness, there was a significant decrease in high alertness with more adults. More adults in the sequence meant a lower proportion of time spent on high alert which corroborates the hypothesis of the negative relationship between vigilance and group size seen among other species⁴⁶.

Finally, in our investigation of how the presence of a novel object related to high alertness, we found a significant negative effect. That said, we rarely captured novel objects in the photos (only six out of 2 021 sequences featured novel objects) and this result may have occurred out of chance as the few observations happened to occur when coyotes were on lower alert. Novel objects in the study were sporadically occurring objects observed to have drifted into view of the cameras. Novel objects also typically occurred in the urban site where coyotes may be more adjusted to the presence of such things and have less of a response. Urban coyotes have also been observed to be more investigative in the presence of novel objects¹⁹, so this data could simply be a further indication of this behavior. This will be an important area of future research to understand whether the likelihood of encountering many more novel objects in an urban ecosystem means that adult coyotes spend even more time supervising pups at the expense of other vital behaviors.

Our study provides evidence for a heightened state of fear-related behavior among urban relative to peri-urban and rural coyote families. As noted, this can come at a cost of adaptive behaviors, as seen in other species^32,39. Because fear is associated with stress and stress can lead to riskier behavior in coyotes and conflict with humans²³, understanding that urban coyotes exhibit significantly more fear daily is important to coexistence practices. In particular, non-lethal coyote management strategies that implement fear-based AC methods have not to our knowledge evaluated what baseline levels of fear exist, or whether adding stress or fear may have compounding negative impacts on coexistence. In our opinion, coexistence programs should consider the efficacy and ethics of programs that ‘stack’ fear upon fear⁴⁷. We observed significantly higher rates of vigilance behavior related to pup guarding, which suggests coyotes may be more fearful during that time. Therefore, there is a risk that using fear-based AC with urban coyotes may create a condition called “trigger stacking”^48,49. In domestic dogs, trigger stacking is known to exacerbate reactivity rather than change or de-escalate behavior⁴⁸. If coyotes might become more reactive due to AC-caused trigger stacking that would be counterintuitive to the goals of coexistence.

While urban coyotes have been characterized as bolder and more exploratory, our results highlight fearfulness (i.e., guarding, pup-attendance, vigilance) as another key element motivating their behavior. If more fear leads to restricting mobility and access to natural foods, this could drive consumption of more easily accessible anthropogenic food, which has historically increased human-coyote conflict. While we examined only three families, one per level of urbanization, the magnitude of difference in behaviors between the sites highlights the need to better understand the baseline ecology and behavioral adaptations of coyotes before applying untested invasive coexistence management techniques. Further study could benefit from exploring behaviors outside of the pup-rearing season and look at direct impacts of such management on coyote and other urban wildlife behaviors.

Study area

Our research was conducted at three sites in and around the City of Calgary, Alberta, Canada (Fig. 1). The study sites rest within the Foothills Parkland Natural Subregions, a hilly area with a mixture of grasslands, shrublands, and forest that lies between the prairies to the east and the foothills to the west⁴⁹. The region is home to a diverse array of plant species and several mammal species from hares (Lepus spp.) to coyotes to bears (Ursus spp.). The region has a relatively dry and cold climate, with a mean temperature of 4.3 °C ranging from − 30 °C to + 30 °C and a mean precipitation of 417 mm of rain and 100 cm of snow⁵⁰. The research reported was covered by Animal Care Certificate number AC20-0160 issued by the University of Calgary Life and Environment Sciences Animal Care Committee for the project “UC Campus Coyote Ecology and Coexistence” on February 23, 2021. The experiments reported in this manuscript were minimally invasive and conducted in accordance with Animal Care Committee guidelines. Clinical trial number: not applicable.

Each study site represented a unique level of human use, with the Campus site (i.e., urban) located within the City of Calgary and surrounded by residential area, the Spyhill site (i.e., peri-urban) located on the northwest edge of the city, at the juncture of urban residential and agricultural land use, and the WA Ranches (i.e., rural) site located approximately 30 km northwest of Calgary, surrounded by ranch land and natural areas. Each site was the core area of a unique coyote family comprised of a breeding pair, one ‘helper,’ and an annual litter of three to eight pups (SM Alexander, unpublished data). Sites were approximately 30 km apart, which is outside the limits of a resident home range⁵¹ and reduced the chances of detecting the same coyote at different sites. Our visual records showed that there was no observed overlap of individuals from different families, even though as Gehrt et al.⁵³ note, this distance can easily allow transient coyotes to cross amongst sites. AC by agencies was only reported to have occurred within the core habitat of Campus coyotes.

Camera-Trap methods

We deployed 27 camera traps (CTs), divided equally by site (21 Reconyx Hyperfire 2 and 6 Cabela’s Outfitter Gen 3). Each camera captured three images at a detected motion, with images continuously captured if motion continued, resulting in a sequence of activity. Images were captured at a rate of one image per second. The resolution of images was 16 megapixels for the Cabela’s cameras and two megapixels for the Reconyx cameras. No differences were noted in the ability of either camera to capture coyote activity, though both would occasionally cease functioning in extremely cold temperatures (i.e., less than − 20 °C). Cameras were set to operate at all times of day and night, only stopping if the batteries died or the SD card storage was filled, but the frequency of camera checks allowed them to run for the most part continuously. The time and date were set on the camera at its placement. Cameras were placed on trees or fences at 30–60 cm off the ground to maximize the field of view for capturing coyotes. Instead of using bait, the cameras were pointed toward known high-coyote-activity areas, such as around the home sites and high-use pathways, as determined from field surveying. The camera locations were purposefully selected as the goal was to capture the highest amount of coyote activity. While a random selection may have captured a more natural range of coyote behaviors throughout their territory, the focus here was specific behaviors at high-use areas within the homesite and comparing these behaviors between sites.

All cameras were in place year-round as part of long-term monitoring of coyotes, but for this project we screened focal images from CTs at post-natal homesites only for the period of May/June 2021 and January to August 2022 from Campus (48 290 images) and May to August 2022 from Spyhill (15 893 images) and WA Ranches (17 259). We focused on the pup-rearing season to capture and compare behavior and activity budgets when we were most likely to see interactions amongst coyote family members at the three sites of interest. We divided the CT photos of coyotes into sequences of activity, using a separation of one minute between a coyote disappearing and reappearing on the screen. Within these sequences we could then determine the proportion of time spent displaying any one behavior (i.e., calculated as the number of captures displaying the behavior over the total captures in the sequence). While we used shorter image capture intervals than other studies, such as 5 min in Wooster et al.⁵⁴ and 10 min in Marion et al.⁴². Our method suited the resolution of our research question; We had no need to try to separate sequences by unique individuals, as known individuals frequented our same site and our exceptionally large photo counts increased replicates when compared to other noted studies.

We developed an ethogram in reference to our CT data and classified fear-related behaviors for coyotes as alertness and pup-rearing behaviors (e.g., den-guarding). The ethogram was founded on previous behavioral research on coyotes⁵⁴, red foxes⁵⁵, and felids⁵⁶ and honed to the study animals during initial reviews of the photos. We describe all behaviors documented and how the behaviors were coded in Table 1. For each CT photo sequence of activity, we documented the following: proportion of time spent in the behavior, site, time of day, pup presence, number of photos, maximum number of adults, novel object presence, date, and camera location. One observer classified all photos into their behavioral categories.

Statistical analyses

To determine whether coyote behavior around pups varied by category (type) amongst study sites, we performed multiple chi-squared tests⁵⁷. For this analysis, we only used images that showed pups to be present (i.e., if photos were adults only, we removed them from this analysis). We compared the frequency of the photos showing the following behaviors across sites: pup photos with and without adult presence, pup play photos with and without adult presence, and pup photos with adults demonstrating a den-guarding posture or not. We used Pearson residuals to determine which behaviors differed significantly by site, visualizing the differences using the vcd package⁵⁸.

To explore whether fear-related behaviors differed across human disturbance categories, we developed a generalized linear model and compared the proportion of photos per sequence demonstrating high alertness. We included variables previously identified to be relevant to behavior: site, pup presence, time of day, sequence length, maximum adult presence, and novel object presence³⁷. We also were interested in how pup presence and time of day might interact with site, so we included an interaction term for those three independent variables. Data were proportional and thus followed a binomial distribution⁵⁷. Given many sequences of activity had proportions of zero for high alert behavior, we used a zero-inflated model using the glmmTMB package⁵⁹. Since our data included multiple images coming from the same cameras and from each site, we followed Zuur and Ieno⁶⁰, using a mixed effects model with a random effect of camera nested within site to account for this level of variation. Following from the previous, our final model was a zero-inflated binomial mixed effects model, described as:

proportion of high alert behavior per sequence ~ site*time of day*pup presence + number of photos in the sequence + maximum number of adults in the sequence + novel object presence in the sequence.

We performed a post-hoc Tukey analysis of multiple comparisons⁶¹ using emmeans package³⁶ to identify pair-wise interactions between model terms. All statistical analyses were performed using R version 3.4.2⁶³.