Species‐specific interactions with apex carnivores yield unique benefits and burdens for mesocarnivores

Abstract

Mesocarnivores navigate a complex risk–reward continuum in ecosystems shared with their apex counterparts, balancing scavenging opportunities with risks of mortality. However, the risks to mesocarnivores in multi‐carnivore systems are not uniform; they can vary with specific apex–meso pairings. Using remote cameras and GPS‐telemetry, we examined space‐use, temporal activity, fine‐scale interactions, and scavenging behaviors of mesocarnivores (red foxes and coyotes) in relation to apex carnivores (wolves and cougars) in northern Yellowstone National Park, USA. Coyote space use was positively linked to wolf presence, while red fox space use was positively linked to cougar presence. Notably, photo‐detection rates of mesocarnivores doubled within 24 h of apex carnivore detections—except for coyotes following cougars. Coyotes were more frequent scavengers at wolf and cougar kills than red foxes, but this behavior came at a cost. Over 60% of wolf‐caused coyote mortalities were linked to wolf kill sites, though wolves rarely consumed the coyotes themselves. In contrast, cougars hunted and consumed coyotes as prey, posing an additional risk of predation. We interpret our findings in light of different hunting strategies and habitat preferences of canid and felid apex carnivores, which create distinct risks for mesocarnivores and drive species‐specific behaviors that can influence trophic dynamics in multi‐carnivore systems.

INTRODUCTION

Intraguild carnivore interactions play a pivotal role in ecosystem function, shaping the abundance, distribution, and behavior of both carnivores and their prey through top‐down effects (Ripple et al., 2014). These interactions, shaped by the complex strategies employed by carnivores across different trophic levels, are particularly important to understand given that human activities have disproportionately disrupted apex carnivores. Humans have driven significant population declines and local extinctions of apex carnivores in some areas, while facilitating their recovery through reintroduction and conservation efforts in others (Chapron et al., 2014; Ripple et al., 2014). Such actions impact entire carnivore communities, as apex carnivores can affect mesocarnivores both directly and indirectly (Prugh & Sivy, 2020). For instance, wolves (Canis lupus) can suppress coyote (C. latrans) populations through lethal removal and resource competition, which in turn reduces the top‐down effects that coyotes exert on red fox (Vulpes vulpes) populations (Levi & Wilmers, 2012).

However, the factors mediating the dynamics between apex and mesocarnivores are complex and often difficult to untangle. While apex carnivores can suppress mesocarnivore populations, they can also benefit them by providing scavenging opportunities from the carcasses of their prey (Brunet et al., 2022; Prugh & Sivy, 2020; Wilmers et al., 2003). As a result, mesocarnivore distributions and behavior in relation to their apex counterparts depend on a delicate balance of risks (injuries/mortalities) and rewards (carcass availability) that vary in both space and time (Ruprecht et al., 2021). For instance, a global review revealed that 30% of mesocarnivore diets were composed of scavenged ungulate carcasses; yet, a nearly identical proportion—about one‐third—of their deaths was attributed to apex carnivores (Prugh & Sivy, 2020). Larger mesocarnivores, in particular, were more reliant on scavenging opportunities, and this facilitation appeared to be more costly in less productive environments, leading to higher risks and mortalities from apex carnivores (Prugh & Sivy, 2020). The relationships between mesocarnivores and apex carnivores are thus shaped by several factors, including the extent of mesocarnivores' reliance on scavenging to supplement their diets, their ability to locate apex carnivore kill sites quickly, and the risks of intraguild predation.

The risks faced by mesocarnivores also depend on how they interact with specific apex carnivores. For example, when mesocarnivores are primarily competing for resources—such as carcasses—their risks typically arise from encounters with apex carnivores at kill sites (Prugh & Sivy, 2020). In contrast, if mesocarnivores are killed by apex carnivores irrespective of these encounters at carcass sites, then their risks are influenced by the habitat preferences and activity patterns of the apex carnivore (Broekhuis, 2015). Such relationships likely drive significant variation in intraguild mortality rates across carnivore family groups. For example, canid mesocarnivores experience mortality from canid apex carnivores at rates five times greater than from felid apex carnivores (Prugh & Sivy, 2020). Thus, the risks incurred by mesocarnivores when associating with their apex counterparts, as well as the benefits they gain, should vary with specific apex–mesocarnivore pairings. However, the mechanisms driving such differences are poorly understood. Such species‐specific interactions can be critical in further mediating mesocarnivore landscapes of fear, and in turn have effects on entire ecosystems.

In the Northern Range of Yellowstone National Park, wolves and cougars (Puma concolor) are abundant during winter when they are supported by a diverse assemblage of herbivores such as elk (Cervus canadensis), bison (Bison bison), and deer (Odocoileus spp.) that congregate at high densities (Houston, 1982; Smith et al., 2020; White et al., 2015). These apex carnivores kill larger ungulate prey during these months (Metz, 2021; Metz et al., 2012; Ruth et al., 2019), and such large carcasses provide scavenging opportunities to a host of mesocarnivores including coyotes and red foxes (Wilmers et al., 2003). Such scavenging opportunities appear to be critical for mesocarnivores in the winter because much of their regular prey (small mammals, ungulate neonates, birds, fruits, and insects) becomes scarce and/or subnivean (Wilmers et al., 2003). Coyotes and red foxes, however, also navigate the threat of lethal encounters with wolves and cougars while sharing similar habitats, especially when accessing carcasses. To examine how mesocarnivores that scavenge from apex carnivore kills navigate these complex interactions, we evaluated the degree to which coyotes and red foxes associated with wolves and cougars at various spatiotemporal scales.

We expected each apex–mesocarnivore species pair to exhibit distinct associations for two reasons. First, long‐term anecdotes from Yellowstone suggest that the mechanisms of lethal interactions for mesocarnivores differ by apex carnivore species. Cougars appear to hunt mesocarnivores as prey, while wolves are more likely to kill them at ungulate‐carcass sites, presumably as a form of resource defense. Therefore, we predicted that mesocarnivores would exhibit less overlap with cougars compared to wolves with respect to diel activity and space use and further exhibit greater avoidance of areas recently utilized by cougars given the increased threats of predation. Additionally, we predicted that the greater size and energetic demands of coyotes would increase their scavenging reliance on apex carnivore kills compared to red foxes, especially during the winter when alternative resources are scarce.

To explore these relationships, we conducted winter camera‐trapping surveys to assess occupancy, diel activity patterns, and fine‐scale spatiotemporal responses of apex and mesocarnivores. We also monitored predation events of GPS‐collared wolves and cougars to quantify mesocarnivore scavenging at kills, as well as direct killings of mesocarnivores. While our primary goal was to study apex–mesocarnivore relationships, it is well established (Prugh & Sivy, 2020; Rabe et al., 2025; Ruth, Boutte, and Hornocker 2019) that dominance hierarchies exist within both apex and mesocarnivore guilds in our system (i.e., wolves are competitively dominant over cougars and coyotes are competitively dominant over red foxes). We therefore included pairwise comparisons for wolf–cougar and coyote–fox in analyses of diel activity patterns and occupancy.

MATERIALS AND METHODS

Study area

We conducted the study in the Northern Range of Yellowstone National Park (Figure 1), an area with elevations ranging from 1500 to 2400 m. The landscape features a mix of rugged, rocky terrain and open, rolling hills. Common vegetation types include Douglas‐fir (Pseudotsuga menziesii), sagebrush (Artemisia spp.), and juniper (Juniperus spp.). The area experiences long, cold winters, with average snow‐water equivalents ranging from 2 to 30 cm (Plumb et al., 2009). Apex carnivores, such as wolves, grizzly bears (Ursus arctos horribilis), black bears (U. americanus), and cougars, are present in relatively high numbers, while coyotes, red foxes, bobcats (Lynx rufus), and martens (Martes americana) are the prominent mesocarnivores (Ruth et al., 2019; Smith et al., 2020).

FIGURE 1.

Locations of camera stations, gradient of habitat covariates, and photo‐capture rates of mesocarnivores (red fox = red, coyote = blue) and apex carnivores (wolf = green, cougar = pink) in northern Yellowstone National Park, USA. Species silhouettes sourced from Phylopic.org under a CC0 1.0 Universal Public Domain Dedication license.

Camera trap surveys

Camera surveys were conducted within a 580‐km² area, defined by a 5‐km buffer surrounding our camera‐trap grid, which was designed to estimate cougar densities. We deployed camera stations in a “checkerboard” pattern to increase the variability in station spacing (Sun et al., 2014), using a grid of 2.3‐km² cells (Anton, 2020). In each selected grid cell, we set up two camera stations at locations chosen to maximize detections, such as natural movement corridors or pinch points, with a minimum distance of 1 km between stations. All camera traps were non‐baited and set such that the passive infrared sensors were aimed at approximately 40 cm above the ground at the targeted game trail. Each camera was configured for long‐range motion detection and recorded 20‐s videos with a one‐minute delay between recordings. See Appendix S1: Section S1 for additional information regarding camera deployments.

Camera stations were active during winter sampling periods (referred to as “sessions”) spanning 15 one‐week‐long occasions from December to March in the years 2020–2021, 2021–2022, and 2022–2023 (hereafter referred to as 2021, 2022, and 2023, respectively). In 2021, we deployed 90 cameras across 60 stations, with half of the stations having a single camera and the other half having two cameras placed opposite each other. For 2022 and 2023, all stations were equipped with two cameras, bringing the total number of cameras to 120. To maintain independence of detections at camera stations for the occupancy and diel activity analyses, records from each camera were combined and repeat detections of the same species within 30 min were omitted (Tanwar et al., 2021). We manually reviewed all videos, identified species, and extracted metadata using the Timelapse2 image analysis software (Greenberg et al., 2019).

Data analyses

Multi‐species occupancy models

To evaluate how habitat covariates and spatial interactions with other carnivores influence the broad space use of apex and mesocarnivores, we constructed multi‐species occupancy models (Rota et al., 2016) using the unmarked R package (Fiske & Richard Chandler, 2011; R Core Team, 2023). Occupancy models incorporate binary presence–absence (1/0) data through likelihood‐based approaches to estimate species occurrence at site i in a latent z (occupancy) “state” model, while accounting for imperfect detections through a separate “detection” model. Multi‐species occupancy models are an extension of the single‐species occupancy model (MacKenzie, 2006) that apply a hierarchical modeling framework to estimate the effect of covariates (first‐order) and species co‐occurrence (second‐order) on occupancy.

While we placed camera stations at least 1 km apart from one another, because wolves, cougars, coyotes, and potentially red foxes typically occupy ranges larger than 1 km², the close spacing of our cameras may violate assumptions of spatial independence. Hence, our occupancy estimates more aptly describe patterns of “space use” rather than distribution (Burton et al., 2015; Kéry & Royle, 2015; Steenweg et al., 2018). For estimating detection probabilities of each species, we considered two covariates: the number of active cameras at a station and snow depth. As the number of active cameras at a station (ranging from 0 to 2 as a product of how many cameras were active and for how long during week‐long occasions) increased, we expected an increase in detection probability. We used mean snow depth at camera stations during each week‐long occasion at a 1‐km² resolution using SNODAS (National Operational Hydrologic Remote Sensing Center, 2004). We expected a negative relationship between snow depth and detection probability, such that increasing snow depth as winter sessions progressed would lead to fewer individuals moving through camera stations, lowering detection probability. For estimating species occupancy, we included topographic roughness, percent tree cover, and average snow depth at a 1‐km² spatial resolution to account for bottom‐up processes, such as habitat and landscape characteristics, that influence carnivore distributions (Cano‐Martínez et al., 2024). These habitat covariates were included because they are known to affect the occupancy and movement of carnivores (Smith et al., 2020). Please refer to Appendix S1: Table S1 for covariate details.

Our primary interest was in estimating species co‐occurrence patterns and the effects of landscape covariates on occupancy, rather than temporal occupancy dynamics. We therefore adopted a “stacked” data structure, where each site × year combination was treated as an independent site (Kéry & Royle, 2015, 2020), resulting in 178 sites (58 in 2021, and 60 in 2022 and 2023). To account for sampling of the same sites across years, we included year as a fixed effect on both detection and occupancy in all models. Using a two‐stage model selection approach (Twining et al., 2024), we first constructed single‐species models to evaluate which covariates best explained detection probability while holding occupancy constant at the null model (Ψ(.)). For each species, we used Akaike's information criterion (AIC) to select the best detection model from four candidate models representing all possible additive combinations of the two detection covariates, including the null model with only year as a fixed effect. Second, we carried over the top detection sub model to select the best first‐order model by considering all possible additive combinations of covariates on occupancy (eight candidate models for each species). We accounted for parameter redundancy by selecting more parsimonious models if they were within 2 AIC units of a model with more parameters (Arnold, 2010; Twining et al., 2024). After selecting the top first‐order model, we added second‐order species interactions to develop multi‐species models that tested the effects of species co‐occurrence on occupancy for every pairwise comparison of the four included species. To limit model complexity, we did not consider third‐ or fourth‐order species interactions. Please refer to Appendix S1: Tables S3–S5 for a list and comparison of candidate models.

Temporal activity patterns

We used the activity package in R to estimate diel activity patterns for wolves, cougars, coyotes, and red foxes using non‐parametric kernel density estimation (Rowcliffe, 2023). To quantify temporal niche partitioning between species and identify potential coexistence strategies that extend beyond broad‐scale occupancy, we used the overlap package in R to assess overlap of diel activity patterns between each species pairing by calculating the area under the combined activity curves, $\widehat{{\mathrm{\Delta }}_4}$ (Ridout & Linkie, 2009). We generated 10,000 bootstrap samples of the activity overlap coefficient, $\widehat{{\mathrm{\Delta }}_4}$, for each species‐pair to estimate 95% confidence intervals of overlap.

Time since apex carnivore detection

We examined fine‐scale mesocarnivore activity in response to apex carnivore movements using a time‐since‐event analysis (Cusack et al., 2017; Davis et al., 2021). The elapsed time between a mesocarnivore detection following an apex carnivore detection was recorded at each camera station for a given session. Subsequently, we combined data across all sessions for each apex–meso pairing. We truncated apex carnivore detections to only include events that were followed by a mesocarnivore detection (i.e., apex carnivore detections followed by another apex carnivore detection were omitted). Detections of mesocarnivores were recorded for 5 days following an apex carnivore detection, with elapsed time differences binned into 24‐h periods. We estimated detection probabilities by dividing the number of mesocarnivore detections in each 24‐h period by the total number of detections across all sessions (Cusack et al., 2017; Davis et al., 2021). These detection probabilities were compared to random iterations (n = 1000) of mesocarnivore detections to derive expected detection probabilities for each 24‐h period by using standard two‐tailed permutation tests (Davis et al., 2021). If mesocarnivores were avoiding apex carnivores through fine‐scale temporal segregation, we would expect coyote and red fox detection probabilities to be significantly lower following an apex carnivore detection. The opposite would apply if they were instead tracking/following wolves and cougars, likely to acquire food by scavenging apex carnivore kills.

Monitoring of apex carnivore predation

To assess mesocarnivore scavenging at apex carnivore kills and direct killing of mesocarnivores, we monitored predation by GPS‐collared wolves (n = 26 from 8 packs) and cougars (n = 17) during 30‐day winter sampling periods (November to March) from 2016 to 2023. Monitored wolves and cougars were fitted with Vectronics or Telonics GPS collars programmed to collect hourly fixes. We searched aggregations of GPS fixes (hereafter “GPS clusters”; Anderson & Lindzey, 2003) to identify kill sites and conducted necropsies on prey remains. At each kill, we ascertained the likely cause of death, the species, sex, age of prey when possible, and the presence of scavengers through observations and sign (e.g., scat, tracks, and hair). We used binomial generalized linear models (GLMs) to quantify mesocarnivore scavenging of apex carnivore kills which allowed us to determine if the odds of scavenging varied by mesocarnivore species, as well as by apex carnivore kill. We also documented mesocarnivores that were killed by apex carnivores at GPS clusters with ungulate prey remains (i.e., during scavenging attempts) as well as independent of ungulate prey (GPS clusters where mesocarnivores were the only prey remains present). To complement our camera‐trap analysis, we truncated this long‐term predation data to scavenging and predation events that occurred during the three winter camera‐trapping sessions, and qualitatively assessed whether the trends during those years markedly differed from the longer term, aggregated data.

RESULTS

Across all three 15‐week winter camera‐trapping sessions, we collectively sampled across 18,617 days. Red foxes were detected at 88% of sites, coyotes at 56%, wolves at 53%, and cougars at 47% (Appendix S1: Figure S1).

Multi‐species occupancy models

Detection probability

Variation in red fox detection probability was best explained by snow depth, with detection declining as snow depth increased (𝛽 = −0.21, 95% CI = −0.34 to −0.08, p = 0.0018; Appendix S1: Figure S2). Holding snow depth at its mean, red fox detection varied by year with marginal detection probabilities of 0.33 (95% CI = 0.29–0.36) in 2021, 0.24 (95% CI = 0.21–0.28) in 2022, and 0.39 (95% CI = 0.34–0.43) in 2023. While neither the number of active cameras nor snow depth significantly affected coyote or wolf detection probability, there was variation between sessions. Coyote detection probability was 0.25 (95% CI = 0.21–0.28) in 2021, 0.14 (95% CI = 0.11–0.18) in 2022, and 0.22 (95% CI = 0.18–0.27) in 2023. Wolf detection probability was 0.17 (95% CI = 0.14–0.21) in 2021, 0.12 (95% CI = 0.09–0.16) in 2022, and 0.21 (95% CI = 0.18–0.26) in 2023. Cougar detection probability did not vary annually but did decline with increasing snow depth (𝛽 = −0.27, 95% CI = −0.51 to −0.03, p = 0.03). See Appendix S1: Table S2 for first‐order detection model comparisons.

Occupancy probability

In the top‐performing first‐order model, habitat covariates only explained significant variation in cougar occupancy, while the null “year‐only” model performed best for red fox, coyote, and wolf. Cougar occupancy probability increased as mean snow depth decreased (𝛽 = −1.06, 95% CI = −1.99 to −0.13, p = 0.025) and roughness increased (𝛽 = 1.4, 95% CI = 0.78–2.03, p < 0.0001; Appendix S1: Figure S3). See Appendix S1: Table S3 for first‐order occupancy model comparisons, Appendix S1: Table S4 for the best‐supported first‐order models, and Appendix S1: Table S5 for first‐ and second‐order model comparisons. After incorporating second‐order species interactions on occupancy, we found that coyote occupancy was positively associated with wolf occupancy (𝛽 = 1.62, 95% CI = 0.81–2.42, p < 0.0001; Appendix S1: Table S6, Figure S4). Coyote occupancy was 0.75 (95% CI = 0.58–0.86) at sites where wolves were present and 0.35 (95% CI = 0.20–0.55) where wolves were absent (Figure 2b). Red fox occupancy was positively associated with cougar occupancy (𝛽 = 1.94, 95% CI = 0.51–3.36, p = 0.0077; Appendix S1: Table S6). Where cougars occurred, red fox occupancy was 0.96 (95% CI = 0.86–0.99), compared to 0.82 (95% CI = 0.63–0.91) where cougars did not occur (Figure 2b). We did not find any significant associations between wolf and cougar occupancy (𝛽 = 0.06, p = 0.89) or between coyote and red fox occupancy (𝛽 = 1.14, p = 0.09).

FIGURE 2.

(a) Apex and mesocarnivore marginal occupancy probabilities with 95% CIs estimated from camera trapping surveys between 2021 and 2023. (b) Coyote occupancy probability for 2021–2023, conditional on wolf presence–absence and red fox occupancy probability for 2021–2023, conditional on cougar presence–absence. Occupancy probabilities were based on the best‐supported, second‐order multi‐species occupancy model. Species silhouettes sourced from Phylopic.org under a CC0 1.0 Universal Public Domain Dedication license.

Temporal activity patterns

Cougars (n = 341 detections) exhibited unimodal diel activity, being most active around dusk peaking at 1800 and least active from morning (0800) to midday (1400) (Figure 3). Wolves (n = 443 detections) had bimodal, crepuscular activity patterns, with peak activity shortly after dawn (~0800–0900) and a secondary peak at dusk (1700–1800) (Figure 3). Coyotes (n = 541 detections) had more consistent activity across the diel cycle and were more diurnal, with peak activity occurring from late morning (0900) to midday (1400) and the lowest activity before dawn (0500) and after dusk (2100) (Figure 3). Red foxes (n = 1520 detections), on the other hand, exhibited highly nocturnal patterns, being most active shortly after dusk (1800–2100) and sustaining high activity through the night (2100–0600) with little activity during the day (0800–1600) (Figure 3). Temporal overlap between apex and mesocarnivores was $\widehat{{\mathrm{\Delta }}_4}$ = 0.86 (CI: 0.81–0.89) for wolf–coyote, 0.68 (CI: 0.61–0.7) for wolf–red fox, 0.81 (CI: 0.74–0.85) for cougar–coyote, and 0.72 (CI: 0.66–0.76) for cougar–red fox across the three sessions (Figure 3a). Overlap between mesocarnivores (red fox‐coyote) was $\widehat{{\mathrm{\Delta }}_4}$ = 0.61 (CI: 0.53–0.62), while overlap between apex carnivores (wolf‐cougar) was $\widehat{{\mathrm{\Delta }}_4}$ = 0.86 (CI: 0.8–0.91) (Figure 3b).

FIGURE 3.

Diel activity patterns with 95% bootstrapped CIs for study carnivores. Mean activity overlap (a) between apex and mesocarnivores, and (b) between apex carnivores and between mesocarnivores. Estimates of diel activity overlap between each species pairing ($\widehat{{\mathrm{\Delta }}_4})$ and the associated CIs are provided in each panel.

Time since apex carnivore detection

Compared to the mean values derived from random iterations of detection probabilities, red foxes and coyotes were both over twice as likely to be detected in the 24‐h period following a wolf detection (red foxes: 0.034 compared to 0.015, p = 0.001; coyotes: 0.05 compared to 0.018; p = 0.001; Figure 4a). However, their detection probabilities following cougar detections differed. While red foxes were over twice as likely to be detected in the 24‐h period following a cougar detection (0.024 compared to 0.011; p = 0.001), coyote detection rates were not affected. Additionally, red foxes were more likely to be detected in the 72‐h period following a wolf detection (0.01 compared to 0.007; p = 0.04) and the 96‐h period following a cougar detection (0.009 compared to 0.004; p = 0.001). Overall, the time interval between detections of a mesocarnivore subsequent to an apex carnivore detection was shortest for wolf–coyote (median = 17.67 h) and longest for cougar–coyote (median = 57.35 h; Appendix S1: Table S7). Foxes showed similar latency to wolf and cougar detections, with medians of 21.11 and 24.69 h, respectively (Figure 4b; Appendix S1: Table S7).

FIGURE 4.

(a) Daily binned observed and expected detection probabilities of mesocarnivores in response to apex carnivore detections at camera stations. Apex–mesocarnivore pairs include wolf–coyote (top‐left), wolf–red fox (top‐right), cougar–coyote (bottom‐left), and cougar–red fox (bottom‐right). Points and lines represent observed mesocarnivore detection probabilities, while violin plots show the distribution of expected mesocarnivore detection probabilities based on random movements (i.e., no positive or negative association). *Denotes significant differences between observed and expected detection probabilities. (b) Median time differences of mesocarnivore detections following apex carnivore detections. Species silhouettes sourced from Phylopic.org under a CC0 1.0 Universal Public Domain Dedication license.

Monitoring of apex carnivore predation

We detected 327 wolf kills and 257 cougar kills through winter predation monitoring (2016–2023). Coyote presence was recorded at 222 (68%) wolf kills and 80 (31%) cougar kills (Figure 5a). Red fox presence was recorded at 67 (20%) wolf kills and 46 (18%) cougar kills (Figure 5a). The odds of coyotes scavenging wolf kills were 8.2 times higher than red foxes, in contrast to the odds of coyotes scavenging cougar kills being just 2.1 times more than red foxes (Appendix S1: Table S8). From this predation monitoring, we documented 18 coyotes and one red fox killed by wolves (Figure 5b). Of the coyotes killed by wolves, 11 (61%) occurred at wolf feeding sites (i.e., ungulates killed or scavenged by wolves), while the red fox mortality did not. Cougars killed eight coyotes and three red foxes during this time (Figure 5b). All of these mesocarnivore carcasses were fully consumed, and none of these mortalities were associated with cougar‐killed ungulate prey sites. Instead, these mesocarnivore remains were generally found in cache piles at their own GPS clusters—suggestive of cougar predation and consumption. Wolf‐killed mesocarnivores were typically not consumed, as signs of consumption were documented at just 6 of 19 carcasses. However, it was unclear if those carcasses were consumed by wolves or other scavengers. We found no major differences in these trends between the longer term data and the data overlapping with our camera‐trapping study (see Appendix S1: Figure S5 for filtered results).

FIGURE 5.

Wolf (n = 327) and cougar (n = 257) kills, surveyed through GPS‐cluster searches of collared wolves and cougars between 2016 and 2023, bearing (a) presence of coyote and red fox scavenging expressed as a proportion and (b) counts of coyote and red fox killed. For methods, see main text. Cougar silhouette sourced from Phylopic.org under a CC0 1.0 Universal Public Domain Dedication license. Wolf and elk carcass graphics by Kira Cassidy.

DISCUSSION

Ecological communities are influenced by the top‐down effects carnivores exert on prey populations, but these effects also extend to other carnivores (Ripple et al., 2014). Intraguild carnivore interactions are complex, in that the negative impacts of competition can be partially offset by the scavenging opportunities competitors provide with their kills. While these trade‐offs are common between apex and mesocarnivores, we found considerable variation across species‐specific pairings.

Red foxes and coyotes were both ubiquitous in our study area and common scavengers at wolf and cougar kills, but exhibited unique spatial and temporal relationships with each apex carnivore (Figure 6). Our multi‐species occupancy models revealed positive associations between the space use of coyotes and wolves, as well as between red foxes and cougars. Further, red foxes and coyotes exhibited opposite diel activity patterns (Figure 3b), as the nocturnal activity of red foxes closely aligned with that of cougars, while coyotes were most active mid‐day when apex carnivores were least active. Their fine‐scale movements also differed in response to apex carnivore presence. Red fox detection probabilities at camera stations doubled in the 24 h following a wolf or cougar detection, but coyotes only displayed this positive association to wolves, not cougars (Figure 4a). Accordingly, we found that coyotes were more frequent scavengers at wolf kills compared to cougar kills. However, scavenging of wolf kills was often costly, as over half the coyote mortalities attributed to wolves during our study occurred at wolf kill sites. Interestingly, we did not find clear evidence of wolves consuming the coyotes they killed, indicating that such lethal actions were most likely a form of resource defense. Conversely, all coyotes killed by cougars occurred as standalone predation events (i.e., independent of ungulate carcass sites) in which cougars consumed the coyotes. The ambush hunting style of cougars, combined with the forested and topographically rough terrain they inhabit, could lead to coyotes being more vulnerable in such habitats (Perrig et al., 2023). In contrast, the open and flatter terrain preferred by wolves could provide more visual and olfactory cues for coyotes to assess (and escape) immediate risks of lethal encounters. Such disparities in apex carnivore hunting strategies seemingly tie back to the positive spatiotemporal association of coyotes with wolves but not cougars, and highlight the nuances of such competition.

FIGURE 6.

Conceptual representation of results from analyses for each apex–mesocarnivore pairing. Mesocarnivores (Coyote = blue, Red fox = red) are shown in the center and apex carnivores around the perimeter (Wolf = Green, Cougar = Pink). For “Occupancy/ Broad Space Use,” the multi‐directional arrow refers to co‐occurrence results from the top second‐order multi‐species occupancy model. “Scavenging” represents the proportion of ungulate kills made by an apex carnivore that were visited by the respective mesocarnivore. “+” represents a positive association/effect, “–” represents a negative association/effect, and “no effect” indicates an insignificant result. Effects are colored green or pink to indicate that they are affiliated with wolf or cougar results, respectively. The size of the “+” or “−” is positively related to the magnitude of the effect observed. Species silhouettes sourced from Phylopic.org under a CC0 1.0 Universal Public Domain Dedication license.

As mesocarnivores adapt their coexistence strategies to different apex carnivore species, their behaviors are likely to shift more dramatically when multiple competitors are present. For example, high winter apex carnivore densities in our study system may explain the midday peak in activity for coyotes, which differed from the more crepuscular activity patterns of coyotes inhabiting adjacent areas of the Greater Yellowstone Ecosystem with lower apex carnivore densities (Smith et al., 2023). In contrast, red foxes appeared less sensitive to the diel activity of apex carnivores. Not only did their peak in activity align with that of cougars but they also displayed the same nocturnal patterns as red foxes across the Greater Yellowstone Ecosystem (Smith et al., 2023). Notably, red foxes exhibited high temporal segregation from coyotes, having the lowest temporal overlap of any species pair (Figure 4). This may be the result of competition between these mesocarnivores, as the larger coyotes can suppress red fox populations (Fedriani et al., 2000; Levi & Wilmers, 2012). Thus, the nocturnal activity of red foxes could result from avoiding peak coyote activity during the day that is in turn mediated by wolf and cougar activity (Figure 4). Temporal sympatry among the carnivore guild appears to be layered, with apex carnivores affecting coyotes, and coyotes affecting red foxes.

Apex and mesocarnivore relationships depend on competition within trophic groups, as spatial and/or temporal responses to such competitors alters the availability of ecological niche space for other carnivores in the community. For example, cougars in Yellowstone changed their winter space use after wolves were reintroduced by increasing their selection for areas with escape terrain (e.g., ruggedness and tree cover) (Ruth et al., 2019). This avoidance of wolves could in turn facilitate coyote space use, providing habitats where encounters with cougars are scarce. Our results support this, as coyote space use, fine‐scale movements, and scavenging rates were more strongly associated with wolves than cougars. Accordingly, if red foxes experience the most severe competition from coyotes, then greater overlap with cougar space use could alleviate such competition. While red foxes scavenged wolf and cougar kills at a similar rate, our analyses revealed substantial overlap between cougar and red fox space use and activity patterns. Mortalities from cougars still occurred, but foxes may be more adept at avoiding immediate risks of cougar predation during encounters through their increased agility, as well as their relatively large foot: body ratio that allow for better maneuverability on top of snow (Van Etten et al., 2007). Coyotes being heavier, might not be as lucky when encountering cougars. Thus, competition within trophic groups likely plays a key role in shaping the apex–mesocarnivore interactions we observed, creating scenarios akin to “the enemy of my enemy is my friend.”

In contrast, competition between apex and mesocarnivores is often driven by scavenging, where apex carnivores take measures to reduce food loss (e.g., carcass defense and caching) and mesocarnivores balance the need for such dietary subsidies with the associated risks of confrontations at carcass sites (Prugh & Sivy, 2020). Thus, the availability of alternative resources for mesocarnivores should influence the extent to which they seek out scavenging opportunities and the overall strength of their relationship with apex carnivores. In northern Yellowstone, the limitations of prey acquisition during winter may increase mesocarnivore reliance on scavenging, particularly for coyotes who were over 8 times more likely to scavenge wolf kills and twice as likely to scavenge cougar kills compared to red foxes (Figure 5; Appendix S1: Table S8; Jensen et al., 2022; Smith et al., 2023). Foxes may be better suited to meet their energetic demands by hunting themselves, as they are commonly observed hunting subnivean rodents by pouncing through the snowpack during winter (D. Stahler personal communication). Therefore, apex carnivores facilitate coyotes and red foxes differently, and these relationships are likely to change during the spring/summer when the availability of rodents, neonatal ungulates, vegetation, and insects increase (Bassing et al., 2025; Jensen et al., 2022). The presence of grizzly and black bears that emerge in summer after winter‐long hibernations can also complicate these relationships because bears often limit carrion consumption for other scavengers (Allen et al., 2014).

While our study highlights the positive aspects of apex and mesocarnivore interactions through scavenging, we did not explore variations in red fox and coyote population trends. In fact, coyote abundance declined by 39% following wolf reintroduction in the Lamar Valley of Yellowstone, an area adjacent to our study area (Berger & Gese, 2007). Additionally, wolf‐coyote interactions decreased over time (between 1997 and 2007) in our study area, indicative of changes to coyote density and/or behavior (e.g., wolf avoidance) (Merkle et al., 2009). Despite benefiting from carrion made available by wolves and cougars, red fox and coyote populations may still experience top‐down regulation via direct killing that maintains them at low densities (Sinclair, 2003). Long‐term studies that estimate population trends of mesocarnivores with respect to apex carnivore densities are required to test such effects.

The magnitude and impact of trophic‐mediated responses related to large carnivore recoveries have become a contentious issue, perhaps nowhere more so than in Yellowstone National Park (Brice et al., 2022; Hobbs et al., 2024; Mech, 2012; Ripple & Beschta, 2012). Although the reintroduction of wolves and recolonization of cougars to this area have likely altered mesocarnivore distributions and abundance, red foxes and coyotes also benefit from living with their dominant counterparts. Overall, red foxes and coyotes in our study area exhibited more positive than negative associations with apex carnivores during winter when small prey and alternative food resources are scant. Such positive associations, possibly a consequence of desperate foraging decisions in winter, can also prove fatal. However, risks to mesocarnivores are uneven across dominant–subordinate carnivore pairings, with apex carnivore behavior changing the drivers of these interactions, adding nuance to the landscape of fear for mesocarnivores.

AUTHOR CONTRIBUTIONS

Wesley Binder, Jack W. Rabe, Daniel R. Stahler, and Stotra Chakrabarti conceptualized the study, raised resources, and oversaw project completion. Wesley Binder, Daniel R. Stahler, Claire Lacey, and Gordon Scott conducted fieldwork and collected the data. Wesley Binder, Jack W. Rabe, and Stotra Chakrabarti developed the methodology and analytical approaches. Wesley Binder, Jack W. Rabe, Zoe Lowe, Gordon Scott, and Eliza King curated the data and performed the analyses, with inputs from Stotra Chakrabarti and Daniel R. Stahler. Wesley Binder, Jack W. Rabe, and Stotra Chakrabarti wrote the manuscript, with help from Gordon Scott.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

Wildlife captures and handling were conducted under National Park Service protocols (IACUC permits IMR_YELL_Stahler_Cougar_2018_A1, IMR_YELL_Smith_Wolf_2012).

Supporting information

Appendix S1.

Binder, Wesley , Rabe Jack W., Lowe Zoe, Scott Gordon, Lacey Claire, King Eliza, Stahler Daniel R., and Chakrabarti Stotra. 2026. “Species‐Specific Interactions with Apex Carnivores Yield Unique Benefits and Burdens for Mesocarnivores.” Ecology 107(3): e70331. 10.1002/ecy.70331

DATA AVAILABILITY STATEMENT

Data and code (Binder, 2025) are available in Zenodo at https://doi.org/10.5281/zenodo.14852260.