Habitat selection, movement patterns, and diel activity patterns are the result of responses to dynamic limiting factors and exemplify trade-offs to maximize fitness. Resources needed to meet individual energetic and physiological requirements generally determine species distribution, given the constraints imposed by inter- and intra-specific interactions (e.g., predator–prey relationships, competition). Species select habitat in a hierarchical process, where the extent to which animals select resources changes spatially and temporally (Rettie and Messier, 2000; Boyce et al., 2003, Bastille-Rousseau et al., 2018). Animals must select areas that satisfy their nutritional requirements while integrating smaller scale seasonal limitations like growth stage or access to resources (Johnson, 1980; Dussault et al., 2005; Hebblewhite and Merrill, 2009). Examples of temporal variation in resource selection are migration, seasonally shifting diet composition (Hebblewhite and Merrill, 2009; Middleton et al., 2013a), and changing resource selection across diel periods (Kohl et al., 2018; Courbin et al., 2019; Smith et al., 2019). Results of tradeoffs made in response to costs can influence nutritional condition and fitness (Christianson and Creel, 2008; Hebblewhite and Merrill, 2009; White et al., 2012); however, resource and habitat selection are often adjusted, particularly with experience and as an anti-predator strategy (Gaynor et al., 2018; Cunningham et al., 2019; Palmer et al., 2021).

Predation risk can influence seasonal habitat use (Mao et al., 2005; Hebblewhite and Merrill, 2009), resource selection (Christianson and Creel, 2008), movements (Ditmer et al., 2018), and diel activity of prey species (Kohl et al., 2018; Smith et al., 2019) and can be present both in time (temporally) and in space (spatially). Risk to prey depends on a variety of factors including predator hunting strategy and the degree of spatiotemporal overlap (Suraci et al., 2022), landscape attributes (Kauffman et al., 2007), seasonal and biological differences in prey vulnerability and risk taking (Winnie and Creel, 2007; Oates et al., 2019), and competition between multiple predator species (Sih et al., 1998). For example, ambush predators use stalking cover and darkness to increase hunting success, therefore, the risk they present may be more concentrated in time and space and more predictable than risk posed by cursorial predators (Preisser et al., 2007).

The perception of risk by prey species is spatially and temporally dynamic (Kohl et al., 2018, 2019, Smith et al., 2019). A risky place indicates an area of increased danger, but the level at which danger persists fluctuates in time. When risk is predictable, prey are more likely to limit exposure using proactive measures, such as grouping-up to dilute individual risk, isolating from a group to reduce detection, avoiding areas of high risk, or adjusting activity patterns (Basille et al., 2015; Creel, 2018; Kohl et al., 2019; Courbin et al., 2019). However, proactive risk mitigation may result in decreased foraging efficiency or nutritional intake (Brown et al., 1988; Creel, 2018). Conversely, reactive responses occur when risk is unpredictable and therefore consistent spatial or temporal avoidance is not worth the energetic costs. Instead, prey flee, hide, increase induced vigilance, or display defenses when encountering predators and may experience elevated hormone-based stress and/or reductions in time spent foraging (Creel, 2018). Recently, it has been debated whether indirect or non-consumptive effects from predation risk might have ecological and demographic consequences of greater magnitude than direct predation (Winnie and Creel, 2017; Say-Sallaz et al., 2019; Sheriff et al., 2020). For example, physiological stress, caused by elevated glucocorticoid levels and/or nutritional deficiencies related to changes in habitat selection or decreased foraging activity may affect pregnancy rates, offspring survival, and/or body condition, potentially influencing population productivity (Christianson and Creel, 2008; Winnie and Creel, 2017; Allen et al., 2022).

Knowledge of spatiotemporal strategies used by large mammalian prey to balance risk is incomplete and inconclusive, particularly in multi-predator systems. Most existing research takes place in protected areas and further research is needed to understand if similar responses occur in unprotected areas, which often experience more complex human–predator–prey relationships (Hebblewhite and Merrill, 2009). Restoration of once extirpated predators may reestablish constraints on prey behavior and population growth by forcing prey species to cope with risk that has been absent from these systems for decades. Because predation is a limiting factor for behavior and resource selection strategies of ungulates (Lima and Dill, 1990; Fortin et al., 2005; Creel et al., 2005; White et al., 2012; Farley et al., 2024), there is a need to assess how prey are affected by reestablishing predators in both protected areas and working landscapes to better understand the dynamic relationships between predators, prey, and their habitats.

In the southwestern United States, Mexican gray wolves (Canis lupus baileyi) were extirpated from the wild by the 1970s and reintroduced to eastern Arizona and western New Mexico by the U.S. Fish and Wildlife Service (USFWS) in 1998 (U.S. Fish and Wildlife Service, 2017). Population growth initially was slow, but in recent years the minimum number of individuals counted between 2010−2023 has increased 414% (U.S. Fish and Wildlife Service, 2017, 2024). As of 2023, there were a minimum of 257 Mexican wolves in the United States—slightly more than 80% of the original recovery goal of 320 (U.S. Fish and Wildlife Service, 2017, 2024). At the beginning of this research there were several suitable regions within the recovery area with no known established packs (USFWS unpublished), however, each subsequent year the occupied range has expanded and the population has shown continuous growth since 2015. Elk are the primary prey of Mexican wolves (Smith et al., 2023; Martinez, 2024) and recent research suggests Mexican wolves influence the foraging and vigilance behavior of elk (Farley et al., 2024; Olson, 2024). Yet, the indirect impact(s) they may have on elk resource selection is unknown.

If predation risk is partially dependent upon predator density (Vucetich et al., 2002) and strength of predator avoidance changes with levels of risk (threat-sensitive hypothesis; Helfman, 1989; Basille et al., 2015), it is important to understand if, and at what scale, elk are adjusting their movement and habitat use patterns in response to the growing population of Mexican wolves. Elk distribution has both management and ecological implications. Elk distribution can influence the nutritional condition and survival of individuals, erosion, post-fire aspen establishment (Fortin et al., 2005), and nutrient cycling (Huntly, 1991). In turn, elk distribution may also impact where wolves expand their range. In this study, we developed multiple indices of Mexican wolf predation risk to investigate the influence of predation risk on diel and seasonal elk habitat selection patterns and spatiotemporal predator avoidance strategies. We predicted elk would: 1) spatiotemporally avoid areas of high wolf use during times of peak wolf activity, 2) continue to select for areas and resources which optimize fitness, and 3) be most responsive to wolf utilization as a risk index, rather than wolf resource selection or landscape characteristics that indicate risk.

This study occurred within the Mexican Wolf Experimental Population Area (MWEPA) in New Mexico and Arizona. The area is predominately public land managed by the U.S. Forest Service and Bureau of Land Management, with private and state trust lands occurring throughout. The western edge is composed of tribal lands of the White Mountain and San Carlos Apache ( Figure 1 ). Elevations range from 1,700 to 3,477 m. Topography consists of rolling hills, sandy washes, moderately steep canyons, rugged slopes, rock cliffs, and elevated mesas. Eight wilderness areas and one primitive area occur within the study area, primarily at higher elevations. Cattle grazing is common throughout the MWEPA, and small towns are scattered throughout along the main highways. Outdoor recreation is popular, particularly during summer months and fall hunting seasons.

Dominant vegetation communities at higher elevations are primarily spruce-fir and mixed conifer forests which include Engelmann spruce (Picea engelmannii), sub-alpine fir (Abies lasiocarpa), Douglas‐fir (Pseudotsuga menziesii), white fir (Abies concolor), and aspen (Populus tremuloides). Ponderosa pine (Pinus ponderosa) and Gambel oak (Quercus gambelii) compose mid-elevation vegetation communities and low elevation areas primarily consist of alligator juniper (Juniperus deppeana), oneseed juniper (J. monosperma), and piñon pine (Pinus edulis). Cool and warm season grasses are interspersed throughout and produce large swaths of high elevation montane grasslands and low elevation plains. Forest fires are common, with large (> 297,844 acre) stand replacing fires occurring in 2011 and 2012. Aside from elk, native ungulates include mule deer (Odocoileus hemionus), Coues white-tailed deer (Odocoileus virginianus couesi), pronghorn (Antilocapra americana), javelina (Pecari tajacu), and Rocky Mountain bighorn sheep (Ovis canadensis canadensis). Predators in addition to Mexican wolves include mountain lion (Puma concolor), black bear (Ursus americanus), coyote (Canis latrans), and bobcat (Lynx rufus).

Climate is generally mild with warm summers and moderate to cold winters. Average annual precipitation is 68.7 ± 12.1 cm (SD) for high elevation areas (≥ 2,684 m) and 36.7 ± 6.7 cm (SD) for low elevation areas (≤ 1,975 m), with snowfall averaging 193.2 ± 105.3 cm (SD) and 44 ± 43.4 cm (SD) in high and low elevations, respectively (30-year normals; PRISM Climate Group, 2021, Western Regional Climate Center, 2019). Approximately half of precipitation falls during monsoon season (June through September). Mean daily high/low temperatures for high elevation areas are 22.5°C/7.7°C in summer and 5.2°C/−7.7°C in winter (PRISM Climate Group, 2021). Mean daily high and low temperatures for lower elevation areas are 30.2°C/11.9°C and 10.6°C/−6.2°C for summer and winter respectively (PRISM Climate Group, 2021).

Adult female elk within the MWEPA employed spatiotemporal avoidance strategies at both a seasonal and diel scale to mitigate predation risk from Mexican wolves. Elk also showed a functional response, suggesting the way elk used risky places depended on their level of exposure to wolves. Our results are consistent with research emphasizing the importance of accounting for temporal variation in predation risk (Lima and Bednekoff, 1999; Creel et al., 2008; Kohl et al., 2018, 2019, Smith et al., 2019), but also reveal the hierarchal nature of changes in elk habitat selection over time by examining variation in spatiotemporal avoidance strategies in different seasons in relation to wolf presence. Once a temporal component is incorporated into habitat selection models including spatial risk metrics, it becomes evident that risk avoidance also occurs at a finer scale in space and time that corresponds with activity budgets and dynamic behavior of both predator and prey (Lima and Bednekoff, 1999; Middleton et al., 2013b; Kohl et al., 2019).

The type of risk which was most influential changed with the inclusion of an interaction with diel period and the risk term, resulting in risky places being the most-supported risk metric across all seasons. Within every season, all top spatiotemporal models contained either a risky places or openness metric. These spatial risk indices are the most predictable predation risk cues for elk as they are relatively static in space. Metrics like relative intensity of wolf use and predicted wolf presence may have been less supported because they represent a coarser temporal scale of risk and wolf use on the landscape is much more dynamic than landscape characteristics, thus less predictable. Predicted risky places are associated with more direct consequences for adult female elk, representing the capture component of predation risk (Hebblewhite et al., 2005), and thus it is most beneficial for elk to mitigate the component of risk which affects fitness the most (Lima and Dill, 1990; Lima and Bednekoff, 1999). Conversely, all other metrics represent various forms of encounter risk, which is a component of predation, but is not necessarily correlated with kill success (Hebblewhite et al., 2005; Kauffman et al., 2007; Moll et al., 2017). While the predicted risky places metric does not capture the nuances and compounding elements involved in predation (e.g., the age and health of predator and prey and the landscape features encountered during pursuit), habitat features significantly influence the outcome (Hebblewhite et al., 2005; Kauffman et al., 2007). Habitat characteristics associated with a higher probability of being killed also tended to be associated with higher availability and better forage quality for elk: pasture/wet meadows and more open areas, aspen stands, burns, and areas closer to roads, where diversity of plants may have often been higher due to seeds dispersed from vehicles and presence of ditches which concentrate water (Creel et al., 2005; Fortin et al., 2005; Spitz et al., 2018). Thus, impediments to escape and early predator detection, such as downed woody debris and thick regenerating vegetation, may make elk vulnerable to predation once encountered by wolves. This was supported by the significant role of fire scars and aspen stands in areas where elk were killed and has been demonstrated in other studies (Hebblewhite et al., 2005; Halofsky and Ripple, 2008). Additionally, in an exploratory analysis of risky places, the top performing model included a term for mean distance to fence, where it was the second most influential variable among predictors, however geospatial fence data were not available for tribal lands within the study area, so models containing this variable were removed from consideration as predictions of risky places needed to be made on tribal lands as well ( Supplementary Material : Risky places index). Although distance to fence was purely an exploratory element in this analysis, it appears to be an important factor in risk, particularly to neonates, and should be considered in future studies if possible. Because wolves primarily occurred at moderate to high elevations during this study, kill site characteristics reflected this influence. Kill site characteristics may vary as wolves expand to lower elevations areas with different vegetation and habitat structure.

When individuals are nutritionally deficient, anti-predator strategies may not be implemented as intensely (Winnie and Creel, 2007, Oates et al., 2019). Our results supported this as adult female elk increased use of risky places at risky times during spring and monsoon seasons, particularly during diel periods associated with foraging (dawn and dusk) and midday. This contrasts to winter and fall behavior where elk only slightly selected and even avoided risky areas at dawn and dusk. Adult female elk may be more vulnerable in spring, and to a lesser degree during monsoon, because the nutritional demands required for late gestation and lactation are high and elk are recovering from winter months with low forage quality (Parker et al., 2009; Huggler et al., 2022). Average digestible protein and energy within this region appear to increase the most in May following snowmelt, but herbaceous biomass is generally low at this time (J. Cain, NMCFWRU, unpublished data), thus most nutritional forage does not become widely available until late June when monsoons begin, particularly during years with low winter snowfall. Average annual precipitation between 2019 and 2021 was 92% lower than the 30-year average (PRISM Climate Group, 2021) and may have exacerbated energetic and nutritional constraints of female elk and influenced the results herein. Because of the nutritional benefits that riskier places might afford, especially during spring and monsoon, female elk appear to be more willing to take chances and make trade-offs during the seasons they can best increase nutritional condition, which can influence fitness. Additionally, wolf activity is lowest during midday and is more predictable in spring due to denning, providing support for both the risky times hypothesis and risk allocation hypothesis, where female elk are likely balancing fitness costs and benefits by exploiting riskier places at less risky times (Lima and Dill, 1990; Lima and Bednekoff, 1999; Creel et al., 2008; Kohl et al., 2018). Furthermore, a concurrent behavioral study observed that during spring and monsoon the probability that an adult female elk would be foraging midday increased as predation risk increased, further supporting our findings (Farley et al., 2024). Moreover, the similar response across risk strata in winter and more pronounced differences in elk response between strata in spring and monsoon provide further evidence of the link between risky places and forage conditions because in winter there is less nutritional benefit in exploiting risky places.

The risk strata analysis revealed a functional response, where the intensity of use of risky places gradually changed as initial wolf exposure of elk increased. Elk with the least amount of wolf exposure used riskier places more intensely at dawn, dusk, and night during spring and monsoon and strongly avoided riskier places midday. In contrast, elk initially exposed to moderate to high wolf presence selected for riskier areas less intensely at dawn, dusk, and night in spring and monsoon and more intensely during midday than other seasons. These results suggest that elk may become more cautious to perceived risky places with prolonged exposure to established wolf presence and support the risk allocation hypothesis (Lima and Bednekoff, 1999). Additionally, a learning curve appears to occur, where the typically more intense selection of risky places by naïve elk decreases at very risky places. Further, this functional response to predation risk has also been observed in mule deer (Huggler et al., 2022). During our study, at dawn and dusk, elk sought areas with good foraging conditions, which typically exhibit characteristics that facilitate success of cursorial predators (Creel et al., 2005; Hebblewhite et al., 2005; Preisser et al., 2007). Additionally, wolf activity was also consistently higher at dawn and dusk. Thus, elk occupying areas with higher wolf presence may have more limited options for selecting perceived safe areas with preferred foraging conditions. This limitation may cause elk to use preferred foraging areas less during times wolves are most active to spatiotemporally mitigate predation. Elk with moderate to high initial wolf-exposure may balance increased predation costs with potential nutritional benefits that may correspond with risky places during spring and monsoon by shifting use of riskier areas during less risky times. Elk with little exposure to wolves likely do not need to temporally shift behavior patterns in order to balance wolf encounters and nutritional intake, thus select for risky places associated with foraging opportunities more intensely at dawn and dusk and prioritize thermal refuge and resting midday, supported by intense avoidance of risky places midday and more midday resting behavior than elk in moderate to high wolf risk areas (Farley et al., 2024).

In all instances, incorporating diel period improved spatial risk models and, in most instances, had more support than models with wolf activity as the temporal component, revealing mixed evidence of elk utilizing areas proportional to perceived risk and wolf activity (Kohl et al., 2018, 2019). The inconsistent response to wolf activity could be due to variability in daily wolf activity patterns across seasons within the MWEPA (French et al., 2022). Activity patterns were relatively variable across years except in spring, likely due to fluctuations in sample size, weather conditions, and human influence (e.g., hazing, recreationists, and livestock operations). Thus, we examined diel period as an alternative representation of temporal risk which is broadly representative of wolf activity but is less variable and related to specific environmental cues like sun angle and ambient temperature. Elk habitat selection in relation to predation risk was likely more responsive to diel period because it is more predictable than anticipating the variable movements of a highly mobile and wide-ranging predator and is more directly related to the behavioral state of elk. Response to predation risk should, at least in part, be related to the primary behavioral state across the diel period. For example, during foraging periods, elk may have to accept higher levels of risk to access forage than they would during diel periods when they are typically resting.

On a finer spatial scale, interaction effects between wolf use and base model terms revealed how areas with more wolf use affected the influence of other landscape characteristics on relative intensity of elk use. However, due to intensity of wolf use representing a temporally broad metric of encounter likelihood, results examining how elk respond to landscape elements differently in relation to wolf use should be interpreted with caution, as cursorial predators often overlap heavily in space with prey (Schmitz et al., 2017) and encounter metrics only represent one component necessary for predation (search–encounter–kill; Lima, 1992; Hebblewhite et al., 2005). Relative intensity of use between intermediate canopy cover often did not differ from closed or open canopy cover, which may have reflected the effects of reclassifying a continuous variable into a discrete one for visualization purposes. However, as wolf use increased, the degree of overlap between all canopy cover levels, and to a lesser degree terrain ruggedness and road use, increased. This supports the idea that seasonal predator use is too broad of a predation risk metric for species that heavily overlap in space and time and suggests elk may respond to increased encounter likelihood by shifting selection reactively, at a finer and more immediate scale, thus causing an averaging effect in selection for specific levels of landscape elements over a season (Creel et al., 2005; Middleton et al., 2013b; Basille et al., 2015). For example, open canopy is associated with increased wolf encounters, but it also provides greater nutritional benefits for elk and is easier to detect wolves in. It is unlikely elk are pursued by wolves every time they forage in open spaces, so instead of avoiding these areas, elk continue to use open spaces and rather react once wolves are detected, which may include fleeing into cover or staying in cover for a few hours longer. The variable use of open spaces dependent on wolf presence would show up as overlapping use of canopy cover types over the course of a season. The control of risk hypothesis (Creel, 2018) suggests anti-predator responses are either proactive or reactive depending on the prey’s ability to perceive, predict, and control risk. Wolves typically travel farther than elk, therefore, wolves may not frequent the same areas as often or encounter the same elk herd/individuals for several days and thus may not be perceived by elk as a persistent threat needing severe mitigation, particularly within the MWEPA as diel patterns in wolf activity are less predictable for some seasons (Schmitz et al., 2004; Preisser et al., 2007; Middleton et al., 2013b). Further research is needed to evaluate spatiotemporally fine-scale wolf–elk interactions and whether potential shifts in selection are significant enough to cause population-level non-consumptive effects.

Across all seasons, elk used riskier places at night, which may have been a product of balancing risk mitigation strategies for two predators with opposite hunting modes. Elk are the primary prey for not only Mexican wolves in the study area, but also for mountain lions (Martinez, 2024). Wolf kill sites within the MWEPA were generally associated with more open areas, consistent with Kohl et al. (2019), which found female elk utilized areas with a higher spatial likelihood of wolf predation (open areas) at night when wolves were less active to avoid mountain lions in forested areas. At night, despite moderate to high wolf activity during most seasons, elk consistently selected for open spaces over closed canopy in areas with high wolf use, except in winter when pack cohesion was greatest (Metz et al., 2011; Benson and Patterson, 2015) and the nutritional reward of open places was most limited, suggesting elk were prioritizing avoidance of mountain lions at night and potentially taking advantage of foraging opportunity. Cursorial predators may encounter elk more but have higher success in killing non-prime age individuals or those in poorer body condition (Metz et al., 2012), whereas lions may have lower encounter rates with elk, but once encountered, kill success is higher and they are less discriminate among age and condition of elk (Hornocker, 1970). Therefore, due to differences in hunting style, wolves may not be perceived by a healthy adult female elk as more lethal or risky than a mountain lion (Morosinotto et al., 2010; Kohl et al., 2019). Additionally, as ambush predators, lions are more reliant on cover and darkness for stealth, which are recognizable environmental cues predictable to prey (Hornocker, 1970; Smith et al., 2019).

Although spatiotemporal avoidance of wolves is exhibited on a seasonal and diel scale, it more proactively occurs and is strongest during seasons when the loss of associated foraging opportunities is lowest (winter, fall) and during time periods where wolf activity is more predictable (i.e., dawn). It is likely energetically inefficient for female elk to spatiotemporally avoid areas of high wolf risk completely during times of high nutritional reward due to lack of consistent spatial/temporal cues to indicate wolf presence. It does, however, seem energetically efficient for adult female elk to proactively mitigate risk from a more lethal, more spatiotemporally predictable predator with fewer nutritional costs associated with avoidance (Kohl et al., 2019). Therefore, when it comes to mitigating risk, it appears elk employ a combination of proactive and reactive responses, and shift strategies seasonally in accordance with physiological demand and reward, supporting the risk allocation hypothesis (Lima and Bednekoff, 1999). Proactive responses can have detrimental effects on nutritional condition if loss of foraging efficiency is not mitigated (Creel, 2018). Reactive responses are linked to more predator–prey encounters and thus result in learned behaviors but also can result in physiological and psychological stress such as increased glucocorticoid levels and decreased body condition (Creel, 2018). Negative effects from both strategies have the potential to affect demographic parameters like survival and pregnancy rates, thus it is necessary to consider how predation pressure, from the suite of predators elk are exposed to, can indirectly affect elk fitness in addition to consumptive effects.

Understanding non-consumptive effects requires knowledge of the mechanisms driving them (Gude et al., 2006). Thus, ongoing research is currently evaluating pregnancy rates, survival, and body condition of elk within the MWEPA to determine if non-consumptive effects are occurring because of anti-predator responses to increased wolf presence (J. Cain, NMCFWRU, unpublished data). It is important to acknowledge however, that elk co-evolved with both wolves and mountain lions and natural selection tends to favor anti-predator strategies that optimize fitness. Shifts in habitat selection and foraging behavior in areas of higher wolf presence and during times of greatest nutritional reward presumably help maintain healthy nutritional condition, however, not avoiding risky areas likely increases encounters and reactive avoidance (Winnie and Creel, 2017, Creel, 2018; Farley et al., 2024). Elk appear to compensate for increased predation pressure by exploiting risky places that likely have greater foraging opportunities at less risky times, particularly during seasons of greatest nutritional demand.

It is important to consider the effects that competing predators have on elk habitat selection. To fully understand the suite of anti-predator responses occurring and their effects on elk populations, future work could evaluate risk from mountain lions, wolves, and humans and with respect to maternal status. Although humans do not hunt elk year-round, hunter harvest was a large source of mortality for elk in this study and the perception of human risk and its interaction with other predators warrants further investigation. Furthermore, research has shown movement rates to be related to risk, which would reveal an additional mitigation strategy elk might employ in relation to predation risk (Basille et al., 2015; Ditmer et al., 2018). Predator–prey relationships are typically complex and dynamic, and it is critical to evaluate multiple representations of spatial and temporal risk. Existing research on the effects of predation risk of cursorial predators is often based on data from protected areas, lacking diversity in human land uses and activities (Moll et al., 2017; Say-Sallaz et al., 2019). This study revealed that in working landscapes, a more predictable diel period metric may better represent temporal risk compared to wolf activity and emphasizes how elk habitat use and response to risk is dynamic and temporally hierarchal at the seasonal and diel scale. Accounting for spatial and temporal differences in use provides insight to managers wishing to anticipate how elk populations may be affected by predators indirectly. It is also important for managers to consider non-consumptive effects and the mechanisms behind them as they can underly demographic parameters and may be less straightforward to quantify or mitigate. Increasing our understanding of prey responses in a variety of ecosystems, especially those which are multiple-use and have complex human–predator–prey–habitat interactions is necessary, particularly as wolves recover throughout the United States and elsewhere.