Intentionally removing predators from an ecosystem for the purpose of increasing prey populations is a common wildlife management strategy (Connolly, 1978; Salo et al., 2010; Peek et al., 2012). However, a growing number of studies question whether such predator removals are effective (Bergstrom, 2017). Given that predator removal efforts continue in many regions throughout the world, it is important to capitalize on ongoing removal programs by experimentally testing their efficacy and providing stakeholders with data needed to make informed management decisions (Bergstrom, 2017). A recent meta-analysis, for example, found that experimental predator removals only resulted in modest increases in prey abundance and/or survival at best, but that most studies lacked rigor in experimental design (Clark and Hebblewhite, 2021). Importantly, just six of the 52 experiments were conducted over a period of time that exceeded the target prey’s generation time (Clark and Hebblewhite, 2021). This mismatch is problematic because such removal schedules may not be long enough or in the locations needed to generate observable effects on prey survival (Connell and Sousa, 1983).

It is likely that both temporal and spatial aspects of predator removal influence response of prey. Temporal aspects can include both time of year and single vs. multiple years of predator control. An observed increase in survival over multiple years could result from (1) an absolute decrease in the total number of predation events in a given year or (2) a decrease in the proportion of animals taken due to increases in population size in subsequent years of predator removal. Both of these scenarios result in the desired outcome of increasing survival in prey populations. In either case, it is likely that time of year when predators are removed will affect likelihood of success. For example, predator removal during late winter may increase likelihood of increased prey survival because that timing disrupts social structure of predators during the time of pair bonding or reproduction and occurs just prior to birthing of prey species (Blejwas et al., 2002). Nonetheless, some predator species refill home ranges vacated by removed predators quickly. For example, under intense coyote (Canis latrans) removal programs, coyotes have been reported to repopulate areas within months to a short number of years (Beasom, 1974; Connolly and Longhurst, 1975; Connolly, 1995), although there are examples where coyotes do not rapidly recolonize (Mahoney, 2017). Because predators can exhibit compensatory natality and increased survival following reductions in density (Knowlton, 1972; Knowlton et al., 1999), it is likely that removal at the most important time of year and across multiple consecutive years has the greatest likelihood of a resulting increase in survival of prey.

Predation rates on domestic livestock have been correlated with density of coyotes on the landscape (Stoddart et al., 2001) leading to predator control programs that are often non-selective and broad in spatial scale (Mitchell et al., 2004). However, wild ungulates do not live and reproduce evenly or randomly across the landscape. Therefore, predator control programs focused on habitat where prey are most vulnerable have the greatest likelihood of success. Further, survival of prey should increase even if predators backfill the vacated space over several months to a few years since individuals that disperse or backfill tend to be young individuals that are less likely to take large prey items compared to resident adults (Harrison, 1992; Watine and Giuliano, 2017). Coyotes have the greatest effect on mule deer (Odocoileus hemionus) during the first few months of life when mule deer are small and unable to escape or during late winter when some individuals have exhausted energy reserves (Bowyer, 1987; Lingle, 2000). Mule deer susceptible to coyotes in late winter are likely individuals with severely depleted energy reserves (often associated with old age). Therefore, this source of mortality is more likely to be compensatory or have little effect on population growth. Consequently, coyote control programs focused on birthing and early-life rearing habitat have the greatest likelihood of benefiting mule deer by decreasing mortality that is more likely to be additive. Indeed, numerous studies indicate that coyote predation is a common source of mortality for mule deer fawns (Bartmann et al., 1992; Whittaker and Lindzey, 1999; Pojar and Bowden, 2004; Bishop et al., 2009). However, in perhaps the largest coyote removal experiment, mule deer fawn survival increased when coyotes were removed, but increases were modest and only evident when alternate prey (i.e., lagomorph) populations were low (Hurley et al., 2011). While the study by Hurley et al. (2011) employed a robust experimental design over six consecutive years, the number of coyotes removed varied drastically from year to year, ranging from a mean of 31.15–79.96 coyotes per 1,000 km² and occurred in a traditional manner over a broad spatial scale. This high variability in the number of coyotes removed combined with the broad spatial scale (i.e., non-targeted) leaves questions about the effectiveness of coyote removal during this early life stage of mule deer.

Mule deer have high cultural and economic value as a primary big game ungulate in western North America but have recently experienced fluctuations in population size (Bleich and Taylor, 1998; Unsworth et al., 1999; Peek et al., 2002; Bergman et al., 2015). As in other ungulate species (Gaillard et al., 1998), low fawn survival and subsequent low recruitment may drive fluctuations in mule deer populations (Peek et al., 2002; Lomas and Bender, 2007). In this study, we explicitly tested the effect of two consecutive years vs. a single year of coyote removal on survival of mule deer fawns in Utah, United States. Further, we attempted to examine the effectiveness of spatially explicit removal (i.e., removal from birthing and rearing habitat). Therefore, the objective of our study was to determine the effect of coyote removal on fawn survival using (1) a crossover experimental design in two comparably sized locales, (2) two consecutive years of control and treatment conditions in both locales, and (3) relatively consistent numbers of coyotes removed across years. Further, we examined the relationship between location of removal relative to birthing habitat and the likelihood of fawn survival. We hypothesized that both spatial and temporal aspects of predator removal would influence survival of neonate mule deer. We predicted that multiple years of predator removal would increase survival of mule deer fawns because reproduction by coyotes would be disrupted, densities would be further reduced, and the recovery of coyotes to pre-control levels would be prolonged. In addition, we predicted that control efforts in birthing habitat would be more effective at reducing fawn mortality than control efforts at a broad spatial scale. Importantly, our research site experienced normal climatic patterns (no extremely harsh winters or dry summers) during the years of the study, which enabled the examination of predation effects that were not confounded by severe weather.

During 2012–2015 we captured 287 adult female mule deer from winter-range locations surrounding Monroe Mountain (Table 1). Range of percent ingesta-free body fat (IFBF) was 2.9%–12.1% and was normally distributed around the mean. Of the 287 females captured, 95% (n = 273) were pregnant. Vaginal Implant Transmitters were inserted into 260 of these 273 pregnant females; 13 females were not used because they were recaptured and known to summer outside the two study areas (n = 5), their vaginas were too small for VIT insertion (n = 5), or they experienced capture-related injury (n = 3). Of the 260 animals that received VITs, 146 moved onto one of our study areas on Monroe Mountain. The remaining females either died before parturition or transitioned onto summer range outside the study areas and could not be included in the study (Table 1).

During late winter and early spring 2012–2015, Wildlife Services removed a total of 166 coyotes. These removals included 47, 34, 38, and 47 in 2012, 2013, 2014, and 2015, respectively. In 2012 and 2013, coyote removal efforts were on the north study area whereas removal efforts focused on the south study area in 2014 and 2015 (Figure 1). Ten coyotes were removed from the buffer area or from outside the study area. Each of these animals had at least part of their home range within the study area as all tracking events to locate a coyote were initiated within the removal study area.

Searches for neonate mule deer associated with expulsion of VITs began on 29 May, when the first VIT was expelled, and continued through early July. Based on expulsion of VITs, we obtained dates of parturition for 140 of the 146 females that remained in the study area during parturition. Five females died prior to parturition, and one female that received a VIT never gave birth. Mean dates of parturition were June 13th, 16th, 15th, and 14th for 2012–2015, respectively. Using VITs and opportunistic searching, we captured 266 neonate mule deer, including 71 sets of twins, between the two study areas. We excluded six neonates from our sample because they were stillborn (n = 2) or their deaths were human-caused (poaching, vehicle strike; n = 4); consequently, 260 neonates were used in final analyses.

We attributed mortality of neonate mule deer to predation, starvation, disease, stillbirth, roadkill, and unknown (Table 2). Predation was the leading cause of mortality accounting for approximately 68% of all mortalities. Coyotes killed 16% of all collared neonates on the north study area and 13% of all neonates monitored on the south study area. Approximately 90% of coyote-related mortality occurred during the first 16 weeks of life for neonates.

Survival of neonate mule deer to 16 weeks averaged 65% across all years. Sex and weight were influential variables in our models (Table 3). Females had a higher likelihood of survival than males. Additionally, survival of neonate mule deer increased as weight increased. None of our models contained uninformative parameters (Arnold, 2010).

Number of consecutive years of treatment (YrofTrtmt) was a prominent variable in our known-fate analyses, appearing in 8 of the top 10 models (Table 3). Likelihood of survival to 16 weeks of age for fawns in a treatment area with no coyote removal was 57.6% ± 5.1% (±SE). In contrast, likelihood of survival to 16 weeks of age increased to 65.4% ± 7.9% and 72.8% ± 5.1% with one and two consecutive years of coyote removal, respectively. Survival curves from a Kaplan–Meier function illustrated the pattern of survival relative to year of treatment (Figure 2).

The location of coyote removal relative to birth sites (HRCKills) appeared in 6 of the top 10 models which accounted for 40% QAICc weight. The support for HRCKills in our models indicated that proximity of coyote removal to fawning locations positively influenced the likelihood of survival of neonate mule deer (Table 3). The probability of a neonate mule deer surviving to 16 weeks increased with increasing numbers of coyotes removed in a deer home range surrounding birth locations (Figure 3).

Removing coyotes had a positive effect on survival of neonate mule deer to 16 weeks of age regardless of the density of alternate prey. Importantly, subsequent years of coyote control in the same area had a greater effect than a single year of control (Figure 2). This result is consistent with a study on pronghorn (Antilocapra americana) that highlighted the need for multi-year coyote removal to increase fawn survival (Smith et al., 1986). Even under intense coyote removal programs, coyotes have been reported to repopulate areas within months or a short number of years (Beasom, 1974; Connolly and Longhurst, 1975; Connolly, 1995). Additionally, some coyote populations have returned to pre-treatment levels through recolonization, compensatory breeding, and increased survival rates following coyote removal (Knowlton, 1972; Knowlton et al., 1999). Our findings underscore the need for consecutive annual removal efforts, with consistent removal among years, to have positive effects on survival of mule deer fawns.

Removing coyotes near birth sites of fawns increased fawn survival, whereas removing coyotes from areas distant from birth sites did not influence survival of neonates. Coyote removal is often conducted during winter months because coyotes leave tracks in the snow that managers can use to locate and remove them. Since coyotes generally use the same areas in winter as they do in summer on Monroe Mountain (Mahoney, 2017), removing coyotes from potential fawning habitat in winter would likely benefit neonate mule deer the following summer. Our results support this prediction. This finding is notable because many coyote removal studies do not consider proximity to prey birth sites when removing coyotes, which may explain the equivocal impacts on fawn survival that have been reported, if some studies removed coyotes far away from prey birth sites. Our results underscore the need for coyote removal programs to target coyotes near fawning habitat in order to be effective, whereas removal programs targeting coyotes distant from fawning habitat likely have no effect on prey populations.

Predation was the leading cause of mortality of neonate mule deer accounting for at least 68% of all mortalities. It is likely this estimate, however, is low and underestimates the actual effects of predators. For example, it is probable that at least some of the cases we assigned to starvation were due to the mother being killed by a predator resulting in the subsequent starvation of the neonate. In fact, we had multiple instances where we were able to determine this was the case by locating the deceased collared mother of the neonate mule deer that had starved. However, we do not know the full extent of this type of mortality because many of our collared neonates were opportunistically captured from unmarked adults. If populations are not limited by resources or there is capacity for the landscape to support a larger population, it is likely that reducing predation, the leading cause of mortality, will lead to an increase in population size.

The availability of alternative prey did not appear in any of our top models suggesting that alternative food was not a driving factor in the patterns of predation on mule deer that we observed. This finding is inconsistent with previous research demonstrating that alternative prey can influence the rate of predation on neonate mule deer (Hamlin et al., 1984). However, Sacks and Neale (2002) demonstrated that small prey comprise only a minor portion of coyote diet when neonate sheep are present on the landscape. We suggest that neonate mule deer are an optimal food source for coyotes due to their large size and relative ease of capture. Therefore, coyotes potentially take neonate mule deer selectively when available. Coyotes did not start killing neonate mule deer until approximately the mean date of parturition each year (Hall, 2018), suggesting that the predators were unaware of the presence of the new food source on the landscape initially. Once coyotes started killing neonates, loss to coyotes became common suggesting that the predators shifted their behavior to target neonate mule deer. Indeed, some generalist predators switch to being a specialist when a potentially preferred prey becomes available (Roughgarden, 1974). A possible alternative explanation for the pattern we observed is that lagomorphs (i.e., alternative prey) were not randomly distributed with greater distribution at low elevations, whereas fawning habitat was at relatively high elevations. If coyotes do not transition up and down the mountain as suggested by Mahoney (2017), coyotes living in areas with mule deer neonates did not have access to lagomorphs, and therefore, shifts in abundance of lagomorphs should not influence taking of neonate mule deer.

Our study focused on spatial aspects of predator control and the effect of two consecutive years vs. a single year of coyote removal on fawn survival. Our results indicate that predator control efforts focused on fawning habitat have the greatest likelihood of leading to increased population growth. This finding is especially meaningful because there were significant sources of error (e.g., variation in the spatial density of predators and prey) that our study could not control. In addition to the spatial effects of predator control efforts, we found that two consecutive years of predator control are better than a single year. While it would be ideal to test the effect of additional consecutive years, it is difficult to maintain optimal experimental conditions for extended durations. As we described in the Introduction, the study by Hurley et al. (2011) took place over 6 years, but coyote removal numbers varied drastically from year to year in that study (we maintain that their study was a feat despite this limitation). Focusing on only 2 years of removal using a crossover design allowed us to maintain consistent removal efforts and carry out our desired goal of experimentally evaluating the effect of removal years, but it also precluded our ability to thoroughly examine population growth in response to our removal efforts given the short timeframe. Although increasing the survival rates of neonate ungulates can increase recruitment and the rate of population growth, it is possible that predator-related mortality is compensatory and not additive—especially for populations that are limited by the availability of nutritional resources. Therefore, it is crucial for future studies to determine the parameters of a prey population where increased fawn survival resulting from multiple consecutive years of consistent coyote removal translates to positive population growth. Further, future studies should determine how long the effect of predator removal lasts after removal efforts cease. Nevertheless, our results are valuable given that we were able to experimentally demonstrate the effect of multiple years vs. a single year of removal on fawn survival in a controlled naturalistic setting.

Managing for robust deer populations is often a high priority for agencies charged with managing wildlife. Our results indicate that coyote control can increase survival of neonate mule deer. Coyote control increased survival when (1) control efforts occurred for multiple consecutive years, and (2) when control efforts occurred in or near fawning habitat. Therefore, efforts to control predators should occur at relatively high elevation with shrubby understory consistent with the location of fawning habitat (Long et al., 2009; Freeman, 2014). Likewise, efforts should occur when populations have room to grow (size is below carrying capacity) to decrease the likelihood of compensatory mortality.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Brigham Young University Institutional Animal Care and Use Committee.

BM, KH, and RL conceived the ideas and designed the study and its methods. JH and EF conducted much of the fieldwork and data collection. JH and RL conducted the statistical analyses and produced the figures. BM and JH drafted the initial version of the manuscript. All authors contributed to the article and approved the submitted version.

Funding was provided by Sportsmen for Fish and Wildlife, Mule Deer Foundation, Utah Division of Wildlife Resources (Contract #126240), Brigham Young University, and Federal Aid in Wildlife Restoration.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.