Survival and cause-specific mortality of coyotes in Wisconsin

Lydia L S Margenau, Robin E Russell, Alexander T Hanrahan, Nathan M Roberts, Jennifer L Price Tack, Daniel J Storm, Survival and cause-specific mortality of coyotes in Wisconsin, Journal of Mammalogy, Volume 104, Issue 4, August 2023, Pages 833–845, https://doi.org/10.1093/jmammal/gyad033

Abstract

Understanding the drivers of population dynamics informs management actions and assures the public that harvest activities are not detrimental to the long-term stability of wildlife populations. We examined the survival and cause-specific mortality of 66 adult coyotes (34 males, 32 females) using GPS radiotelemetry in southwestern Wisconsin during October 2016 to March 2020. We paired our study with a literature review of coyote survival and mortality across the United States and Canada, focusing on the geographical distribution of studies, demographic aspects of survival, and the level of exploitation by humans on coyote populations. In Wisconsin, annual survival did not differ between sexes or across years but did vary among seasons and social statuses. The relative risk for a coyote dying was higher during the winter compared to the summer. A transient coyote had a higher relative risk of mortality compared to a resident coyote. Mean annual survival probability (sexes combined) was higher for a year-long resident compared to a year-long transient. The predominant sources of known mortality (n = 37) were harvest (83.8%) and vehicle collisions (13.5%). For our literature review, we identified 56 studies estimating coyote survival or mortality from 1971 to 2021 spanning the geographic range of coyotes. We found no distinct temporal or regional patterns in survival probability or the proportion of human-induced mortality, although fewer studies originated from the northeast region of the United States. Additionally, we detected weak correlation between survival probability and proportion of human-induced mortality, suggesting coyote harvest may be compensatory. Although our findings indicate that the Wisconsin coyote population had relatively higher human-induced mortality than populations in other regions, these mortality rates appear to be sustainable for this population under current landscape and habitat conditions.

Comprender los factores que gatillan las dinámicas poblacionales sirve de base para las medidas de gestión y garantiza que las actividades de captura no son perjudiciales para la estabilidad a largo plazo de las poblaciones de fauna. Examinamos la supervivencia y la mortalidad por causas específicas de 66 coyotes adultos (34 machos y 32 hembras) mediante radio telemetría por GPS, en el suroeste de Wisconsin, desde octubre de 2016 hasta marzo de 2020. Combinamos nuestro estudio con una revisión bibliográfica sobre la supervivencia y la mortalidad del coyote en Estados Unidos y Canadá, y nos enfocamos en la distribución geográfica de los estudios, los aspectos demográficos de supervivencia, y el nivel de explotación por parte del hombre de las poblaciones de coyotes. En Wisconsin, la supervivencia anual no difiere entre sexos ni a través de los años, pero sí varía según las estaciones y el estatus social. El riesgo relativo de que un coyote muera durante el invierno en comparación con el verano. Un coyote en estado de tránsito tiene un mayor riesgo relativo de mortalidad en comparación con un coyote residente. La probabilidad media de supervivencia anual (ambos sexos) para un coyote residente durante un año en comparación con un para un coyote en estado de tránsito durante un año. Las fuentes de mortalidad predominantes conocidas (n = 37) fueron las capturas (83.8%) y las colisiones con vehículos (13.5%), lo que dio lugar a una incidencia acumulada para las capturas y para las colisiones con vehículos en todos los coyotes. Para la revisión bibliográfica, identificamos 56 estudios que estimaban la supervivencia o mortalidad del coyote desde 1971 hasta 2021, abarcando toda el área de distribución geográfica de esta especie. No identificamos patrones temporales o regionales distintivos en la probabilidad de supervivencia o en la proporción de mortalidad provocada por el hombre, aunque el menor número de estudios se iniciaron en la región noreste de Estados Unidos. Además, observamos una leve correlación entre la probabilidad de supervivencia y la proporción de mortalidad provocada por el hombre, lo que sugiere que la captura de coyotes puede ser compensatoria. Si bien nuestros hallazgos indican que la población de coyotes en Wisconsin tuvo una mortalidad provocada por el hombre relativamente más alta que las poblaciones en otras regiones, estos índices de mortalidad parecen mantenerse en esta población, en particular en las condiciones actuales de entorno y hábitat.

Coyotes (Canis latrans) are the most widely distributed and abundant mammalian predator in North America. Before predator removal efforts began in North America, gray wolves (C. lupus) and red wolves (C. rufus) inhabited most of the forested landscapes, restricting coyotes to open savannas, prairies, and arid regions in the western two-thirds of North America ( Jackson 1961; de Vos 1964; Gipson 1978). With near total extirpation of wolves in the conterminous United States and increase in timber harvest and agricultural land use, highly adaptable coyotes expanded their geographic range, now extending from northern Alaska and the Yukon and Northwest territories of Canada to as far south as Central America and spanning from the Pacific Ocean to Atlantic Ocean coastlines of North America ( Bekoff and Wells 1986; Parker 1995; Laliberte and Ripple 2004; Hody and Kays 2018). Currently, coyotes inhabit and are managed through harvest within every state/province in the continental United States and Canada. This population expansion has precipitated an increase in human–coyote conflict ( Siemer et al. 2007; White and Gehrt 2009; Gehrt and Riley 2010), genetic hybridization with the endangered red wolf ( McCarley 1962; Kelly et al. 1999) and eastern gray wolf ( Lehman et al. 1991; Kays et al. 2010; Wheeldon et al. 2010; Monzón et al. 2014), and conflict with hunters over perceived effects on white-tailed deer ( Kilgo et al. 2012; Jackson and Ditchkoff 2013; Gulsby et al. 2015; Conner et al. 2016), upland game birds ( Guthrey and Beasom 1977; Guthrey 1995), and waterfowl populations ( Sooter 1946; Sovada et al. 1995; Reynolds et al. 2001; Brown 2007).

Although coyotes are often perceived negatively by humans and are viewed as a nuisance species in many jurisdictions ( Kellert 1985; Frank et al. 2016; Nardi et al. 2020; Vaske and Sponarski 2021), coyotes play important ecological roles. Through interference competition and predation, coyotes limit mesocarnivore populations, generating a trophic cascade for prey species, such as increasing avian diversity and abundance ( Harrison et al. 1989; Sovada et al. 1995; Crooks and Soulé 1999). Coyote predation on rodents and lagomorphs contributes to a reduction of agricultural crop damage caused by these prey species ( Henke and Bryant 1999; Crimmins et al. 2012; Windberg and Mitchell 2013). As scavengers, coyotes remove carcasses, potentially aiding in the reduction of zoonotic disease transmission ( Szcodronski and Cross 2021).

Coyote harvest regulations are almost universally unrestrictive ( Supplementary Data SD1 ). This is exemplified by Wisconsin, where coyotes were subject to a bounty until 1963, after which they were managed as a furbearer game species. Current harvest regulations in Wisconsin include open year-round hunting, a trapping season from fall through mid-winter, and no harvest quota or bag limit. Trappers in Wisconsin harvest approximately 10,000–15,000 animals annually, with an additional 35,000–60,000 harvested by hunters ( Kitchell 2020). Coyote populations in Wisconsin, like most jurisdictions, are not monitored intensively; thus, information is lacking with respect to the influence of harvest and harvest regulations on coyote populations.

Understanding the drivers of population dynamics informs how harvest regulations affect wildlife populations and assures the public that harvest activities are not detrimental to the long-term stability of wildlife populations. Hunting and trapping regulations specify equipment used, timing and duration of harvest seasons, limits on individual and cumulative maximum harvests, standards for trapper education and training, licensing and reporting requirements, and a variety of other specific requirements depending on jurisdiction and management protocols in place ( White et al. 2015). The wide assortment of regulatory mechanisms gives management agencies the flexibility to adapt and respond to changes in coyote abundance.

In this paper, we aim to estimate factors influencing the survival of coyotes in Wisconsin and compare survival of Wisconsin coyotes to survival in coyote populations across their geographic range. First, we investigated coyote survival within a southwestern Wisconsin population. Specifically, we estimated annual survival, determined potential effects of sex, social status, and season on survival, and explored cause-specific mortality of coyotes. We predicted that survival estimates would be similar among sexes but differ across social status and season. Next, we reviewed the existing scientific literature on coyote survival and mortality in the United States and Canada. We summarized the reported effects of human exploitation on coyote survival and evaluated whether those effects differed by geographical region. We predicted that human exploitation would be the dominant cause of coyote mortality and coyote survival would be highly variable across populations.

Materials and Methods

Study area

We primarily captured coyotes in forested lands in northern Iowa and Grant counties, Wisconsin, United States, although collared coyotes moved throughout southwestern Wisconsin ( Fig. 1). Trapping occurred within the Western Coulee and Ridges ecological landscape, defined by highly eroded, unglaciated topography with steep valleys and ridges, high-gradient headwater streams, and large rivers with extensive, complex floodplains and terraces ( Wisconsin Department of Natural Resources 2015). This region contained a mix of oak–hickory (Quercus spp.–Carya spp.) and maple–basswood (Acer spp.–Tilia spp.) forests (41%), agriculture (36%), and grasslands (14%) within a predominantly privately owned landscape. The climate of the Western Coulees and Ridges ecological landscape is similar to other ecological landscapes in southern Wisconsin with a mean growing season of 145 days and mean annual temperature of 6.5°C. The average minimum January and maximum August temperatures ranged from −17.5°C to 27.3°C, respectively, with a mean annual precipitation of 82.8 cm and snowfall of 109.2 cm ( Wisconsin Department of Natural Resources 2015).

Land cover and capture locations of coyotes in northern Iowa and Grant counties, Wisconsin, United States. The capture region is indicated in the inset map by the hollow, rectangular polygon, while the shaded counties indicate the broader study area where GPS telemetry locations were collected in Wisconsin.

Coyote capture and radiotelemetry

We captured coyotes from October 2016 through January 2019, in coordination with local trappers voluntarily reporting coyotes to be collared and released and through a trapper payment system offering $100 per uninjured coyote of adequate size to collar. Additionally, project personnel captured coyotes during fall and winter using coyote cable restraints and two-coiled foothold traps with offset jaws (MB-550-RC and #2 Bridger Dogless Modified [laminated], Minnesota Trapline Products, Inc., Pennock, Minnesota). We used intramuscular injections of ketamine (4 mg/kg of body weight), dexmedetomidine (0.02 mg/kg of body weight), and butorphanol (0.4 mg/kg of body weight) to anesthetize coyotes during handling. Atipamezole (0.2 mg/kg) was used as an antagonist to reverse the effects of the dexmedetomidine when handling was complete but not prior to 20 min following the initial injections. We applied uniquely numbered ear tags to each individual and recorded weight, body length, and the general condition of each coyote. We determined age as: juvenile (

To determine cause of death, we investigated collars that were transmitting a mortality alert following >12 h of immobility. Investigations occurred as soon as possible following the mortality alert, and we attempted to retrieve the collar and carcass if possible. For nonharvest mortalities, we conducted site investigations and, if necessary, field or laboratory necropsies to gather evidence for assigning cause of mortality. Harvest mortalities were reported by hunters or trappers, who were interviewed for information on the harvest circumstances (e.g., harvest date, time, location, and method).

We designated the social status of individuals as resident or transient based on variation in movement patterns ( Gese et al. 1988; Berger and Gese 2007) and differences in home range size during the six-month seasonal periods of summer (April–September) and winter (October–March) using a kernel density estimator in adehabitatHR ( Calenge 2006). While assessing assumptions of normality, we found the distribution of coyote home range size to be bimodal ( Kamler and Gipson 2000; Hinton et al. 2015). We designated coyotes with a home range size >30 km 2 as transient based on the bimodal distribution observed and the difference in movement patterns of those individuals relative to resident coyotes. Coyotes that exhibited extended periods of home range contraction or expansion during a six-month seasonal period were designated as a ‘mix’ social status, with movement characteristics of both transients and residents during the season. We identified coyotes with

Survival analysis

We analyzed our data using a Bayesian version of the proportional hazard regression model for grouped (discrete) data to evaluate the time-to-event (mortality) as a function of covariates ( Prentice and Gloeckler 1978). This model allows for the comparison of groups with respect to their hazard rates, which is the risk of mortality given that the individual has survived to a specific time, using hazard ratios. The continuous hazard model can be described as h ( t | X ) = h 0 ( t ) exp ⁡ ( β X ) = h 0 ( t ) e x p ( ∑ k = 1 K β k X k ) ⁠ , where h ( t | X ) is the expected hazard rate at time t conditional on X ⁠ , h 0 ( ⋅ ) is the baseline hazard (nonparametric), β is the vector of unknown regression coefficients (parametric), and X is the matrix of K covariates. Simply, h ( t | X ) is the product of the baseline hazard and the exponential function of the linear combination of the covariates. Proportional regression hazard models assume: (1) an independence of survival times between distinct individuals in the sample; (2) a multiplicative relationship between the covariates and the hazard; and (3) a constant hazard ratio over time.

The survival analysis in the hazard model converts the hazard ratio to survival times through distributions, such as exponential, Weibull, or log-normal distributions. We opted to use the exponential distribution to describe the survival function as S ( t | X ) = e x p < − H 0 ( t ) e x p ( β X ) >⁠ , where H 0 ( t ) is the cumulative baseline hazard. After discretization, the likelihood of the survival function can be written as S ( s | r X ) = e x p < − ∑ r s e x p ( γ + β X ) >⁠ , where S ( s | r X ) is the rate of survival to time s given survival to time r and X covariates, γ is the intercept value, and β is a vector of estimated coefficients. We estimated γ and β using NIMBLE ( de Valpine et al. 2017; NIMBLE Team 2022) in R ( R Core Team 2020; Supplementary Data SD2 ). Priors for γ and β were chosen to be noninformative, normally distributed with a mean of zero and an inverse variance of 0.001. We ran three chains for 100,000 iterations and a burn-in of 50,000 iterations. Parameters were checked for convergence using Ȓ values ( Gelman and Rubin 1992).

The analysis was based on survival histories for October 2016 to March 2020, with staggered entry of individuals. Individuals for which fate could not be determined and individuals alive at the end of the monitoring period were treated as random censoring events. We evaluated eight time-to-event models including an intercept-only model (null model); four univariate models evaluating the effects of sex, social status, season (summer [April–September] or winter [October–March]); and year (continuous time trend); and three additive effects models of sex and season, social status and season, and sex, social status, and season on coyote survival. Additionally, we attempted to model year as a categorical effect, but the models failed to converge, and were thus excluded from further analysis. We conducted model selection using the Watanabe Akaike Information Criteria (WAIC; Watanabe 2010). Additionally, we estimated mortality-specific (harvest and vehicle collision) hazard rates for 1-year post-collaring as a cumulative incidence function (CIF; Heisey and Patterson 2006) using the subset of coyotes with known causes of death.

Literature review

We implemented a snowballing procedure ( Wohlin 2014) for conducting our systematic literature review of coyote survival and cause-specific mortality from 17 September to 18 October 2021. We first identified our start set literature using several search engines including AGRICOLA, PROQUEST, and Google Scholar. Search terms included coyote, Canis latrans, survival, mortality, demography, and population. We classified literature under the categories publication date, type of article (peer-reviewed, theses and dissertations, or gray literature [e.g., book chapter, conference proceedings]), study area (state/province), study objective, monitoring method (VHF telemetry, GPS telemetry, ear-tagging only, or scat collection), and survival estimation method. Following the examination of the start set for inclusion, we implemented several iterations of backward snowballing (i.e., used the reference/literature cited list to identify new papers to include) and forward snowballing (i.e., identified new papers to include based on papers citing the current paper being examined) to produce the final set of literature included within the review. We synthesized literature content to provide a summary of coyote survival and mortality estimates in the United States and Canada during the past 50+ years. We calculated overall range-wide mean estimates for survival and human-induced mortality using the subset of studies reporting sample sizes alongside the parameter estimates. We used descriptive statistics to explore spatiotemporal patterns in survival and mortality with the focus on the four United States census regions of West, Midwest, South, Northeast, and an additional region encompassing Canada, which tended to reflect similar coyote management regulations ( Supplementary Data SD1 ) and historical coyote range expansions. Additionally, we used Kendall’s rank correlation coefficient (Kendall’s tau) to measure the association between survival probability and the proportion of total mortality attributable to humans across studies ( Kendall and Gibbons 1990; Newson 2002).

Results

Wisconsin coyote survival

Between October 2016 and January 2019, we captured 74 coyotes (38 males, 36 females) in Iowa and Grant counties, Wisconsin. We captured seven coyotes in 2016, 34 coyotes in 2017, 23 coyotes in 2018, and 10 coyotes in 2019. Of the 74 coyotes captured, we GPS radio-collared 68 coyotes (35 males, 33 females). One male was initially captured and collared in 2017 and later re-captured and re-collared in 2019, with the collar still functioning at time of recapture. Our sample was comprised predominantly of adult individuals ( Supplementary Data SD3 ), except for two juvenile individuals (one male, one female) that were excluded from statistical analyses.

We assigned cause of death to 37 mortality events (20 males, 17 females). Human-related mortalities accounted for 97.3% of coyote deaths, with harvest (shooting = 29, trapping = two) being the predominant cause (83.8%) followed by vehicle collisions (13.5%) and natural causes (2.7%). We recorded 33 (89.2%) mortality events in the winter, with the remaining four (10.8%) mortalities occurring in the summer. We recorded loss of transmitter signals and GPS radio collar failure for 20 (30.3%) coyotes resulting in right censorship in the survival analysis, although seven individuals were later harvested by hunters. Nine (13.6%) individuals remained alive and with functioning collars at the end of the monitoring period. The cumulative incidence at day 365 (one-year post-collaring) was 0.47 (95% CI: 0.37–0.58) for harvest, 0.06 (95% CI: 0.02–0.13) for vehicle collision, and 0.01 (95% CI: 0.00–0.04) for natural causes ( Fig. 2).

Cumulative incidence function for cause-specific mortalities (harvest and vehicle collision) for a coyote, one-year post-collaring (365 days).

We classified social status for coyotes during each seasonal period of GPS monitoring from winter 2016–2017 to winter 2019–2020. We identified 15 coyotes as switching seasonal social status at least once during their monitoring period. We assigned seasonal social statuses to 82 resident coyotes, 58 transient coyotes, 16 mix coyotes, and two unknowns. For analysis, we combined transient, mix, and unknown into a singular ‘transient’ factor for comparison against resident coyote status. Of the 37 coyotes assigned cause of death, we observed three individuals categorized as mix social status during time of death. One individual died due to natural causes and displayed movement patterns more characteristic of a resident coyote during the latter portion of winter 2018–2019. The other two coyotes were harvested during winter 2017–2018 and winter 2018–2019 while exhibiting potential transient and dispersing behaviors at the time of death.

The best-supported survival model included a seasonal and social status effect on survival probability ( Table 1). The relative risk for a coyote dying during the winter compared to the summer was 2.53 (95% CI: 1.70–3.92), reflecting a higher probability of mortality over the winter period. Residents had a higher probability of survival compared to transients, with a relative risk of mortality of 2.19 (95% CI: 1.54–3.29) for a transient compared to a resident. The estimated survival for a coyote over a 180-day summer period was 0.86 (95% CI: 0.83–0.93) for a resident versus 0.76 (95% CI: 0.68–0.84) for a transient ( Fig. 3). The estimated survival for a resident coyote over a 180-day winter period was 0.73 (95% CI: 0.65–0.80) versus 0.51 (95% CI: 0.43–0.58) for a transient.

Model selection of proportional hazard regression models for coyotes in southwestern Wisconsin. Models are ranked based on Watanabe Akaike Information Criteria (WAIC) relative differences between the top-ranked model and each other model (ΔWAIC), WAIC model weights (w_i), and the number of parameters (k). Parameters included sex (male or female), social status (resident or transient [including mix and unknown]), season (winter [October–March] or summer [April–September]), and year (continuous time trend).

Model selection of proportional hazard regression models for coyotes in southwestern Wisconsin. Models are ranked based on Watanabe Akaike Information Criteria (WAIC) relative differences between the top-ranked model and each other model (ΔWAIC), WAIC model weights (w_i), and the number of parameters (k). Parameters included sex (male or female), social status (resident or transient [including mix and unknown]), season (winter [October–March] or summer [April–September]), and year (continuous time trend).

Estimated survival for 180 days of a coyote collared at the beginning of summer (April–September) versus the beginning of winter (October–March) for a resident and transient.

The additive effects of season and social status resulted in four annual (365 days) survival estimates. The estimated annual survival for a year-long resident coyote was 0.65 (95% CI: 0.55–0.73), while a year-long transient had an estimated annual survival of 0.38 (95% CI: 0.31–0.46). A coyote that transitioned from a summer resident to a winter transient had an estimated survival of 0.44 (95% CI: 0.38–0.51). A coyote that transitioned from a summer transient to a winter resident had an estimated survival of 0.55 (95% CI: 0.48–0.63).

While the second-best survival model included an additional sex-specific additive effect on survival probability with social status and season, the sex-only model performed the worst out of the models tested ( Table 1). Annual survival estimates were only nominally different between males (0.45 [95% CI: 0.33–0.58]) and females (0.52 [95% CI: 0.42–0.63]).

Literature review

We identified 56 studies estimating coyote survival or mortality in the United States and Canada from 1971 to 2021 ( Supplementary Data SD4A ). We were unable to access the complete manuscript or article of seven additional studies presumed to estimate coyote survival or mortality identified during our snowballing procedure, and these studies were thus excluded from further examination ( Supplementary Data SD4B ). We excluded the use of abstract previews and summarizations, paraphrasing, and quotations from referencing literature as a method for validating coyote estimates within studies. Additionally, we excluded one study ( Prugh et al. 2005), which provided survival estimation results in graphical format only without providing exact point estimates. Of the 56 studies in the literature set, 29 were peer-reviewed journal articles, 24 were theses or dissertations, and three were gray literature. We identified five major study objective areas within the literature: 43 studies investigated demographic objectives; 14 studies addressed spatial–ecological objectives; eight studies incorporated movement–ecological objectives; five studies had disease ecology objectives; and four studies focused on new sampling techniques or statistical methodologies. Additional areas of research focus included intraspecies interactions and social dynamics (n = six studies), interspecies interactions with other carnivores (n = seven studies), and human–wildlife interactions (n = eight studies). Survival estimates were included in 44 of the studies, while cause-specific mortality was estimated in 43 of the studies. The geographic distribution of studies extended across the United States and Canada, reflecting coverage of the current range of coyotes ( Fig. 4).

frequency of estimates from the state or province region." />

Location of studies providing coyote survival and mortality estimates, based on a review of literature from 1971 to 2021 ( Supplementary Data SD4 ). Color intensity represents frequency of estimates from the state or province region.

In the 44 studies reporting coyote survival estimates, we identified six field sampling and monitoring techniques used to derive samples for survival analysis. Very high frequency (VHF) radio collars were the predominant sampling technique (n = 32 studies), followed by ear-tagging only (n = three studies). GPS radio collar monitoring, a combination of GPS and VHF radio collar monitoring, scat survey, and carcass retrievals occurred within two studies each.

A variety of survival estimation procedures accompanied the suite of sampling techniques, reflecting temporal trends in increasing computational power and development of novel statistical software. Life tables, Brownie tag-recovery models ( Brownie et al. 1978), the Chapman–Robson method ( Chapman and Robson 1960), and the Trent–Rongstad method ( Trent and Rongstad 1974) were predominantly used in the 1970s and 1980s to estimate coyote survival using carcass retrievals, ear-tag recoveries, or VHF radio collar data. In the late 1980s, studies shifted toward using staggered entry Kaplan–Meier nonparametric techniques ( Kaplan and Meier 1958; Pollock et al. 1989) and Program MICROMORT ( Heisey and Fuller 1985) allowing for the censorship of animals with unknown fates and incorporation of individual covariates in modeling frameworks. The development of Program MARK ( White and Burnham 1999) allowed for increased accessibility in using known fates models for survival estimation during the early 2000s. Modern approaches for coyote survival estimation of GPS radio collar data have transitioned toward using modifications of the Cox proportional hazards regression model ( Therneau and Grambsch 2000), allowing for the estimation of hazard rates alongside survival estimates. Recently, mark–recapture studies with scat collection have been used as an alternative noninvasive sampling technique for the estimation of (apparent) survival rates using Jolly–Seber (JS) models ( Jolly 1965; Seber 1965) and Cormack–Jolly–Seber (CJS) models ( Cormack 1964; Jolly 1965; Seber 1965).

We report our review of coyote survival using three categories (pooled, juvenile, or adult) defined by the sample used for analysis within each study. Studies either provided a pooled estimate incorporating both juveniles and adults within the sample for analysis ( Supplementary Data SD5, Fig. A ) or provided age-specific survival estimates for the juvenile ( Supplementary Data SD5, Fig. B ) and adult ( Supplementary Data SD5, Fig. C ) age classes. Within the 24 pooled studies, annual survival ranged from 0.20 ( Bogan 2004) to 0.92 ( Gese 2005; Seidler and Gese 2012) across coyote populations, with an overall range-wide mean annual survival of 0.65 ( Supplementary Data SD5, Fig. A ). Among the nine studies reporting juvenile coyote survival, estimates ranged from 0.00 ( Pence et al. 1983; subpopulation infected with sarcoptic mange) to 0.74 ( Harrison 1992; resident subpopulation; Supplementary Data SD5, Fig. B ). Due to variation in the timing of collaring of juveniles and age classification of juveniles among studies, we did not calculate an overall range-wide mean annual survival estimate. Within the 19 studies reporting adult coyote survival, estimates ranged from 0.23 ( Pence et al. 1983; subpopulation infected with sarcoptic mange) to 0.87 ( Gese et al. 1989), with an overall range-wide mean annual survival of 0.58 ( Supplementary Data SD5, Fig. C ).

Within the 43 studies referencing mortality rates, we noted 1,351 coyote mortalities from 1966 to 2018 ( Supplementary Data SD5, Fig. D ). We identified 10 specific causes of death in the literature: vehicle (automobile and snowmobile) collision (n = 31 studies); shooting/hunting (n = 30 studies); trapping/snaring (n = 22 studies); disease/parasite (n = nine studies); predation (n = five studies); toxicants (n = four studies); starvation (n = four studies); intraspecific mortality (n = two studies); hunting dogs (n = two studies); and unknown causes due to carcass deterioration (n = 24 studies). The percentage of human-induced mortalities represented a combination of mortalities due to vehicle collisions, shooting/hunting, trapping/snaring, hunting dogs, and toxicants and ranged from 22% (n = seven; Holzman et al. 1992) to 100% (n = five, 14, and 9 in Kalmer and Gipson 2000; Crête et al. 2001; and Grinder and Krausman 2001, respectively). Overall, human-induced mortalities represented 77% of all coyote mortalities in the literature ( Supplementary Data SD5, Fig. D ). Additionally, we detected a weak correlation (τ = −0.12, P = 0.42) between survival probability and the proportion of human-induced mortality within the subset of studies (n = 22 studies) that reported both metrics for adult and pooled coyote samples.

We found no distinct temporal or regional patterns in survival probability or the proportion of human-induced mortality ( Table 2). Survival and mortality estimate ranges overlapped across all regions, although human-induced mortality in Canada had less variability. We found limited information regarding coyote survival and mortality from the northeast region, with estimates available for only two states. Additionally, investigation into coyote survival and mortality lagged in the northeast region, with monitoring efforts beginning 6–17 years after other regions and the first publication not until the 1990s.

Regional minimum and maximum annual survival and human-induced mortality estimates for coyotes reported from literature.

Regional minimum and maximum annual survival and human-induced mortality estimates for coyotes reported from literature.

Discussion

The survival of coyotes varies across their geographic range. The annual survival rates of coyotes in Wisconsin were within the range of estimates reported in other studies across North America. In accord with our prediction, the Wisconsin coyote population experienced seasonal variation in survival, with higher survival during the summer pup-rearing season. Temporal patterns of mortality are consistent across the geographic range (e.g., Chamberlain and Leopold 2001; Van Deelen and Gosselink 2006; Berger and Gese 2007). Seasonal variation in survival is further compounded by social status, with transients at higher risk than residents. Dispersal of juvenile and yearling coyotes usually begins in autumn and continues throughout the winter ( Voigt and Berg 1987), with individuals voluntarily leaving their natal territory due to mate competition and/or resource competition ( Gese et al. 1996). Dispersers and transients tend to make large movements that may increase hunter and trapper encounters within unfamiliar habitats ( Bekoff and Wells 1986), resulting in higher mortality rates than residents (e.g., Andelt 1985; Harrison 1992; Berger and Gese 2007).

Seasonal variation in survival is also associated with seasonal differences in hunting and trapping effort associated with environmental conditions and coyote pelt quality. Hunting effort is typically highest during the winter season when vegetative cover is limited and when opportunistic harvest by hunters pursuing ungulates and gallinaceous birds is greatest ( Chamberlain and Leopold 2001; Van Deelen and Gosselink 2006). Previous coyote research in Wisconsin noted high levels of opportunistic hunting during White-tailed Deer (Odocoileus virginianus) gun-hunting season ( Smith 1984), although we only observed five coyotes harvested during the 9-day firearm deer hunting season in Wisconsin. Additionally, trapping effort for coyotes is typically highest during the winter season, when it coincides with the furbearer trapping season in most states and provinces and when coyotes have prime pelts ( Supplementary Data SD1 ; Voigt and Berg 1987). Trends in trapper harvest are strongly influenced by furbearer abundance ( Allen et al. 2019), the number of trappers ( Ahlers et al. 2016; Bauder et al. 2020b), and trapper effort and success ( Ruette et al. 2003; Ahlers et al. 2016; Bauder et al. 2020a). During the 2020–2021 trapping season, 34.4% of Wisconsin trappers actively pursued coyotes ( Dhuey and Rossler 2021), representing the third highest-targeted species behind raccoons (Procyon lotor) and muskrats (Ondatra zibethicus). While one study in Alberta, Canada, found that survival of coyotes varied with local pelt sales and prices ( Pruss 2002), another study in Illinois reported annual trapper harvest for coyotes was not strongly affected by socioeconomic or environmental factors ( Bauder et al. 2020a). In Wisconsin, annual trapper and hunter harvest for coyotes have increased in recent decades ( Kitchell 2020) with nominal influence by annual inflation-adjusted pelt prices.

Sex is typically not an important source of variation in studies of Coyote survival (e.g., Windberg et al. 1985; Grinder and Krausman 2001; Wheeldon 2020), which is consistent with our findings. This finding contrasts with other harvested carnivores–including polar bears (Ursus maritimus; Derocher et al. 1997); cougars (Puma concolor; Cooley et al. 2009); Eurasian lynx (Lynx lynx; Nilsen et al. 2012); bobcats (Lynx rufus; Allen et al. 2018)–where males are generally at higher risk than females to hunting pressure and mortality. Harvest selectivity may arise due to hunter preferences (e.g., trophy hunting), opportunities to be selective via management regulations (e.g., quotas, season lengths, harvest methods), population demography (i.e., abundance, sex–age structure), or differential risks caused by variability in individual characteristics (e.g., movement patterns, morphology, social status, risk aversion; Mettler and Shivik 2007; Bischof et al. 2009; Mysterud 2011). Harvest selectivity is unlikely to be occurring within the Wisconsin population of coyotes due to nonrestrictive management regulations (year-round hunting season, unlimited quota), limited desire/ability for hunters to select based on secondary sex characteristics, and negative perceptions (e.g., nuisance, varmints, predators) of coyotes by hunters.

Sex-biased age structures may suggest differences in mortality risks between sexes under varying levels of exploitation ( Caughley 1977; Fuller et al. 1985; Lloyd 1998; Krofel et al. 2012). Balanced sex ratios typically exist within lightly exploited populations ( Gese et al. 1989; Windberg 1995), while adult female-biased population structures usually appear in heavily exploited populations ( Lloyd 1998; Jackson 2014). This population characteristic may be driven more so by age rather than sex, with increased vulnerability of juvenile males to human-induced mortality due to higher dispersal rates ( Windberg 1985; Chamberlain and Leopold 2001; Van Deelen and Gosselink 2006). Yet, some studies found no differences in survival rates due to age ( Holzman et al. 1992; Grinder and Krausman 2001; Wheeldon 2020). We were unable to explore age-specific survival in the Wisconsin population due to limited monitoring of juveniles from our opportunistic sampling of coyotes.

While coyotes die of various causes, most deaths are attributed to anthropogenic causes including hunting, hounding, trapping, snaring, poisoning, and vehicle collisions. Historically in Wisconsin, coyotes were probably preyed upon predominantly by gray wolves and cougars ( Jackson 1961), but with predator control, humans have become the dominant source of mortality with few naturally occurring deaths ( Smith 1984). We attributed 97.3% of coyote deaths to humans, which was higher (91st percentile) than the mean across literature (77%). Gray wolves in Wisconsin also experience a high proportion of human-induced mortality ( Wiedenhoeft et al. 2020), with vehicle collisions (40%) and illegal poaching (31%) the predominant sources of mortality from 2019 to 2020. Few instances of coyote–vehicle collision have previously been reported in Wisconsin ( Jackson 1961), although we identified vehicle collisions as the cause of death for 13.5% of study animals. While vehicle collision mortality rates do not appear to be related to urban association within either natal or adult home ranges ( Riley et al. 2003; Zepeda et al. 2021), highway mitigation structures, such as exclusion fencing and wildlife crossing, have been known to reduce mortalities ( Gilhooly et al. 2019).

In implementing our literature review, we acknowledge several limitations within the data and our interpretations. The search engines and keywords selected for our literature review protocol may not have generated an exhaustive list of all publications including estimates of coyote survival and mortality. Studies implementing telemetry monitoring often have small sample sizes due to high logistical costs, which may result in low power within the survival analysis and high uncertainty. We note that our literature review consists of several publications reporting estimates using data collected within similar spatial extents (e.g., Piñon Canyon Maneuver Site, Las Animas County, Colorado; Gese et al. 1989; Gese 2005; Seidler and Gese 2012) and/or data from overlapping study periods (e.g., 1995–1997 and 1995–2000; Kamler and Gipson 2000, 2004). We opted to use descriptive rather than quantitative analysis to describe patterns in the literature review due to confounding effects inherent in the spatiotemporal scale of the data. These confounding effects reflect shifts in harvest regulations, population dynamics, and range expansion and are coupled with study-specific random effects (e.g., monitoring method, estimation method, sample group) that make evaluating causal inferences problematic.

Coyote population dynamics are highly plastic, allowing for a variety of survival strategies across their geographic range. Under high levels of exploitation, coyote populations usually have a younger age structure, lower survival, increased number of yearlings reproducing, increased litter size, and relatively small packs ( Gese et al. 1989; Windberg 1995; Crabtree and Sheldon 1999). In contrast, unexploited coyote populations tend to have an older age structure, higher survival rates, lower reproductive rates, and large pack sizes. While we lack population estimates for coyotes in Wisconsin, the population is considered stable or increasing based on population indices from winter track counts, annual mammal survey counts, and harvest summaries ( Dhuey 2018; Kitchell 2019, 2020), suggesting that the observed low survival rates are likely coupled with increased reproduction and immigration. These demographic responses to human harvest align with the compensatory mortality and natality hypothesis which predicts that harvest mortality causes density-dependent responses (decreases in natural mortality, increases in reproduction) by reducing competition for resources ( Connell 1978; Swenson 1985; Boyce et al. 1999). Additionally, population losses from hunting may also be compensated by increased immigration from adjacent areas ( Robinson et al. 2008), even when compensatory reproduction is moderate ( Kilgo et al. 2017).

We found no relationship across the literature between survival rates and the proportion of human-induced mortality, further supporting the assertion that human-induced mortality is a compensatory component of population regulation in coyotes ( Sterling et al. 1983). Due to the demographic plasticity of coyotes, a net survival of about 33–38% has been suggested as a minimum to maintain population stability ( Knowlton 1972; Nellis and Keith 1976). Another study proposed needing only 10% of juveniles to survive and reproduce to maintain population of coyotes ( Knowlton et al. 1999). The Wisconsin population survival rates suggest that current levels of exploitation are likely sustainable ( Crabtree and Sheldon 1999). Coyote populations broadly seem resilient to liberal hunting and trapper regulations, that is, year-long season and unlimited tags. While harvest influences various demographic parameters, the compensatory mechanisms buffer those changes from having population-level effects.

The lack of relationship we found between harvest and populations is also supported by results of management studies that intentionally try to reduce populations of coyotes, that is, to minimize depredation risk to livestock ( Conner et al. 2008; Eklund et al. 2017). Lethal removal studies have seen minimal declines in population size and quick recovery of the population to pre-control levels due to density-dependent responses ( Conner et al. 1998; Knowlton et al. 1999). For example, Kilgo et al. (2014) monitored a population of coyotes in the southeastern United States before and during 3 years of lethal control and observed the population recovered each year via immigration and reproductive compensation. These findings agree with results of control simulation studies that show lethal removal may result in marginal and short-term population reduction when dominant, resident coyotes are targeted for removal at small spatial scales ( Conner et al. 2008). Given the infeasibility of reducing populations even using intentional and intensive lethal removals, we do not anticipate that changes to harvest regulations would lead to population declines in Wisconsin. Hunting season on coyotes in Wisconsin is already liberal with a year-long season and unlimited tags for those with a hunting license. The trapping season is limited to the fall and winter to align with trapping of other furbearer species, but given the reduced pelt prices for coyotes outside this season, we would not expect significant increases in trapping if the season were extended. Similarly, our literature review did not identify a pattern between a more liberal trapping season on coyotes, that is, extending to year-long, and reductions in population sizes.

Supplementary Data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.—Description of coyote hunting and trapping seasons in the continental United States and Canadian provinces and territories.

Supplementary Data SD2.—NIMBLE model R code for a Bayesian proportional hazard regression model for grouped (discrete) data to evaluate the time-to-event as a function of covariates and a Bayesian proportional hazard regression model for evaluating mortality-specific cumulative incidence functions.

Supplementary Data SD3.—Description of coyotes outfitted with GPS collars during a four-year study on survival in Wisconsin. Each animal was given a unique animal ID, sexed (male or female), and aged (adult or juvenile) at time of capture and outfitted with a GPS collar programmed to emit a mortality signal. Location data were collected at 3-h intervals and extended until either a mortality incidence, collar failure, or end of the monitoring period (4/1/2020).

Supplementary Data SD4.—Literature citations containing information on survival or cause-specific mortality in North American coyotes.

Supplementary Data SD5.—Forest plots describing literature estimating coyote survival and mortality in the United States and Canada from 1971 to 2021.

Acknowledgments

We thank D. M. MacFarland and S. Rossler for their constructive comments on earlier drafts of this manuscript. We thank the Associate Editor and reviewers for assistance in revising this manuscript. We thank S. Bundick, W. Ellarson, N. Forman, M. Hunsaker, D. Jarosinski, T. Johannes, T. Klein, K. Luukkonen, and H. Manninen for their work collaring coyotes and conducting mortality investigations. We are especially grateful to landowners for allowing access to their properties for coyote capture and to the volunteer trappers who captured coyotes for this study. Lastly, we thank the hunters and trappers who cooperated with us by reporting their harvest and providing the necessary harvest information. The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.

Funding

Funding was provided by the Federal Aid in Wildlife Restoration Program (Project W-160-P).

Conflict of Interest

The authors declare that they have no conflict of interest.