Research

INTRODUCTION

The coyote (Canis latrans) is a common example of an urban-adapted species that is also the largest carnivore that is common in cities across North America (Magle et al. 2019, Schell et al. 2020). Coyotes thrive in urban areas largely by avoiding interactions with humans (Mowry et al. 2020, Drake et al. 2021) while benefitting from reduced competition with other predators (Prugh et al. 2009), less human persecution in urban compared to rural areas (Collins and Kays 2011), and abundant urban food resources, such as rodents, garbage, compost, and fruit trees (Fedriani et al. 2001, Murray et al. 2015a, Sugden et al. 2021). Urban coyotes can potentially also improve human quality of life by limiting populations of feral cats that prey on native biodiversity (Crooks and Soulé 1999, Gehrt et al. 2013, Bonacic et al. 2019), supporting a sense of connection with nature (Cox and Gaston 2018), and providing aesthetic enjoyment that is often inherent in seeing wild animals (Soulsbury and White 2015). For these reasons, many people living in cities tolerate, and even appreciate, urban coyotes (Soulsbury and White 2015, Sponarski et al. 2018). However, over the past two decades, there have been increasing reports of bold and aggressive interactions between urban coyotes and people that are indicative of human-coyote conflict and reduce the tolerance of people for populations of urban coyotes (Baker and Timm 2017, Poessel et al. 2017, Draheim et al. 2019). A better understanding of the circumstances associated with conflict could inform approaches to coyote management and public education to support human-coyote coexistence in urban areas.

The level of conflict between humans and coyotes can be assessed in two broad ways: coyote behaviors, such as boldness or aggression, associated with conflict; and human perception of conflict, indicated by people’s concern about coyotes (Table 1). Coyote behaviors that contribute to conflict are often studied retroactively on the basis of factors associated with attacks, which represent the highest level of human-coyote conflict (White and Gehrt 2009, Baker and Timm 2017). Coyote attacks on pets are usually attributed to predation or the defense of territories or dens (Gehrt et al. 2013, Poessel et al. 2017, Nation and St. Clair 2019). Attacks on humans are rare but typically generate substantial media attention and degrade public tolerance (Carbyn 1989, Alexander and Quinn 2011, Draheim et al. 2019). These attacks are often preceded by elevated bold or aggressive behavior, which can be caused by increasing coyote habituation to people (Baker and Timm 2017) and associated food conditioning (Lukasik and Alexander 2011). Assessing the factors associated with bold coyote behavior, which refers to their tendency to approach or interact with people or pets, can help mitigate human-coyote conflict before it occurs at its highest level in the form of attacks.

Human perceptions of coyotes may be positive, neutral, or negative, with negative perceptions typically relating to concern of harm to themselves, children, or pets when they see or interact with a coyote. Negative perceptions of coyotes, which are associated with conflict, may not align with the actual risk of a coyote attack but still reduce public tolerance of coyotes and, consequently, public attitudes toward various forms of wildlife management and policy (Sponarski et al. 2018, Draheim et al. 2019). When attacks on people occur or attacks on pets are prevalent, the public demand for lethal management typically increases with rising negative perceptions of coyotes (Baker and Timm 2017, Draheim et al. 2019). These circumstances make the perception of risk by the public, as indicated by their level of concern about coyotes, an informative metric for coyote management and one for which correlates might also be studied to guide proactive, non-lethal management efforts to minimize potential conflict.

Reports of encounters between humans and coyotes that describe coyote behavior and human perceptions can be examined with spatial, temporal, and contextual variables via explanatory models. Spatial variables used to predict negative interactions between people and wildlife can help identify high-conflict areas, especially when measured at relevant scales to support management actions (Delsink et al. 2013, van Bommel et al. 2020). For example, knowing which areas have low, moderate, or high probabilities of interactions between humans and black bears (Ursus americanus) allows for management resources to be more efficiently allocated to improve coexistence (Merkle et al. 2011). Temporal variables may apply to scales that range from diel, through seasonal, to inter-annual, again with the goal of predicting when human-wildlife conflict is most likely to occur (Morehouse and Boyce 2017, Soulsbury 2020) so that it can be mitigated more effectively. Additional information for understanding conflict includes contextual variables, such as the presence of pets or children, off-leash dogs, and poor health of individual animals (Poessel et al. 2013, Murray et al. 2015b, Olson et al. 2015), which influence both coyote behavior and human perceptions.

The importance of these spatial, temporal and contextual variables for predicting coyote behavior and human perceptions of conflict can be advanced with large, long-term datasets collected by community members on public reporting interfaces. Such databases allow researchers to gather information over many years and large geographic areas, while simultaneously engaging and educating members of the public (Weckel et al. 2010, Frigerio et al. 2018). Previous studies of this sort have collected voluntary reports of coyote activity using public surveys (Weckel et al. 2010), city reporting databases (Lukasik and Alexander 2011, Poessel et al. 2013), websites (Wine et al. 2015, Mowry et al. 2020), and mobile phone apps (Mueller et al. 2019, Drake et al. 2021). Past analyses of these datasets have shown that reporting varies across levels of urban development, land cover types, and coyote seasons (e.g., breeding, pup rearing, dispersal; Weckel et al. 2010, Poessel et al. 2013), as well as with human demographic variables, such as household income and education (Wine et al. 2015, Mowry et al. 2020). Studies that have used public reports to assess different levels of conflict related to coyote behavior have shown that bold coyote behavior is most prominent during the coyote pup-rearing season (Lukasik and Alexander 2011, Drake et al. 2021). These studies also found more negative interactions in areas where coyotes consume more anthropogenic food (Lukasik and Alexander 2011) and when coyotes were encountered farther from roads (Drake et al. 2021). Although these studies have provided foundational information that can be applied in cities across North America to facilitate human-coyote coexistence, to date, community reports have not been used to examine changes in interactions over time. Therefore, extensive and long-term data can provide invaluable information for the additional quantification of the specific spatial, temporal, and contextual variables associated with conflict-indicative coyote behavior and human perceptions about coyotes.

In this study, we used over 9000 community science reports collected over 10 years from a website in Edmonton, Canada, to develop two indicators of human-coyote conflict (Table 1): coyote boldness toward people or pets, based on the reported coyote behavior, and human concern about coyotes, which we inferred from the descriptions of coyotes expressed by study participants. We then used spatial, temporal, and contextual information from the reports to construct exploratory models with the goal of identifying correlates of higher conflict scores for each of these indicators (coyote boldness and human concern). By identifying the predictors of bold behavior by coyotes and greater human concern about coyotes, we hope to support proactive coyote management and effective public education. In turn, these tools could facilitate sustainable coexistence of humans and coyotes to maximize the ecosystem services, including public appreciation, provided by urban coyote populations throughout North America.

METHODS

Study area

This study occurred in Edmonton, Alberta, Canada (latitude 53.54728, longitude -113.50068), which has an area of 684 km² and a population of 976,223 based on the 2019 City of Edmonton Municipal Census (https://www.edmonton.ca/city_government/facts_figures/municipal-census-results). A combination of large area and moderate population density makes Edmonton one of the most sprawling cities in North America, with large areas of undeveloped land (City of Edmonton 2018). Edmonton has warm summers (June to August daily average = 16.7 °C) and cold winters (December to March daily average = -9.7 °C; Environment and Climate Change Canada 2018). The city is bisected by the North Saskatchewan River valley and several large ravines, which form a network of minimally developed natural areas that provide abundant habitat for coyotes and other wildlife (Fig. 1).

Report collection

Beginning in September 2010, members of the public were able to voluntarily report coyote sightings or encounters through a web-based platform on the Edmonton Urban Coyote Project website (https://www.edmontonurbancoyotes.ca/reportsighting.php). We promoted the website opportunistically during media interviews, public lectures, and social media posts, as well as through word of mouth, on labels attached to wildlife cameras in the city, and via a link on the City of Edmonton website (https://www.edmonton.ca/residential_neighbourhoods/pets_wildlife/Coyotes.aspx) that was added in 2019.

When submitting a report, participants were asked to provide the location, nearest road intersection of the report location, date, and time of day. The website included a map interface to allow people to precisely locate their report by placing a pin on the map. Time of day was submitted by reporters using a drop-down menu with the option to select either hourly times between 5 AM and midnight or a general time window (dawn, morning, afternoon, evening, or night). Participants were also asked to specify whether their report was a “sighting,” defined as an observation of a coyote at a distance with no interaction, or “encounter,” defined as an interaction with a coyote at close range. Reporters were invited to provide free-form comments, as well as their name and contact information. The high frequency and extensiveness of these comments led to our post hoc decision to quantify and interpret the information in them. For the N = 3366 reports that did not include map coordinates, we determined them post hoc on the basis of the reported nearest street intersection and other information in the comments (e.g., if a specific park or building was named). To encourage participation, no registration or login was required.

Extraction of boldness, concern, and contextual variables from reports

Most reports (96.8%, N = 8859) included optional comments with further details about the human-coyote interaction, including information on coyote behavior, the participant’s perception of coyotes, and various contextual factors. We reviewed the database to remove duplicate entries and spam reports, removed the names and contact information of reporters, and then recruited a team of 30 volunteers and undergraduate students who read and classified the comments in each report, following an online form and standardized protocol contained in Appendix 1. Despite clear instructions and training materials for volunteers, there was potential for variation in the interpretation of reporter comments. Thus, for every 100 reports that each volunteer or undergraduate classified, one author (JJF) randomly assessed 10 reports to ensure that the protocol was appropriately followed. To further assess the repeatability between report classifiers, JJF randomly selected and blindly re-classified 100 reports drawn randomly from the entire database. We assessed inter-observer agreement by calculating the percentage of reports that generated the same classifications for each of the seven variables extracted with the classification form.

For reports with comments that described coyote behavior, volunteer report classifiers assessed the degree of boldness on a scale from one (ran away) to nine (made physical contact with pets or people). We later simplified these categories into a four-point ordinal scale of coyote boldness, using avoidance, indifferent, bold, and aggressive as characterizations of coyote behavior (Table 2). We classified human concern about coyotes from reports where participants expressed their perceptions about coyotes to form a three-point scale from the explicit presence of words conveying positive (e.g., beautiful), neutral (e.g., curious or not scared), or negative (e.g., scared) emotions or attitudes (Table 3). We also classified report comments for the presence of five contextual variables: (1) the human activity occurring at the time of the report (e.g., walking, cycling, driving); (2) the presence or mention of vulnerable individuals (children, dogs, or cats); (3) if dogs present were leashed or off leash; (4) the number of coyotes observed; and (5) any mention of the reporter’s interpretation of coyote health status (e.g., mangy).

Spatial and temporal variable collection

To quantify the geospatial setting of each report, we imported report locations into ArcGIS Pro v2.7 (Fig. 1). We excluded reports that were located outside of Edmonton city limits or in recently annexed but undeveloped rural land, and we identified our study area by generating a minimum convex polygon around the remaining report locations. Land cover types within our study area were classified by using geospatial data from the City of Edmonton Urban Planning Land and Vegetation Inventory (uPLVI) database, a high-resolution database that uses remotely sensed imagery and Softcopy photogrammetry to identify land cover types for urban land use decisions (City of Edmonton 2018). For our study, we binned uPLVI land cover classifications into six land cover types representing various degrees of human development and coyote habitat quality (Table 4), which were comparable to those used in similar studies in the American cities of New York (Weckel et al. 2010), Denver (Poessel et al. 2013), and Atlanta (Mowry et al. 2020).

Because coyote boldness and human concern may be affected by a combination of site-specific conditions (van Bommel et al. 2020) and broader landscape characteristics (Murray et al. 2015b, Wine et al. 2015), we measured land cover at five different scales (within 100, 200, 400, 800, and 1600 m radii of each report). Land cover was calculated as the proportional area of each land cover type within the circular area defined by each scale. Proportional land cover measurements were then centered log-ratio transformed; this transformation minimizes the multicollinearity among different land cover proportions, which can otherwise be negatively correlated with each other because they are constrained to a constant sum (i.e., 100% cover; Aitchison 1982). To compare the distribution of reports across different land cover types, we also assigned a single land cover category to each report on the basis of the category with the greatest proportional area within a 100-m radius of the report.

Building density and road distance have previously been associated with human-coyote encounters (Wine et al. 2015, Drake et al. 2021); therefore, we determined building density based on the proportional area of building footprints within each of the five scales around each report (Statistics Canada 2019). We also measured the distance from each report to the nearest road from the single line street network geospatial database from the City of Edmonton. For road distance, we applied an exponential distance decay function (e^-0.002d, where d = meters to the nearest road) to confine values between zero (far from road) and one (on road; Nielsen et al. 2009). All spatial variables were measured in raster format with a 10 x 10-meter cell size.

To support temporal analyses, we measured changes in reporting, coyote boldness, and human concern across years, months, and time of day, as well as across the coyote seasons of breeding (1 January to 30 April), pup rearing (1 May to 31 August), or dispersal (1 September to 31 December). We manually categorized time of day into either day (after sunrise and before sunset) or night (before sunrise or after sunset). Sunrise and sunset times were specific to Edmonton and were adjusted for seasonal variation.

Statistical methods

We began by categorizing spatial, temporal, and contextual patterns in report submissions. For land cover types, we estimated the expected number of reports on the basis of the total proportion of that land cover type within the study area. We then applied Pearson’s chi square test to determine whether reports occurred more or less frequently than expected in each land cover type. To assess how reporting varied over time, we summarized the number of reports in each of the biological coyote seasons for each year from 2012 to 2021, the percentage of reports during each month, and the number of reports from day and nighttime. For each contextual variable we determined the number of reports assigned to categories within each variable.

To identify the best spatial and temporal predictors of coyote boldness and human concern, we used an exploratory modelling framework (Tredennick et al. 2021) based on ordered logistic regression with the clm function in the R package ordinal (Christensen 2019). Time of day and contextual variables were strongly correlated with each other (Table A1.1, Appendix 3), so we excluded these variables from our models and examined them separately (see below). We used a pseudo-optimized multiple scale approach (McGarigal et al. 2016) to select which of the five measurement scales for each land cover variable and building density fit best for a multivariate model. This approach involved conducting univariate models for each variable and then retaining the scale with the lowest Akaike’s information criterion value (AIC; Burnham and Anderson 2004; Table A3.2 in Appendix 3), which we did separately for coyote boldness and human concern. If a variable’s best-fit scale did not improve on the AIC of the null model, we excluded that variable from further analyses. We then assessed correlations between the remaining variables using Spearman’s rank correlation coefficient (Tables A3.3 and A3.4 in Appendix 3), and for any pairs of variables where r > 0.6, we removed the variable that produced a higher AIC value in univariate models. All numerical variables were mean centered and scaled with a standard deviation of 1.

For each of coyote boldness and human concern, we constructed global models (Table A3.5 in Appendix 3) that included each of the non-correlated spatial variables, year, and coyote biological season (using breeding season as the reference) as additive effects. We included interaction terms for year and each of the spatial variables to test if temporal changes in the response variables were associated with specific spatial factors in the urban environment. In models of coyote boldness, we also included interaction terms for biological season and each of natural and modified open areas to test for seasonal changes in coyote behavior that might be associated with denning in these less-developed areas (Dodge and Kashian 2013). Because we were primarily interested in maximizing the explanatory power of our models to explain each of boldness and human concern (Tredennick et al. 2021), we used AIC-based model selection with the dredge function from the package MuMIn (Barton 2022) to identify the variables and interactions that were retained in the top models (ΔAIC < 2). To determine whether reports of coyote boldness or human concern had changed over the course of our 10-year dataset, we explored these relationships in more detail using linear regressions predicting the percentage of reports within each of the ordinal scores as a function of year.

We explored the effects of time of day and contextual factors on each of coyote boldness and human concern using Pearson’s chi square tests of independence (Weckel et al. 2010), followed by standardized residual post hoc tests (chisq.posthoc.test.package; Ebbert 2019) to determine which levels of each factor were most strongly associated with boldness or concern. For each of these analyses, we used as a reference category the reports for which the contextual variable could not be determined from reporter comments. We also used chi square tests to determine whether reports that identified bolder coyote behavior also expressed more human concern, and adjusted alpha values for each residual test with Holm’s correction for multiple comparisons (MacDonald and Gardner 2000). We conducted all statistical analyses in R version 4.1.3 (R Core Team 2022) and considered effects to be significant if 95% confidence intervals did not overlap zero or if p values < 0.05.

RESULTS

Reporting patterns

From 2 September 2010 to 31 December 2021, 11,239 reports were submitted on the Edmonton Urban Coyote project website. Of these, we removed 1722 spam or duplicate reports, 256 reports that were outside of Edmonton city limits, and 127 reports from 2010 and 2011 because of limited reporting in these years. The resulting dataset included 9134 unique and spatially explicit coyote reports between 1 January 2012 and 31 December 2021. Of the 100 reports that were re-classified to assess classification repeatability, inter-rater agreement for each variable ranged from 85% to 96% (Table A1.1 in Appendix 1), suggesting that our training produced high consistency, but not perfect repeatability, of interpretations among reporters.

Reports were widely distributed across the city and unevenly spread across land cover types (χ²₄ = 1564, p < 0.001; Fig. 1). Based on the proportion of each land cover type within our study area, we received more reports than expected in residential (59.1%), mowed grass (12.2%), and natural land cover (10.9%) areas and fewer than expected in commercial (11.1%) and modified open areas (6.7%). Reporting increased over years, and, within years, was consistently higher in the breeding and dispersal seasons (Fig. 2A-B). Reports were also 1.4 times more common during the day than at night (Fig. 2C). Human activity was discernible in 48.1% of reports and mostly involved walking (19.1%), being in a home or yard (18.4%), or driving (9.1%; Fig. 3). Vulnerable individuals (mostly dogs, 20.9%) were present or mentioned in 30.8% of reports, and a subset of those reports identified dogs as leashed (11.7%) or off-leash (9.4%). Most reports involved a single coyote (59.4%), and a subset of reports identified coyote health as healthy (13.9%) or unhealthy (6.0%).

Coyote boldness, as determined from the reported coyote behaviors in 3540 reports, was most commonly reported as avoidant or indifferent, followed by bold and aggressive (Table 2). In contrast, human concern, based on the perceptions that people expressed about coyotes in 1060 reports, indicated that negativity toward coyotes was much more common than neutral (3.6 times) or positive responses (4.9 times; Table 3). Reports that mentioned physical contact between people or pets and coyotes consisted mostly of dog attacks (N = 85), followed by cat depredations (N = 50); in only one report did a coyote contact a human, which occurred while the coyote attempted to take a sled from a child. For the 653 reports for which both boldness and human concern could be classified, the two variables were significantly related (χ²₆ = 56.3, p < 0.001), with reports of bold or aggressive behavior being more likely to express negative perceptions of coyotes (Fig. 4).

Spatiotemporal predictors of coyote boldness and human concern

Ordered logistic regression analysis revealed a suite of spatial and temporal variables that explained each of coyote boldness and human concern (Fig. 5). We present results from only the top models for these response variables (Fig. 5) because there was little variation among coefficient and confidence estimates within the full set of top-ranked models (ΔAIC_c < 2; Tables A3.6 and A3.7 in Appendix 3). The top model predicting coyote boldness indicated that the likelihood of bolder behavior was higher during the pup-rearing season and in areas with higher proportions of mowed land cover (within a 100 m radius), but lower closer to roads and in areas with greater building density (within 200 m). The significant interaction term in this model indicated that boldness was higher during the pup-rearing season, especially in areas with more modified open land cover (400 m buffer size; Fig. A3.1 in Appendix 3). The top model for human concern indicated that a higher likelihood of human concern was associated with increases in the proportion of residential area (within 800 m) and modified open land cover (within 1600 m), as well as during the dispersal season (Fig. 5).

Models for both coyote boldness and human concern indicated a significant increase in the likelihood of human-coyote conflict over years (Fig. 5). None of the interaction terms with year were significant when we considered only the single best model for each response variable (Table A3.6, Table A3.7 in Appendix 3); however, for models predicting human concern, the negative interaction term between residential area and year was retained in 19 of 20 top models and was significant in 13 of these. This interaction indicated that concern was generally higher in residential areas in early years, but increasing levels of concern in non-residential areas reduced the magnitude of the effect of residential area over time (Fig. A3.2 in Appendix 3). Although several other variables and interaction terms appeared in some of the top models (ΔAIC_c < 2) for boldness and concern, their effects were not significant in any of these (Tables A3.6 and A3.7 in Appendix 3).

We examined temporal changes in greater detail by evaluating the percentage of reports within each of the ordinal scores for each year (Fig. 6). Specifically, the percentage of reports describing bold behavior increased significantly (β = 2.19, p < 0.001), whereas avoidance behavior decreased (β = -1.82, p < 0.001), although there were no differences in the percentage of reports describing indifferent (β = -0.71, p = 0.21) and aggressive behavior (β = 0.24, p = 0.16). Similarly, negative perceptions about coyotes became more common over years (β = 1.07, p = 0.072) and positive perceptions became less common (β = -1.07, p = 0.005), with no change in the proportion of neutral perceptions (β = 0.002, p = 0.997).

Analysis of diel patterns in coyote boldness showed that indifferent behavior was significantly more common during the day and avoidance behavior was significantly more common at night (χ²₂ = 30.1, p < 0.001; Fig. A2.1 in Appendix 2). However, human concern did not differ between day and night (χ²₂ = 1.09, p = 0.58).

Contextual influences on boldness and concern

All five contextual variables were significantly related to coyote boldness (Fig. 7, Table A2.3 in Appendix 2). Bold behavior was described more frequently when reporters were walking, when dogs were mentioned, and when two or three coyotes were present. Aggressive coyote behavior was reported more frequently than expected when cats or dogs were mentioned, when dogs were off-leash, and when two or more than three coyotes were observed. The least threatening coyote behaviors, avoidance and indifference, occurred mostly in reports when people were driving, cycling, or in their home or yard; when only one coyote was observed; and when coyotes were perceived as healthy.

Most contextual variables were also related to human concern, demonstrating that they were important factors affecting the perceived risk presented by coyotes (Fig. 7, Table A2.3 in Appendix 2). Concern was more frequently reported when dogs, children, or multiple vulnerable individuals were mentioned; conversely, reports that did not mention any vulnerable individuals expressed less concern. Perceptions were more likely to be positive when only one coyote was observed and when the coyote(s) were described as healthy.

DISCUSSION

Human-coyote conflict is increasing in urban areas throughout North America (White and Gehrt 2009, Baker and Timm 2017), creating a need to better understand the circumstances surrounding conflict to inform proactive management strategies. We used data from a 10-year database of citizen reports to explore which spatial, temporal, and contextual predictors best explained the degree of boldness in descriptions of coyote behaviors and concern in perceptions of coyotes expressed by reporters. We found that descriptions of coyote boldness increased in areas with more mowed grass, with lower building density, and as distance to roads increased; over time and in the pup-rearing season; and when reporters were walking, mentioned a cat or dog, and when more coyotes were present. Human concern was greater in areas with higher proportions of residential and modified open land cover; over time and during the dispersal season; and when reporters mentioned vulnerable individuals. Both boldness and human concern increased significantly from 2012 to 2021.

Spatiotemporal patterns in coyote boldness

We found that coyote boldness was higher in less-developed areas that were mowed or otherwise not naturally vegetated, similar to what was described for open relative to natural areas in Denver, Colorado (Poessel et al. 2013) and in managed clearings in North Carolina (Wine et al. 2015). Collectively, these results suggest that conflict-indicative coyote behaviors are most prevalent in spaces that are at the interface of natural and developed urban areas. Similar interfaces between peri-urban and rural or rural and wildland areas frequently create human-wildlife conflicts in many other species (König et al. 2020). The pattern we observed may arise because coyotes in open areas are visible at greater distances and may thus appear to be bolder; alternatively, bolder animals may be more likely to occupy areas with less vegetation cover, as has been reported for brown bears (Ursus arctos; Bombieri et al. 2021). This behavior might be expected of coyotes owing to their evolution in the arid southwest of the North American continent (Hody and Kays 2018). Our spatial variables were most explanatory of boldness when measured at smaller spatial scales (≤ 400 m radii from reports; Table A3.2 in Appendix 3), suggesting that boldness is driven by site-specific factors, such as proximity to vegetation cover, territorial boundaries, or den sites. A further influence on this pattern could be a non-random distribution of human observations, such as would occur if dog-walkers targeted similar habitat types.

Seasonally, boldness was significantly more likely during the summer pup-rearing season, which is also consistent with other studies (White and Gehrt 2009, Lukasik and Alexander 2011). The fact that fewer reports were submitted during this period, despite a time of greater outdoor activity by people in our study area, suggests that coyotes avoid humans and pets during pup rearing but behave more aggressively when interactions occur. Aggression by coyotes during the pup-rearing season presumably reflects defense of pups from perceived threats posed by humans or dogs (Bombieri et al. 2018). Many reports described coyotes rushing out of cover to bite large dogs on their hamstrings, suggestive of defensive behavior. There were also more reports of cat depredations during the pup-rearing season (29 of 50 total), probably caused by a combination of coyotes seeking food for their pups and generally greater numbers of free-roaming cats in the summer that might, in turn, be hunting naïve young birds and rodents (Nation and St. Clair 2019).

Boldness during the pup-rearing season was particularly associated with modified open areas, which may have been associated with denning by coyotes. Dens in modified open areas, such as utility corridors, meadows within the city’s ravine system, or agricultural land near and within the city, have less vegetative cover than dens in natural areas, which might reduce opportunities for avoidant behavior and increase defensive aggression. Indeed, reports from natural areas were more likely to describe avoidance (28.9%) than aggression (12.9%) relative to reports from modified open areas (20.3% avoidance and 20.7% aggression; Table A2.1 in Appendix 2). An alternative hypothesis is that coyotes denning in more disturbed modified open areas may be more prone to boldness because of repeated exposure and thus habituation to humans and their pets in these areas (Young et al. 2019). In either case, our study shows that coyote behavior in modified open areas during the pup-rearing season may often present a risk to the safety of humans and their pets.

We found that coyote boldness increased over the 10-year reporting period, potentially explaining the mechanism for rising coyote attacks on people in North America (White and Gehrt 2009, Baker and Timm 2017). Both patterns may reflect the greater boldness of urban relative to rural coyotes that others have reported and attributed to reduced persecution by people, repeated benign interactions with humans, and access to anthropogenic food (Breck et al. 2019, Young et al. 2019, Brooks et al. 2020). There is evidence that coyote boldness toward humans is passed from parents to offspring (Schell et al. 2018) and increases with greater exposure to people (Young et al. 2019), both of which could increase coyote boldness over time to accelerate boldness-driven conflict. Separate from these interactions with people, higher coyote population density within cities may lead to intraspecific competition that favors bolder individuals (Bateman and Fleming 2012). Our top ordinal regression models did not include any interactions between spatial variables and year, suggesting that these changes in boldness were relatively consistent across the urban environment. Despite increases in boldness over time, we did not find a similar increase in aggressive behavior, possibly because the most aggressive individuals were targeted for removal by city managers, which is an effective means of reducing conflict in the short term (Baker 2007, Farrar 2007, Breck et al. 2017).

Spatiotemporal patterns in human concern

We classified human perceptions of coyotes described in reports as positive, neutral, and negative to create a metric of human concern that increased with the amount of residential area within 800 m of the report. This observation is similar to previous findings that people are less tolerant of coyotes near their homes despite being generally tolerant of coyotes in cities (Bonnell and Breck 2017, Drake et al. 2020). A similar pattern occurred for cougars (Puma concolor) at a rural-wildland interface in Alberta (Knopff et al. 2016). Such effects show how human concern may not align with the actual risk of a coyote behaving boldly or aggressively; concern was higher in residential areas, but boldness was negatively associated with building density and road proximity, which were correlated with residential area in our study (Table A3.3 in Appendix 3).

Human concern about coyotes was also higher in areas with more modified open land cover, where bold interactions are more likely during the pup-rearing season. This land cover may also be disproportionally responsible for the positive correlation we found between reports that described bolder coyotes and greater human concern. Such a correlation might be amplified if people gain awareness over time of the bold or aggressive coyote interactions that are more common in those areas, which has previously been shown to increase the risk people perceive from coyotes (Sponarski et al. 2018, Draheim et al. 2019). Again, dog walkers may have also been more likely to target these areas and dogs were often associated with reports of coyote aggression (below). People may respond most to the visibility of coyotes, which is likely highest during the fall dispersal season when our database received more reports and higher levels of concern. Interestingly, human concern correlated with land cover predictors measured at larger scales (≥ 800 m radii; Table A3.2 in Appendix 3) than did the measures of boldness derived from descriptions of coyote behavior. This difference may reveal that coyote behavior and human perceptions are affected by spatial variables at different scales, or, simply, that coyote behavior is more readily described for animals that are nearby.

Like boldness, human concern about coyotes increased over the 10-year reporting period (Figs. 5 and 6), and our study demonstrates that these perceptual changes may be quite nuanced across the urban landscape. For example, concern was initially higher in areas with more residential land cover, but this effect was reduced over time because of growing human concern across most other land cover types. Whereas some studies have predicted that humans will habituate to the presence of urban coyotes over time (Lawrence and Krausman 2011, Jackman and Rutberg 2015), as has been reported in Alaskan brown bears (Smith et al. 2005), our findings support suggestions by others that this does not occur with carnivores if people fear for their safety (Williams et al. 2002, Kaltenborn et al. 2006). Expecting that human concern will naturally decrease as coyotes become more prevalent in cities may not be an effective strategy to facilitate coexistence, because public concern about any carnivore may typically increase with increasing carnivore prevalence, with negative interactions, or if conflict is emphasized by the media (Lute and Carter 2020).

Contextual factors affecting coyote boldness and human concern

Coyote boldness and human concern were each strongly explained by several contextual variables. Reports describing the presence or mention of vulnerable individuals exhibited higher scores for both coyote boldness and human concern. The mention of dogs or cats was associated with descriptions of bolder and more aggressive coyote behavior, supporting the findings of others that human-coyote interactions involving pets are more likely to cause conflict (Poessel et al. 2013, Baker and Timm 2017). Coyotes were described as “aggressive” less often when dogs were leashed (14.6%) compared to when they were described as off leash (32.5%; Table A2.1 in Appendix 2), but bold behavior was more common when dogs were leashed (39.2%) than off leash (22.3%). This result suggests that leashed dogs may still engender coyote behavior that is associated with conflict even if they are less likely to be attacked than when they are off leash. Many of these interactions during the pup-rearing season may have involved escorting behavior, wherein coyotes do not attack, but follow people with leashed dogs out of areas near their den sites. We did not find a significant relationship between the presence or mention of children and coyote boldness or aggression. Nonetheless, human concern was significantly higher when reports mentioned dogs, children, or multiple vulnerable individuals, perhaps because occasional coyote attacks on children are well publicized and can lead to serious injuries (Carbyn 1989, White and Gehrt 2009, Alexander and Quinn 2011, Baker and Timm 2017). A higher perceived risk of coyote attacks on pets and children likely reduces tolerance for coyotes in cities (Draheim et al. 2019).

With respect to other contextual variables, coyotes were described as being bolder when they were observed with other coyotes and when people were walking, whereas expressions of human concern were lower when people were driving or when only a single coyote was mentioned. Coyotes in poor health may be more conflict-prone (Murray at el. 2015b) and people were more likely to perceive healthy coyotes positively. However, we found no evidence that people found unhealthy animals to be bolder or feared them more. In fact, coyotes perceived to be healthy were more likely to be described as avoidant or indifferent. It could be that reporters were less likely to notice and characterize a coyote’s health in encounters where coyotes were behaving boldly or aggressively.

Limitations

Our study had several limitations. First, reports were collected non-randomly and non-independently, which introduces several potential biases inherent to most community reporting databases (Poessel et al. 2013, Sullivan et al. 2014). These biases include greater tendencies for repeat reporting by some residents with particularly strong views about coyotes, uneven advertising of the reporting database across neighborhoods or over years, potentially higher reporting from affluent neighborhoods with higher education levels (Wine et al. 2015, Mowry et al. 2020), and varying visibility of coyotes across seasons, time of day, or land cover types because of differences in vegetative cover, human activity, and daylight (Quinn 1995, Poessel et al. 2013). We attempted to mitigate these effects by focusing on measures of coyote boldness and human concern, rather than spatiotemporal influences on the number or distribution of reports, but interactions with contextual variables and habitat types are likely. We attempted to overcome spatial and temporal autocorrelation in the reports that could contribute to Type 1 statistical errors by restricting analyses to those with large sample sizes and verifying modelling results with chi-square tests (Table A2.3 in Appendix 2). Despite these precautions, our post hoc method of quantifying coyote boldness and human concern from a community reporting database cannot be compared to empirical behavioral observations of animals (e.g., Breck et al. 2019) or randomized public surveys (e.g., Drake et al. 2020). Urban coyote management in Edmonton may have also affected our findings. A suite of management tools has been applied by civic personnel over the course of our study. These included public education via websites and media, warnings about attractants at specific residences, signage about aggressive coyotes in particular areas, high-intensity hazing by wildlife professionals at sites with recurrent complaints (since 2018), and targeted removal of aggressive individual coyotes. In 2021, a community-based hazing program was initiated in some of Edmonton’s neighborhoods (Lajeunesse 2023). These actions undoubtedly affected our results; for example, recurrent actions to reduce coyote occupancy of, or boldness in, residential areas likely contributed to the lower boldness that we identified in these areas. Despite various actions by wildlife managers, both boldness and human concern increased over time in the public reports we examined.

Management implications

Despite some limitations, our findings do support the increased implementation of several management actions in our study area and elsewhere by identifying areas, seasons, and contexts that are associated with higher rates of human-coyote conflict. The rise in both coyote boldness and human concern we reported from 2012 to 2021 may continue in cities across North America without more effective and extensive implementation of strategies to reduce negative human-coyote interactions (Kaltenborn et al. 2006, Lute and Carter 2020), facilitate positive wildlife experiences (Kretser et al. 2009), and increase public knowledge about risk reduction (Riley and Decker 2000). Spatially, coyote boldness was generally higher in less developed and open areas (Poessel et al. 2013, Wine et al. 2015) where managers might emphasize public education and protection of pets via leashing and fencing (Draheim et al. 2019). Managers could also target these areas for initiatives to train community members to haze coyotes that behave boldly (Bonnell and Breck 2017, Breck et al. 2017, Lajeunesse 2023). Both awareness campaigns and community-based hazing programs might target conflict-prone areas in and near residential neighborhoods, where human concern of coyotes is highest, and especially during the breeding season to prevent denning in these areas and proactively limit negative interactions (Bonnell and Breck 2017). Coyote attractants of food and shelter should be more actively identified, secured, and removed in residential areas (Murray and St. Clair 2017). When these steps fail and aggressive coyotes pose ongoing threats to the security of leashed pets and people, targeted removal may be both necessary and effective at reducing conflict (Baker 2007, Farrar 2007, Breck et al. 2017). Although lethal removal is controversial (Breck et al. 2017, Draheim et al. 2019), its use can prevent attacks on pets or people that generate substantial media attention and erode public tolerance of coyotes (Alexander and Quinn 2011, Alexander and Quinn 2012, Draheim et al. 2019). Finally, proactive management of coyote-human conflict is especially important where it potentially endangers children, such as in parks, schoolyards, and playgrounds, which may require greater attention to the surrounding areas (e.g., residential neighborhoods).

Proactive management is made more possible with rapid information about the locations and contexts of bold coyote behavior, such as can be provided by public reporting databases. Our study demonstrated the utility of a community reporting interface for understanding human-coyote interactions, and we suggest that similar efforts should be integrated within urban wildlife management programs across North America. Increased specificity in reporting interfaces on apps or websites that are built with collaboration among managers, social scientists, and wildlife ecologists could increase both quantity and quality of the information collected from reporters (e.g., likert scales for the degree of reporter concern or degree of coyote aggression). In turn, reported information could increase the ability of managers to target education, policy, and enforcement to facilitate human-coyote coexistence in cities across North America.

LITERATURE CITED

Aitchison, J. 1982. The statistical analysis of compositional data. Journal of the Royal Statistical Society: Series B 44(2):139-177. https://doi.org/10.1111/j.2517-6161.1982.tb01195.x

Alexander, S. M., and M. S. Quinn. 2011. Coyote (Canis latrans) interactions with humans and pets reported in the Canadian print media (1995-2010). Human Dimensions of Wildlife 16:(5):345-359. https://doi.org/10.1080/10871209.2011.599050

Alexander, S. M., and M. S. Quinn. 2012. Portrayal of interactions between humans and coyotes (Canis latrans): content analysis of Canadian print media (1998-2010). Cities and the Environment 4(1):9. https://doi.org/10.15365/cate.4172012

Baker, R. O. 2007. A review of successful urban coyote management programs implemented to prevent or reduce attacks on humans and pets in southern California. Pages 382-392 in Proceedings of the 12th Wildlife Damage Management Conference.

Baker, R. O., and R. M. Timm. 2017. Coyote attacks on humans, 1970-2015: implications for reducing the risks. Human-Wildlife Interactions 11(2):3. https://doi.org/10.26077/jy37-s271

Barton, K. 2022. Package MuMIn: multi-model inference. Version 1.46.0. https://cran.hafro.is/web/packages/MuMIn/MuMIn.pdf

Bateman, P. W., and P. A. Fleming. 2012. Big city life: carnivores in urban environments. Journal of Zoology 287(1):1-23. https://doi.org/10.1111/j.1469-7998.2011.00887.x

Bombieri, G., M. d. M. Delgado, L. F. Russo, P. J. Garrote, J. V. Lopez-Bao, J. M. Fedriani, and V. Penteriani. 2018. Patterns of wild carnivore attacks on humans in urban areas. Scientific Reports 8:17728. https://doi.org/10.1038/s41598-018-36034-7

Bombieri, G., V. Penteriani, M. d. M. Delgado, C. Groff, L. Pedrotti, and K. Jerina. 2021. Towards understanding bold behaviour of large carnivores: the case of brown bears in human‐modified landscapes. Animal Conservation 24(5):783-797. https://doi.org/10.1111/acv.12680

Bonacic, C., R. Almuna, and J. T. Ibarra. 2019. Biodiversity conservation requires management of feral domestic animals. Trends in Ecology & Evolution 34(8):683-686. https://doi.org/10.1016/j.tree.2019.05.002

Bonnell, M. A., and S. W. Breck. 2017. Using resident-based hazing programs to reduce human-coyote conflicts in urban environments. Human-Wildlife Interactions 11(2):5. https://doi.org/10.26077/ab7k-6j25

Breck, S. W., S. A. Poessel, and M. A. Bonnell. 2017. Evaluating lethal and nonlethal management options for urban coyotes. Human-Wildlife Interactions 11(2):4. https://doi.org/10.1038%2Fs41598-019-38543-5

Breck, S. W., S. A. Poessel, P. Mahoney, and J. K. Young. 2019. The intrepid urban coyote: a comparison of bold and exploratory behavior in coyotes from urban and rural environments. Scientific Reports 9:2104. https://doi.org/10.1038/s41598-019-38543-5

Brooks, J., R. Kays, and B. Hare. 2020. Coyotes living near cities are bolder: implications for dog evolution and human-wildlife conflict. Behaviour 157(3-4):289-313. https://doi.org/10.1163/1568539X-bja10002

Burnham, K. P., and D. R. Anderson. 2004. Multimodel inference: understanding AIC and BIC in model selection. Sociological Methods & Research 33(2):261-304. https://doi.org/10.1177/0049124104268644

Carbyn, L. N. 1989. Coyote attacks on children in western North America. Wildlife Society Bulletin 17:444-446.

Christensen, R. H. B. 2019. Ordinal: regression models for ordinal data. http://cran.uni-muenster.de/web/packages/ordinal/ordinal.pdf

City of Edmonton. 2018. Urban primary land and vegetation inventory (uPLVI). Spatial data, draft 2018 edition. Edmonton, Alberta, Canada.

Collins, C., and R. Kays. 2011. Causes of mortality in North American populations of large and medium-sized mammals. Animal Conservation 14(5):474-483. https://doi.org/10.1111/j.1469-1795.2011.00458.x

Cox, D. T. C., and K. J. Gaston. 2018. Human-nature interactions and the consequences and drivers of provisioning wildlife. Philosophical Transactions of the Royal Society B 373:20170092. https://doi.org/10.1098/rstb.2017.0092

Crooks, K. R., and M. E. Soulé. 1999. Mesopredator release and avifaunal extinctions in a fragmented system. Nature 400:563-566. https://doi.org/10.1038/23028

Delsink, A., A. T. Vanak, S. Ferreira, and R. Slotow. 2013. Biologically relevant scales in large mammal management policies. Biological Conservation 167:116-126. https://doi.org/10.1016/j.biocon.2013.07.035

Dodge, W. B., and D. M. Kashian. 2013. Recent distribution of coyotes across an urban landscape in southeastern Michigan. Journal of Fish and Wildlife Management 4(2):377-385. https://doi.org/10.3996/062013-JFWM-040

Draheim, M. M., E. C. M. Parsons, S. A. Crate, and L. L. Rockwood. 2019. Public perspectives on the management of urban coyotes. Journal of Urban Ecology 5(1):juz003. https://doi.org/10.1093/jue/juz003

Drake, D., S. Dubay, and M. L. Allen. 2021. Evaluating human-coyote encounters in an urban landscape using citizen science. Journal of Urban Ecology 7(1):juaa032. https://doi.org/10.1093/jue/juaa032

Drake, M. D., M. N. Peterson, E. H. Griffith, C. Olfenbuttel, C. S. DePerno, and C. E. Moorman. 2020. How urban identity, affect, and knowledge predict perceptions about coyotes and their management. Anthrozoös 33(1):5-19. https://doi.org/10.1080/08927936.2020.1694302

Ebbert, D. 2019. Package chisq.posthoc.test: a post hoc analysis for pearson’s chi-squared test for count data. Version 0.1.2.

Environment and Climate Change Canada. 2018. Monthly climate summary 1981-2010 Edmonton city center weather station. Government of Canada, Ottawa, Ontario, Canada.

Farrar, R. O. 2007. Assessing the impact of urban coyote on people and pets in Austin, Travis County, Texas. Pages 334-341 in Proceedings of the 12th Wildlife Damage Management Conference.

Fedriani, J. M., T. K. Fuller, and R. M. Sauvajot. 2001. Does availability of anthropogenic food enhance densities of omnivorous mammals? An example with coyotes in southern California. Ecography 24(3):325-331. https://doi.org/10.1034/j.1600-0587.2001.240310.x

Fox, J., and J. Hong. 2009. Effect displays in R for multinomial and proportional-odds logit models: extensions to the Effects package. Journal of Statistical Software 32(1):1-24. https://doi.org/10.18637/jss.v032.i01

Frigerio, D., P. Pipek, S. Kimmig, S. Winter, J. Melzheimer, L. Diblíková, B. Wachter, and A. Richter. 2018. Citizen science and wildlife biology: synergies and challenges. Ethology 124(6):365-377. https://doi.org/10.1111/eth.12746

Gehrt, S. D., E. C. Wilson, J. L. Brown, and C. Anchor. 2013. Population ecology of free-roaming cats and interference competition by coyotes in urban parks. PLoS ONE 8(9):e75718. https://doi.org/10.1371/journal.pone.0075718

Hody, J. W., and R. Kays. 2018. Mapping the expansion of coyotes (Canis latrans) across North and Central America. ZooKeys 759:81-97. https://doi.org/10.3897/zookeys.759.15149

Jackman, J. L., and A. T. Rutberg. 2015. Shifts in attitudes toward coyotes on the urbanized east coast: the Cape Cod experience, 2005-2012. Human Dimensions of Wildlife 20(4):333-348. https://doi.org/10.1080/10871209.2015.1027973

Kaltenborn, B. P., T. Bjerke, and J. Nyahongo. 2006. Living with problem animals - self-reported fear of potentially dangerous species in the Serengeti Region, Tanzania. Human Dimensions of Wildlife 11(6):397-409. https://doi.org/10.1080/10871200600984323

Knopff, A. A., K. H. Knopff, and C. C. St. Clair. 2016. Tolerance for cougars diminished by high perception of risk. Ecology and Society 21(4):33. https://doi.org/10.5751/ES-08933-210433

König, H. J., C. Kiffner, S. Kramer‐Schadt, C. Fürst, O. Keuling, and A. T. Ford. 2020. Human-wildlife coexistence in a changing world. Conservation Biology 34(4):786-794. https://doi.org/10.1111/cobi.13513

Kretser, H. E., P. D. Curtis, J. D. Francis, R. J. Pendall, and B. A. Knuth. 2009. Factors affecting perceptions of human-wildlife interactions in residential areas of northern New York and implications for conservation. Human Dimensions of Wildlife 14(2):102-118. https://doi.org/10.1080/10871200802695594

Lajeunesse, G. 2022. Two programs for use of aversive conditioning to manage bold urban coyotes. Thesis. University of Alberta, Edmonton, Alberta, Canada.

Lawrence, S. E., and P. R. Krausman. 2011. Reactions of the public to urban coyotes (Canis latrans). Southwestern Naturalist 56(3):404-409. https://doi.org/10.1894/N03-MLK-06.1

Lukasik, V. M., and S. M. Alexander. 2011. Human-coyote interactions in Calgary, Alberta. Human Dimensions of Wildlife 16(2):114-127. https://doi.org/10.1080/10871209.2011.544014

Lute, M. L., and N. H. Carter. 2020. Are we coexisting with carnivores in the American West? Frontiers in Ecology and Evolution 8:48. https://doi.org/10.3389/fevo.2020.00048

MacDonald, P. L., and R. C. Gardner. 2000. Type I error rate comparisons of post hoc procedures for I j chi-square tables. Educational and Psychological Measurement 60(5):735-754. https://doi.org/10.1177/00131640021970871

Magle, S. B., M. Fidino, E. W. Lehrer, T. Gallo, M. P. Mulligan, M. J. Ríos, A. A. Ahlers, J. Angstmann, A. Belaire, B. Dugelby, et al. 2019. Advancing urban wildlife research through a multi‐city collaboration. Frontiers in Ecology and the Environment 17(4):232-239. https://doi.org/10.1002/fee.2030

McGarigal, K., H. Y. Wan, K. A. Zeller, B. C. Timm, and S. A. Cushman. 2016. Multi-scale habitat selection modeling: a review and outlook. Landscape Ecology 31:1161-1175. https://doi.org/10.1007/s10980-016-0374-x

Merkle, J. A., P. R. Krausman, N. J. Decesare, and J. J. Jonkel. 2011. Predicting spatial distribution of human-black bear interactions in urban areas. Journal of Wildlife Management 75(5):1121-1127. https://doi.org/10.1002/jwmg.153

Morehouse, A. T., and M. S. Boyce. 2017. Troublemaking carnivores: conflicts with humans in a diverse assemblage of large carnivores. Ecology and Society 22(3):4. https://doi.org/10.5751/ES-09415-220304

Mowry, C. B., A. Lee, Z. P. Taylor, N. Hamid, S. Whitney, M. Heneghen, J. Russell, and L. A. Wilson. 2020. Using community science data to investigate urban coyotes (Canis latrans) in Atlanta, Georgia, USA. Human Dimensions of Wildlife 26(2):163-178. https://doi.org/10.1080/10871209.2020.1806415

Mueller, M. A., D. Drake, and M. L. Allen. 2019. Using citizen science to inform urban canid management. Landscape and Urban Planning 189:362-371. https://doi.org/10.1016/j.landurbplan.2019.04.023

Murray, M., A. Cembrowski, A. D. M. Latham, V. M. Lukasik, S. Pruss, and C. C. St. Clair. 2015a. Greater consumption of protein-poor anthropogenic food by urban relative to rural coyotes increases diet breadth and potential for human-wildlife conflict. Ecography 38(12):1235-1242. https://doi.org/10.1111/ecog.01128

Murray, M., M. A. Edwards, B. Abercrombie, and C. C. St. Clair. 2015b. Poor health is associated with use of anthropogenic resources in an urban carnivore. Proceedings of the Royal Society B 282:20150009. https://doi.org/10.1098/rspb.2015.0009

Murray, M. H., and C. C. St. Clair. 2017. Predictable features attract urban coyotes to residential yards. Journal of Wildlife Management 81(4):593-600. https://doi.org/10.1002/jwmg.21223

Nation, P. N., and C. C. St. Clair. 2019. A forensic pathology investigation of dismembered domestic cats: coyotes or cults? Veterinary Pathology 56(3):444-451. https://doi.org/10.1177/0300985819827968

Nielsen, S. E., G. B. Stenhouse, and J. Cranston. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5:38-60.

Olson, E. R., T. R. Van Deelen, A. P. Wydeven, S. J. Ventura, and D. M. MacFarland. 2015. Characterizing wolf-human conflicts in Wisconsin, USA. Wildlife Society Bulletin 39(4):676-688. https://doi.org/10.1002/wsb.606

Poessel, S. A., S. W. Breck, T. L. Teel, S. Shwiff, K. R. Crooks, and L. Angeloni. 2013. Patterns of human-coyote conflicts in the Denver Metropolitan Area. Journal of Wildlife Management 77(2):297-305. https://doi.org/10.1002/jwmg.454

Poessel, S. A., E. M. Gese, and J. K. Young. 2017. Environmental factors influencing the occurrence of coyotes and conflicts in urban areas. Landscape and Urban Planning 157:259-269. https://doi.org/10.1016/j.landurbplan.2016.05.022

Prugh, L. R., C. J. Stoner, C. W. Epps, W. T. Bean, W. J. Ripple, A. S. Laliberte, and J. S. Brashares. 2009. The rise of the mesopredator. Bioscience 59(9):779-791. https://doi.org/10.1525/bio.2009.59.9.9

Quinn, T. 1995. Using public sighting information to investigate coyote use of urban habitat. Journal of Wildlife Management 59(2):238-245. https://doi.org/10.2307/3808936

R Core Team. 2022. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Riley, S. J., and D. J. Decker. 2000. Risk perception as a factor in wildlife stakeholder acceptance capacity for cougars in Montana. Human Dimensions of Wildlife 5(3):50-62. https://doi.org/10.1080/10871200009359187

Schell, C. J., L. A. Stanton, J. K. Young, L. M. Angeloni, J. E. Lambert, S. W. Breck, and M. H. Murray. 2020. The evolutionary consequences of human-wildlife conflict in cities. Evolutionary Applications 14(1):178-197. https://doi.org/10.1111/eva.13131

Schell, C. J., J. K. Young, E. V. Lonsdorf, R. M. Santymire, and J. M. Mateo. 2018. Parental habituation to human disturbance over time reduces fear of humans in coyote offspring. Ecology and Evolution 8(24):12965-12980. https://doi.org/10.1002/ece3.4741

Smith, T. S., S. Herrero, and T. D. DeBruyn. 2005. Alaskan brown bears, humans, and habituation. Ursus 16(1):1-10. https://doi.org/10.2192/1537-6176(2005)016[0001:ABBHAH]2.0.CO;2

Soulsbury, C. D. 2020. Temporal patterns of human-fox interactions as revealed from internet searches. Human Dimensions of Wildlife 25(1):70-81. https://doi.org/10.1080/10871209.2019.1666322

Soulsbury, C. D., and P. C. L. White. 2015. Human-wildlife interactions in urban areas: a review of conflicts, benefits and opportunities. Wildlife Research 42(7):541-553. https://doi.org/10.1071/WR14229

Sponarski, C. C., C. Miller, and J. J. Vaske. 2018. Perceived risks and coyote management in an urban setting. Journal of Urban Ecology 4(1):juy025. https://doi.org/10.1093/jue/juy025

Statistics Canada. 2019. The open database of buildings. Government of Canada, Ottawa, Ontario, Canada. https://www.statcan.gc.ca/en/lode/databases/odb

Sugden, S., M. Murray, M. A. Edwards, and C. C. St. Clair. 2021. Inter-population differences in coyote diet and niche width along an urban-suburban-rural gradient. Journal of Urban Ecology 7(1):juab034. https://doi.org/10.1093/jue/juab034

Sullivan, B. L., J. L. Aycrigg, J. H. Barry, R. E. Bonney, N. Bruns, C. B. Cooper, T. Damoulas, A. A. Dhondt, T. Dietterich, A. Farnsworth, et al. 2014. The eBird enterprise: an integrated approach to development and application of citizen science. Biological Conservation 169:31-40. https://doi.org/10.1016/j.biocon.2013.11.003

Tredennick, A. T., G. Hooker, S. P. Ellner, and P. B. Adler. 2021. A practical guide to selecting models for exploration, inference, and prediction in ecology. Ecology 102(6):e03336. https://doi.org/10.1002/ecy.3336

van Bommel, J. K., M. Badry, A. T. Ford, T. Golumbia, and A. C. Burton. 2020. Predicting human-carnivore conflict at the urban-wildland interface. Global Ecology and Conservation 24:e01322. https://doi.org/10.1016/j.gecco.2020.e01322

Venables, W. N., and B. D. Ripley. 2002. Modern applied statistics with S. Fourth edition. Springer, New York, New York, USA. https://doi.org/10.1007/978-0-387-21706-2

Weckel, M. E., D. Mack, C. Nagy, R. Christie, and A. Wincorn. 2010. Using citizen science to map human-coyote interaction in suburban New York, USA. Journal of Wildlife Management 74(5):1163-1171. https://doi.org/10.2193/2008-512

White, L. A., and S. D. Gehrt. 2009. Coyote attacks on humans in the United States and Canada. Human Dimensions of Wildlife 14(6):419-432. https://doi.org/10.1080/10871200903055326

Williams C. K., G. Ericsson, and T. A. Heberlein. 2002. A quantitative summary of attitudes toward wolves and their reintroduction (1972-2000). Wildlife Society Bulletin 30(2):575-584.

Wine, S., S. A. Gagné, and R. K. Meentemeyer. 2015. Understanding human-coyote encounters in urban ecosystems using citizen science data: what do socioeconomics tell us? Environmental Management 55:159-170. https://doi.org/10.1007/s00267-014-0373-0

Young, J. K., E. Hammill, and S. W. Breck. 2019. Interactions with humans shape coyote responses to hazing. Scientific Reports 9:20046. https://doi.org/10.1038/s41598-019-56524-6

Copyright © 2023 by the author(s). Published here under license by The Resilience Alliance. This article is under a Creative Commons Attribution 4.0 International License. You may share and adapt the work provided the original author and source are credited, you indicate whether any changes were made, and you include a link to the license.