Patterns of wild carnivore attacks on humans in urban areas
Patterns of wild carnivore attacks on humans in urban areas
Mar?a del Mar Delgado
Luca Francesco Russo
Pedro Jos? Garrote
Jos? Vicente L?pez-Bao
Jos? M. Fedriani
OPEN Attacks by wild carnivores on humans represent an increasing problem in urban areas across North America and their frequency is expected to rise following urban expansion towards carnivore habitats. Here, we analyzed records of carnivore attacks on humans in urban areas of the U.S. and Canada between 1980 and 2016 to analyze the general patterns of the attacks, as well as describe the landscape structure and, for those attacks occurring at night, the light conditions at the site of the attacks. We found that several behavioral and landscape-related factors were recurrent elements in the attacks recorded. The species for which the attack locations were available (coyote and black bear) attacked in areas with different conditions of landscape structure and artificial light. Specifically, black bears attacked more frequently in areas with abundant and aggregated vegetation cover and scarce buildings and roads, while coyotes attacked in a broader range of landscape conditions. At night, black bears attacked in generally darker areas than coyotes. By providing a comprehensive perspective of the phenomenon, this study will improve our understanding of how effective strategies aimed at reducing the frequency of risky encounters in urban areas should be developed.
Recent years have witnessed an increase in conflicts between humans and wild carnivores in North American
urban areas (i.e., populated places, defined by the U.S. Geological Survey (https://www.usgs.gov/) as ?a place or
area with clustered or scattered buildings and a permanent human population (city, settlement, town, village)?)1,2.
These conflicts include property damage, anthropogenic food consumption, livestock and pet attacks and, more
rarely, attacks on people1,3?5. Increasing overlap between human and carnivore habitats may be behind this
trend6,7. On the one hand, some populations of carnivores have expanded their range due to the improved human
attitudes and stricter protection in recent years. On the other hand, the rapid expansion of urban areas into
landscapes inhabited by these species is causing large areas of natural patches to surround or be incorporated into
urban areas6,8,9. These natural patches provide carnivores with suitable habitats (e.g., abundance of prey and
shelter) in close proximity to, or even inside, human developments. This, together with the ability of some carnivores
to use anthropogenic resources (e.g., non-seasonal and high-caloric anthropogenic food) and thrive in highly
human-modified landscapes may lead to increased conflictual interactions10?12.
Even though attacks on humans in urban areas are rare and mainly result in minor injuries, they often elicit
lethal responses towards the animals considered responsible for the attack and decrease public tolerance towards
these species, subsequently influencing management and conservation actions6,13,14. Therefore, both humans and
carnivores lose when such incidents happen and, because of this, reducing the occurrence of such attacks in urban
areas should be considered a priority for authorities. For this reason, rigorous analysis of attack scenarios aimed
at identifying the factors which may drive risky human-carnivore encounters can provide decision-makers with
Only a handful of studies have focused on wild carnivore attacks on humans in urban areas1,15,16. These studies
have analyzed coyote Canis latrans attacks only and have highlighted that changes in human behaviors (e.g.,
management of attractants and pet supervision) can play a crucial role in reducing the number of attacks. However,
several other factors need to be taken into consideration when analyzing attack triggers and scenarios. For
example, information regarding the characteristics of the natural and human environment at the site of the attack, as
well as the conditions of artificial illumination for those incidents that occurred at night, might turn out to be
crucial for understanding the dynamics of the attacks and for the development of management actions aimed at
reducing the risk of dangerous encounters. Moreover, although until recent years coyotes have been almost the
only species responsible for attacks on humans in urban areas, the current increase in the number of attacks by
other wild carnivores17 highlights the need for a more comprehensive approach encompassing all carnivore
species occurring in urban landscapes.
Here, we analyzed the scenarios of carnivore attacks on humans that occurred in urban areas across the U.S.
and Canada during the last 36 years (from 1980 to 2016). We first studied temporal patterns of the attacks at
different scales (i.e., circadian, seasonal and annual) and general patterns related to various factors such as age and
sex of the victims, party composition, location and scenario of the attacks. Further, we examined the structure of
the landscape (i.e., abundance and structure of vegetation, abundance of buildings and roads) at the attack sites
and assessed whether differences in attack patterns between species exist. Specifically, following what found in
previous studies on other kinds of conflicts4,5,18, we hypothesized that species which are mostly forest-obligate
and generally avoid humans will mainly attack under landscape conditions characterized by high vegetation
cover and the fewest human infrastructures, whereas we expected landscape structure to not be relevant for those
species which are known to reside in urban environments and tolerate human presence. Finally, for those attacks
occurring at night, we explored whether (and how) light conditions might influence the occurrence of an attack.
Specifically, we hypothesized that a higher number of attacks will occur in dark areas.
Results and Discussion
General patterns of the attacks. Most of the attacks occurred in California (n = 66, 37%), followed by
Colorado (n = 16, 9%), British Columbia (n = 13, 7%) and the other jurisdictions (47%) (Fig.?1 and Supplemental
Fig.?S1a). The number of attacks recorded in urban areas has increased over time, with similar trends for the
different species (Supplemental Fig.?S2a). Spring and summer were the seasons showing the highest rates of attack
(Supplemental Fig.?S2b). Coyotes attacked uniformly throughout the year, with a slight peak during the spring,
bear attacks were rare during winter and cougars attacked more often during spring and summer. This seasonal
pattern conforms to the species? biology and confirms what was previously shown in other studies17,19?21. Indeed:
(a) bears are generally hibernating during winter; and (b) coyotes are rearing their pups during spring, when we
observed a slight increase in attacks, and thus they might be in search of additional food and defending their dens
during this period1, which makes them more likely to be involved in aggressive encounters with humans and
Most of the attacks occurred during the day, especially those involving coyotes and cougars (Supplemental
Fig.?S3a). This outcome is likely the result of the daily activity of humans in urban areas. Moreover, although
coyotes in urban environments have been shown to change their activity patterns to crepuscular and nocturnal
to avoid humans15, in many cities they have become habituated to people and, consequently, they might have lost
their avoidance behavior and returned to being active during the day15. On the other hand, black bears tend to be
mostly active at night to avoid humans23,24.
In general, children (<13 y. o.) were attacked less often than adults (Supplemental Fig.?S2c), with a trend
towards younger individuals (n = 34 attacks between 0 and 3 years old, n = 16 between 4 and 7 years old and
n = 14 between 8 and 11 years old; no attacks were recorded on 9-, 10- and 12-year-old children). Coyotes and
cougars attacked children and adults almost equally, while bears attacked considerably more adults than children
(Supplemental Fig.?S2c). This difference is probably related to the reasons triggering an attack. Indeed, brown
and black bears were mainly involved in attacks related to dog presence and anthropogenic food (related to food
and trash handling), two scenarios that primarily involve adults, whereas most of the predatory attacks in which
victims were prevalently children25 were carried out by cougars and coyotes. Additionally, bears attacked more
frequently at night, when children are less likely to be found outside than adults?. These patterns also reflect
differences in the species? ecology. While bears are omnivores, cougars are strictly carnivore and coyotes, although
they are known to forage on other resources as well26,27, are also mainly carnivore. Consequently, we can expect
cougars and coyotes to be involved in predatory attacks (and, therefore, attack children) more likely than bears.
The presence of dogs at the moment of the attack was the most prevalent scenario, followed by attacks related
to anthropogenic food, predatory motivation and other kinds of scenarios (Supplemental Fig.?S2d). Cougar and
polar bear attacks were all predatory. Victims of predatory attacks were mainly children (84%), and coyotes were
responsible for the majority (63%) of these attacks. People involved in attacks related to dog presence were all
adults, which represented the majority of the victims of food-related attacks as well. The high incidence of night
attacks when the presence of a dog is involved is probably linked to the late walks that dog owners take in urban
areas due to their work schedules and locally hot temperatures during the day28,29.
There was only a slight difference between the number of male and female victims (Supplemental Fig.?S3b).
Most of the victims of black bear attacks were alone, while coyotes attacked unaccompanied people and children
in a party nearly equally (Supplemental Fig.?S3c).
Landscape structure and artificial light at the site of the attacks. The exact location of the attacks
was available for coyotes and black bears only (nblack bear = 22, ncoyote = 47) and, of these attacks, 15 occurred
at night (nblack bear = 7, ncoyote = 8). Our results were consistent with our initial hypothesis. Indeed, the PCA
(Supplemental Tables?S1A and S1B) showed a clear difference between attacks by coyotes and black bears in
terms of landscape structure (Fig.?2). On one hand, black bear attacks occurred in areas with specific landscape
conditions, i.e. (a) few buildings and roads, and (b) dense vegetation cover, which is in line with the ecology of the
species both in wildlands and urban areas11,30,31, as well as with previous studies which have analyzed the spatial
distribution of other types of human-black bear conflicts2,4,7,32. These studies suggested that the probability of
conflicts with this species was correlated with proximity to large forest patches and intermediate housing
densities. This is probably related to the fact that black bears are predominately a forest obligate species30,32, although
We found that those attacks that occurred at night took place in areas with a relatively low amount of artificial
light (radiance always <6.00 W/cm2 * sr), with black bear attacks occurring in particularly dark areas (radiance
values lower than 1.00 W/cm2 * sr; Supplemental Table?1C, Fig.?3). This outcome might be related to the recorded
abundance of vegetation cover at the locations of black bear attacks. Indeed, we can expect that areas with high
vegetation cover are also characterized by lower artificial light than intensely urbanized sectors.
Are there solutions for this increasing conflict? The role of human behavior in the attacks. Dog
presence: It is noteworthy that in at least 20 (66%) of the 33 attacks related to the presence of dogs, humans were
not the first target. In these incidents, either the carnivore targeted the dog first and the owner intervened in its
defense (most of the cases, 80%, n = 16) or the dog confronted the carnivore, with the owner being subsequently
involved in the encounter. Improved public education by local authorities on how to behave with dogs in areas
frequented by wild carnivores would certainly help increase public awareness and thus reduce the occurrence of
these incidents. As also previously suggested14,16,17, keeping dogs on-leash while out walking in areas with
carnivores would reduce the number of risky encounters. In the case of coyotes, scaring off the animal with the help of
objects has been recommended by wildlife services16,19,34. Similarly, keeping dogs inside or in a well-fenced shelter
in the yard might help to avoid predatory attempts when the owner is not directly taking care of their dog16. Our
results suggest that, while for coyotes these precautions should always be taken, i.e. independent of the landscape
structure and light conditions, in areas where black bears are present they might be particularly important when
the vegetation cover is high and the density of human buildings is low. Additionally, particular attention should
be taken at night, especially in areas where artificial illumination is scarce.
Attractants management: The insufficient management of anthropogenic food such as pet food, bird feeders
and garbage, both in private properties and public parks, together with the practice of wildlife feeding, are already
known to be among the most common causes of human-carnivore conflicts15,19,20,35. The proper management
of attractants is even more important within urban areas, due to the high number of people potentially exposed
to a risky encounter with a wild carnivore. Although significant effort has been made to inform and educate the
public on how to reduce attractants, and wildlife feeding has been forbidden in many cities19,36, the increasing
trend of attacks indicates that current efforts might not be sufficient and more resources should be invested in
preventive actions. Additionally, while education and regulations alone might have little effect on changing human
behaviour37,38, combining these actions with proactive enforcement (e.g. increased patrolling and application of
warnings) might prove to be more efficient in altering human behaviour11,38.
Predatory attacks on children: Lone children are the preferred target of coyote, cougar and black bear
predatory attacks. This kind of attack is the most dangerous and has already been documented in previous studies15,19,25.
When outside, both in yards and green spaces, children should be continuously supervised by an adult, at a
minimum, and never left alone. The presence of an adult may help to reduce the chances of a child being attacked.
Additionally, fencing yards and playgrounds in areas where carnivores are present may be an effective precaution
to increase child safety.
The role of landscape planning. Assuming that both human and carnivore populations will continue to rise in
the future, we should expect an increasing overlap between urban areas and carnivore ranges and, therefore,
an increase in the number of attacks. The sprawl of human developments towards natural habitats is rapidly
rising and residential housing is expected to increase across the landscape, due to homebuyers? preferences for
single-family detached homes10,39. Moreover, the recent trend towards ?greener? and wildlife-friendly urban
landscape design is leading urban planners to promote the inclusion of natural patches and wildlife habitat
requirements into the urban matrix9,40,41, which may create optimal habitats for some carnivore species. The presence of
green spaces and the recent spread of the practice of ?wildlife gardening? (i.e., employment of a series of practices
aimed at increasing wildlife in gardens) have been shown to provide important benefits to both human health and
wildlife biodiversity6,40,42. However, practices such as keeping dense vegetation and fruit-trees in yards and green
areas, as well as leaving bird feeders outside, are likely to attract wild carnivores and, consequently, may increase
the probability of a risky encounter5,11,32. These practices should then be avoided in urban areas with resident
carnivore populations and/or located near carnivore habitats. Instead, reducing thick vegetation (e.g., dense forests
or bushes) to increase visibility and prevent carnivores from using it as shelter, as well as the implementation of
fences and improved artificial illumination systems in green areas and yards, can effectively result in increasing
both human and pet safety (see also5,15,34).
Similarly, in areas scheduled for development, urban planners and homebuyers should be informed of the
risk that low-density developments (i.e., sparse housing developments which incorporate large wildland areas)
might involve10. These kinds of developments, which also include ex-urban and suburban areas, have already
been shown to favor the colonization of urban areas by wild carnivores5,33 and present a higher concentration
of human-wildlife conflicts, especially when situated in proximity to natural areas4,6,10,14. In this sense, in terms
of land use, our findings support the ?land sparing? model, which favors high-density developments in order to
preserve wildland43,44. This kind of development might be an effective way not only to minimize habitat
fragmentation in general, but also to exclude carnivores from urban areas by separating human developments from
wildlife habitats and, thus, reduce the occurrence of negative interactions with these species. Finally, we suggest
that further studies should investigate whether the attacks are more likely to occur in specific areas within the
areas used by the species. This fine-scale analysis would require radiotagging of urban carnivores, which will
allow comparing the characteristics of the urban sites where attacks may occur (our results) vs. the areas selected
by these species.
Several behavioral and landscape-related factors were recurrent elements in the attacks recorded in North
American urban areas. Therefore, effective strategies aimed at benefitting both humans and carnivores will need
to combine carnivore knowledge, citizen education and landscape planning. Specifically: (
) because different
species attack under different conditions, management plans should be developed according to the species
occurring in a given area and generalizations should be avoided; (
) education actions should provide the public with
practical information on how to avoid conflicts and how to behave in case of an encounter with a wild carnivore;
) landscape planners should work to develop plans able to balance human health, wildlife conservation and
conflict risk. Specific landscape modifications and design should thus be employed both in already existing urban
green areas and when planning new urban areas.
Collection of records of carnivore attacks on humans in urban areas. We collected reports of wild
carnivore attacks on humans resulting in physical injury or death that occurred in urban areas in the United States
and Canada from 1980 to 2016. We used attack reports included in the database used in Penteriani et al.17 by only
selecting the attacks which occurred in urban areas within the above-mentioned study area and time period.
We included attack reports starting in 1980 because attacks were poorly documented before that year. We then
updated the database by adding reports from the years 2014 to 2016.
Our search included the following species: brown bear/grizzly Ursus arctos, black bear Ursus americanus, polar
bear Ursus maritimus, cougar Puma concolor, grey wolf Canis lupus and coyote Canis latrans. We attempted to
exclude attacks by rabid animals from this work because their behavior is likely atypical. Records of attacks were
collected from unpublished reports and PhD/MS theses, webpages, books and scientific articles. To complete
the dataset, we also collected news reports from online newspapers. To do this, for each species and area, we
searched on an annual basis for news articles on Google using the combination of the following terms: ?species
name? + ?attack?, ?species name? + ?attack? + ?human? and ?species name? + ?attack? + ?State/province name? + ?year?.
Because we used several sources, some of the attacks recurred repeatedly during the search, but we used
information such as date, location and sex/age of the victims to prevent duplicate records in the dataset. Furthermore, we
were able to obtain additional information concerning attacks from the Florida Fish and Wildlife Conservation
Commission, Washington Department of Fish and Wildlife, New Mexico Department of Game and Fish and New
York Department of Environmental Conservation. This information allowed us to (
) verify if the information
we had recorded about the attacks was correct, (
) obtain the exact location of the attacks recorded and (
new attack reports (if any).
We collected a total of 177 attacks, of which 63% were by coyotes (n = 101), 27% by black bears (n = 44), 7%
by cougars Puma concolor (n = 12), 2% by polar bears Ursus maritimus (n = 4) and 1% by grizzlies Ursus arctos
(n = 2) (Supplemental Fig.?S1b).
For each attack, we recorded the following information: (
) carnivore species; (
) year; (
) month; (
location of the attack; (
) time of the day, which we classified into three categories: twilight, day, night; (
of the attack within the urban area: inside home, near home, playground/park, school, others (examples of other
locations include: outside a hotel, parking lot, golf course, university campus, on the street); (
) sex and age of the
) sex and age of the victim; (
) party composition, simplified into three categories: (a) victim alone,
(b) child ?from 0 to 13 years old? in a party of adults, and (c) party of adults, i.e., people >13 years old; (
of the attack, i.e., injuries or death; and (
) scenario, i.e., the main factor that could have triggered the attack. We
defined four scenarios: (a) predatory, i.e. when the carnivore deliberately attacked and/or killed a human with the
presumed purpose of consuming it. Specifically, we considered predatory only those cases where: (
) the human
was treated as food (i.e., the person is dragged by the carnivore far from the attack site to a more hidden location
such as a forest patch or bushes); (
) the body (of both live and dead victims) is covered with leaves and soil;
) after its death, the victim is partially consumed; and/or (
) a carnivore has been found near the body25; (b)
dog-related, i.e., one or more dogs present; (c) anthropogenic food-related, e.g., a carnivore reported feeding on
anthropogenic food at the time of the attack or an individual known to be food-conditioned or intentionally fed
by humans; and (d) other scenarios, i.e., female with young, aggressive reaction after a sudden encounter, food/
territory defense or a wounded animal.
Characterization of the landscape structure at the site of the attacks. To describe the landscape
structure of the attack site, we selected only those attacks for which the exact location was available (39% of the
total attacks recorded; an estimated maximum error of ca. 100 m was accepted). The exact locations of the attacks
were obtained from the U.S. departmental agencies mentioned above and other sources reporting the precise site
of the attack (i.e., providing the address, coordinates or the name of the park/school where the attack took place).
We uploaded the coordinates of each attack into the Google Earth Pro application and selected a plot of 1
km2 centered at the point of the attack. We considered 1 km2 to be a good trade-off between the accuracy of the
location points recorded and the aim of our work, which was to analyze landscape structure in the immediate
vicinity of the attack site. Because of the dynamic structure of both natural and human landscapes, for each attack
we searched for the map of the year when the attack took place. When the map of the year was not available, we
used a map from within 3 years preceding or following the attack. Once the satellite images for each attack
location were extracted, we analyzed them by using the image processing software Photoshop CS6 and calculated 5
landscape parameters to both quantify vegetation structure and describe the degree of aggregation of the
vegetation in our plots: the area (in m2) occupied by (
) vegetation (trees and shrubs), (
) buildings and (
) roads; (
the vegetation patch density (PD), defined as the number of vegetation patches (i.e. homogeneous areas occupied
by vegetation) per unit area, where high values of PD mean a high number of patches per area unit (i.e., highly
fragmented vegetation); and (
) the mean patch size in the form of area-weighted mean patch size (AREA_AM),
which equals the sum, across all patches, of the patch area multiplied by the proportional abundance of the patch
(i.e., patch area divided by the sum of all patch areas). The area-weighted mean patch size (AREA_AM) is less
sensitive to small patches than simple mean size and provides a better overall measure of subdivision45. PD and
AREA_AM were calculated using the area of vegetation and the number of vegetation patches in each plot
following Mcgarigal et al.45. Both metrics were calculated at the class level (i.e., vegetation level), as our landscape area
was constant throughout all 1 km2 plots.
Collection of information related to artificial light during night attacks. For those attacks that
occurred at night and for which we had the exact location (nblack bear = 7, ncoyote = 8), we analyzed the amount of
artificial light available near the attack site. Specifically, we extracted a map of artificial light at the attack site (as
with the landscape parameters, we considered an area of 1 km2 centered at the point of the attack) from the website
https://www.lightpollutionmap.info/. This website provides a world atlas of artificial night sky radiance (in W/cm2
* sr, where W = watt and sr = steradian or square radian, i.e. the International System of Units of solid angles
that quantifies planar angles, which is used to measure the luminous intensity of a light source)46, where different
ranges of radiance are represented by different colors. Specifically, low values of radiance correspond to lower
amounts of artificial light, with radiance values <0.25 considered as a typical moonless night sky background
far from the Milky Way, zodiacal and artificial light (artificial sky brightness <1% of the natural background)46.
The atlas includes maps from 2010 to 2017 and, as with the landscape metrics, when the map of the year was not
available we used a map from within 3 years preceding or following the attack. Once the images were extracted,
we calculated the area occupied by each color (i.e. range of radiance) using Photoshop CS6 and calculated the
mean radiance of each map.
Data analysis. Landscape structure and artificial light at the site of the attacks. Since the landscape
parameters estimated were correlated (Spearman rank correlation rs always >0.70, P < 0.001), after log-transforming
PD and AREA_AM, we ran a Principal Component Analysis (PCA) including the 5 variables. We then built a
set of competing models which included the number of attacks per species as the response variable and
principal component 1 (PC1) and principal component 2 (PC2) obtained from the PCA as explanatory variables. We
finally built a second set of models, again with the number of attacks per species as the response variable, but now
including radiance (i.e., our proxy of the artificial light conditions) as the explanatory variable. In both sets of
models, as our response variable was categorical and had 2 levels (i.e., either attacks by coyotes or attacks by black
bears), we built Generalized Linear Models (GLMs) with a binomial distribution. We performed model selection
based on the Akaike?s Information Criterion corrected for small sample sizes (AICc; Burnham & Anderson)47 and
calculated two additional statistics for each model: ?AICc and AICc weights, which indicate the probability that
the model selected was the best among the competing candidates48. We considered models with ?AICc values
lower than 2 as competitive. All statistical analyses were performed using R 3.2.5 statistical software49.
We want to thank all officers of the U.S. departmental agencies and carnivore experts who provided us with
the information about the attacks. Among them, special thanks go to Sarah Barrett, Jennifer Brown, Jennifer
Montoya, Tom Chester and Helen McGinnis. We would also like to thank Rui Lourenco for the helpful advice
concerning the statistical analyses. We thank two referees for their valuable comments, which helped us to
improve the manuscript. M.M.D. and J.V.L.B. were supported by a Spanish Ramon y Cajal research contract
(M.M.D: RYC-2014-16263; J.V.L.B: RYC-2015-18932) from the Spanish Ministry of Economy, Industry and
Competitiveness. VP was financially supported by the Excellence Project CGL2017-82782-P financed by the
Spanish Ministry of Economy, Industry and Competitiveness (MINECO), the Agencia Estatal de Investigacion
and the Fondo Europeo de Desarrollo Regional of the European Union.
G.B., V.P. and M.M.D. initiated and conceived the study, G.B. and V.P. prepared the database with the collaboration
of L.F.R., P.J.G. and J.M.F., G.B., M.M.D. and V.P. analyzed the data, G.B., V.P. and M.M.D. wrote the manuscript.
All authors (G.B., M.M.D., L.F.R., P.J.G., J.V.L., J.M.F. and V.P.) commented on the manuscript draft.
Competing Interests: The authors declare no competing interests. Publisher?s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article?s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article?s Creative Commons license and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
1. Baker , R. O. & Timm , R. M. Coyote attacks on humans. 1970-2015: implications for reducing the risks . 11 , 120 - 132 ( 2017 ).
2. Evans , M. J. , Hawley , J. E. , Rego , P. W. & Rittenhouse , T. A. G. Exurban land use facilitates human-black bear conflicts . J. Wildl. Manage . 78 , 1477 - 1485 ( 2014 ).
3. Don Carlos , A. W. , Bright , A. D. , Teel , T. L. & Vaske , J. J. Human-black bear conflict in urban areas: An integrated approach to management response . Hum. Dimens. Wildl . 14 , 174 - 184 ( 2009 ).
4. Merkle , J. A. , Krausman , P. R. , Decesare , N. J. & Jonkel , J. J. Predicting spatial distribution of human-black bear interactions in urban areas . J. Wildl. Manage . 75 , 1121 - 1127 ( 2011 ).
5. Poessel , S. A. , Gese , E. M. & Young , J. K. Environmental factors influencing the occurrence of coyotes and conflicts in urban areas . Landsc. Urban Plan . 157 , 259 - 269 ( 2017 ).
6. Soulsbury , C. D. & White , P. C. L. Human-wildlife interactions in urban areas: a review of conflicts, benefits'and opportunities . Wildl. Res . 3330 , https://doi.org/10.1071/WR14229 ( 2015 ).
7. Baruch-Mordo , S. , Breck , S. W. , Wilson, K. R. & Theobald , D. M. Spatiotemporal Distribution of Black Bear-Human Conflicts in Colorado, USA . J. Wildl . Manage. 72 , 1853 - 1862 ( 2008 ).
8. Kabisch , N. & Haase , D. Green spaces of European cities revisited for 1990-2006 . Landsc. Urban Plan. 110 , 113 - 122 ( 2013 ).
9. Adams , L. W. Urban wildlife ecology and conservation: a brief history of discipline . Urban Ecosyst . 8 ( 2 ), 139 - 156 ( 2005 ).
10. Kretser , H. E. , Sullivan , P. J. & Knuth , B. A. Housing density as an indicator of spatial patterns of reported human-wildlife interactions in Northern New York . Landsc. Urban Plan. 84 , 282 - 292 ( 2008 ).
11. Lewis , D. L. et al. Foraging ecology of black bears in urban environments: guidance for human-bear conflict mitigation . Ecosphere 6 , art141 ( 2015 ).
12. Blecha , K. A. , Boone , R. B. & Alldredge , M. W. Hunger mediates apex predator's risk avoidance response in wildland-urban interface . J. Anim. Ecol . 609 - 622 , https://doi.org/10.1111/ 1365 - 2656 .12801 ( 2018 ).
13. L?e , J. & R?skaft , E. Large Carnivores and Human Safety : A Review. AMBIO A J. Hum . Environ. 33 , 283 - 288 ( 2004 ).
14. Poessel , S. A. , Breck , S. , Teel , T. L. , Shwif , S. & Crooks , K. R. Patterns of Human - Coyote Conflicts in the Denver Metropolitan Area . J. Wildl. Manage . 77 , 297 - 305 ( 2013 ).
15. Timm , R. M. , Baker , R. O. , Bennett , J. R. & Coolahan , C. C. Coyote Attacks : An Increasing Suburban Problem . Trans. North Am. Wildl. Nat. Resour. Conf . 69 , 67 - 88 ( 2004 ).
16. Timm , R. M. & Baker , R. O. A history of urban coyote problems . Wildl. Damage Manag. Conf. Proc ( 2007 ).
17. Penteriani , V. et al. Human behaviour can trigger large carnivore attacks in developed countries . Sci. Rep . 6 , 20552 ( 2016 ).
18. Lukasik , V. M. & Alexander , S. M. Human-Coyote interactions in Calgary, Alberta . Hum. Dimens. Wildl. 16 , 114 - 127 ( 2011 ).
19. White , L. A. & Gehrt , S. D. Coyote attacks on humans in the United States and Canada . Hum. Dimens. Wildl . 14 , 419 - 432 ( 2009 ).
20. Herrero , S. & Higgins , A. Human injuries inflicted by bears in Alberta: 1960 - 98 . Ursus (Knoxville) 14 , 44 - 54 ( 2003 ).
21. Baruch-Mordo , S. et al. Stochasticity in natural forage production affects use of urban areas by black bears: Implications to management of human-bear conflicts . PLoS One 9 , 1 - 10 ( 2014 ).
22. Morey , P. S. , Gense , E. M. & Gehrt , S. D. Spatial and temporal variation in the diet of coyotes in the Chicago metropolitan area . Am. Midl. Nat . 158 , 147 - 161 ( 2007 ).
23. Merkle , J. A. , Robinson , H. S. , Krausman , P. R. & Alaback , P. Food availability and foraging near human developments by black bears . J. Mammal . 94 , 378 - 385 ( 2013 ).
24. Beckmann , J. P. & Berger , J. Rapid ecological and behavioural changes in carnivores: the responses of black bears (Ursus americanus) to altered food . J. Zool . 261 , S0952836903004126 ( 2003 ).
25. Penteriani , V. et al. Humans as prey: coping with large carnivore attacks using a predator - prey interaction perspective . Human-Wildlife Interact . 11 , 192 - 207 ( 2017 ).
26. Fedriani , J. M. , Fuller , T. K. & Sauvajot , R. M. Does availability of anthropogenic food enhance densities of omnivorous mammals? An example with coyotes in southern California . Ecography (Cop.) . 24 , 325 - 331 ( 2001 ).
27. Poessel , S. A. , Mock , E. C. & Breck , S. W. Coyote (Canis latrans) diet in an urban environment: variation relative to pet conflicts, housing density, and season . Can. J. Zool . 95 , 287 - 297 ( 2017 ).
28. Lee , H. , Shepley , M. & Huang , C. Evaluation of off-leash dog parks in Texas and Florida: A study of use patterns, user satisfaction, and perception . Lanscape Urban Plan . 92 , 314 - 324 ( 2009 ).
29. Cutt , H. , Giles-Corti , B. , Knuiman , M. & Burke , V. Dog ownership, health and physical activity: A critical review of the literature . Heal. Place 13 , 261 - 272 ( 2007 ).
30. Tri , A. N. et al. Habitat use by American black bears in the urban-wildland interface of the Mid-Atlantic, USA . Ursus 27 , 45 - 56 ( 2016 ).
31. Johnson , H. E. et al. Shifting perceptions of risk and reward: Dynamic selection for human development by black bears in the western United States . Biol. Conserv . 187 , 164 - 172 ( 2015 ).
32. McFadden-Hiller , J. E. , Beyer , D. E. & Belant , J. L. Spatial Distribution of Black Bear Incident Reports in Michigan . PLoS One 11 , e0154474 ( 2016 ).
33. Bateman , P. W. & Fleming , P. A. Big city life: Carnivores in urban environments . J. Zool . 287 , 1 - 23 ( 2012 ).
34. U.S. Department of Agriculture, W. S. Urban and Suburban Coyotes . Agriculture ( 2002 ).
35. Young , S. & Malpeli , K. Coyote Ecology and Conflicts with Humans across the Urban- Wildland Gradient: Identifying the Potential Impacts of Changing Land Use . 1 - 15 ( 2015 ).
36. Gore , M. L. , Knuth , B. A. , Curtis , P. D. & Shanahan , J. E. Education programs for reducing American black bear-human conflict: indicators of success? Ursus 17 , 75 - 80 ( 2006 ).
37. Gore , M. L. , Knuth , B. A. , Scherer , C. W. & Curtis , P. D. Evaluating a conservation investment designed to reduce human-wildlife conflict . Conserv. Lett. 1 , 136 - 145 ( 2008 ).
38. Baruch-Mordo , S. , Breck , S. W. , Wilson, K. R. & Broderick , J. The Carrot or the Stick? Evaluation of Education and Enforcement as Management Tools for Human-Wildlife Conflicts . PLoS One 6 , e15681 ( 2011 ).
39. Vogt , C. A. & Marans , R. W. Natural resources and open space in the residential decision process: A study of recent movers to fringe counties in southeast Michigan . Landsc. Urban Plan . 69 , 255 - 269 ( 2004 ).
40. Aronson , M. F. J. et al. Biodiversity in the city: key challenges for urban green space management . Front. Ecol. Environ . 15 , 189 - 196 ( 2017 ).
41. Pataki , D. E. Grand challenges in urban ecology . Front. Ecol. Evol. 3 , 1 - 6 ( 2015 ).
42. Gaston , K. J. , Smith , R. M. , Thompson , K. & Warren , P. H. Urban domestic gardens (II): Experimental tests of methods for increasing biodiversity . Biodivers. Conserv . 14 , 395 - 413 ( 2005 ).
43. Lin , B. B. & Fuller , R. A. FORUM : Sharing or sparing? How should we grow the world's cities? J. Appl. Ecol . 50 , 1161 - 1168 ( 2013 ).
44. Phalan , B. , Onial , M. , Balmford , A. & Green , R. E. Reconciling food production and biodiversity conservation: Land sharing and land sparing compared . Science ( 80 -.). 333 , 1289 - 1291 ( 2011 ).
45. Mcgarigal , K. , Cushman , S. A. , Neel , M. C. & Ene , E. FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps ( 2002 ).
46. Falchi , F. et al. The new world atlas of artificial night sky brightness . Sci. Adv . 2 , e1600377 - e1600377 ( 2016 ).
47. Burnham , K. P. & Anderson , D. R. Model Selection and Multimodel Inference : A Practical Information-Theoretic Approach (2nd ed). Ecological Modelling 172 ( 2002 ).
48. Gelman , A. & Hill , J. Data Analysis Using Regression and Multilevel/Hierarchical Models . Cambridge Univ. Press Analytical, 625 ( 2006 ).
49. R Foundation for Statistical Computing. R: A Language and Environment for Statistical Computing ( 2018 ).
50. QGIS Development Team . QGIS Geographic Information System. Open Source Geospatial Foundation ( 2016 ).
51. Boundary Files , 2011 Census. Statistics Canada Catalogue92-160 -X Available at, http://www.statcan.gc.ca/pub/92-160-g/ 92 -160- g2016002-eng.htm.
52. United States Census Bureau. Available at, https://www.census.gov/geo/maps-data/data/cbf/cbf_state.html. ( 2016 ).
53. Google Earth Pro. Available at, https://www.google.com/earth/ ( 2017 ).
Additional Information Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-018-36034-7.