Mesocarnivore landscape use along a gradient of urban, rural, and forest cover

Introduction

Habitat loss and fragmentation is one of the greatest threats to global biodiversity (Butchart et al., 2010; Seto, Guneralp & Hutyra, 2012). Urbanization and agricultural development are among the chief drivers of this phenomenon (Imhoff et al., 2004; Aronson et al., 2014), but a suite of human commensal species also benefit from anthropogenic landscapes (Baker & Harris, 2007; Soulsbury & White, 2016). This dichotomy is particularly true for mammalian carnivores, some of which are frequently extirpated while others reach their highest population densities in anthropogenic landscapes. Many mammalian carnivores benefit either from human-derived resources or release from predation in anthropogenic landscapes (Pickett et al., 2001; Prange, Gehrt & Wiggers, 2003; Gaston et al., 2005). The variable species response to human disturbance can complicate our understanding of community structure and landscape use of the urban-rural-forest matrix (Lesmeister et al., 2015; Moss, Alldredge & Pauli, 2016).

Species most susceptible to anthropogenic disturbance include those with large home ranges, low population densities, or low fecundity (Pimm et al., 2014). Apex predators, which frequently share these traits, have declined globally in fragmented and urbanized systems (Crooks & Soulé, 1999; Hollings et al., 2016) while many mesopredators with faster life histories, smaller home ranges, and higher fecundities have increased their distribution and abundance (Prugh et al., 2009; Santini et al., 2019). The ability of species to persist in anthropogenic environments is also a product of human attitudes that lead to the persecution of species, such as apex predators, that are thought to be dangerous or damaging to human interests (Kansky & Knight, 2014).

The local extirpation of apex predators can release herbivorous prey species and mesopredators (Ritchie & Johnson, 2009; Prugh et al., 2009). Although mesopredators can benefit from anthropogenic subsidies and predation release (Pickett et al., 2001; Prange, Gehrt & Wiggers, 2003; Gaston et al., 2005), they incur substantial risk in human-dominated landscapes (Woods, Mcdonald & Harris, 2003; Baker et al., 2004; Blanchoud, Farrugia & Mouchel, 2004). Risks in urban landscapes include increased vehicle traffic, lethal removals, and domesticated pets (Lowry, Lill & Wong, 2013; Soulsbury & White, 2016). Animals living within urban environments may consequently alter daily activity patterns compared to their rural or forested counterparts due to human activity during diurnal hours (Grinder & Krausman, 2001; George & Crooks, 2006). Individual species within diverse carnivore communities, such as bobcat (Lynx rufus), coyote (Canis latrans), gray fox (Urocyon cinereoargenteus), opossum (Didelphis virginiana), raccoon (Procyon lotor), and striped skunk (Mephitis mephitis), may partition time to maximize resource gain and reduce energy consumption by avoiding human areas during daylight hours (Schuette et al., 2013; Wang, Allen & Wilmers, 2015).

Here we combined camera traps across a landscape gradient of urban, rural, and forested land cover with occupancy models to quantify effects of the multi-scaled landscape features on landscape use by mesopredators, including bobcat, coyote, gray fox, opossum, raccoon, and striped skunk, while accounting for imperfect detection (Lindstedt, Miller & Buskirk, 1986; Crooks, 2002). Within this guild of mesopredators, we expected some species would be human commensals while others would have less affinity towards human development. We expected to classify species as human avoiders, human adapters, or human exploiters (Blair, 1996; Riggio et al., 2018). Species that were positively associated with natural habitats and negatively associated with human developments were classified as human avoiders. Species that were negatively associated with natural habitats and positively associated with urbanization were classified as human exploiters. Species that did not show strong preference for either natural or human habitats were classified as human adapters. We assessed the response of each species to land cover type across three spatial scales (100 m, 500 m, and 1000 m) to determine the most important scale of response. Spatial scale is important when predicting species use (e.g., Gompper et al., 2016), and we expected that smaller species would respond to differences in cover types at a finer scale (Jetz et al., 2004).

Based on previous research, we predicted that the larger-bodied mesopredators (coyote and particularly bobcat) would be human avoiders that were negatively associated with anthropogenic disturbance (Atwood, Weeks & Gehring, 2004; Clare, Anderson & Macfarland, 2015; Lesmeister et al., 2015; Flores-Morales et al., 2019). In many regions raccoon and opossum have been found associated with urban habitats that provide human-subsidized resources and reduced risk from natural predators, so we expected them to be human exploiters in our study area (Prange, Gehrt & Wiggers, 2003; Bateman & Fleming, 2012). We predicted that gray fox and striped skunk would be human adapters with limited affinity for, or avoidance, of anthropogenic landscapes. Striped skunks are known to use human structures as denning sites (Theimer et al., 2017) and are commonly associated with rural human settlements (Lesmeister et al., 2015). Gray fox is commonly described as a species that associates with human development (Riley, 2006; Lesmeister et al., 2015), but that association can vary greatly between study areas and may relate more to site specific factors such as interference competition for prey or the prevalence of free-ranging dogs (Morin et al., 2018). We anticipated modest responses to forests, grasslands, and wetlands among species because all are capable of exploiting a variety of vegetation types, with the exception of raccoon, which we expected would be substantially associated with water cover (Henner et al., 2004).

Materials and Methods

Study area

We conducted a camera trap survey across urban, rural, and forested cover types in and around Corvallis, Oregon, USA (Fig. 1). We considered urban camera stations those that were within the city limits of Corvallis. Camera traps defined as rural were placed at sites outside city limits but remained on lands with human dwellings with the appropriate permission from landowners. The forest cover type was restricted to the MacDonald-Dunn Research Forest that bordered Corvallis to the northwest and contained no human dwellings.

According to the U.S. Census Bureau (2018), Corvallis had a population of 58,641 and covers 54.67 km². The population density was 1,547.2/km², with an average housing unit density of 665.4/km² (US Census Bureau, 2018). Corvallis, and the rest of the Willamette Valley in Oregon, is located in the dry-summer temperate climate zone, or cool-summer Mediterranean (Kottek et al., 2006). The climate was characterized by year-round mild temperatures with warm, dry summers and cool, wet winters. Average yearly precipitation in Corvallis was 110.9 cm and an average high temperature of 63.4°F (17.4 °C) and an average low temperature of 41.9°F (5.5 °C). The average high and low temperatures in January were 47.0°F (8.3 °C) and 33.6°F (0.9 °C) respectively, and average precipitation of 16.4 cm (US Climate Data, 1981-2010). The average high and low temperatures in July were 81.2°F (27.3 °C) and 51.8°F (11.0 °C) respectively, and average precipitation of 1.4 cm (US Climate Data, 1981-2010).

The McDonald-Dunn Research Forest was managed by Oregon State University and open to public recreation. The forest was over 4450 ha and rested in the eastern foothills of Oregon’s coastal mountain range, extending into the western Willamette Valley (Edson & Wing, 2011). Historically, much of the lowland MacDonald-Dunn Research Forest was dominated by oak savannas, but due to disturbance and other environmental conditions the forest was dominated by an overstory of Douglas fir (Pseudotsuga menziesii) (Fletcher et al., 2005). Other common native species include big-leaf maple (Acer macrophyllum), vine maple (A. circinatum), Oregon grape (Mahonia aquifolium), and western sword fern (Polystichum munitum). The McDonald-Dunn Research Forest was subject to four different forest management types (even-aged, two-storied, uneven-aged, and reserved old-growth stands) (OSU Research Forests). The forest was not open to public vehicle access but does receive over 175,000 recreation visits per year.

Results

Our 47 cameras were operational for 11,914 trap nights over the 41-week study period yielding 23,988 images of 14 identified vertebrate species. Columbian black-tailed deer (Odocoileus hemionus columbianus) were the most commonly detected species (n = 5371) at 42 sites. Raccoon was the most commonly detected carnivore (n = 1964) at 29 sites. There were 1680 detections of opossum at 31 sites, 471 detections of gray fox at 19 sites, 415 detections of striped skunk at 18 sites, 171 detections of coyote at 13 sites, and 134 detections of bobcat at 20 sites. We detected cougar (Puma concolor) 30 times at eight sites. Black bear (Ursus americanus) was detected once at a single site.

At urban camera sites, opossum was detected at the greatest proportion (89%) of sites, followed by raccoon at 78% of sites. Gray fox and striped skunk were both detected at 33% of urban sites. Coyote was detected at 22% and bobcat was detected at 17% of the urban cameras. At rural camera sites, raccoon was detected at the greatest proportion of sites, 79%. Bobcat and opossum were detected at 71% of sites while gray fox was detected at 50% of rural sites. Coyote and striped skunk were both detected at 29% of rural sites. In the MacDonald-Dunn Research Forest, striped skunk was detected at the most cameras, 53%. Bobcat and gray fox were detected at 46% and 40% of sites, respectively. Coyote and opossum were both detected at 33% of MacDonald-Dunn Research Forest, sites. Raccoon was detected at the lowest number of forested sites, 27%.

Temporal activity patterns

For our temporal activity analysis, we used only detections that were greater than 30-minutes apart based on the findings of Ridout & Linkie (2009). Our focal species were detected most frequently during nocturnal hours, but the proportion of daylight detections varied by species (Fig. 2). Across all non-human species, 139 of 3580 detections (3.88%) were between 0800–1800. Out of 3589 detections of human, 3119 (86.9%) were between the hours of 0800-1800. Coyote and bobcat, the two largest of the six focal species, were the species most active during the day. These two species were also likely to favor natural areas over urbanized ones. The species that took advantage of urbanization, like raccoon and opossum, had much less daytime activity. Coyote had the most daytime activity relative to the other mesopredator species in our study with 25 out of 136 (18.4%) detections between 0800–1800. Striped skunk and opossum were the least likely to be detected during these hours with fewer than 2% of detections from 0800–1800. In general, the smaller species such as gray fox, opossum, raccoon, and striped skunk had much less diurnal activity overlap with larger species.

Discussion

We explored mesopredator occupancy in human-dominated landscapes across a gradient of urban, rural, and forested environments. Mammalian carnivores can be sensitive to urbanization and have varying responses along an urban to rural gradient (Goad et al., 2014; Salek, Drahnikova & Tkadlec, 2015; Wang, Allen & Wilmers, 2015). The species-level responses to anthropogenic disturbance varied tremendously, with some species being human exploiters tied to urban cover, others being human adapters that were neither positively nor negatively affected by disturbance and agriculture, and human avoider species that were negatively associated with anthropogenic landscape features. The individual species results of our study were largely supported by some established literature from other regions. However, some of our results were contradictory to expectations based on previous research. For example, we found that grassland cover was a more important landscape feature than forest cover for both bobcat and gray fox, whereas in other studies, forest cover was a better predictor of landscape use (Tucker, Clark & Gosselink, 2008; Lesmeister et al., 2015). We found that raccoon and opossum were human exploiters, and in contrast to other studies, striped skunk landscape use was negatively associated with urban development.

We were able to identify opossum and raccoon as human exploiters due to their high use of urban cover and areas near structures (Fig. 3, Table 2). Opossum was common in both urban and rural developments, had negative relationships with forested cover, and increased use in areas with high urban cover density, suggesting that they can exploit much more densely populated areas than other species. Similarly, raccoon occupancy was greatest in urban areas and lowest in forested sites. This is consistent with previous results suggesting that opossum (Markovchick-Nicholls et al., 2008; Wright, Burt & Jackson, 2012) and raccoon (Prange, Gehrt & Wiggers, 2003; Prange, Gehrt & Wiggers, 2004) are competent cohabiters with humans that exploit urban areas. Although we do not have direct data on their population density, the patterns of occupancy that we observed strongly suggest that raccoon and opossum population density is substantially higher in urban areas. The occupancy of opossum and raccoon was strongly associated with human structures, and opossum was positively associated with urban cover even when accounting for proximity to human structures (Table 2). Similarly, opossum had a substantially stronger negative relationship with forest cover, suggesting that raccoon, unlike opossum, were not actively avoiding forested areas (Table 2). Collectively, these results suggest that while both species are associated with humans, opossum have greater affinity for urban environments. Presumably that affinity for urban environments is due to a combination of anthropogenic subsidies (Bozek, Prange & Gehrt, 2007) and reduced predation risk from larger predators (Lesmeister et al., 2015). The association of raccoon and opossum with urban environments was expected, but we did not find the predicted association with natural water sources that was found by others to be an important predictor of opossum and raccoon occupancy (Stuewer, 1943; Gehrt & Fritzell, 1998; Gehrt & Fritzell, 1999; Fidino, Lehrer & Magle, 2016). However, Fidino, Lehrer & Magle (2016) also noted that anthropogenic water sources may allow opossum to colonize previously uninhabitable urban patches. We did not include small human-derived water sources as a covariate in our models because they can be ephemeral and not detectable in land cover databases.

We expected striped skunk to be a human adapted species; but our results suggest that they are human avoiders with higher use of forest cover and low use of human development. Our findings are similar to some studies (Stout & Sonenshine, 1974; Bixler & Gittleman, 2000), but differ from other studies that found striped skunk home ranges oriented along roads and levees and that this species associates with structures for resting and denning (Frey & Conover, 2006; Theimer et al., 2017). Our study area experiences less volatile temperatures compared to those in the study areas of Frey & Conover (2006) and Theimer et al. (2017), which may mean less incentive to use human structures to help in thermoregulation. Dissuading depredation from larger species by their defense mechanism and warning coloration (Walton & Larivière, 1994; Fisher & Stankowich, 2018) may also make forested cover more viable for striped skunk compared to species of similar body size.

As expected based on previous studies (Lesmeister et al., 2015), bobcat tended to be human avoiders in this study with a negative association with urban cover (Fig. 4). Previous studies have shown that when bobcat inhabit urbanized areas their home ranges consist primarily of natural areas (Riley et al., 2003; Riley, 2006). Bobcat appear to decline in response to increasing urbanization proximity and intensity (Tigas, Van Vuren & Sauvajot, 2002; George & Crooks, 2006; Ordeñana et al., 2010). Crooks (2002) suggests that bobcat appear to be particularly sensitive to fragmentation and the probability of bobcat occurrence is greater in larger reserves. In George & Crooks (2006) bobcat was less likely to be detected on cameras with high human activity. Cameras we placed in the Macdonald-Dunn Research Forest were on trails or roads that were commonly frequented by humans. If bobcat was avoiding areas with human activity, we may have observed a greater association with forest cover if we surveyed more forest sites away from human-use areas. We noted that while bobcat use was greater in less developed areas there were some detections in highly developed areas, although these events were infrequent similar to that of Goad et al. (2014).

Our results did not support our prediction that coyote would be human avoiders. Rather, our results suggest that coyote preferred natural areas like grassland and water cover, and urban features had little effect. Thusly, it was difficult to define coyote as solely human avoiders or exploiters, so we classified them as human adapters. We found that coyote were positively associated with unpaved roads, suggesting modest association with the lightest form of anthropogenic disturbance, but coyote was also positively associated with water and forest cover. Other studies have shown coyote frequently exploit urban environments due to their ability to take advantage of anthropogenic features for food, water, and shelter (Chamberlain, Lovell & Leopold, 2000; Kays, Gompper & Ray, 2008; Gehrt, Riley & Cypher, 2010). However, previous research also suggests that coyote need core wildland habitats (Grinder & Krausman, 2001; Gehrt, Anchor & White, 2009). There was modest support that forest cover was a positive predictor for coyote use, but that effect was less important than we predicted when compared to other covariates such as water cover (Riley et al., 2003; Tucker, Clark & Gosselink, 2008; Gehrt, Anchor & White, 2009; Magle et al., 2014).

In our study area, gray fox appeared to be neither strictly human exploiters nor avoiders. Following the precedent set by our coyote results and classification, we classified gray fox as human adapters. In the literature, there are conflicting gray fox results across various landscapes about how the species is positioned on the exploiter-avoider continuum. Results from Riley (2006) suggest that gray fox reach higher densities in and around urbanized areas, whereas Markovchick-Nicholls et al. (2008) found that gray fox were negatively associated with roads, much like our results. Lesmeister et al. (2015) noted that landscape use of gray fox may be driven by both interference competition by bobcat and coyote and the availability of small mammal prey. Our results suggested a negative relationship with human developments, but the support is weak enough that we did not feel it would be appropriate to classify this species as a human avoider in this study area. In contrast to results from Lesmeister et al. (2015), we found no evidence that forest cover was a significant predictor for gray fox occurrence, in either direction, but in our study area, free-ranging dogs were not as prevalent as in southern Illinois, which may have a depressing effect on gray fox landscape use (Morin et al., 2018). We observed that grassland cover was positively associated with gray fox occurrence, and previous studies found increased abundance of gray fox in areas with high levels of fragmentation of forested and grassland cover (Fuller & Cypher, 2004; Cooper, Nielsen & McDonald, 2012). Corvallis and the surrounding landscape have moderate levels of fragmentation from urban development, timber harvesting, and agricultural fields.

We have characterized the habitat associations of mesocarnivores, but mesocarnivore occupancy can also be influenced by competition and predation. We would be remiss to overlook the importance of cougars as the apex predator in carnivore communities. Throughout our survey effort, there were only 30 detections of cougar, and they were largely confined to the MacDonald-Dunn Research Forest. Based on previous studies (Chamberlain, Lovell & Leopold, 2000; O’connell et al., 2006; Cove et al., 2012; Lesmeister et al., 2015; Gompper et al., 2016), we expected forest cover to be a stronger predictor of landscape occupancy for mesopredators, but many of these studies occurred in areas without cougar where forested areas posed less risk. Species interactions with coyote may also be influencing the distribution of gray fox, which appear to be avoiding forested areas where coyote were more common to reduce intraguild conflict and interference competition (Temple, Chamberlain & Conner, 2010; Levi & Wilmers, 2012; Lesmeister et al., 2015). In another urban-rural carnivore study, raccoon distribution was negatively affected by coyote distribution while being positively correlated with predictors such as road density and distance to urban center (Randa & Yunger, 2006).

Our findings supported the predictions that mesopredators would have limited activity during daylight hours. Previous studies have also documented increased nocturnal activity by mesocarnivores (Riley et al., 2003; Wang, Allen & Wilmers, 2015). There appears to be a distinction between the activity patterns of species that preferred natural areas versus urban areas. However, we were unable to determine if this was primarily driven by body size or human activity, or if both were contributing factors. Both bobcat and coyote appeared to keep their circadian patterns as detections of them were common around dawn and dusk. Species with the most urban use, raccoon and opossum, showed less daytime activity (0600-1800) presumably to avoid increased human activity in these areas during these hours. Many studies (Larivière & Messier, 1998; Neiswenter, Dowler & Young, 2010) have documented the nocturnal nature of striped skunk, and that was reflected in our results as well.

Conclusions

As urbanization continues, it is important to understand how a species might respond to increased urban development and the factors that might put a species at increased risk of decline to avoid the decline of once-common carnivores that has been observed in other species (Purvis et al., 2000; Gompper & Hackett, 2005; Gaston & Fuller, 2007). The taxa in this study have large ranges spanning much of North America, yet it is common for studies in different regions to observe a variety of species-specific relationships with land cover and human development. We provide information on how some common mesocarnivores use landscapes across a gradient of urban to rural to forested environments in the context of the vertebrate community and forest type of the mesic Pacific Northwest forest of the United States. Our results differed from several studies of this guild in other regions of North America, highlighting the importance of ecological context for landscape use along a gradient of urban to natural land cover. The level of human tolerance and species interactions may have also influenced our observed landscape use. For example, gray fox was typically associated with forest and grassland cover, but competitors such as bobcat (Lesmeister et al., 2015) and interference competition in areas of high coyote abundance can influence gray fox distribution (Crooks, 2002; Farias et al., 2005; Temple, Chamberlain & Conner, 2010; Ordeñana et al., 2010). Our results inform how mesopredator communities respond to anthropogenic disturbance in a distinct context, which together with other research can help elucidate the mechanisms underlying their observed distributions and inform their conservation and management. Management actions that impact these species are likely to be most effective if based on locally derived information.

Supplemental Information