Mammal Activity and Species Richness is Reduced in Areas with Recreationists, in Palos Verdes Peninsula Habitat Fragments Open Access

As the human population continues to grow, urban development is reducing available habitat for wildlife (George and Crooks 2006). Los Angeles County, California, USA, is one location where a growing urban area is encroaching into natural landscapes. Due to the proximity of residential developments and natural open spaces, many people access wildlife habitat in and around Los Angeles County. Thus, Los Angeles County is an ideal place to study the effects of urbanization and human presence on wildlife species (Tigas et al. 2002; Ordenana et al. 2010; George and Crooks 2006; Riley et al. 2003; Benson, Sikich and Riley 2016; Ng et al 2003). The Palos Verdes Peninsula (PVP), a coastal landform southwest of the center of the city of Los Angeles, provides a unique opportunity to augment previous studies of the urban wildlife of the Los Angeles area (Riley et al. 2003; Ng et al. 2004; Ordeñana 2010), as the mammal species present have not been surveyed and their activity patterns in relation to humans have not been studied. Sage scrub canyons with varying levels of human recreation in PVP are a natural experimental system to study effects of recreation on wildlife habitat within an urban matrix.

Many non-consumptive recreationists believe they have little to no impact on wildlife species, but several studies indicate that this assumption is incorrect (Larson et al. 2016; Taylor and Knight 2003). Recreational disturbances, such as trail creation and use, can cause harm to wildlife by direct mortality, habitat alteration, or by altering wildlife behaviors (Kays et al. 2017; Reilly et al. 2017). Outdoor recreation in these habitat areas places humans and wildlife in a space where interaction is possible, altering the behavior of wildlife, changing the temporal and spatial distribution of mammals and putting recreationists at risk of harm (Benson et al. 2016; Kays et al. 2017). For example, coyotes restrict activity times in areas with high recreation (George and Crooks 2006; Benson et al., 2016) and the presence of dogs along trails reduces small mammal activity (Lenth et al. 2008). Recreational trails in habitat fragments throughout the PVP are used by hikers, dog-walkers and mountain bikers in addition to wildlife and provide opportunities for interactions. Land managers of the PVP and other urban habitat fragments can create better informed management plans if the impacts of recreationists on mammal wildlife species, such as temporal and spatial alterations in wildlife behavior, are known (Larson et al. 2016; Riley et al. 2016; Kays et al. 2017).

Additionally, interactions between humans and wildlife can lead to conflict that puts humans and wildlife at risk of harm (White and Gehrt 2009). Los Angeles County has one of the highest sympatric populations of humans and coyotes in the United States (Elliot et al. 2016). The coyote (Canis latrans) is the largest bodied mammalian predator on the PVP, making it a key urban focal study species to determine risk of conflict and is of concern to PVP residents (White and Gehrt 2009; Ordeñana 2010; pers. obs.). Wildlife and recreationists must be both present in the same place and time for a conflict instance to occur. However, some wildlife species, such as coyotes, have adapted well to urban areas through behavioral shifts towards nocturnality, while most recreation occurs diurnally, possibly limiting interactions (Grinder and Krausman 2015; Gehrt 2011; Ordeñana 2010).

This study aims to identify which mammal species are present in two habitat fragments in PVP, assess how mammal community activity and species richness are potentially affected by human recreation, and evaluate the possibility of interactions between mammal wildlife species and recreationists. I addressed these questions by placing game cameras in protected PVP canyons of sage scrub habitat fragments of similar size, shape, and isolation, but with or without recreation. I hypothesized that (I) canyons with recreation will have a reduced number of mammal species detected compared to those without, (II) canyons with recreation will have reduced mammal species activity measured by detection rate, (III) mammals in the recreation canyons will display more nocturnal activity while recreationists will be present during the day. The goal of this study is to inform natural resource managers of the PVP of Los Angeles, California and provide baseline information to initiate more complex studies of wildlife in urban matrices using the PVP.

Mammalian responses to recreation were studied in two undeveloped canyons surrounded by suburban development (Margate and Agua Amarga Canyon), situated in the PVP, a coastal suburban landform located in the southwest corner of Los Angeles County, CA USA (Fig. 1). Both Margate and Agua Amarga Canyons have steep hillsides with a dry creek bed running the length of each canyon from west to east. Margate and Agua Amarga canyon are separated by two km and are located on the western side of the PVP (Fig. 1). The dominant plant species in both canyons include native chapparal plants such as lemonadeberry, Rhus integrifolia and California sagebrush, Artemisia californica, in addition to the invasive common fennel, Foeniculum vulgare, and black mustard, Brassica nigra. Human presence in Margate Canyon is limited to a municipal services access trail. Agua Amarga Canyon, however, has an established trail for recreation running parallel to the creek bed on one side of the canyon and experiences high levels of hiking, bicycling and walking of domestic dogs, Canis familiaris. These similar canyons serve as effective sites to study the effects of recreation within urban fragments on the behavioral ecology of mammals due to the disparity in human activity between canyons. The Palos Verdes Peninsula itself is an excellent system for studying urban wildlife in open space surrounded by urban matrix as it has over 25 open space areas of varying visitor-use levels, size, and isolation that are all separated from any large contiguous areas of habitat by the heavy development of the Los Angeles metropolitan area.

Game cameras (Bushnell Model #119537, Overland Park, Kansas, USA) were installed in pairs along a perpendicular axis of both canyon's length with one camera in the creek bed and one camera on the trail to monitor mammal and recreation usage. One camera pair was located near the western mouth of both canyons and another pair was located further east in both canyons relatively near a point of access, for a total of 8 camera installations between the two canyons. Given camera limitations, subsampling the two canyons with four cameras for within-site replication was chosen over more canyons with lower within-site replication, reducing the importance of the placement of an individual camera within the canyons.

Cameras were installed at (33.767418, -118.407337 Agua Amarga Western Creek), (33.766818, -118.406139 Agua Amarga Western Trail), (33.767926, -118.400342 Agua Amarga Eastern Creek), (33.767579, -118.400528 Agua Amarga Eastern Trail), (33.785458, -118.409238 Margate Eastern Creek), (33.785237, -118.408921 Margate Eastern Trail), (33.785000, -118.411250 Margate Western Creek), and (33.784798, -118.411085 Margate Western Trail). Camera locations were selected at natural constrictions in the canyons to reduce the ability of mammals to travel past the camera without being detected and were placed in areas with vegetation communities representative of the canyon in which they were placed. Cameras were angled downwards and along the length of the travel path to minimize the recording of public visitors' faces while maximizing mammal and recreationist detections. Placing game cameras along trails instead of random placement has been shown to not affect community richness or composition conclusions given sufficient sampling effort (more than 1400 trap nights) (Cusack 2015). Cameras in this study were deployed for 1503 camera trap nights (i.e., 24-hr periods per camera). Cameras were deployed from 3 May 2017 to 18 November 2017. This project was permitted through the City of Palos Verdes Estates (permit number 023252) and given permission to proceed by the Palos Verdes Peninsula Land Conservancy when on Conservancy land.

Cameras were set to record a 30 sec video per motion detected trigger with a delay of 10 sec between triggers. Batteries and SD cards were replaced biweekly, and videos were manually sorted as mammal detections, recreation detections (human or domestic dog) or no detection. A detection was categorized as at least one mammal in frame; mammals were not double counted if they left and re-entered the frame in the same video, and multiple individuals in the same video counted as multiple detections. The same protocol was used for scoring recreation detections. The number of detections per species at each camera location was recorded for each biweekly period, as day (color images) or night (infrared images). This ensured that the cutoff between day and night was as consistent as possible throughout changing seasons. Additionally, this method may be more effective than using time to designate night and day for studies utilizing game cameras through multiple seasons as it is reliant on the intensity of light itself.

Detections from the western and eastern cameras for creek bed and trail were combined by canyon in the analysis to reduce the effect of individual camera placement within each canyon. For example, data from the Margate western and eastern creek bed cameras were combined to create reference creek bed in the final dataset. Additionally, for some analyses data from all cameras in each canyon were combined to create the recreation and reference categories. To control for uneven sampling between sites due to camera failures, I used the minimum number of trap days without failures, 161 d to scale the data from each camera. Average trap days per camera was 188 trap d with a standard deviation of 15 d.

I compared the total number of dog and human detections (‘recreation’) between the two canyons and the number of total wildlife detections between canyons using a Chi-square test (α = 0.05) as detections were independent, mutually exclusive counts. I compared the proportion of detections that were nocturnal between the recreation and reference canyons using a t-test (α = 0.05). I compared the mammal average species richness for every two-week segment (each camera deployment) of the full data collection period between the two canyons using a t-test (α = 0.05).

I recorded 652 detections of mammalian wildlife species and 8,117 detections of domestic dogs and humans between both canyons. Only 29 human photos were recorded over a 200-d period in the reference canyon, while a total of 8,088 human and domestic dog detections were recorded in the recreation canyon, demonstrating the canyons were effective treatments for recreation (χ² = 8001, p < 0.01). The high number of dog detections, 5116, in the recreation canyon is partially due to one dog walker who walked 8-12 dogs past the Agua Amarga Western Creek bed camera multiple times per week.

The reference canyon had a total species richness count of 9 species and mean of 2.46 species observed per two-wk period. The recreation canyon had 6 species total and a mean of 0.54 species detected per two-wk period. There was a lower species richness per two-wk data collection period in the recreation canyon (t = 5.69, p < 0.01). Species observed in all sites were coyote (Canis latrans), feral cat (Felis catus), raccoon (Procyon lotor), desert cottontail rabbit, (Sylvilagus audubonii), striped skunk (Mephitis mephitis), and fox squirrel (Sciurus niger) (Fig. 1). Mammals detected in only the reference canyon were the Virginia opossum (Didelphis virginiana), black rat (Rattus rattus), and red fox (Vulpes vulpes). There were no species detected only in the recreation canyon. Two mesopredators, raccoons and feral cats accounted for the majority of detections (41% and 26%, respectively). Black rat was the third most detected species (12%), followed by desert cottontail (7%), striped skunk (5%), opossum (4%), fox squirrel (3%), coyote (1%), and red fox (1%).

The number of mammal detections was 13.5 times greater in the reference canyon as compared to the recreation canyon. Coyotes were observed in both canyons with no significant difference in activity between the two canyons, (χ² = 0.14, p = 0.71). All other mammal species displayed higher activity in the reference canyon (p < 0.05; Table 1). Raccoon and striped skunks showed the greatest difference in activity between canyons, with 28 and 31 times more detections in the reference canyon, respectively. Nocturnal mammal activity occurred much more frequently than diurnal activity overall, (χ² = 271.42, p < 0.01; Table 1) with 537 nocturnal mammal detections compared to 116 mammal detections during the day. Black rat, Virginia opossum, red fox and striped skunk were only detected at night. Fox squirrel was the only strictly diurnal mammal. Raccoon detections occurred at night except for four daytime detections in the reference canyon and were thus more nocturnal (reference canyon: χ² = 187.32, p < 0.01) and (recreation: χ² = 8.00 p < 0.01). Feral cats were detected at night more often than during the day in both canyons, (recreation: χ² = 14.22, p < 0.01) and (reference: χ² = 10.8, p < 0.01). For desert cottontail rabbit, diurnal activity was more frequent in the reference canyon (χ² = 11.57, p < 0.01), while in the recreation canyon there was no significant difference in time of activity (χ² = 0.66, p = 0.41). Coyotes displayed greater nocturnal activity in the recreation but not in the reference canyon, (χ² = 4.00, p = 0.045 and χ² = 0.33, p = 0.56 respectively). The only daytime detection of a coyote occurred along the trail in the reference canyon. The only non-solitary coyotes detected were a pair in the recreation canyon along the trail at night. Overall, 99.5% of all recreationist activity occurred during the day while 82% of mammal detections occurred at night. Between the two canyons, there was no difference in the proportion of all mammal detections that occurred at night (t = 0.12, p = 0.906).

As human development continues to encroach on natural open spaces, human recreation demand and the habitat needs of wildlife will continue to overlap, necessitating knowledge on how to reduce wildlife-recreationist conflict. I detected reduced mammal activity and fewer mammal species where recreation was prevalent, building upon previous similar findings in other regions (Fig. 2; Tucker et al. 2018; Larson et al. 2016). There was 13.5 times the amount of mammal activity in the reference canyon, suggesting a reduction of animal activity due to human presence, a trend found in a global meta-analysis across 57 mammal species (Tucker 2018). Reductions in activity due to human presence could reduce the likelihood of animals moving between fragments, possibly leading to reduced genetic diversity in these small remnant populations (Tucker 2018). Untangling whether this reduction in activity is due to reduced population sizes or reduced levels of individual activity is warranted for future investigation.

The mammal species I detected were a relatively even combination of native and nonnative species, representing a novel mammal assemblage. For example, while the gray fox, Urocyon cinereoargenteus is native to Southern California and present in other nature preserves in the PVP the non-native red fox was present in both study canyons. Red foxes could fill the same functional role as gray foxes or lead to novel biotic interactions. Uncertain ecological consequences make determining the conservation outlook for these novel assemblages challenging (Kowarik 2011). Furthermore, these novel mammal assemblages can alter remaining native vegetation and wildlife communities in these habitat fragments (Chase 2006).

Raccoons and feral cats represented 67% of wildlife detections in the reference canyon and had higher activity in the reference canyon compared to the recreation canyon. Feral cat populations are especially concerning as they have been shown to be effective predators of species of conservation concern (Fitzgerald 1979; Chase 2006). In Agua Amarga Canyon, the possible increase in feral cat predation could be harmful to local populations of cactus wren, Campylorhynchus brunneicapillus, a species of local conservation concern (Dalkey 2016) and California gnatcatcher, Polioptila californica, a federally threatened species and California species of special concern (CDFW 2020). Thus, mesopredators are contributing the majority of detections in these canyons and are an important group to understand to better manage urban preserves for native songbird persistence (Crooks 2002; Crooks and Soulé 1999; Cove et al. 2014).

Wildlife diel activity patterns did not differ significantly from expected patterns for each species and detections were dominated by nocturnal and crepuscular species. Thus, serious human-wildlife conflict seems unlikely in the study canyons, as 82% of wildlife detections occurred at night while 99.5% of recreationist detections occurred during the day and 98% of wildlife detections were either raccoons or a species of smaller body size (Fig. 3).

This study utilized the natural exclusion of human recreation in the reference canyon to investigate how recreation affects the species richness and activity patterns of mammals. I acknowledge strong limitations to this preliminary study due to limited resources. Thus, the sample size of two canyons and eight cameras limits the generalizability of these results to urban wildlife and recreation generally. Sampling many areas with a gradient of human activity would reduce the likelihood that observed effects were due to some intrinsic characteristic of the canyon as opposed to human activity level. Other differences between the two canyons that could be alternative drivers of the observed patterns in mammal species and activity detected include active habitat restoration work in the recreation canyon and the slightly larger size of the recreation canyon. Longer sampling periods allowing for stronger analyses at the species level would reduce the dominating effect of the high number of feral cat and raccoon detections. Given these limitations, my conclusions are preliminary but consistent with previous findings on reduced species richness (Larson et al. 2016) and activity (Tucker et al. 2018) from human activity. Lastly, this study represents the first published study of urban wildlife in the PVP and there is significant possibility for very robust studies of urban wildlife using PVP open space areas.

Palos Verdes Peninsula has strong potential to serve as a study system for wildlife in open space, as there are more than 25 open space areas of varying size and isolation. Sampling many open space areas and utilizing techniques such as occupancy modeling to account for imperfect detection (Mackenzie et al. 2002) would improve the strength of inference of future related studies. Questions that should be investigated in PVP preserves include how wildlife use areas with and without recreation within the same urban preserve and if they travel between reserves to avoid human activity (Atwood et al. 2016), to better understand how to reduce impact on mammal species without excluding recreationists from the last open spaces left in the PVP.

Mammal species richness and activity were reduced in a canyon with recreation as compared to a canyon without recreational activity in the PVP. The mammal communities of both canyons represent a novel assemblage of both native and non-native mammals. Mesopredators were by far the most detected group of mammals, which is concerning given the presence of two songbirds of conservation concern in the PVP. This preliminary study is limited by sampling constraints but is consistent with previous findings on urban wildlife responses to human recreation. The PVP, a matrix of habitat fragments and development isolated from large protected natural areas, provides a suitable system for understanding how wildlife persist within an urban-habitat matrix.

Dr. Theodore Stankowich and Rita Collins provided equipment. Tommy Jankowski and Brandon Nash provided support as field assistants. Dr. Travis Gallo and Dr. Christy McCain were both important help with the manuscript. I recognize the Palos Verdes Peninsula Land Conservancy for allowing me to place cameras on Conservancy land.