https://cpw.cvlcollections.org/files/original/734b9cb7cd8f0d232210cd477692230d.pdf 33bd2f2bfc1b2050077986138982e560 PDF Text Text The research in this publication was partially or fully funded by Colorado Parks and Wildlife. Dan Prenzlow, Director, Colorado Parks and Wildlife • Parks and Wildlife Commission: Marvin McDaniel, Chair • Carrie Besnette Hauser, Vice-Chair Marie Haskett, Secretary • Taishya Adams • Betsy Blecha • Charles Garcia • Dallas May • Duke Phillips, IV • Luke B. Schafer • James Jay Tutchton • Eden Vardy �Received: 27 February 2018 | Revised: 6 June 2018 | Accepted: 26 June 2018 DOI: 10.1002/ece3.4382 ORIGINAL RESEARCH Sharing the same slope: Behavioral responses of a threatened mesocarnivore to motorized and nonmotorized winter recreation Lucretia E. Olson1 | John R. Squires1 | Elizabeth K. Roberts2 | Jacob S. Ivan3 | Mark Hebblewhite4 1 Rocky Mountain Research Station, United States Forest Service, Missoula, Montana Abstract 2 Winter recreation is a widely popular activity and is expected to increase due to White River National Forest, United States Forest Service, Glenwood Springs, Colorado 3 Colorado Parks and Wildlife, Fort Collins, Colorado 4 Wildlife Biology Program, Department of Ecosystem and Conservation Sciences, W.A. Franke College of Forestry and Conservation, University of Montana, Missoula, Montana Correspondence Lucretia E. Olson, Rocky Mountain Research Station, United States Forest Service, Missoula, Montana. Email: lucretiaolson@fs.fed.us Funding information Rocky Mountain Research Station; Vail Associates Inc.; Colorado BLM state office; U.S. Forest Service R2 Regional Office Renewable Resources Department; 10th Mountain Huts; Colorado Department of Transportation changes in recreation technology and human population growth. Wildlife are frequently negatively impacted by winter recreation, however, through displacement from habitat, alteration of activity patterns, or changes in movement behavior. We studied impacts of dispersed and developed winter recreation on Canada lynx (Lynx canadensis) at their southwestern range periphery in Colorado, USA. We used GPS collars to track movements of 18 adult lynx over 4 years, coupled with GPS devices that logged 2,839 unique recreation tracks to provide a detailed spatial estimate of recreation intensity. We assessed changes in lynx spatial and temporal patterns in response to motorized and nonmotorized recreation, as well as differences in movement rate and path tortuosity. We found that lynx decreased their movement rate in areas with high-­intensity back-­country skiing and snowmobiling, and adjusted their temporal patterns so that they were more active at night in areas with high-­intensity recreation. We did not find consistent evidence of spatial avoidance of recreation: lynx exhibited some avoidance of areas with motorized recreation, but selected areas in close proximity to nonmotorized recreation trails. Lynx appeared to avoid high-­ intensity developed ski resorts, however, especially when recreation was most intense. We conclude that lynx in our study areas did not exhibit strong negative responses to dispersed recreation, but instead altered their behavior and temporal patterns in a nuanced response to recreation, perhaps to decrease direct interactions with recreationists. However, based on observed avoidance of developed recreation, there may be a threshold of human disturbance above which lynx cannot coexist with winter recreation. KEYWORDS anthropogenic disturbance, Lynx canadensis, ski resorts, snowmobiles, space use, winter recreation This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2018 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. Ecology and Evolution. 2018;8:8555–8572. � www.ecolevol.org | 8555 �8556 | 1 | I NTRO D U C TI O N OLSON et al. 2013), also differ from dispersed recreation, which requires little infrastructure and minimally affects existing forest conditions. Winter recreation is an important economic contributor to com- The number of participants and total days spent on winter recre- munities in temperate or subarctic regions (Töglhofer, Eigner, & ation is projected to increase over the next several decades (White Prettenthaler, 2011; White & Stynes, 2008; Zhang, Cai, & Ni, 2006). et al., 2016), and coupled with the potential of climate-­induced re- Due to technological advancements in snowmobiles, back-­country duction in persistent and deep snow, research is needed to charac- skis and other outdoor equipment coupled with growing human terize the effects of winter recreation on endangered or threatened populations, the footprint of winter recreation continues to expand. species that are snow-­associated (Larson et al., 2016). Canada lynx Persistent snow-­covered areas, however, are decreasing in spatial (Lynx canadensis), a threatened species in the continental United and temporal extent due to climate change, which will likely result States, is of particular concern because of its relative rarity as well in increased recreation intensity in the areas that remain (Brammer, as its adaptation to and reliance on deep snow to limit competition Samson, & Humphries, 2015; Elsasser & Messerli, 2001; Scott, from other predators during winter (Buskirk, Ruggiero, & Krebs, Dawson, & Jones, 2008). Increased human disturbance to species 1999). Winter recreation may cause increased energy expenditure already stressed by a changing climate may exacerbate negative ef- if lynx are repeatedly disturbed, as well as lost hunting opportuni- fects (Hughes, 2003; Riordan & Rundel, 2014). Therefore, a better ties since lynx are a stalking, sit-­and-­wait predator (Nellis & Keith, understanding of the response of animals to winter recreation in 1968). Western Colorado, USA is an excellent study location for this these ecosystems is critical. question, with an abundance of both dispersed and developed rec- While recreational use of an area is generally assumed to be reation, as well as a resident lynx population. The Vail Pass Winter more compatible with species’ conservation than consumptive ac- Recreation Area in Colorado has 50 miles of established groomed tivities such as development or resource extraction, animals’ per- trails as well as a ski-­hut system for dispersed recreation; this area ceived risk from recreation can lead to behavioral tradeoffs such is subject to intense recreation and receives roughly 35,000 visitors as increased vigilance and decreased feeding, mating, or parental per winter (U.S.D.A. Forest Service 2015). Colorado also has 30 de- care activities (Frid & Dill, 2002; Larson, Reed, Merenlender, & veloped ski resorts (National Ski Areas Association 2016) which co- Crooks, 2016). Snow-­based recreation may also have a greater neg- incide with lynx distribution and received roughly 13 million visitors ative affect on wildlife compared to aquatic or summer-­terrestrial in 2016 (Blevins, 2016). sports (Larson et al., 2016), with changes in space or temporal use The goal of our study was to understand the impact of winter of an area frequently observed. Moose (Alces alces) and mountain recreation on Canada lynx. We used the movement ecology para- caribou (Rangifer tarandus caribou), for example, were spatially dis- digm (Nathan et al., 2008) to frame our investigation of winter rec- placed from suitable habitat by the presence of snowmobile rec- reation impacts on lynx behavior. Specifically, we examined (a) the reation (Harris, Nielson, Rinaldi, & Lohuis, 2013; Seip, Johnson, & impact of dispersed recreation intensity on various metrics of lynx Watts, 2007), while mountain goats (Oreamnos americanus; Richard behavior, (b) the extent to which lynx spatially and temporally avoid & Côté, 2016) and black grouse (Tetrao tetrix) avoided developed dispersed recreation, and (c) the extent to which lynx spatially and ski areas (Patthey, Wirthner, Signorell, & Arlettaz, 2008). Behavioral temporally avoid high-­intensity developed recreation. We exam- responses such as changes in activity or movement have also been ined behavioral metrics including movement speed and movement observed; for instance, moose in Wyoming remained bedded or tortuosity that we hypothesized might be influenced by recreation fed less frequently in response to snowmobile activity (Colescott and could lead to increased energy expenditure or reduction in & Gillingham, 1998). hunting success. We hypothesized that lynx would increase speed Impacts of recreation on animals can also vary depending on and decrease tortuosity if disturbed by recreation, in an effort to whether activities are motorized or nonmotorized, dispersed or de- spend as little time as possible in areas with more recreation and veloped, and low or high intensity. While motorized recreation is fre- as a flight response to disturbance (Stewart et al., 2016). We also quently considered a source of disturbance (Goldstein, Poe, Suring, hypothesized that lynx would adjust their space use to avoid areas Nielson, & McDonald, 2010; Olliff, Legg, & Kaeding, 1999), nonmo- with high recreation intensity, or their temporal habits to avoid ac- torized recreation has also been shown to elicit negative responses tivity during high recreation-­intensity daylight hours, if disturbed in animals, and may even do so to a greater degree than motorized by winter recreation. recreation (Harris et al., 2013; Larson et al., 2016; Stankowich, 2008). Although snowmobiles generate high noise levels, they may be perceived as less of a threat than human voices by species con- 2 | M E TH O DS ditioned to fear persecution (Bowles, 1995). Additionally, many species are able to seek isolated refugia from snowmobilers, whereas nonmotorized recreationists may access remote areas with higher 2.1 | Study area elevations, dense canopies, and nongroomed trails (Olson et al., Our study area was located in western Colorado, USA, at two 2017). Developed ski resorts, which include considerable infrastruc- locations of high recreation activity (Figure 1). The northern Vail ture, tree removal, and continuous maintenance (Rixen & Rolando, Pass study area was on public lands administered by the White �| OLSON et al. River National Forest and the San Isabel National Forest, in the northern Sawatch and Mosquito mountain ranges (approximate centroid coordinates 106.30°W, 39.45°N). The San Juan study 8557 2.2 | Quantifying movements of winter recreationists area was on public lands administered by the Uncompahgre and To provide a spatially and temporally detailed sample of winter San Juan National Forests, and the Bureau of Land Management, recreation intensity, we stationed technicians at parking areas and in and around the towns of Silverton and Ophir (approximate trailheads during winters of 2010–2013 to distribute small (5 × 8 cm) centroid 107.88°W, 37.82°N). Winter recreation occurred on GPS units to be worn around the upper arm or affixed to a back-­pack both sites between end of December and early April, at eleva- (Qstarz International Co., Ltd., model BT-­Q1300, Position accuracy tions of 2,000 m to 4,300 m and with approximately 380 cm to <10 m). Recreationists were informed that participation was volun- 1,000 cm of annual snowfall (National Oceanic and Atmospheric tary and no personally identifiable information was collected, and Adminstration 2017). offered a map of their day’s movement as incentive to carry the GPS Recreation at the Vail Pass location is intense, and includes ap- unit; response from recreationists was positive, with an acceptance proximately 35,000 visitors per winter season, the majority of which rate of approximately 90% (Miller, 2016). We recorded four types are motorized and concentrated along groomed trails or suggested of recreation activity (snowmobile-­assisted ski [hereafter hybrid], routes, while approximately 11,000 are packed-­trail cross-­country snowmobile, back-­country ski or snowboard, and packed trail cross-­ skiers and snowshoers using the 10th Mountain Hut back-­country country ski or snowshoe [hereafter packed-­trail ski]). Only one GPS hut system (U.S.D.A. Forest Service 2015). A developed ski resort unit was given to groups with multiple people to ensure independ- is also near the Vail Pass study area; this resort is among the top ence among recreation tracks, although some people and/or groups 10 largest ski resorts in Colorado. It encompasses approximately may have been sampled more than once during the course of a win- 10.1 km2 of skiable area and has 23 chairlifts (USDA Forest Service ter, or across winters. Locations of recreationists were recorded at National Visitor Use Monitoring Results, 2016). In the San Juan study 5-­s intervals. If GPS units remained stationary, further locations area, recreation is primarily dispersed back-­country ski and snow- were not collected until the device detected movement. Detailed board use, with some motorized recreation concentrated primarily descriptions of our methods to quantify movements of recreation- in high-­elevation areas. The developed ski resort we focused on in ists can be found in Miller, Vaske, Squires, Olson, and Roberts (2016) the San Juan study area has 8.1 km2 of skiable area and 18 chairlifts and Olson et al. (2017). (telluride.com, 2017). To quantify recreation intensity, we converted GPS point locations into density rasters using the Point Density tool in ArcGIS (ESRI, Redlands, CA, USA). To determine the best scale at which to look for lynx response to recreation, we considered GPS point densities in a circular neighborhood at radii of 30 m, 100 m, 500 m, and 1 km, chosen to reflect arbitrary distances at which lynx could reasonably be expected to respond to the sight and sound of recreationists. Upon examining the distribution of the data, only the 1 km scale had enough nonzero values to allow accurate estimation of regression parameters for all lynx (Kutner et al., 2005). We therefore considered only the 1 km scale for all analyses, and our conclusions are relevant to how lynx perceive recreation within 1 km distances. Density rasters were calculated separately for each recreation type and year (2010–2013). To ensure that lynx response was temporally coincident with recreation, we matched year of recreation intensity with year of lynx data collection; we did not attempt to temporally match recreation and lynx movement at any finer resolution than year due to limitations in sample size when lynx and recreation were paired by day or week. Thus, our analysis is prefaced on the assumption that lynx response is to seasonal intensity of recreation, rather than to direct presence of recreationists. We also deployed infrared and magnetic trail counters at recreation portals and trail crossings throughout the study area to provide an index of recreation intensity on lynx home ranges independent F I G U R E 1 Location of Canada lynx and recreation study areas in western Colorado, USA. Canada lynx home ranges are shown in white, recreation 100% minimum convex polygons are shown in dark gray. Inset shows location of Colorado in United States of GPS tracks. Counters were in place between January and March, infrared counters affixed to trees approximately 1.5 m above the snow, and magnetic counters buried beneath the snow in trail center. Counter data was summed across the entire season and divided by �8558 | OLSON et al. the number of days each counter was deployed to provide an index environmental and recreation covariates, which were averaged sep- of use measured as hits/day for each counter. arately for day and night periods. We first summarized lynx movement into temporal day (~0800:1700 hr) and night (~1700:0700 hr) periods using the “sunriset” command in R package “maptools”, 2.3 | Lynx data collection which calculates the changing actual sunrise and sunset times for We trapped lynx in areas of high recreation or proximity to ski re- each day based on a given geographic location (Bivand & Lewin-­Koh, sorts where previous survey work indicated they were present. Lynx 2016). We considered two movement metrics as response variables were captured using a specially designed box trap (Kolbe, Squires, & for this analysis: movement rate (distance traveled [km] per unit time Parker, 2003) to minimize probability of injury. Traps were checked [hr] of only “active” points averaged across temporal period) and daily and animals were handled in accordance with International movement tortuosity (straight line distance from first to last point in Animal Care and Use Committee (IACUC) permit AUP-­062-­ a temporal period divided by summed distance between all points in 13MHWB-­122013. Adult (≥3 years) lynx were fitted with Sirtrack a period; Benhamou, 2004). satellite store on board GPS collars (210–230 g) with conventional Environmental predictor variables included proportion forest/ VHF radio transmitters and a drop-­off mechanism. Collars were pro- nonforest, a binary variable based on landcover categories from grammed to obtain a GPS location every 20 min, 24 hr per day in the National Land Cover Database (NLCD; Homer et al., 2015), av- 2010, 2012, and 2013 and every 30 min, 24 hr per day, every other eraged across all locations in a temporal period (mean: 0.9, range: day in 2011. We considered the potential for scale dependency is- 0.1–1.0), canopy cover (percent per pixel tree canopy density, sues between collars with different fix-­rates (i.e., 20 min vs 30 min; mean: 44.8%, range: 14.0%–67.4%, NLCD; Homer et al., 2015), Pépin, Adrados, Mann, & Janeau, 2004); however, movement rate and Euclidian distance to forest edge (shortest straight line dis- (step-­length [km]/step duration [hr]) between collars with different tance from a lynx GPS point to an NLCD forest landcover category, duty cycles were similar (mean rate 30-­min duty cycle = 0.33 km/ mean: 141.8 m, range: 7.4–711.9 m). Recreation variables included hr, 95% CI = −0.24–0.91 km/hr, n = 3 lynx-­years; 20-­min duty 1 km recreation intensity for the four types of recreation (hybrid, cycle = 0.39 km/hr, 95% CI = 0.12–0.66, n = 17 lynx-­years), and thus backcountry ski, snowmobile, packed-­trail ski), and indicator vari- we felt confident in treating all collars similarly for analysis. Average ables for weekday/weekend, day/night, study area (San Juan/Vail), fix-­rate across collars was 84%; collars were programmed to auto- and sex. All pairwise correlations between predictor variables matically drop off after June 1st. were <0.60, and variance inflation factor was <2.0, indicating no We focused on lynx movement ecology (Nathan et al., 2008) col- multi-­collinearity. lected during peak winter recreation to maximize the potential to As a first step in the model-­building process, we considered measure responses to disturbance (January–March). We evaluated three vegetation-­only models (i.e., single-­variable models using en- movements within 95% minimum convex polygon (MCPs) use areas vironmental covariates listed above) fit to each of the two response for each lynx and excluded movements outside of these areas since variables to control for the influence of habitat on lynx behavior. We animals performing exploratory movements may differ in behavior selected the best performing vegetation model for each response from those on stationary home ranges (Abrahms et al., 2017; Pépin, variable and carried this base habitat model into the second step Adrados, Janeau, Joachim, & Mann, 2008). To split lynx movement where we considered 10 candidate models that we hypothesized into relevant behavior categories, we categorized lynx GPS locations would test the influence of winter recreation intensity on lynx be- as “active” or “stationary” using parameters determined from five havioral response (Supporting Information Table S1). stationary GPS collars (3,464 GPS locations, mean = 693 pts/collar, We used a mixed-­effects linear regression model for movement SD = 528) deployed under field conditions. Based on these station- rate and tortuosity, with Lynx ID as a random intercept to control ary collars we calculated step length (straight line distance between for the nonindependent nature of GPS points within a single lynx two successive GPS points) and turn angles (relative turn angle be- (Gillies et al., 2006). We considered the inclusion of study year as tween the vector from points t and t−1 and the vector from points t an additional random intercept to account for differences between and t + 1) for stationary GPS points. We then used this distribution years, but, as the majority of lynx only occurred in the dataset for a of distances and turn angles to determine threshold values to distin- single year, the addition of this parameter did not noticeably effect guish active from stationary states for collared lynx. We categorized model estimates, and thus we omitted it for model simplicity. We lynx locations as “stationary” if GPS points were ≤27.02 m from the standardized (xi − x̄ ∕SD) all continuous covariates for ease of model o fitting and interpretation. We ranked models using AICc (Akaike, previous point (70th percentile) or had turn angles between 174 and 180 o (90th percentile; Hurford, 2009). 1974) and considered the best performing to have the lowest AICc. We evaluated model fit using Q-­Q plots and scatterplots of fitted 2.4 | Do lynx change movement behavior based on intensity of winter recreation? values versus residuals to verify the linear model assumptions of normality and homoscedasticity (Kutner et al., 2005), and calculated marginal r 2 to assess the variability explained by the fixed effects To determine the impact of recreation intensity on lynx movement of the top-­performing model (Barton, 2015). All models were fitted behavior, we modeled the response of lynx to a combination of using the lme4 package in R (Bates et al., 2015). �| OLSON et al. 8559 predicted the outcome on the withheld partition. Since our response 2.5 | Do lynx avoid high intensity dispersed recreation? variable was binary, we calculated the area under the curve (AUC) To test whether lynx spatially avoided recreation within home this provides a metric ranging between random predictive ability ranges, we compared measures of recreation at lynx GPS points to a (0.5) and perfect model prediction (1.0; Boyce, Vernier, Nielsen, & sample of random locations within a lynx’s MCP home range (i.e., a Schmiegelow, 2002). of the receiver operating characteristic for each fold of the data; third-­order used-­available resource selection function (RSF) design; Additionally, to test specifically whether lynx adjusted their Johnson, 1980; Manly et al., 2002). We generated random locations activity in response to temporal differences in recreation, we ex- at a ratio of 1 use to 2 random, determined 1 km recreation intensity amined the relationship between recreation intensity and activity and percent canopy cover at all locations, and assigned a random state (moving or stationary) at lynx GPS points. We used a GLMM “hour” value to each random point to allow temporal comparisons. to predict whether a point was active or stationary in response to Additionally, since much of the dispersed recreation on our study the interaction of 1 km recreation intensity and time of day (day areas took place on groomed or user-­established trails (Miller, 2016; ~0800:1700 hr, night ~1700:0700 hr), with lynx ID included as a Olson et al., 2017), we hypothesized that lynx might respond more random intercept and a binomial link (Gillies et al., 2006). We also strongly to recreation when near to these high-­use areas. Since evaluated whether temporal response differed based on study area many trails are user-­established, and therefore no spatial data ex- by creating a combined variable of study area and temporal period ists for them, we created trail features for each recreation type (hy- (San Juan day, San Juan night, Vail day, Vail night), to allow separate brid, back-­country ski, snowmobile, and packed-­trail ski) from the estimation of a temporal response at each study area. We performed high-­intensity areas (>25th percentile) delineated by a 100 m point separate models for each recreation type (Supporting Information density recreation raster. We measured the distance from each lynx Table S3), standardized recreation intensity metrics for ease of GPS point and random location to the nearest trail of each recrea- model fitting, and fitted models using the lme4 package in R (Bates tion type. Using this distance to trail value, we created a binary vari- et al., 2015). We cross-­validated top-­performing models as above to able for whether a point was near or far from a trail using thresholds assess model fit (Hastie et al., 2001). of 250 m, 500 m, and 1 km. We also considered the possibility that Finally, we tested whether lynx change their response to recre- influence from a trail would attenuate nonlinearly with distance, and ation depending on how much of it is available (i.e., a functional re- thus created a decay function (e−α/distance) where α was a constant sponse; Mysterud & Ims, 1998). Functional responses can help reveal equal to 50, 100, 250, 500, or 2,500 (Carpenter, Aldridge, & Boyce, response thresholds which may be difficult to detect from individual 2010; Lesmerises, Johnson, & St-­L aurent, 2016). We tested each selection or avoidance (Mysterud & Ims, 1998). For each individual, scale of both covariates in univariate models, and kept the binary or we calculated mean 1 km recreation intensity at used versus avail- decay variable that had the lowest AICc for each recreation type to able locations within home ranges (Holbrook et al., 2017; Laforge, represent response to trails in candidate models. Brook, van Beest, Bayne, & McLoughlin, 2015). We then tested for We then constructed a set of 10 GLMM logistic regression functional responses by modeling use as a function of availability (Gillies et al., 2006) candidate models per recreation type to test lynx for each recreation type. We considered linear and quadratic mod- spatial response to dispersed recreation intensity and high-­intensity els, and used likelihood-­ratio tests to determine which model form trails (Supporting Information Table S2). We considered univariate, best fit the data (Kutner et al., 2005). A functional response, indi- additive, and interactive effects of canopy cover, to test whether cating disproportional changes in use in response to availability, was canopy cover influenced lynx selection or avoidance of recreation, supported when the quadratic form was best-­fitting, or when the since lynx are closely tied to dense forest cover (Holbrook, Squires, slope of the linear response did not equal 1 (Holbrook et al., 2017; Olson, DeCesare, & Lawrence, 2017; Squires, Decesare, Kolbe, & Mysterud & Ims, 1998). In addition, for each lynx, we calculated the Ruggiero, 2010). We considered an interaction between recreation selection ratio (mean use/mean available) for each type of recreation metrics and study area to test lynx response to differences in the (Manly et al., 2002; Mysterud & Ims, 1998). Selection ratios below 1 quality of recreation between study areas (see Methods), and an in- indicate avoidance (use less than availability), while selection ratios teraction between recreation and time of day (day ~0800:1700 hr, above 1 indicate selection (use greater than availability). We plotted night ~1700:0700 hr) to determine whether lynx exhibited a tempo- selection ratio for each individual against average recreation avail- ral response to recreation intensity or recreation trails. We included ability in the home range to better visualize differences in patterns Lynx ID as a random intercept to control for repeated GPS locations of lynx selection for recreation with changing availability. among lynx, and weighted observations to create an equal contribution between the unbalanced used to available samples (Gillies et al., 2006). We standardized covariates as above and used AICc to select 2.6 | Do lynx avoid developed recreation? the best performing model for each recreation type. We assessed Two developed ski areas were adjacent to lynx home ranges in our model fit using fivefold cross-­validation of the best model for each study areas. As an initial test to determine the impact that such recreation type (Hastie, Tibshirani, & Friedman, 2001). We split the permanent, spatially concentrated centers of recreation activity data into five equal partitions, re-­fit models on four partitions and had on Canada lynx space use, we performed a simple bootstrap �8560 | OLSON et al. TA B L E 1 Coefficients (β) and confidence intervals (95% CI) for top-­performing models of Canada lynx movement rate and movement tortuosity in response to recreation intensity at a 1 km scale and other covariates in western Colorado, USA, 2010–2013 Covariate Lower CI β Upper CI the likelihood of lynx use of the ski area (Supporting Information Table S4). We fit models using the lme4 package in R (Bates et al., 2015) and ranked models according to AICc. We validated the best-­ performing model using fivefold cross-­validation, as detailed above (Hastie et al., 2001). Movement rate Hybrid1k the week, as well as interactions that we hypothesized might impact 0.02 0.00 0.05 Back-country ski1k −0.04 −0.06 −0.02 Snowmobile1k −0.02 −0.05 0.00 Packed-­trail ski1k 0.01 −0.01 0.03 Proportion forest −0.03 −0.05 −0.01 0.22 0.15 0.29 −0.01 −0.03 0.00 Lynx moved a mean of 8.0 km per day (SD: 4.9 km), at an average rate Back-­country ski1k 0.00 −0.02 0.02 of 0.63 km/hr (SD: 0.27 km/hr). Snowmobile1k 0.00 −0.01 0.02 Sex Movement Tortuosity Hybrid1k 3 | R E S U LT S We captured 18 individual lynx (9 males, 9 females) from 2010 to 2013, with four individuals captured in two successive years, for a total of 22 yearly lynx home ranges. We collected a total of 34,405 GPS points (average: 1,720/lynx, SD: 1100) from January to March. We collected a total of 2,839 tracks from recreationists (2010: Packed-­trail ski1k −0.01 −0.02 0.01 n = 350, 2011: n = 1015, 2012: n = 651, and 2013: n = 823). Although Proportion forest −0.02 −0.04 −0.01 all lynx were captured in areas used by winter recreationists, recre- Night −0.05 −0.07 −0.02 Covariates whose 95% CI did not overlap 0 are bolded. ation intensity was highly variable across lynx (Appendix A: Figure A1). Hybrid recreation occurred on 13 yearly lynx MCP home ranges, snowmobile on 17, and backcountry ski and packed-­trail ski on 19. Recreation availability also differed between the two study areas; comparison to test whether individual lynx entered ski areas less mean number of unique GPS tracks recorded on lynx home ranges than random expectation (Manly, 2007; Manly et al., 2002). We sam- was: hybrid) 27.8 Vail (SD = 55.4), 3.0 San Juan (SD = 4.5); back-­ pled random locations distributed across each lynx’s 95% MCP home country ski) 11.8 Vail (SD = 10.4), 71.2 San Juan (SD = 51.3); snow- range (sample size equal to the total GPS locations collected for each mobile) 32.6 Vail (SD = 49.7), 20.7 San Juan (SD = 28.5); packed-­trail lynx) 1,000 times with replacement; at each iteration, we recorded ski) 7.9 Vail (SD = 11.2), 66.4 San Juan (SD = 51.9). The mean length the number of random locations inside the ski area boundary. We of all recreation tracks combined within home ranges was 10.9 km/ then calculated the 2.5 and 97.5 percentile from the bootstrap dis- km2 (SD = 24.0) for Vail and 9.7 km/km2 (SD = 6.7) for San Juan tribution for each lynx, and compared that to the actual number of (Appendix A: Table A1). Trail counters had a mean of 35.9 hits/ GPS points inside the ski area boundary; a value outside either of day (SD = 26.8) for Vail and 18.4 hits/day (SD = 10.8) for San Juan these percentiles indicated avoidance or preference, respectively (Appendix A: Table A1). (Manly, 2007). Next, we tested whether factors associated with the intensity of human use of the ski area influenced the probability of lynx use. For this analysis, we included all lynx points collected from January to June to evaluate if lynx use changed with decreased winter recre- 3.1 | Do lynx change movement behavior based on intensity of winter recreation? Both lynx movement rate and movement tortuosity were best mod- ation. We modeled whether a lynx GPS point was in or out of the ski eled by a combination of recreation and environmental variables area boundary as a function of day of the week (weekend or week- (Supporting Information Table S1). Lynx movement rates were a day), since weekday use has been shown to be less intense than week- function of proportion forest, recreation intensity, and sex (Table 1), end, as well as time of day, since daylight hours receive more use than with a marginal r 2 of 0.12. Lynx slowed their movement rate in the dark (Olson et al., 2017). We also considered month as a continuous presence of greater snowmobile and back-­country ski activity. For variable, from February to June, since use of the ski area should de- example, predicted female lynx movement rate was 0.47 km/hr crease with later months as snowpack decreases. Finally, we included (SD = 0.03) with no recreation within 1 km, 0.22 km/hr (SD = 0.07) canopy cover at each GPS location to control for differences in veg- at maximum observed back-­country ski intensity (equivalent to ap- etation inside and outside the ski area boundary. We used GLMMs proximately 66 recreation tracks/km2 in a season), and 0.25 km/ with individual lynx ID as a random effect (Gillies et al., 2006). hr (SD = 0.11) at maximum snowmobile intensity (approximately We considered a candidate model set of 11 models to evaluate 188 tracks/km2 in a season). Conversely, at high hybrid and packed-­ lynx use of developed ski areas (Supporting Information Table S4). trail ski intensities, lynx generally moved faster, although the confi- All models contained a base structure of canopy cover to account dence interval for packed-­trail ski slightly overlapped zero, indicating for habitat differences inside or outside of the ski area boundary. We that lynx movement rate was not as strongly related to packed-­trail evaluated additive combinations of month, time of day, and day of ski intensity. Modeled movement rate was 0.47 km/hr (SD = 0.03) �| OLSON et al. 8561 with no recreation within 1 km, 0.92 km/hr (SD = 0.21) at maximum in areas with greater packed-­trail ski and hybrid ski, but not back-­ observed hybrid intensity (equivalent to approximately 232 tracks/ country ski or snowmobile (Figure 3; Supporting Information Table km2 in a season), and 0.56 km/hr (SD = 0.08) at maximum observed S5). Cross-­validation of selected models, however, found relatively packed-­trail ski intensity (equivalent to approximately 115 tracks/ poor model predictive performance (hybrid: AUC = 0.56, SD = 0.01, km2 in a season). In addition, female lynx moved more slowly than back-­country ski: AUC = 0.57, SD = 0.01, snowmobile: AUC = 0.57, males, and movement rate was slower with greater proportion forest. SD = 0.01, packed-­trail ski: AUC = 0.57, SD = 0.01), indicating that The top performing model for tortuosity included recreation, pro- though lynx temporal activity was influenced by recreation intensity portion forest, and time of day (Table 1), with a marginal r2 of 0.02, in- and study area, much of the variation in temporal activity remained dicating that this model explained little of the variation in tortuosity. unexplained. Based on beta coefficient confidence intervals, recreation intensity We found support for functional responses to hybrid (Likelihood was not an important predictor of tortuosity; lynx movement was Ratio Test p = 0.01, r 2 = 0.78), snowmobile (LRT p = 0.05, r2 = 0.49), more tortuous with greater proportion forest and during the night. and packed-­trail ski recreation (LRT p = 0.01, r2 = 0.83), with mean use best modeled by a quadratic response to mean availability; 3.2 | Do lynx avoid high intensity dispersed recreation? back-­country ski intensity did not support a functional response (LRT p = 0.27, β 0: 219.70, 95% CI: −164.57–603.96, β1: 0.60, 95% CI: −0.13–1.32, r2 = 0.14). Lynx used areas with hybrid and snowmobile Lynx space use within home ranges was better predicted with the recreation in proportion to availability when recreation intensity was addition of dispersed recreation covariates than with canopy cover low; however, as recreation intensity increased, lynx use appeared alone; recreation intensity was most predictive for motorized rec- to decrease. For packed-­trail ski intensity, lynx use appeared to be reation, and distance to trails most predictive for nonmotorized proportional to availability at low and high intensity, but greater recreation (Supporting Information Table S2). The interaction than availability at moderate intensities, while use of areas with with time of day was not ranked highly for any type of recreation, indicating little support for the hypothesis that lynx temporally adjusted their space use in response to recreation (Supporting Information Table S2). Cross-­validation of each model indicated acceptable model fit (hybrid: AUC = 0.74, SD = 0.01, back-­country ski: AUC = 0.74, SD = 0.01, snowmobile: AUC = 0.74, SD = 0.01, and packed-­t rail ski: AUC = 0.75, SD = 0.01). Lynx in the Vail study area avoided areas with greater snowmobile recreation intensity, while lynx in the San Juan study area were more likely to use them (Figure 2; Table 2); greater hybrid recreation intensity was consistently avoided at both study areas, although the effect was stronger in the San Juan than in Vail. Lynx selected areas within 250 m of back-­country ski trails and within 500 m of packed-­trail ski trails; an interaction with study area was supported for back-­country skiing, indicating that the selection for areas near to trails was stronger in the San Juan study area, while an interaction with canopy cover was selected for packed-­trail skiing, indicating that lynx were less affected by trail proximity in areas with greater canopy cover, and more likely to be influenced by trails when cover was low (Figure 2; Table 2). In addition, for all forms of recreation, the predicted probability of lynx presence was always greater with greater canopy cover; this effect tended to be stronger than that of recreation (Table 2). Lynx temporal activity in response to recreation intensity was best modeled when allowed to vary with study area and time of day (Supporting Information Table S3). In general, at areas with no recreation tracks within 1 km (i.e., recreation intensity = 0), the proportion of time that lynx spent active was fairly similar during the day and night and across study areas (Figure 3). As recreation intensity of any type increased, however, lynx activity decreased during the day and increased at night in the Vail study area. Conversely, lynx in the San Juan study area were less active during the day than night F I G U R E 2 Predicted relative probability of Canada lynx presence in response to four recreation types: snowmobile-­assisted hybrid ski, back-­country ski, snowmobile, and packed-­trail ski, in western Colorado, USA, 2010–2013. Lynx (a) avoided greater hybrid intensity, (b) were more likely to be present near (<250 m) back-­country ski trails in the San Juan study area, (c) avoided greater snowmobile intensity in the Vail study area but not the San Juan study area, and (d) were more likely to be present near (<500 m) packed-­trail ski trails, particularly in areas with low canopy cover. Predictions were generated for each recreation type from multivariate general linear mixed-­effects models by holding all other covariates at their mean (see Table 2) �8562 | OLSON et al. TA B L E 2 Coefficients (β) and confidence intervals (95% CI) from top-­performing models of Canada lynx use versus availability in response to recreation intensity at a 1 km scale or proximity to a recreation trail, canopy cover, and study area in western Colorado, USA, 2010-­2013 Covariate Coefficient Lower CI Hybrid Intensity1k Study Area Hybrid1k:Area Random Effect 15.8% (SD: 8.9%), ranging from 3.9% to 27.4% (n = 951 total lynx locations inside ski area boundaries, n = 22,524 lynx locations outside ski area boundaries). Of these nine lynx-­years near developed recreation, 8 had fewer locations inside ski area boundaries than expected, indicat- Upper CI ing avoidance, while one did not differ from random (Table 3; Figure 5). The top supported model of lynx-­use in developed ski areas Hybrid Canopy Cover On average, lynx yearly 95% MCP home ranges overlapped ski areas −1.23 −1.48 −0.97 included all covariates and an interaction between month and day 1.05 1.03 1.06 of the week (Supporting Information Table S4), and fivefold cross-­ −0.27 −0.43 −0.11 1.10 0.84 1.36 Var: 0.03 SD: 0.17 validation indicated the model had adequate predictive ability (mean AUC = 0.77, SD = 0.02). Lynx were more likely to enter the ski area boundary during night than day (Table 4). Additionally, an interaction between month and day of week was strongly supported: on Back-­country Ski Back-country Ski Trail250 m 0.57 0.53 0.62 Canopy Cover 1.04 1.02 1.06 Study Area −0.27 −0.45 −0.09 BC-SkiTrail:Area −0.61 −0.74 −0.49 Random Effect Var: 0.04 SD: 0.20 weekends, lynx use of ski areas increased with month as months became warmer and winter recreation declined, so that predicted use on June weekends was 4.7 times that of February weekends. During weekdays the effect of month was less pronounced, with predicted June use only 1.1 times greater than predicted February use. Canopy cover was weakly lower at lynx locations inside the ski area than outside, although the model with only canopy cover ranked second to Snowmobile last among the candidate models, indicating that habitat alone was Snowmobile Intensity1k 0.19 0.17 0.21 Canopy Cover 1.06 1.04 1.08 Study Area −0.42 −0.57 −0.27 Snowmobile1k:Area −0.38 −0.42 −0.34 Random Effect Var: 0.03 SD: 0.16 a poor predictor of lynx use of the ski area (Supporting Information Table S4). 4 | D I S CU S S I O N Our results demonstrate a nuanced response of Canada lynx to Packed-­trail Ski winter recreation, ranging from avoidance of developed ski resorts Packed-trail Ski Trail500 m 0.84 0.80 0.89 Canopy Cover 1.10 1.08 1.12 skiing. Taken together, lynx spatial and behavioral responses to the PT-SkiTrail:Canopy −0.37 −0.42 −0.32 gradient of recreation recorded in our study may suggest a tol- Random Effect Var: 0.08 SD: 0.28 Covariates whose 95% CI did not overlap 0 are bolded. to tolerance of nonmotorized back-­country skiing and packed-­t rail erance threshold, with little disturbance from low and moderate intensity recreation but increasing disturbance when intensity exceeds a given level. For instance, evidence of temporal avoidance of recreation was most marked in the high-­intensity Vail study back-­country skiing were proportional to availability. Among our area. Functional responses of use versus availability also indicated sampled lynx, most had relatively low recreation availability, how- little evidence of avoidance when recreation availability was low, ever, so that the few individuals with high availabilities may have but consistent avoidance for the two lynx with the greatest rec- driven results. The plotted selection ratios (Figure 4), which allowed reation availability. Lynx also consistently avoided developed ski better visualization of selection across the range of recreation inten- resorts, especially at times when recreation was most intense. In sity, demonstrated no consistent pattern of selection or avoidance at areas with low and moderate recreation intensity, lynx exhibited low availabilities, but showed that the only two lynx with consistent spatial tolerance coupled with behavioral modifications that al- avoidance also had the highest recreation availabilities for three out lowed lynx and dispersed recreation to co-­o ccur. Based on these of four types of recreation (Figure 4, highlighted boxes). results, it appears that dispersed winter recreation at the low to moderate intensities found in western Colorado does not provoke 3.3 | Do lynx avoid developed recreation? We captured lynx near two ski resorts, one near the Vail Pass Winter a strong negative response in Canada lynx, but that high-­intensity dispersed or developed recreation may provide enough of a disturbance to elicit lynx avoidance. Recreation area and the other in the San Juan Mountains. Five unique Animal responses to human disturbance often vary depending on lynx, two captured in successive years, had home ranges adjacent to the type of disturbance activity. Wild reindeer fled longer distances the ski area near the Vail Pass study area, while one unique lynx, cap- when disturbed by skiers than snowmobiles (Reimers, Eftestøl, & tured in two successive years, was adjacent in the San Juan study area. Colman, 2003), while moose exhibited short-­term disturbance from �| OLSON et al. 8563 spatially avoid the ski area (Figure 5), and to temporally adjust their activity to avoid high traffic times when they did go near it, even after controlling for differences in vegetation between the ski area and its surroundings. Lynx avoidance of intensely used ski resorts also supports the idea of a threshold level of tolerance toward human disturbance, with lynx able to adjust their space use or behavior in the presence of most dispersed recreation, but unable to tolerate high levels of human use that occur at a resort. Developed resorts in Colorado are intensively used, and are also subject to frequent motorized trail grooming and maintenance. Combined, this may represent unacceptably high levels of human disturbance for lynx. Other species have shown similar avoidance of developed ski areas, including mountain goats (Richard & Côté, 2016), reindeer (Nellemann et al., 2010), and alpine black grouse (Patthey et al., 2008). Behaviorally, lynx tended to move more slowly in areas with greater snowmobile and back-­country ski intensity, while their movement tortuosity remained unchanged. This behavioral change F I G U R E 3 Predicted differences in temporal lynx activity patterns in response to four types of recreation intensity (snowmobile-­assisted hybrid ski, back-­country ski, snowmobile, and packed-­trail ski) in western Colorado, USA, 2010–2013. Line type represents day or night at the Vail or San Juan study areas. Proportion of time spent active was similar at low recreation intensities, but diverged as recreation intensity increased, with activity primarily lowest during the day at the Vail study area may indicate that lynx perceive a threat from human disturbance, and respond by hiding or moving more cautiously (Tablado & Jenni, 2017), but do not change their foraging behavior by either stopping completely or moving directly out of an area. While increased movement rate or flight responses are common indications of disturbance (Arlettaz et al., 2015; Reimers et al., 2003), hiding or freezing are also common behavioral responses to threats (Tablado & Jenni, 2017). Lynx may rely on their cryptic coloration to protect them from no- skiers (Neumann, Ericsson, & Dettki, 2010) and avoidance of areas tice, thus saving themselves from a more energetically costly flight with high snowmobile trail density (Harris et al., 2013). Lynx did not response (Stankowich & Blumstein, 2005). While lynx did not exhibit appear to exhibit a consistent response to all dispersed recreation strong temporal avoidance of recreation, they adjusted the propor- types, although some consistent differences between motorized tion of time they spent active in areas with greater recreation, par- and nonmotorized recreation types emerged. Areas with greater ticularly in the high-­intensity Vail study area, in which they were less nonmotorized recreation intensity (i.e., back-­country and packed-­ active during the day. Rather than leave high intensity areas during trail ski,) were selected by lynx, while areas with greater snowmo- the day, lynx may simply become less active and more cautious, bile and hybrid recreation intensity were generally avoided. This waiting for the disturbance to decline and increasing their activity pattern may reflect similarities between the habitat preferences of at night. Temporal avoidance is frequently observed in response to lynx and skiers, and habitat differences between lynx and motor- human disturbance, and has been demonstrated in predators such ized recreation. For instance, nonmotorized recreation is frequently as coyotes (Canis latrans) and bobcats (Lynx rufus) (Reilly, Beier, & located in high elevation areas, with dense canopy cover and steep Sonderegger, 2016; Riley et al., 2003). slopes (Olson et al., 2017), habitat which is likely favored by forest-­ While we focused our study on some of the most heavily dwelling Canada lynx, which prefer areas with multi-­storied forest recreated landscapes in Colorado, collected during peak winter and high horizontal cover (Holbrook et al., 2017; Squires et al., 2010). recreation to maximize the potential to measure responses to Motorized recreation such as snowmobiling, however, usually takes disturbance, the lack of consistent avoidance may suggest that place on groomed trails or forest roads, which are placed in areas of low to moderate dispersed recreation at our study areas was open forest and gentle topography to allow safer fast travel (Olson not intense enough to elicit a strong population-­l evel response et al., 2017), and which is not as hospitable to lynx. from lynx. Response to recreation can vary at the level of the Developed recreation may be more likely to have an effect on individual, often depending on an individual’s age, sex, repro- animals, given the high intensity, large infrastructure, and frequent ductive status, or other factors (Lesmerises & St-­L aurent, 2017; maintenance requirements of large ski resorts (Rixen & Rolando, Tablado & Jenni, 2017). The results of our functional response 2013). For example, Pacific marten (Martes caurina) have been analysis indicate that lynx in our study varied in their selec- shown to be negatively influenced by ski resorts through habitat tion or avoidance of recreation, with differences both between fragmentation and reduced occupancy and density during the win- individuals in response to the same type of recreation, and ter season (Slauson, Zielinski, & Schwartz, 2017). Lynx also appeared within individuals given different types of recreation (Figure 4). to be affected by developed recreation, although our sample size for Interestingly, the two lynx in areas with the greatest amount of this analysis was small. Lynx near developed ski resorts appeared to recreation also demonstrated the most consistent avoidance. �8564 | OLSON et al. F I G U R E 4 Individual selection ratios (mean use/mean availability; colored bars) compared to mean relative recreation availability (black dots) for each recreation type (a) snowmobile-­assisted hybrid ski, (b) back-­country ski, (c) snowmobile, (d) packed-­trail ski. For ease of interpretation, selection ratio-­1 is shown, so that avoidance is below 0, selection above 0, and availability values are relative (each divided by the maximum availability value) to force the range between 0 and 1. Lynx in each study area are indicated by dark bars (San Juan) or light bars (Vail). Stafford Female and Stafford Male (names outlined in boxes), from the Vail study area, are the only lynx to consistently avoid all types of recreation, and have the highest availability for 3 out of the 4 recreation types Lynx ID Study area Year Lower CI Upper CI Breckenridge Female 2010 Vail 2010 78 466 544 Breckenridge Female 2011 Vail 2011 116 322 392 Climax Female 2010 Vail 2010 11 87 126 N Climax Male 2011 Vail 2011 29 141 182 Stafford Female 2010 Vail 2010 92 243 295 Stafford Female 2011 Vail 2011 366 329 396 Stafford Male 2011 Vail 2011 204 618 700 Ophir Male 2012 San Juan 2012 0 87 127 Ophir Male 2013 San Juan 2013 55 134 181 Lynx that avoided ski areas (actual GPS points less than the lower 2.5% bootstrapped value) are bolded. TA B L E 3 Summary of the number of Canada lynx GPS locations (N) inside two ski resort boundaries in western Colorado, USA, compared to the bootstrapped 95% confidence values (CI) for each lynx �| OLSON et al. 8565 F I G U R E 5 The spatial distribution of Canada lynx GPS points (red dots) around the ski resort in the Vail Pass study area, in western Colorado, USA, 2010–2013; lynx locations indicate avoidance of this heavily recreated area. Individual ski runs are the light-­colored lines in the center of the picture, while the resort infrastructure is toward the bottom right. Lower intensity dispersed skiing (blue lines) along a groomed trail to a back-­country hut is shown to the right of the ski area; lynx did not exhibit spatial avoidance of this type of use These lynx had the highest availability of recreation intensity for three out of the four recreation types (Figure 4), and were located in the Vail study area, which had extremely high TA B L E 4 Coefficients (β) and confidence intervals (95% CI) from the top-­performing model of lynx use of ski resorts in western Colorado, USA, 2010–2013 intensity use, approximately 35,000 recreationists per winter (U.S.D.A. Forest Service, 2015), as well as a large developed ski area. Thus, recreation intensity in our study area may be low enough for the majority of lynx to ignore and spatially coexist with, but an intensity threshold may exist above which dispersed recreation cannot be tolerated by lynx. We recognize that behavior responses are not necessarily expressed in changes in population demography and adult survivorship. However, the population of lynx in Colorado has been recover- Intercept Month Weekend Night Canopy Cover Month:Weekend Coefficient SE Lower CI Upper CI −4.20 0.55 −5.28 −3.12 0.18 0.03 0.11 0.24 −2.00 0.31 −2.61 −1.39 0.14 0.07s 0.01 0.28 −0.07 0.03 −0.14 0.00 0.38 0.07 0.25 0.52 Covariates whose 95% CI did not overlap 0 are bolded. ing since reintroduction in 1999, and is currently estimated at 200 to 300 individuals (Martin, 2013). Our sample likely represents approximately 6%–9% of the entire lynx population in their home ranges, but will tolerate extremely urban areas, pos- Colorado, and the majority of resident lynx at each study area, sibly because of a correlated increase in prey availability (Bouyer and illustrates that a sizeable portion of the population is sub- et al., 2014, 2015). Similarly, Canada lynx in Riding Mountain ject to disturbance from recreation. National Park, Canada, tended to have high probabilities of occur- Carnivores are often reported to be particularly sensitive to rence in the less disturbed park interior, but highest occurrence anthropogenic disturbance because of their need for large con- near a town with intense winter recreation yet close proximity tiguous home ranges and a tendency to draw human persecution to highly suitable hare habitat (Montgomery, Roloff, Millspaugh, (Carroll, Noss, & Paquet, 2001; Noss, Quigley, Hornocker, Merrill, & Nylen-­N emetchek, 2014). Lynx in our study also failed to ex- & Paquet, 1996; Woodroffe, 2000). However, both Canada lynx hibit strong behavioral avoidance from low to moderate intensity and Eurasian lynx (Lynx lynx) have demonstrated a high degree dispersed recreation, instead appearing to segregate themselves of tolerance to human presence. For example, Eurasian lynx in from high-­intensity motorized recreation and to adjust their tem- Norway show a preference for low levels of human disturbance in poral and movement patterns. �8566 | OLSON et al. Given the nature of our study design, we were unable to eval- in our study. However, lynx residing in more heavily recreated uate the potential for second order (i.e., home-­r ange placement; landscapes left stronger response signatures, culminating in fairly Johnson, 1980) avoidance of recreation. The arrangement of home strong avoidance of the most intensely recreated landscapes, such ranges to avoid recreation or human disturbance has been ob- as commercial ski areas. While behavioral changes do not neces- served in northern mountain woodland caribou (Rangifer tarandus sarily reflect impacts on survival or reproductive success, if lynx caribou) in response to human infrastructure (Polfus, Hebblewhite, conservation is an important goal in these areas, implementation & Heinemeyer, 2011), but was not found in capercaillie (Tetrao of programs to alleviate recreation intensity may be considered. urogallus) in response to outdoor recreation (Coppes, Ehrlacher, Alternatively, intense recreation could be administratively concen- Thiel, Suchant, & Braunisch, 2017) or red deer (Cervus elaphus) in trated, leaving space for lynx conservation measures in adjacent response to recreation infrastructure (Coppes, Burghardt, Hagen, regions. Suchant, & Braunisch, 2017). It is possible that some lynx may have already exhibited avoidance of recreation, through the selection of home ranges that do not overlap with recreation. If this is the AC K N OW L E D G M E N T S case, the lynx that were trapped for our study may be habituated We thank the U.S. Department of Agriculture, U.S. Forest Service, to recreation, and the continued occupancy of these territories in subsequent lynx generations may not be assured. Alternatively, use of high recreation intensity areas may be a function of limited habitat distribution in high elevation linear valleys, rather than habituation to recreation per se. Since it is possible that the lack of strong response to recreation we found represents the result of an ongoing strategy of avoidance by lynx sensitive to recreation, we recommend long-­term monitoring of lynx occupancy near heavily recreated areas to ensure that lynx are not negatively impacted by recreation (with the assumption that continued occupancy reflects a lack of detrimental demographic effects). Further, to thoroughly and White River National Forest for funding this work. Additional funding and support was provided by the Rocky Mountain Research Station, Vail Associates Inc., Colorado BLM state office, U.S. Forest Service R2 Regional Office Renewable Resources Department, 10th Mountain Huts, and Colorado Department of Transportation. Special thanks to the many field technicians that contributed to this project, the participants who volunteered to carry the GPS units, and the local FS offices for providing logistical support and information about the area. evaluate causal impacts of recreation to lynx in Colorado, we sug- C O N FL I C T O F I N T E R E S T gest continued research to measure demographic (and/or behav- None declared. ioral) responses to experimental manipulation of user access and density. AU T H O R S C O N T R I B U T I O N 5 | CO N C LU S I O N S J. Squires and E. Roberts conceived the concepts, J. Squires, E. Roberts, J. Ivan, L. Olson, and M. Hebblewhite designed the meth- We evaluated a gradient of human disturbance from winter rec- odology, L. Olson performed the data analyses, J. Squires, M. reation, from intensively used developed ski areas to low-­intensity Hebblewhite, and J. Ivan consulted on data analyses, L Olson led the dispersed back-­country recreation. In keeping with this range of writing of the manuscript, all authors contributed critically to the disturbance, we found a range of Canada lynx responses to winter drafts. All authors gave final approval for publication. recreation, from avoidance of developed ski resorts to tolerance of nonmotorized back-­country skiing and packed-­t rail skiing. Lynx may tolerate low to moderate levels of dispersed winter recreation (similar to levels we sampled at the San Juan Study Area) via behavioral modifications governing activity levels, activity timing, DATA AC C E S S I B I L I T Y Data available from the figshare repository: https://doi.org/10.6084/ m9.figshare.6452807.v1 and locations of various activities. Thus, recreation management such as trail closures, visitor limitation, etc., may convey little benefit to species conservation in areas with low to moderate levels of dispersed recreation. We found less spatial avoidance of nonmo- ORCID Lucretia E. Olson http://orcid.org/0000-0002-5703-3351 torized recreation compared to motorized, although lynx response varied by study area, and lynx exhibited a behavioral response to both motorized and nonmotorized recreation. Our results support the conclusion that lynx, as evidenced by changes in space use and behavior, were not uniformly negatively influenced by dispersed winter recreation at the low to moderate intensities found REFERENCES Abrahms, B., Seidel, D. P., Dougherty, E., Hazen, E. L., Bograd, S. J., Wilson, A. M., … Getz, W. M. (2017). Suite of simple metrics reveals common movement syndromes across vertebrate taxa. Movement Ecology, 5(1), 1–11. �OLSON et al. Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control, 19(6), 716–723. https://doi. org/10.1109/TAC.1974.1100705 Arlettaz, R., Nusslé, S., Baltic, M., Vogel, P., Palme, R., Jenni-Eiermann, S., … Genoud, M. (2015). Disturbance of wildlife by outdoor winter recreation: Allostatic stress response and altered activity–energy budgets. Ecological Applications, 25(5), 1197–1212. https://doi. org/10.1890/14-1141.1 Barton, K. (2015). MuMIn: Multi-­model inference. R Package Version, 1(15), 1. Bates, D., Mächler, M., Bolker, B., Walker, S. (2015). Fitting linear mixed-­effects models using lme4. Journal of Statistical Software, 67, 1–48. Benhamou, S. (2004). How to reliably estimate the tortuosity of an animal’s path: Straightness, sinuosity, or fractal dimension? Journal of Theoretical Biology, 229(2), 209–220. https://doi.org/10.1016/j. jtbi.2004.03.016 Bivand, R., & Lewin-Koh, N. (2016). maptools: Tools for reading and handling spatial objects. R Package version 0.8-40. Blevins, J. (2016). Colorado ski areas set visitor record, pass 13 million for first time. The Denver Post, https://www.denverpost. com/2016/06/09/colorado-ski-areas-set-visitor-record-pass-13-million-for-first-time/, Accessed 19 June 2017. Bouyer, Y., Gervasi, V., Poncin, P., Beudels-Jamar, R. C., Odden, J., & Linnell, J. D. C. (2014). Tolerance to anthropogenic disturbance by a large carnivore: The case of Eurasian lynx in south-­eastern Norway. Animal Conservation, 18, 271–278. Bouyer, Y., San Martin, G., Poncin, P., Beudels-Jamar, R. C., Odden, J., & Linnell, J. D. C. (2015). Eurasian lynx habitat selection in human-­m odified landscape in Norway: Effects of different human habitat modifications and behavioral states. Biological Conservation, 191, 291–299. https://doi.org/10.1016/j. biocon.2015.07.007 Bowles, A. E. (1995). Responses of wildlife to noise. In R. L. Knight, & K. J. Gutzwiller (Eds.), Wildlife and recreationists: Coexistence through management and research (pp. 109–156). Washington, DC: Island Press. Boyce, M. S., Vernier, P. R., Nielsen, S. E., & Schmiegelow, F. K. (2002). Evaluating resource selection functions. Ecological Modelling, 157, 281–300. https://doi.org/10.1016/ S0304-3800(02)00200-4 Brammer, J. R., Samson, J., & Humphries, M. M. (2015). Declining availability of outdoor skating in Canada. Nature Climate Change, 5(1), 2–4. https://doi.org/10.1038/nclimate2465 Buskirk, S. W., Ruggiero, L. F., & Krebs, C. J. (1999). Habitat fragmentation and interspecific competition: Implications for lynx conservation. In L. F. Ruggiero et al., (Eds.), Ecology and conservation of lynx in the United States. General Technical Report RMRS-GTR-30WWW (pp. 83–100). Fort Collins, CO: U.S. Department of Agriculture, Forest Servce, Rocky Mountain Research Station. https://doi.org/10.2737/ RMRS-GTR-30 Carpenter, J., Aldridge, C., & Boyce, M. S. (2010). Sage-­grouse habitat selection during winter in Alberta. Journal of Wildlife Management, 74(8), 1806–1814. https://doi.org/10.2193/2009-368 Carroll, C., Noss, R. F., & Paquet, P. C. (2001). Carnivores as focal species for conservation planning in the rocky mountain region. Ecological Applications, 11(4), 961–980. https://doi. org/10.1890/1051-0761(2001)011[0961:CAFSFC]2.0.CO;2 Colescott, J. H., & Gillingham, M. P. (1998). Reaction of moose (Alces alces) to snowmobile traffic in the Greys River valley, Wyoming. Alces, 34(2), 329–338. Coppes, J., Burghardt, F., Hagen, R., Suchant, R., & Braunisch, V. (2017). Human recreation affects spatio-­temporal habitat use patterns in red deer (Cervus elaphus). PLoS One, 12(5), e0175134. https://doi. org/10.1371/journal.pone.0175134 | 8567 Coppes, J., Ehrlacher, J., Thiel, D., Suchant, R., & Braunisch, V. (2017). Outdoor recreation causes effective habitat reduction in capercaillie Tetrao urogallus: A major threat for geographically restricted populations. Journal of Avian Biology, 48(12), 1583–1594. https://doi. org/10.1111/jav.01239 Elsasser, H., & Messerli, P. (2001). The vulnerability of the snow industry in the Swiss Alps. Mountain Research and Development, 21(4), 335–339. https://doi. org/10.1659/0276-4741(2001)021[0335:TVOTSI]2.0.CO;2 Frid, A., & Dill, L. (2002). Human-­c aused disturbance stimuli as a form of predation risk. Ecology and Society, 6(1), 11. [online] URL: http:// www.consecol.org/vol6/iss1/art11/ Gillies, C. S., Hebblewhite, M., Nielsen, S. E., Krawchuk, M. A., Aldridge, C. L., Frair, J. L., … Jerde, C. L. (2006). Application of random effects to the study of resource selection by animals. Journal of Animal Ecology, 75(4), 887–898. https://doi.org/10.1111/j.1365-2656.2006.01106.x Goldstein, M. I., Poe, A. J., Suring, L. H., Nielson, R. M., McDonald, T. L. (2010). Brown bear den habitat and winter recreation in South-­ Central Alaska. Journal of Wildlife Management, 74(1), 35–42. https:// doi.org/10.2193/2008-490 Harris, G., Nielson, R. M., Rinaldi, T., & Lohuis, T. (2013). Effects of winter recreation on northern ungulates with focus on moose (Alces alces) and snowmobiles. European Journal of Wildlife Research, 60(1), 45–58. Hastie, T., Tibshirani, R., & Friedman, J. H. (2001). The elements of statistical learning: Data mining, inference, and prediction. New York, NY: Springer-Verlag. https://doi.org/10.1007/978-0-387-21606-5 Holbrook, J. D., Squires, J. R., Olson, L. E., DeCesare, N. J., & Lawrence, R. L. (2017). Understanding and predicting habitat for wildlife conservation: The case of Canada lynx at the range periphery. Ecosphere, 8(9), e01939. https://doi.org/10.1002/ecs2.1939 Homer, C. G., Dewitz, J., Yang, L., Jin, S., Danielson, P., Xian, G., … Megown, K. (2015). Completion of the 2011 National Land Cover Database for the conterminous United States-­Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, 81(5), 345–354. Hughes, T. P. (2003). Climate change, human impacts, and the resilience of coral reefs. Science, 301(August), 929–933. https://doi. org/10.1126/science.1085046 Hurford, A. (2009). GPS measurement error gives rise to spurious 180 degree turning angles and strong directional biases in animal movement data. PLoS One, 4(5), e5632. https://doi.org/10.1371/journal. pone.0005632 Johnson, D. H. (1980). The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61(1), 65–71. https://doi.org/10.2307/1937156 Kolbe, J. A., Squires, J. R., & Parker, T. W. (2003). An effective box trap for capturing lynx. Wildlife Society Bulletin, 31(4), 1–6. Kutner, M. H. (2005). Applied linear statistical models (5th ed.). Boston, MA: McGraw-Hill Irwin. Laforge, M. P., Brook, R. K., van Beest, F. M., Bayne, E. M., & McLoughlin, P. D. (2015). Grain-­dependent functional responses in habitat selection. Landscape Ecology, 31, 855–863. Larson, C. L., Reed, S. E., Merenlender, A. M., & Crooks, K. R. (2016). Effects of recreation on animals revealed as widespread through a global systematic review. PLoS One, 11(12): e0167259. http://doi. org/10.1371/journal.pone.0167259 Lesmerises, F., Johnson, C. J., & St-Laurent, M.-H. (2016). Refuge or predation risk? Alternate ways to perceive hiker disturbance based on maternal state of female caribou. Ecology and Evolution, 7(3), 845–854. Lesmerises, R., & St-Laurent, M.-H. (2017). Not accounting for interindividual variability can mask habitat selection patterns: A case study on black bears. Oecologia, 185(3), 415–425. https://doi.org/10.1007/ s00442-017-3939-8 Manly, B. F. J. (2007). Randomization, bootstrap and Monte Carlo methods in biology (3rd ed.). Boca Raton, FL: Chapman & Hall/CRC. �8568 | Manly, B. F. J., McDonald, L. L., Thomas, D. L., McDonald, T. L., Erickson, W. P. (2002). Resource selection by animals: Statistical design and analysis for field studies (2nd ed.). New York, NY: Kluwer Academic Publishers. Martin, J. (2013). A Rocky Mountain success story. U.S. Fish & Wildlife Service. Retrieved from https://www.fws.gov/endangered/map/ ESA_success_stories/CO/CO_story2/ Miller, A. D. (2016). Recreation conflict and management options in the Vail Pass Winter Recreation Area, Colorado, USA. (Unpublished master’s thesis). Colorado State University, Fort Collins, Colorado. Miller, A. D., Vaske, J. J., Squires, J. R., Olson, L. E., & Roberts, E. K. (2017). Does zoning winter recreationists reduce recreation conflict? Environmental Management, 59, 50–67. Montgomery, R. A., Roloff, G. J., Millspaugh, J. J., & Nylen-Nemetchek, M. (2014). Living amidst a sea of agriculture: Predicting the occurrence of Canada lynx within an ecological island. Wildlife Biology, 20(3), 145–154. https://doi.org/10.2981/wlb.13108 Mysterud, A., & Ims, R. (1998). Functional responses in habitat use: Availability influences relative use in\rtrade-­off situations. Ecology, 79(4), 1435–1441. https://doi.org/10.1890/0012-9658(1998)079[1435:FRIHUA]2.0.CO;2 Nathan, R., Getz, W. M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D., & Smouse, P. E. (2008). A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences of the United States of America, 105(49), 19052–19059. https://doi.org/10.1073/pnas.0800375105 National Oceanic and Atmospheric Adminstration (2017). National Weather Service. Retrieved from https://www.weather.gov/bou/ co_snowpack National Ski Areas Association (2016). Number of ski areas operating per state during 2015/16 season. Retrieved from http://www.nsaa.org/ media/275041/1516_Areas_per_state.pdf Nellemann, C., Vistnes, I., Jordhøy, P., Støen, O.-G., Kaltenborn, B. P., Hanssen, F., & Helgesen, R. (2010). Effects of recreational cabins, trails and their removal for restoration of reindeer winter ranges. Restoration Ecology, 18(6), 873–881. https://doi. org/10.1111/j.1526-100X.2009.00517.x Nellis, C. H., & Keith, L. B. (1968). Hunting activities and success of lynxes in Alberta. The Journal of Wildlife Management, 32(4), 718–722. https://doi.org/10.2307/3799545 Neumann, W., Ericsson, G., & Dettki, H. (2010). Does off-­trail backcountry skiing disturb moose? European Journal of Wildlife Research, 56(4), 513–518. https://doi.org/10.1007/s10344-009-0340-x Noss, R. F., Quigley, H. B., Hornocker, M. G., Merrill, T., & Paquet, P. C. (1996). Conservation biology and carnivore conservation in the Rocky Mountains. Conservation Biology, 10(4), 949–963. https://doi. org/10.1046/j.1523-1739.1996.10040949.x Olliff, T., Legg, K., & Kaeding, B. (1999). Effects of winter recreation on wildlife of the Greater Yellowstone Area: A literature review and assessment. Wyoming, MT: Yellowstone National Park. Olson, L. E., Squires, J. R., Roberts, E. K., Miller, A. D., Ivan, J. S., & Hebblewhite, M. (2017). Modeling large-­scale winter recreation terrain selection with implications for recreation management and wildlife. Applied Geography, 86, 66–91. https://doi.org/10.1016/j. apgeog.2017.06.023 Patthey, P., Wirthner, S., Signorell, N., & Arlettaz, R. (2008). Impact of outdoor winter sports on the abundance of a key indicator species of alpine ecosystems. Journal of Applied Ecology, 45, 1704–1711. https:// doi.org/10.1111/j.1365-2664.2008.01547.x Pépin, D., Adrados, C., Janeau, G., Joachim, J., & Mann, C. (2008). Individual variation in migratory and exploratory movements and habitat use by adult red deer (Cervus elaphus L.) in a mountainous temperate forest. Ecological Research, 23(6), 1005–1013. https://doi. org/10.1007/s11284-008-0468-2 Pépin, D., Adrados, C., Mann, C., & Janeau, G. (2004). Assessing real daily distance traveled by ungulates using differential GPS locations. Journal of Mammalogy, 85(4), 774–780. https://doi.org/10.1644/BER-022 OLSON et al. Polfus, J. L., Hebblewhite, M., & Heinemeyer, K. (2011). Identifying indirect habitat loss and avoidance of human infrastructure by northern mountain woodland caribou. Biological Conservation, 144(11), 2637– 2646. https://doi.org/10.1016/j.biocon.2011.07.023 Reilly, M., Beier, P., & Sonderegger, D. L. (2016). Spatial and temporal response of wildlife to recreational activities in the San Francisco Bay ecoregion. Biological Conservation, 207, 117–126. Reimers, E., Eftestøl, S., & Colman, J. E. (2003). Behavior responses of wild reindeer to direct provocation by a snowmobile or skier. Journal of Wildlife Management, 67(4), 747–754. https://doi. org/10.2307/3802681 Richard, J. H., & Côté, S. D. (2016). Space use analyses suggest avoidance of a ski area by mountain goats. The Journal of Wildlife Management, 80(3), 387–395. https://doi.org/10.1002/jwmg.1028 Riley, S. P. D., Sauvajot, R. M., Fuller, T. K., York, E. C., Kamradt, D. A., Bromley, C., & Wayne, R. K. (2003). Effects of urbanization and habitat fragmentation on bobcats and coyotes in southern California. Conservation Biology, 17(2), 566–576. https://doi. org/10.1046/j.1523-1739.2003.01458.x Riordan, E. C., & Rundel, P. W. (2014). Land use compounds habitat losses under projected climate change in a threatened California ecosystem. PLoS One, 9(1), e86487. https://doi.org/10.1371/journal. pone.0086487 Rixen, C., & Rolando, A. (Eds.) (2013). The impacts of skiing and related winter recreational activities on mountain environments. Sharjah: Bentham Books. Scott, D., Dawson, J., & Jones, B. (2008). Climate change vulnerability of the US Northeast winter recreation-­ tourism sector. Mitigation and Adaptation Strategies for Global Change, 13(5–6), 577–596. https:// doi.org/10.1007/s11027-007-9136-z Seip, D. R., Johnson, C. J., & Watts, G. S. (2007). Displacement of mountain caribou from winter habitat by snowmobiles. Journal of Wildlife Management, 71(5), 1539–1544. https://doi.org/10.2193/2006-387 Slauson, K. M., Zielinski, W. J., & Schwartz, M. K. (2017). Ski areas affect Pacific marten movement, habitat use, and density. The Journal of Wildlife Management, 81, 892–904. Squires, J. R., Decesare, N. J., Kolbe, J. A., & Ruggiero, L. F. (2010). Seasonal resource selection of Canada lynx in managed forests of the Northern Rocky Mountains. Journal of Wildlife Management, 74(8), 1648–1660. https://doi.org/10.2193/2009-184 Stankowich, T. (2008). Ungulate flight responses to human disturbance: A review and meta-­analysis. Biological Conservation, 141(9), 2159– 2173. https://doi.org/10.1016/j.biocon.2008.06.026 Stankowich, T., & Blumstein, D. T. (2005). Fear in animals: A meta-­analysis and review of risk assessment. Proceedings of the Royal Society B: Biological Sciences, 272(1581), 2627–2634. https://doi.org/10.1098/ rspb.2005.3251 Stewart, F. E. C., Heim, N. A., Clevenger, A. P., Paczkowski, J., Volpe, J. P., & Fisher, J. T. (2016). Wolverine behavior varies spatially with anthropogenic footprint: Implications for conservation and inferences about declines. Ecology and Evolution, 6(5), 1493–1503. https://doi. org/10.1002/ece3.1921 Tablado, Z., & Jenni, L. (2017). Determinants of uncertainty in wildlife responses to human disturbance. Biological Reviews, 92(1), 216–233. https://doi.org/10.1111/brv.12224 Töglhofer, C., Eigner, F., & Prettenthaler, F. (2011). Impacts of snow conditions on tourism demand in Austrian ski areas. Climate Research, 46(1), 1–14. https://doi.org/10.3354/cr00939 U.S.D.A. Forest Service (2015). White river national forest official webpage on vail pass winter recreation area. Retrieved from http://www.fs.usda.gov/recarea/whiteriver/recreation/ recarea/?recid=41445&actid=92 White, E. M., & Stynes, D. J. (2008). National forest visitor spending averages and the influence of trip-­t ype and recreation activity. Journal of Forestry, 106(February), 17–24. �| OLSON et al. White, E. M., Bowker, J. M., Askew, A. E., Langner, L. L., Arnold, J. R., English, D. B. K. (2016). Federal outdoor recreation trends: Effects on economic opportunities. Gen. Tech. Rep. PNW-GTR-945. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Station. 46 p. Woodroffe, R. (2000). Predators and people: Using human densities to interpret declines of large carnivores. Animal Conservation, 3, 165– 173. https://doi.org/10.1111/j.1469-1795.2000.tb00241.x Zhang, X., Cai, L., & Ni, S. (2006). Effect of the ice-­snow sports travels on the national economy and social development. China Winter Sports, 5, 29. S U P P O R T I N G I N FO R M AT I O N Additional supporting information may be found online in the Supporting Information section at the end of the article. How to cite this article: Olson LE, Squires JR, Roberts EK, Ivan JS, Hebblewhite M. Sharing the same slope: Behavioral responses of a threatened mesocarnivore to motorized and nonmotorized winter recreation. Ecol Evol. 2018;8:8555– 8572. https://doi.org/10.1002/ece3.4382 8569 �8570 | OLSON et al. APPENDIX F I G U R E A 1 Maps showing the location of yearly Canada lynx 95% minimum convex polygon home ranges (black lines), overlapped with recreation tracks (colored lines; green = snowmobile-­assisted hybrid skiing, blue = back-­country ski, orange = snowmobile, purple = packed-­trail ski) created by recreationists carrying handheld GPS devices in western Colorado, USA, 2010–2013. Gray polygon indicates a ski area boundary. Panel A shows lynx home ranges at the Vail study area, panel B shows the San Juan study area. Background image credit: Esri, DeLorme, USGS, NPS �| OLSON et al. 8571 TA B L E A 1 Summary statistics for recreation intensity on each Canada lynx’s yearly 95% minimum convex polygon home range (n = 22) in western Colorado, USA, 2010–2013. For each lynx’s yearly home range (HR), the number of recreation tracks that we recorded of each type in home ranges is given (#), along with total length of track (km) of each type recorded in each home range, the home range size (km2), the density of all recreation tracks combined (linear km of recreation tracks/km2 home range area), and the average (Avg) and standard deviation (SD) of trail counter hits per day at all trail counters within each lynx’s home range # Recreation tracks on HR Study Area Indv ID Hyb BKSki Vail Stafford Female 2011 134 27 Vail Turquoise Female 2013 4 Vail Stafford Female 2010 Vail Snmb km of Recreation tracks on HR HR size (km2) Track density km/km2 Avg counter hits/day SD counter hits/day PTSki Hyb BCski Snmb PTSki 126 2 1182 170 1079 17 31.37 78.00 53.11 47.29 13 25 24 32 44 300 100 45.03 10.57 65.67 0.0 0 0 2 0 0 0 11 0 14.72 0.74 86.27 74.61 Breckenridge Female 2011 0 3 0 1 0 12 0 4 34.56 0.47 16.83 5.94 Vail Breckenridge Female 2010 0 0 0 0 0 0 0 0 25.07 0.00 0.00 0.0 Vail Climax Female 2010 0 0 0 0 0 0 0 0 35.09 0.00 12.81 9.82 Vail Stafford Male 2011 132 24 123 3 494 167 435 21 92.14 12.11 24.16 28.42 Vail Turquoise Male 2013 4 20 25 29 41 59 410 157 135.43 4.93 46.85 25.07 Vail Half Moon Male 2013 4 14 25 18 23 41 163 66 175.41 1.67 34.38 25.33 Vail Climax Male 2011 0 17 0 2 0 41 0 6 85.06 0.55 18.69 12.33 San Juan Molas Female 2012 7 53 43 57 122 222 1660 215 106.77 20.79 11.21 4.86 San Juan Cement Female 2012 1 37 0 5 4 138 0 27 48.68 3.47 4.10 2.77 San Juan Hope Lake Female 2013 1 64 6 172 3 226 98 1118 67.36 21.45 30.25 22.07 San Juan Animas Female 2012 1 4 1 4 1 16 40 22 79.62 1.00 8.11 1.14 San Juan Ironton Female 2012 0 22 1 48 0 39 3 187 43.20 5.30 24.42 19.95 San Juan South Mineral Male 2012 15 107 95 116 415 483 3269 569 663.02 7.14 11.38 9.88 San Juan Cement Male 2012 2 102 3 8 20 328 10 40 104.52 3.80 5.87 7.57 �8572 | TA B L E A 1 OLSON et al. (Countinued) # Recreation tracks on HR Study Area Indv ID Hyb km of Recreation tracks on HR BKSki Snmb PTSki Hyb BCski Snmb PTSki HR size (km2) Track density km/km2 Avg counter hits/day SD counter hits/day San Juan Molas Male 2012 7 69 43 56 122 247 1660 226 157.63 14.31 10.81 4.85 San Juan Ophir Male 2012 2 184 29 70 35 883 508 267 175.51 9.64 29.10 19.80 San Juan Ironton Male 2012 0 22 1 48 0 35 3 186 42.26 5.32 27.45 16.08 San Juan South Mineral Male 2013 0 71 6 127 0 251 41 962 75.22 16.67 34.50 22.60 San Juan Ophir Male 2013 0 127 20 86 0 658 469 279 200.30 7.02 24.15 27.32 � https://cpw.cvlcollections.org/files/original/625917c1c6efe7f38f3a4fbc6e3ed059.pdf 5c7754a65de0998ce5ada8f8c3defd21 PDF Text Text 1 2 3 4 5 6 Appendix Figure 1: Maps showing the location of yearly Canada lynx 95% minimum convex polygon home ranges (black lines), overlapped with recreation tracks (colored lines; green=snowmobile-assisted hybrid skiing, blue=back-country ski, orange=snowmobile, purple=packed-trail ski) created by recreationists carrying handheld GPS devices. Gray polygon indicates a ski area boundary. Panel A shows lynx home ranges at the Vail study area, panel B is the San Juan study area. Background image credit: Esri, DeLorme, USGS, NPS. 7 9 1 �10 12 2 �13 14 15 16 17 18 19 20 21 Appendix Table 1: Candidate models considered for Canada lynx movement behavior (movement speed, movement tortuosity) in western Colorado, USA, 2010-2013. Models are ranked by AICc, with the best performing listed first in bold. Number of model parameters (K), difference in AICc values (ΔAICc), weight of individual models (AICcWT), and model log likelihood (LL) are shown. Model parameters include all 4 recreation intensity covariates (Rec; an additive formula of snowmobile, back-country ski, snowmobile-assisted hybrid ski, and packed-trail ski intensity at the 1km scale), sex of the individual (Sex), proportion of forested landcover type (Prop.Forest), weekday/weekend (DOW), day/night (TOD), and study area (Area; Vail or San Juan). Movement Speed Rec+Sex+Prop.Forest Rec*DOW+Prop.Forest Rec+ DOW +Prop.Forest Rec+Area+Prop.Forest Rec+Prop.Forest Rec*Prop.Forest Red+TOD+Prop.Forest Rec*Area+Prop.Forest Rec* TOD +Prop.Forest Prop.Forest Null K 9 13 9 9 8 12 9 13 13 4 3 AICc 591.55 603.13 608.64 612.43 613.02 614.87 614.95 617.5 618.33 622.94 633.57 ΔAICc 0.00 11.58 17.09 20.88 21.47 23.31 23.40 25.95 26.78 31.39 42.01 AICcWt 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 LL -286.70 -288.41 -295.24 -297.14 -298.45 -295.30 -298.40 -295.59 -296.01 -307.45 -313.77 Tortuosity Rec+TOD+Prop.Forest Rec * TOD +Prop.Forest Rec+Area+Prop.Forest Prop.Forest Rec +DOW+Prop.Forest Rec +Prop.Forest Rec *Area+Prop.Forest Rec +Sex+Prop.Forest Rec *DOW+Prop.Forest Rec*Prop.Forest Null K 9 13 9 4 9 8 13 9 13 12 3 AICc -213.29 -210.55 -206.01 -205.98 -205.22 -202.02 -200.48 -200.16 -199.62 -197.46 -195.70 ΔAICc 0.00 2.74 7.28 7.31 8.07 11.27 12.81 13.13 13.67 15.82 17.59 AICcWt 0.76 0.19 0.02 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 LL 115.72 118.43 112.08 107.01 111.68 109.07 113.40 109.15 112.96 110.87 100.86 22 3 �23 24 25 26 27 28 29 30 31 Appendix Table 2: Candidate models predicting Canada lynx selection (i.e., a third-order usedavailable resource selection function (RSF) design; Johnson, 1980; Manly et al., 2002) as a function of recreation intensity in western Colorado, USA, 2010-2013. Within recreation type, models are ranked by AICc, with the best performing listed first in bold. Type of recreation modeled (Mode), number of model parameters (K), difference in AICc values (ΔAICc), and model log likelihood (LL) are shown. Model parameters include recreation intensity measured at the 1km scale (Intensity1k), within 250m or not of high-intensity use trails (Trail250), within 500m or not of high-intensity use trails (Trail500), percent canopy cover (Canopy), study area (Area; Vail or San Juan), and day or night (Time). Mode Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Hybrid Model Hyb_Intensity1k*Area + Canopy Hyb_Intensity1k*Canopy Hyb_Intensity1k + Canopy Hyb_Intensity1k*Time + Canopy Hyb_Trail1k*Area Hyb_Trail1k*Canopy Canopy cover Hyb_Trail1k Hyb_Trail1k*Time Hyb_Intensity1k Null K 6 5 4 6 6 5 3 3 5 3 2 AICc 81556.20 81566.40 81654.50 81656.79 81660.60 81672.79 81766.20 95188.30 95190.59 95213.28 95394.90 ΔAICc 0.00 10.20 104.40 116.59 126.65 210.00 98.30 100.59 104.40 116.59 126.65 LL -40772.10 -40778.20 -40823.30 -40822.40 -40824.30 -40831.39 -40880.10 -47591.15 -47590.29 -47603.64 -47695.50 Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski Back-country Ski BC-Ski_Trail250*Area BC-Ski_Trail250+Canopy BC-Ski_Trail250*Canopy BC-Ski_Intensity1k*Area + Canopy BC-Ski_Intensity1k*Canopy BC-Ski_Intensity1k*Time + Canopy BC-Ski_Intensity1k + Canopy Canopy cover BC-Ski_Trail250 BC-Ski_Trail250*Time BC-Ski_Intensity1k Null 6 4 5 6 5 6 4 3 3 5 3 2 81198.67 81296.20 81296.40 81363.50 81594.20 81615.30 81619.00 81766.20 94475.08 94476.03 94768.85 95394.90 0.00 97.53 97.73 164.83 395.53 416.63 420.33 567.53 13276.41 13277.36 13570.18 14196.23 -40593.33 -40644.10 -40643.20 -40792.10 -40593.33 -40644.10 -40643.20 -40675.70 -40792.10 -40801.65 -40805.50 -40880.10 Snowmobile Snowmobile Snowmobile Snowmobile Snowmobile Snowmobile Snmb_Intensity1k*Area + Canopy Snmb_Trail1k*Canopy Snmb_Trail1k*Area Snmb_Trail1k+Canopy Snmb_Intensity1k*Canopy Snmb_Intensity1k*Time + Canopy 6 5 6 4 5 6 81350.50 81529.85 81564.13 81605.84 81682.10 81686.31 0.00 179.35 213.63 255.34 331.60 335.81 -40669.20 -40759.93 -40776.06 -40798.92 -40836.00 -40837.15 4 �Snowmobile Snowmobile Snowmobile Snowmobile Snowmobile Snowmobile Snmb_Intensity1k + Canopy Canopy cover Snmb_Trail1k*Time Snmb_Trail1k Snmb_Intensity1k Null 4 3 5 3 3 2 81689.60 81766.20 95210.63 95225.86 95391.61 95394.90 339.10 415.70 1827.00 13860.13 13875.36 14041.11 -40840.80 -40880.10 -41585.70 -47600.31 -47609.93 -47695.50 Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski Packed-trail Ski PT-Ski_Trail500*Canopy PT-Ski_Trail500*Area PT-Ski_Trail500+Canopy PT-Ski_Intensity1k*Canopy PT-Ski_Intensity1k*Area + Canopy PT-Ski_Intensity1k*Time + Canopy PT-Ski_Intensity1k + Canopy Canopy cover PT-Ski_Trail500 PT-Ski_Trail500*Time PT-Ski_Intensity1k Null 5 6 4 5 6 6 4 3 3 5 3 2 80345.95 80465.46 80548.53 81364.70 81448.80 81453.22 81459.10 81766.20 93326.41 93330.11 94516.62 95394.90 69.64 189.15 202.58 1088.39 1172.49 1176.91 1182.79 1489.89 12980.46 12984.16 14170.67 15118.59 -40168.00 -40226.70 -40270.26 -40677.30 -40718.40 -40720.60 -40725.60 -40880.10 -46660.21 -46660.05 -47255.31 -47695.50 32 5 �33 34 35 36 37 38 39 40 41 42 Appendix Table 3: Candidate models considered for proportion of time Canada lynx spent active versus stationary in response to temporal period for four types of winter recreation in western Colorado, USA, 2010-2013. Model parameters include time of day (TOD; day or night) and study area (Area; Vail or San Juan). Two candidate models for each recreation type were considered: an interaction of time of day with recreation intensity (TOD Interaction), and an interaction with study area plus time of day (Vail day, Vail night, San Juan day, San Juan night) with recreation intensity (Area+TOD Interaction). Within recreation type, models are ranked by AICc, with the best performing listed first in bold. Type of recreation (Mode), number of model parameters (K), difference in AICc values (ΔAICc), and model log likelihood (LL) are shown. The model including the effect of area was more supported in all cases. Mode Model K AICc ΔAICc LL Hybrid Hybrid Area+TOD Interaction TOD Interaction 9 5 47194.15 47242.35 0.00 48.20 -23588.07 -23616.17 Hybrid Null 2 47282.60 88.45 -23639.30 Back-country Ski Area+TOD Interaction 9 47184.78 0 -23583.4 Back-country Ski TOD Interaction 5 47226.69 41.91 -23608.3 Back-country Ski Null 2 47282.60 97.82 -23639.30 Snowmobile Area+TOD Interaction 9 47176.31 0.00 -23579.15 Snowmobile TOD Interaction 5 47226.82 50.51 -23608.41 Snowmobile Null 2 47282.60 106.29 -23639.30 Packed-trail Ski Area+TOD Interaction 9 47127.53 0.00 -23554.76 Packed-trail Ski TOD Interaction 5 47179.02 51.49 -23584.51 Packed-trail Ski Null 2 47282.60 155.07 -23639.30 43 44 6 �45 46 47 48 49 Appendix Table 4: Candidate models predicting the presence of Canada lynx GPS locations inside ski area boundaries in western Colorado, USA, 2010-2013. Number of model parameters (K), difference in AICc values (ΔAICc), and model log likelihood (LL) are shown. Model parameters include a continuous variable of Month, weekday/weekend (DOW), day/night (TOD), and percent canopy cover (Canopy). The top-performing model is given in bold. Model Structure K AICc ΔAICc LL Month*DOW+TOD+Canopy Month+DOW*TOD+ Canopy 7 7 6724.74 6738.43 0 13.69 -3355.37 -3362.21 Month*TOD+DOW+ Canopy 7 6744.68 19.94 -3365.34 Month+TOD+DOW+ Canopy 6 6755.06 30.32 -3371.53 Month+DOW+ Canopy 5 6757.33 32.59 -3373.66 Month+TOD+ Canopy 5 6777.96 53.22 -3383.98 Month+ Canopy 4 6780.19 55.45 -3386.09 DOW*TOD+ Canopy 6 6814.1 89.36 -3401.05 DOW+ Canopy 4 6830.11 105.37 -3411.05 Canopy 3 6853.32 128.57 -3423.66 TOD+ Canopy 4 6853.86 129.11 -3422.93 Null 2 6861.640 136.86 -3428.82 50 51 7 �Stafford Female 2011 Turquoise Female 2013 Stafford Female 2010 Breckenridge Female 2011 Breckenridge Female 2010 Climax Female 2010 Stafford Male 2011 Turquoise Male 2013 Half Moon Male 2013 Climax Male 2011 Molas Female 2012 Cement Female 2012 Hope Lake Female 2013 Animas Female 2012 Ironton Female 2012 South Mineral Male 2012 Cement Male 2012 Molas Male 2012 Ophir Male 2012 Ironton Male 2012 South Mineral Male 2013 Ophir Male 2013 134 4 0 0 0 0 132 4 4 0 7 1 1 1 0 15 2 7 2 0 0 0 27 13 0 3 0 0 24 20 14 17 53 37 64 4 22 107 102 69 184 22 71 127 126 25 2 0 0 0 123 25 25 0 43 0 6 1 1 95 3 43 29 1 6 20 km of Recreation Tracks on HR 2 24 0 1 0 0 3 29 18 2 57 5 172 4 48 116 8 56 70 48 127 86 1182 32 0 0 0 0 494 41 23 0 122 4 3 1 0 415 20 122 35 0 0 0 170 44 0 12 0 0 167 59 41 41 222 138 226 16 39 483 328 247 883 35 251 658 1079 300 11 0 0 0 435 410 163 0 1660 0 98 40 3 3269 10 1660 508 3 41 469 PTSki Vail Vail Vail Vail Vail Vail Vail Vail Vail Vail San Juan San Juan San Juan San Juan San Juan San Juan San Juan San Juan San Juan San Juan San Juan San Juan # Recreation Tracks on HR Snmb Indv ID BCski Study Area Hyb 56 PTSki 55 Snmb 54 BKSki 53 Appendix Table 5: Summary statistics for recreation intensity on each Canada lynx’s yearly 95% minimum convex polygon home range (n=22) in western Colorado, USA, 2010-2013. For each lynx’s yearly home range (HR), the number of recreation tracks that we recorded of each type in home ranges is given (#), along with total length of track (km) of each type recorded in each home range, the home range size (km2), the density of all recreation tracks combined (linear km of recreation tracks/km2 home range area), and the average (Avg) and standard deviation (SD) of trail counter hits per day at all trail counters within each lynx’s home range. Hyb 52 17 100 0 4 0 0 21 157 66 6 215 27 1118 22 187 569 40 226 267 186 962 279 HR size (km2) Track density km/km2 Avg counter hits/day SD counter hits/day 31.37 45.03 14.72 34.56 25.07 35.09 92.14 135.43 175.41 85.06 106.77 48.68 67.36 79.62 43.20 663.02 104.52 157.63 175.51 42.26 75.22 200.30 78.00 10.57 0.74 0.47 0.00 0.00 12.11 4.93 1.67 0.55 20.79 3.47 21.45 1.00 5.30 7.14 3.80 14.31 9.64 5.32 16.67 7.02 53.11 65.67 86.27 16.83 0.00 12.81 24.16 46.85 34.38 18.69 11.21 4.10 30.25 8.11 24.42 11.38 5.87 10.81 29.10 27.45 34.50 24.15 47.29 0.0 74.61 5.94 0.0 9.82 28.42 25.07 25.33 12.33 4.86 2.77 22.07 1.14 19.95 9.88 7.57 4.85 19.80 16.08 22.60 27.32 57 8 �58 59 60 61 62 63 64 Appendix Table 6: Coefficients (β) and confidence intervals (95% CI) for proportion of time Canada lynx spent active versus stationary in response to an interaction of temporal period and study area (Vail day, Vail night, San Juan day, San Juan night) with recreation intensity at the 1km scale. Each recreation type was modeled separately, the reference group for each model was ‘San Juan day’. Covariates whose 95% CI did not overlap 0 are bolded. Model predictions are visualized in the manuscript in Figure 5. Hybrid San Juan night Vail day Vail night Hybrid San Juan night:Hybrid Vail day:Hybrid Vail night:Hybrid Random Effect Back-country Ski San Juan night Vail day Vail night Back-country Ski San Juan night:Ski Vail day:Ski Vail night:Ski Random Effect Snowmobile San Juan night Vail day Vail night Snowmobile San Juan night:Snmb Vail day:Snmb Vail night:Snmb Random Effect Packed-Trail Ski San Juan night Vail day Vail night Packed-trail Ski San Juan night:PTSki Vail day:PTSki Vail night:PTSki Random Effect β Lower 95% Upper 95% 0.12 -0.24 0.19 -0.68 0.75 0.65 0.67 Var: 0.05 0.06 -0.47 -0.04 -1.06 0.31 0.27 0.29 SD: 0.23 0.18 -0.02 0.41 -0.3 1.2 1.03 1.05 0.07 -0.32 0.14 0.08 -0.03 -0.18 -0.06 Var: 0.06 0.02 -0.56 -0.09 0.05 -0.08 -0.32 -0.16 SD: 0.24 0.12 -0.09 0.37 0.12 0.01 -0.04 0.05 0.06 -0.32 0.12 0.06 0.03 -0.13 -0.06 Var: 0.05 0.02 -0.54 -0.09 0.02 -0.02 -0.22 -0.14 SD: 0.23 0.11 -0.1 0.34 0.1 0.08 -0.04 0.03 0.06 -0.35 0.26 0.03 0.1 -0.27 0.4 Var: 0.06 0.01 -0.62 0.01 -0.01 0.06 -0.66 0.07 SD: 0.25 0.1 -0.09 0.51 0.06 0.15 0.13 0.72 9 � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal Articles Description An account of the resource CPW peer-reviewed journal publications Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Sharing the same slope: behavioral responses of a threatened mesocarnivore to motorized and nonmotorized winter recreation Description An account of the resource <span>Winter recreation is a widely popular activity and is expected to increase due to changes in recreation technology and human population growth. Wildlife are frequently negatively impacted by winter recreation, however, through displacement from habitat, alteration of activity patterns, or changes in movement behavior. We studied impacts of dispersed and developed winter recreation on Canada lynx (</span><i>Lynx canadensis</i><span>) at their southwestern range periphery in Colorado, USA. We used GPS collars to track movements of 18 adult lynx over 4&nbsp;years, coupled with GPS devices that logged 2,839 unique recreation tracks to provide a detailed spatial estimate of recreation intensity. We assessed changes in lynx spatial and temporal patterns in response to motorized and nonmotorized recreation, as well as differences in movement rate and path tortuosity. We found that lynx decreased their movement rate in areas with high-intensity back-country skiing and snowmobiling, and adjusted their temporal patterns so that they were more active at night in areas with high-intensity recreation. We did not find consistent evidence of spatial avoidance of recreation: lynx exhibited some avoidance of areas with motorized recreation, but selected areas in close proximity to nonmotorized recreation trails. Lynx appeared to avoid high-intensity developed ski resorts, however, especially when recreation was most intense. We conclude that lynx in our study areas did not exhibit strong negative responses to dispersed recreation, but instead altered their behavior and temporal patterns in a nuanced response to recreation, perhaps to decrease direct interactions with recreationists. However, based on observed avoidance of developed recreation, there may be a threshold of human disturbance above which lynx cannot coexist with winter recreation.</span> Bibliographic Citation A bibliographic reference for the resource. Recommended practice is to include sufficient bibliographic detail to identify the resource as unambiguously as possible. Olson, L. E., J. R. Squires, E. K. Roberts, J. S. Ivan, and M. Hebblewhite. 2018. Sharing the same slope: behavioral responses of a threatened mesocarnivore to motorized and nonmotorized winter recreation. Ecology and Evolution 8:8555-8572. <a href="https://doi.org/10.1002/ece3.4382" target="_blank" rel="noreferrer noopener">https://doi.org/10.1002/ece3.4382</a> Creator An entity primarily responsible for making the resource Olson, Lucretia E. Squires, John R. Roberts, Elizabeth K. Ivan, Jacob S. Hebblewhite, Mark Subject The topic of the resource Anthropogenic disturbance Canada lynx Ski resorts Snowmobiles Space use Winter recreation Extent The size or duration of the resource. 18 pages Date Created Date of creation of the resource. 2018-07-30 Rights Information about rights held in and over the resource <a href="http://rightsstatements.org/vocab/InC-NC/1.0/" target="_blank" rel="noreferrer noopener">In Copyright - Non-Commercial Use Permitted</a> <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank" rel="noreferrer noopener">Attribution 4.0 International (CC BY 4.0)</a> Format The file format, physical medium, or dimensions of the resource application/pdf Language A language of the resource English Is Part Of A related resource in which the described resource is physically or logically included. Ecology and Evolution Type The nature or genre of the resource Article