Introduction

Several herds of southern mountain and boreal caribou (Rangifer tarandus) are facing imminent extirpation and the ultimate cause of their decline is believed to be anthropogenic disturbance (McLoughlin et al. 2003; Festa-Bianchet et al. 2011; Boutin et al. 2012; Johnson et al. 2015). Recent anthropogenic activities have extensively increased early seral stage habitats in caribou ranges and concurrently increased the density of moose (Alces alces), deer (Odocoileus spp.), and elk (Cervus elaphus) that thrive in early seral habitats (Serrouya et al. 2011). Known as apparent competition, the abundance of alternate prey within caribou ranges has resulted in a numerical response in wolves (Canis lupus) and possibly other predators such as bears and cougars (Kinley and Apps 2001) believed to be responsible for low caribou survival and recruitment (James et al. 2004; DeCesare et al. 2010). Historically, southern mountain and boreal caribou population coexisted with wolves at low densities by spacing themselves out within mature forest, thereby effectively reducing predation (Bergerud et al. 1984; Seip 1991).

Linear features from oil and gas, and forestry sectors including seismic lines, pipelines, and access roads represent more than 90 % of the industrial footprint within southern mountain and boreal caribou ranges (Sorensen et al. 2008). In addition to supporting apparent competition by increasing early seral habitats, linear features negatively impact caribou via direct habitat loss from cleared areas that convert mature forests into early seral habitats (Dyer et al. 2001), displacement and avoidance of anthropogenic features (Cameron et al. 1992; Nellemann and Cameron 1996; James and Stuart-Smith 2000), increased energy costs associated with avoidance of habitat disturbance (Bradshaw et al. 1998; Murphy and Curatolot 1987), improved travel efficiency for predators and humans (Latham et al. 2011; DeCesare 2012; Shanley et al. 2013), increased caribou–predator encounters (Whittington et al. 2011), and increased mortality from vehicle collisions (Wolfe et al. 2000).

Of particular importance, 5–15 m-wide conventional seismic lines established between the 1950s and early 2000s, hereafter referred to as legacy seismic lines, are pervasive across the boreal forest and natural regeneration of these seismic lines is slow (Oberg 2001; Lee and Boutin 2006; van Rensen et al. 2015). In Alberta, there are currently no restoration standards for legacy seismic lines because the intent was that these seismic lines would naturally regenerate to their pre-disturbance state. However, natural regeneration of legacy seismic lines is impeded by ground compaction and altered local hydrology caused by mechanical damage during their establishment, and by the ongoing physical damage and compaction from off-highway vehicles (OHV) traffic (Lee and Boutin 2006; van Rensen et al. 2015). Even though tree recruitment in boreal upland forest usually occurs within several years after disturbance (Chen and Wang 2006), natural regeneration on seismic lines is slow (Lee and Boutin 2006). Recent studies have shown that only 8 % of seismic lines within spruce forests recovered to at least 75 % cover of woody vegetation after 35 years (Lee and Boutin 2006) and that approximately one-third of legacy seismic lines in north-eastern Alberta failed to regenerate within 50 years (van Rensen et al. 2015).

Given the extent and pervasiveness of seismic lines within caribou ranges (Dyer et al. 2002; Lee and Boutin 2006), the high density of legacy seismic lines associated with the loss of mature habitat and increase of early seral habitat supporting high densities of alternate prey for wolves, and given that the federal government of Canada has set a 65 % disturbed habitat threshold within caribou ranges as part of the caribou recovery strategy (Environment Canada 2012), restoration of legacy seismic lines within caribou ranges is a priority (Alberta Woodland Caribou Recovery Team 2005). In most caribou ranges, the 65 % threshold has already been surpassed (Komers and Stanojevic 2013), and active management is therefore necessary to meet the planned recovery targets for the recovery of threatened populations in Alberta (Alberta Woodland Caribou Recovery Team 2005). However, the restoration of all legacy seismic lines is currently unrealistic because of the magnitude of these disturbances within caribou ranges and the costs associated with restoration (Schneider et al. 2010). A triage-type approach is therefore advocated (Bottrill et al. 2008; Schneider et al. 2010).

Our objective was to determine factors that best explained levels of motorized OHV use on legacy seismic lines intersecting roads to aid land managers in mitigating the negative impacts of OHV use within caribou ranges. By understanding factors associated with human use of legacy seismic lines, we can help deter human motorized use of seismic lines in areas prioritized for caribou recovery (Nellemann et al. 2010; Finnegan et al. 2013). We hypothesized that levels of OHV use on legacy seismic lines would be best explained by (1) local topography and vegetation attributes of seismic lines associated with ease-of-travel, either alone or in combination with broad-scale landscape attributes linked to (2) industrial activity, (3) recreation access, or (4) hunting (Table 1). We therefore developed four a priori hypotheses based on local topography and vegetation attributes of seismic lines, the presence of wildlife signs (tracks and trails), and the density and proximity of surrounding anthropogenic factors (i.e., road density, proximity to human settlements, and density of oil and gas facilities). We predicted that OHV use of legacy seismic lines would (1) be deterred by high vegetation, steep slopes, and wet soils, (2) be associated with high densities of oil and gas facilities and roads, and the proximity of young (0–15 years old) regenerating cutblocks, (3) be associated with the proximity of towns and campgrounds as a surrogate for recreation access, and (4) that OHV use would be different among administrative wildlife management units (WMU) and increase with the presence of ungulate signs present on seismic lines during the ungulate hunting season. To test our hypotheses (Table 1), we evaluated levels of OHV use on legacy seismic lines within the range of southern mountain and boreal caribou herds in west-central Alberta, Canada.

Table 1 Working hypotheses and predictions proposed to investigate human motorized use (OHV) of legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014 using cumulative mixed link ordinal regression
Full size table

Methods

Study Area

We investigated levels of OHV use within the foothills region of the A la Peche (ALP) and Little Smoky (LSM) herds of west-central Alberta, Canada, during the Summers of 2013 and 2014 (Fig. 1). The ALP range extends east from the British Columbia—Alberta border toward highway 40 (53°18′N, 119°20′W) south of Grande Cache and the LSM range extends east from highway 40 adjacent to the ALP range toward highway 32 (53°49′N, 118°29′W; Fig. 1). Elevation ranges between 681 and 3320 m with a gradient of lowland bogs and fens to upland mixed woods and conifer forests. Forested habitats are dominated by lodgepole pine (Pinus contorta), black spruce (Picea mariana), white spruce (Picea glauca), aspen (Populus tremuloides), and balsam poplar (Populus balsamifera). The ALP range includes Pierre’s Grey Provincial Park where hunting is prohibited, and part of the Willmore Wilderness protected area where hunting is permitted (Fig. 1). Lower elevations of the ALP range and the entire LSM range are extensively altered by anthropogenic activities associated with oil and gas exploration, forestry, and recreational activities. Respectively, 17 % of the ALP range is within 250 m of legacy seismic lines, while more than 83 % the LSM range is within 250 m (Sorensen et al. 2008). Within these caribou ranges, oil and gas development and forest harvesting began in the 1950s and has increased steadily since the early 2000s (White et al. 2011). Hunting, trapping, and fishing are other common anthropogenic activities in the region. The ALP herd is part of the southern mountain ecotype (Environment Canada 2014), and, the central mountain caribou population designable unit (COSEWIC 2014), is listed as “endangered” by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) and as “threatened” by the federal Species at Risk Act (SARA; Environment Canada 2014). The LSM herd is part of the boreal caribou ecotype and designatable unit and is listed as threatened by COSEWIC and the federal SARA (COSEWIC 2011; Environment Canada 2012). Southern mountain caribou typically winter in the foothills of the Rocky Mountains and migrate to their summer ranges in the mountains where they calve at high elevation (Saher and Schmiegelow 2005; Environment Canada 2014). However, the ALP herd has mostly abandoned their traditional winter range in the foothills and now uses a reduced range area at high elevations year-round (COSEWIC 2014). In comparison, boreal caribou inhabit low-elevation boreal forest throughout the year (COSEWIC 2011). Other ungulates in the area include deer, moose, elk, and big horn sheep (Ovis canadensis). Predators include wolves (Canis lupus), black bears (Ursus americanus), grizzly bears (Ursus arctos), cougars (Puma concolor), and wolverine (Gulo gulo).

Fig. 1
figure 1

Overview of the study area in Alberta, Canada, showing legacy seismic lines, subplots visited between June and October 2013 and 2014, communities, administrative WMU, major highways, and Parks (Jasper National Park, Willmore Wilderness, Wildland Parks, and Provincial Parks) in the vicinity, and within the ALP and LSM caribou ranges

Full size image

Attributes of Legacy Seismic Lines

We used base map data provided by the Government of Alberta under the Alberta Open Government License and a geographical information system (GIS) with a random number generator in ArcGIS 10.2.2 to subset legacy seismic lines that intersected roads in the ALP and LSM caribou ranges (http://data.alberta.ca/licence; Environmental Systems Research Institute [ESRI] 2014). We conducted field-based site investigations between June and October 2013 and 2014 (Fig. 1). On each legacy seismic line, we recorded information on at least one, and up to three subplots (10 m2) located 0, 100, and 500 m from the road intersection (Supplemental Material S1 Figure S1). We were unable to record information at 27 of the 100 m subplots and at 79 of the 500 m subplots because cutblocks effectively “erased” these seismic lines before we reached all distances (i.e., the seismic lines footprint became unclear). At each subplot, we recorded the observed level of OHV use (zero, light, moderate, and high OHV use; HUse), the maximum height of woody vegetation (m; VegHeight), the surface vegetation cover (%; VegCov), the prevalent soil type and the terrain wetness (fSoilWet; Table 2). We defined levels of OHV use as zero (no observable signs of OHV use, light (flattened vegetation), moderate (visible tracks), and high (deep ruts indicating heavy traffic). Although categorizing levels of OHV use is inherently subjective, we were interested in factors that could explain the relative difference among levels of OHV use in the study area rather than OHV use itself. Still, as an effort to minimize bias among field recorders, we (1) conducted extensive field training for all technicians as a group, and (2) assessed the usefulness of a random effect for recorders in our model structure. We measured the maximum height of woody vegetation using a marked 4-m tent pole held at the base of the tallest trees within each subplot. For trees > 5 m, the pole was used as a visual aid to estimate tree height. We estimated the surface vegetation cover within each subplot ocularly. We categorized each subplot within a wetness index (Wet, MesicWet, MesicDry, Dry) based on field observations of the water table (where applicable) and predominant soil type: we associated rock, gravel, and sandy soils to dry conditions, loam to mesic conditions, and clay and organic soils to wet conditions. We recorded the presence of wildlife trails, and when present we recorded the ungulate species of wildlife tracks and pellets (UngulateSigns). At each subplot investigated we determined the slope angle using data derived from a 5-m resolution digital elevation model build from LiDar-derived products provided by Alberta Environment and Parks, Government of Alberta (Slope), and also generated broad-scale landscape GIS-derived attributes associated with human activities: The density of oil and gas facilities and roads within a 5-km radius moving window (DensityOilGas and RdDensity), the Euclidean distance from the legacy seismic line and road intersection to the nearest young regenerating cutblock (0–15 year old; DistCuts), and the nearest campground (DistRecCamps). Base map data were provided by the Government of Alberta under the Alberta Open Government License. Forest harvest data were provided by West Fraser and Weyerhaeuser. We also calculated the traveling distance (on-road distance) to the nearest town (population > 1000) using the least cost path tool in ArcGIS 10.2.2 (DistTown; Environmental Systems Research Institute [ESRI] 2014). We only considered young regenerating cutblocks because cutblocks older than 15 years old have reached provincial “free-to-grow” standards and no longer require the attention of forest companies (Weyerhaeuser pers. comm). Finally, we also categorized each legacy seismic line within three administrative WMU that have their own hunting regulations because hunting pressure can vary according to the availability of special licenses and tags allotted (WMU; http://aep.alberta.ca/fish-wildlife/fishing-hunting-trapping/hunting-alberta/wildlife-management-units.aspx). See Table 2 for a full description of covariates.

Table 2 Field-based and GIS-based variables used to investigate human motorized use (OHV) of legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014 using cumulative mixed link ordinal regression
Full size table

Statistical Analyses

We used ordinal logistic regression to investigate the relationship among four levels of human motorized use of legacy seismic lines (none, light, moderate, and high), the level of re-vegetation, local topography, and broad-scale landscape attributes related to human activities (industrial vs. recreation) using cumulative mixed link models (CLMMs) from the ordinal package in R (Table 2; Christensen 2013; R Core Team 2014). The CLMM package allowed us to account for the spatial autocorrelation among subplots on legacy seismic lines, autocorrelation among seismic lines, and the different field recorders (Breslow and Clayton 1993; Zuur et al. 2009). CLMMs are also commonly known as proportional odds models and are appropriate for ranked ordinal-dependent variables (i.e., ranked OHV use levels; McCullagh 1980). We compared model fit between the logit link suggested for uniformly distributed outcomes and the log–log link suggested for distributions with increasing rank probabilities using Akaike Information Criterion (AIC). We determined the appropriate autocorrelation structure on the saturated model using ANOVA and assessed goodness-of-fit using profile likelihood of the fixed-effect model (Smith and McKenna 2012; Christensen 2013).

We used an information theoretic model selection approach with multiple working hypotheses based on AIC to obtain probabilities of OHV use from vegetation, local topography, and broad-scale landscape attributes linked to human activities (Burnham and Anderson 2002). We assessed collinearity using variance inflation factors (VIFs) (Zuur et al. 2010) and standardized density and distance variables to improve model convergence. We first tested for the appropriate threshold structure (flexible, symmetric, or equidistant) among levels of OHV use for the saturated model using AIC, and subsequently used the threshold structure with the lowest AIC for our hypothesis-driven model selection (Table 1; Supplemental Material S2 Table 1). We also assessed potential differences in OHV use between the ALP and LSM ranges a priori in the saturated model (Supplemental Material S2 Table 2). We ranked candidate models according to AIC and used Akaike weights (ω) to determine the hypotheses that best explained the level of OHV use on legacy seismic lines for the ALP and LSM ranges (Table 3).

Table 3 AIC, delta AIC (Δ AIC), AIC weight (ω), and k (number of parameters) for the candidate models investigating human motorized use (OHV) of legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014 using cumulative mixed link ordinal regression
Full size table

For the best-selected model, we calculated the McFadden pseudo R 2 and the odds ratio (OR) for each model variable as a measure of effect size (Table 4; McFadden 1974; Grimes and Schulz 2008; Le and Marcus 2012; Smith and McKenna 2012). McFadden pseudo R 2 was obtained by comparing the log likelihood of univariate models to the log likelihood of the null model (McFadden 1974).

Table 4 Parameter estimates (β), lower (LCI), and upper (UCI) 95 % confidence interval, Z-statistic (Z-Value), P-value (P-value), and OR for the best selected CLMM used to investigate human motorized use (OHV) of legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014
Full size table

Post-Hoc Piecewise Regression

Using results from the best-selected model, we used piecewise regression to define breakpoints in vegetation height where the relationship between the probabilities of “Zero Use” and “High Use” of legacy seismic lines from OHV changed. Although the relationship between OHV use and vegetation height was continuous for each level of OHV use, breakpoints identified for “Zero”-use and “High”-use categories allowed us to identify a threshold in vegetation heights, where OHV use was likely impeded. We partitioned the probabilities of OHV use into two categories based on whether the observed level of OHV use was “Zero” or “High” and used linear regression to plot each of the predicted OHV use levels against vegetation height (Tables 4 and 5; Fig. 2). We used the piecewise regression function from the SiZer package in R with 1000 bootstrap replicates and α = 0.05 (Fig. 2; Sonderegger 2012; R Core Team 2014). We conducted all statistical analyses in program R (R v 3.1.1; R Core Team 2014).

Table 5 Parameter estimates (β), lower (LCI), and upper (UCI) 95 % confidence interval, Z-statistic (Z-value), P-value (P-value), and OR for the second best selected CLMM used to investigate human motorized use (OHV) of legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014
Full size table
Fig. 2
figure 2

Probability of OHV use being “Zero” a, “Light” b, “Moderate” c, and “High” d as a function of vegetation height (m) on legacy seismic lines in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013–2014 using CLMM (Tables 3 and 4). The vertical dashed lines represent the vegetation height where the probability of OHV use being zero (4.3 m; panel A) and the probability of high OHV use (2.4 m; panel D) changed in relation to vegetation height according to post-hoc piecewise regression

Full size image

Results

We surveyed 66 legacy seismic lines in the ALP and 202 in the LSM caribou ranges for a total of 699 10 m2 subplots. While no OHV use was observed on 58 % of legacy seismic lines visited in the ALP (61 %) and LSM (57 %) ranges, we observed low, moderate, and high OHV use on 27, 8, and 6 % of legacy seismic lines assessed within the ALP and LSM caribou ranges. When fitting the CLMMs, the logit link performed better and the flexible threshold structure was best (Supplemental Material S2 Table 1). We found no difference in the log-likelihood of saturated models including a random effect for field recorders, subplot levels, and subplots nested within seismic lines and therefore only retained the random effect at the subplot level in subsequent models (saturated model σ 2 = 9.7, SD = 3.1). We also found no difference between herd ranges and therefore combined data from the ALP and LSM ranges in models (Supplemental Material S2 Table 2). VegHeight and VegCov were correlated (r: 0.5), and because measures of vegetation heights were more objective than percent cover estimates and more useful as a potential management tool, we performed model selection using the VegHeight variable only. See Supplemental Material S3 for model selection using VegCov. All VIFs of variables included in the same models were <3. Profile likelihood of the saturated cumulative link model are shown in Supplemental Material S4 (Christensen 2013).

Contrary to our hypotheses, we found no support for associating levels of OHV use of legacy seismic lines to industrial, recreation access, or hunting activities (Hypotheses 2–4; Tables 1 and 3). The best selected model with an Akaike weight (ω) of 0.5 was the model in support of local topography and vegetation attributes of legacy seismic lines (Hypothesis 1; Tables 1, 3, and 4). The second best selected model was in support of the hunting hypothesis (4a) and had a ω of 0.3 and a Δ AIC of 1.0 (Tables 1 and 3). However, even in the model associated with the hunting hypothesis (4a; Tables 1 and 3), the only variables with coefficients and 95 % confidence intervals that did not overlap zero were associated with the local topography and vegetation attributes of legacy seismic lines (Tables 4 and 5). Distance to nearest towns, recreation campground, young regenerating cutblocks, and the density of roads and active oil and gas facilities did not explain levels of OHV use on legacy seismic lines in the ALP and LSM caribou ranges (Table 3). We also found no difference in the level of OHV use of legacy seismic lines according to administrative WMU, and the presence of ungulate signs on seismic lines in autumn did not explain the level of OHV use (Tables 3 and 5).

High vegetation (VegHeight) and high percent cover of surface vegetation on legacy seismic lines (VegCov) were associated with low OHV use of seismic lines (Tables 3–5; Supplemental Material S3, Fig. 2). Slope angle was not associated with the level of OHV use of legacy seismic lines, but wet seismic lines were used less than expected (Tables 3–5). The percent cover of surface vegetation on legacy seismic lines had the highest pseudo R 2 followed by the height of vegetation, soil wetness, and slope (pseudo R 2 VegCov: 0.03, VegHeight: 0.01, fSoilWet-Wet: 0.004, Slope: 0.0002).

Post-Hoc Piecewise Regression

According to piecewise regression, the relationship between vegetation height and the probability of OHV use being zero changed at a vegetation height of 4.3 m, while the relationship between vegetation height and a high probability of OHV use changed at a vegetation height of 2.4 m (Fig. 2). At a vegetation height below 4.3 m, the probability of OHV use being zero increased by approximately 9 % for every 1 m increase in vegetation height, while at a vegetation height greater than 4.3 m the probability of OHV use being zero increased by approximately 2 % with every 1 m increase in vegetation height, a 78 % decrease in response to 1 m increase in vegetation height (Fig. 2a). At a vegetation height below 2.3 m, the probability of OHV use being high decreased by approximately 2 % for every 1 m increase in vegetation height, while at vegetation heights greater than 2.4 m the probability of OHV use being high decreased by <1 % with every 1 m increase in vegetation height (Fig. 2d).

Discussion

Using data collected in the ALP and LSM caribou ranges of Alberta, Canada, between June and October 2013 and 2014, we demonstrated that OHV use of legacy seismic lines was mainly associated with local topography and vegetation attributes of seismic lines rather than hypothesized broad-scale landscape attributes related to industrial activity, recreation access, or hunting. OHV use of legacy seismic lines was highest on low vegetation, dry seismic lines, and contrary to our hypotheses we found no support for models associated with industrial or recreational activities. Overall, our results suggest that within the ALP and LSM ranges, ease-of-travel best predicts OHV use of legacy seismic lines.

Environmental assessments of OHV use are common (see, e.g., Barton and Holmes 2007; Groom et al. 2007; Pierskalla et al. 2011; Steenhof et al. 2014). However, our study is unique in that it is the first to investigate factors associated with different degrees of use on legacy seismic lines by OHV users. Among the variables investigated, vegetation height and percent surface vegetation cover were more associated with OHV use of seismic lines, and piecewise regression indicated that once the vegetation on seismic lines regenerated to a height between 2.4 and 4.3 m, OHV use was minimal. Investigating natural regeneration of legacy seismic lines, van Rensen et al. (2015) argued that seismic lines with vegetation that had regenerated to a height of at least 3 m were more likely to fully regenerate because a 3 m vegetation height was more representative of trees rather than shrubs. Our results are in accordance with van Rensen et al. (2015), and suggest that once legacy seismic lines have regenerated to a height of at least 2.4 m, seismic lines become less attractive to OHV users and are therefore less subject to soil erosion, compaction, active clearing, and vegetation damage.

As expected (Table 1), OHV use was high on legacy seismic lines with low percent surface vegetation cover and low vegetation height. OHV use damages vegetation, impedes seed germination, and increases soil compaction and erosion (Lee and Boutin 2006; Pierskalla et al. 2011; van Rensen et al. 2015). It is therefore likely that soil compaction, erosion, active clearing for trap lines and other OHV use during winter and summer, and damage to young trees, shrubs, and vegetative growth from OHV use impede vegetative re-growth and therefore perpetuate high OHV use on low vegetation legacy seismic lines, effectively creating a positive feedback loop.

Along with low vegetation, high OHV use was associated with dry seismic lines, supporting our hypothesis that ease-of-travel is one of the main factors explaining OHV use of legacy seismic lines. Even though van Rensen et al. (2015) found that legacy seismic lines with wet soil types are the slowest to regenerate, it is also likely that increased OHV use on dry seismic lines hinders natural regeneration of well-traveled legacy seismic lines. Our results indicate that to improve revegetation rates on dry seismic lines, legislative approaches to prohibit use within specific areas, or the use of physical barriers such as boulders and rocks, soil berms, or tree-jams that would impede OHV access could be considered. However, the restoration of legacy seismic lines with wet soil types might require additional restoration measures such as mounding and tree-planting, for example.

Unlike Lee and Boutin (2006), we found no support for increased use of legacy seismic lines with proximity of forestry activities, or with increased density of oil and gas facilities and roads. Our results indicate that across our study area, industrial facilities and regenerating cutblocks are mainly accessible by roads, and that facilities that are only accessible via OHV are relatively rare. Also, even though recreational OHV users might specifically seek areas that are difficult to access with the use of regular vehicles, we found no support for increased OHV use with increased distance from towns, and with the proximity of recreational campgrounds. Future research might benefit from investigating legacy seismic lines in isolated areas because identifying a decay function in travel distances from towns and on seismic lines for OHV users could be useful to target restoration of legacy seismic lines. We also found no difference in the level of OHV use among administrative WMU, and with the presence of ungulate signs. It is therefore possible that our recreation access-related and hunting-related variables failed to capture important landscape features targetted among these different OHV user groups. For example, we were unable to assess the influence of unofficial camp sites that are likely frequented by recreational users, and it is likely that official campgrounds fail to adequately describe OHV use in our area. It is also likely that the ability to park near seismic lines intersecting roads influence OHV use of seismic lines. The low R 2 obtained from variables included in models associated with the ease-of-travel hypothesis suggest that other factors, along with local topography and vegetation attributes of seismic lines, likely play an important role in OHV use of seismic lines. We therefore recommend that future research aimed at understanding the use of legacy seismic lines by OHV user groups incorporate user-specific structured surveys (e.g., Flood 2005; Shanley et al. 2013). User-specific surveys (i.e., surveys targeted at recreational users) may shed light onto factors influencing OHV use of legacy seismic lines, and may also help identify under-represented natural attractions such as waterfalls or sand dunes in the vicinity of high-use seismic lines.

To our knowledge, this is the first study that attempts to explain OHV use of legacy seismic lines based on seismic line attributes associated with vegetation, local topography, and broad-scale landscape attributes linked to human activities. Using cumulative mixed link ordinal regression, we were able to successfully identify variables associated with the probability of zero, light, moderate, and high OHV use of legacy seismic lines. Our results can therefore be used to target restoration of legacy seismic lines with high probability of OHV use that are also located within caribou ranges. We found that once vegetation on seismic lines regenerated to a height between 2.4 and 4.3 m, OHV use was minimal, and because a triage-type approach is necessary to target restoration of early seral habitats within caribou ranges to impede OHV use, restoration of dry, low vegetation (<2.4 m) seismic lines could be prioritized. Once vegetation on these seismic lines grows taller than 2.4–4.3 m, these seismic lines could be re-attributed to a low priority. To facilitate targeted restoration of legacy seismic lines, further research should investigate the usefulness of fine-grained resolution tools such as LiDar and sattelite technology to identify seismic lines prefered by OHV users using local topography and seismic line attributes.

Management Implications

As part of the federal caribou recovery strategies, a minimum of 65 % undisturbed habitat within boreal and southern mountain caribou ranges, including low elevation and connectivity corridors, is required to achieve a 60 % probability of self-sustaining populations (Environment Canada 2012, 2014). However, this threshold has been surpassed in the ALP and LSM ranges (Environment Canada 2011; Komers and Stanojevic 2013), largely due to linear features such as access roads and legacy seismic lines. Our results indicate that OHV use may be delaying vegetation regrowth on dry, low-vegetation seismic lines within the ALP and LSM caribou ranges. Impeding OHV use on a selected portion of legacy seismic lines within caribou ranges might be an attractive, low-cost option to expedite caribou habitat recovery and meet federal recovery targets. Based on our results and previous findings (van Rensen et al. 2015), dry seismic lines with vegetation heights of <2.4 m could be prioritized for restoration within caribou ranges to mimimise the deterimental effects of OHV use on seismic line regeneration. While managing recreational use of OHV is challenging (Brooks and Champ 2006; Kuehn et al. 2011), techniques that can successfully deter OHV use of legacy seismic lines could include (1) legal restrictions on OHV use within specific areas (Kuehn et al. 2011; Pierskalla et al. 2011; Shanley et al. 2013), (2) creating designated OHV trails outside of identified caribou habitat (Nellemann et al. 2010; Shanley et al. 2013), (3) use of educational tactics such as posters and kiosks, and partnership with OHV assocations to improve awareness and promote care of widlife species and natural systems (Kuehn et al. 2011; Pierskalla et al. 2011), and (4) physical barriers such as boulders and rocks, soil berms, or tree-jams to impede access of legacy seismic lines with high potential for OHV use in caribou ranges. Although costly, tilling, mounding, and replanting of seismic lines could also be used as tools to restore low-vegetation dry seismic lines within caribou ranges. Considering that OHV use was observed on >40 % of legacy seismic lines within the ALP and LSM ranges, and considering that natural regeneration is impeded by soil compaction, erosion, active clearing, and vegetation damage from OHV use, management actions that reduce OHV use of legacy seismic lines could benefit caribou recovery in Alberta, Canada, and elsewhere.