Cole Burton, Christopher Beirne, Catherine Sun, Erin Tattersall, Joanna Burgar, Wildlife Coexistence Lab, Department of Forest Resources Management, University of British Columbia & Jason Fisher, School of Environmental Studies, University of Victoria, and Ecosystem Management Unit, InnoTech Alberta
July 2020
Executive Summary
Increasing rates of species endangerment and a growing human footprint necessitate ambitious efforts to protect and recover threatened species. This need is acute in working landscapes, where resource development must be compatible with the persistence of wildlife populations dependent on habitats that overlap resources of interest. An important example is the conservation of woodland caribou in western Canadian boreal forests, where declines of caribou populations have been linked to habitat disturbance from oil and gas extraction. In particular, seismic exploration lines cut extensively through these forests have been linked to altered predator-prey dynamics that result in unsustainable wolf predation on caribou. Accordingly, a key focus of caribou recovery efforts is the restoration of seismic lines in caribou habitats, with a goal of reducing line use by wolves and other predators, and thus restoring mammal community dynamics that are compatible with caribou conservation.
The Algar Caribou Habitat Restoration Program was a pioneering, industry-led initiative to restore legacy seismic lines in a portion of the Algar caribou herd range within the East Side Athabasca River population. Between 2012-2015, the Program treated 386 km of seismic lines, including 148 km receiving active restoration treatments (mounding, woody debris, tree planting) and 192 km designated for natural regeneration protection (i.e. passive restoration). We initiated the Algar Wildlife Monitoring Project in late 2015 to monitor wildlife responses to the restoration treatments (vegetation recovery was monitored by a separate program). We developed an experimental sampling design to assess the use of seismic lines by caribou, their predators, and other medium- and large-bodied mammals using noninvasive camera trap surveys. We deployed 73 camera trap stations across 5 sampling strata: actively restored lines, passively restored lines, unrestored lines left open as experimental controls or for human use, and off-line areas. Camera trap sampling concluded in November 2019, for a total sampling effort of 74,076 camera trap-days (averaging 1,015 days per station).
All motion-triggered images were processed to identify species and classify animal behaviours, while daily timelapse images were used to quantify snow cover and vegetation phenology. We developed a data management system to meet emerging camera trap metadata standards and facilitate efficient, repeatable analyses. We obtained 7,354 independent detections of medium- and large-bodied vertebrates, including 19 mammal species, as well as humans and several bird species. The most commonly detected mammals were white-tailed deer, black bear, and snowshoe hare, with an intermediate number of detections of wolf, moose, and caribou (see report section 4). Spatial patterns of detections showed some segregation of focal species according to major habitat preferences, e.g. caribou in lowlands and white-tailed deer in uplands. Temporal patterns suggested some trends in detections over the sampling period, most notably decreases in detections of wolves and coyotes, and increases in detections of caribou and white-tailed deer. We conducted a series of statistical analyses to rigorously evaluate species behavioural and population responses to restoration and other factors (sections 5-9).
We used generalized linear mixed models to estimate differences in habitat use by the focal species (caribou, wolf, black bear, white-tailed deer, and moose) across the 5 restoration treatment categories, while controlling for other covariates like habitat and seasonality. Models using data from the first 3 years of monitoring (without off-line sites) indicated that the short-term responses to restoration were muted, with no avoidance of restored lines by wolves or bears, but an indication that white-tailed deer used actively restored lines less frequently (published in Tattersall et al. 2020b, https://doi.org/10.1016/j.biocon.2019.108295; Appendix 7). We updated these models using the full 4 years of camera trap data and incorporating off-line samples as undisturbed references. Results from the updated analysis were broadly similar but highlighted that among the five focal species, only caribou did not prefer seismic lines over off-line areas, regardless of whether the lines were restored or not (section 5.1). More encouragingly, the models showed some subtle albeit mixed signals of restoration effectiveness: wolves used restored lines (active and passive) less than human-use lines, bears used passively restored lines less than other lines, deer and moose used actively restored lines less, and caribou used actively restored sites more than unrestored controls.
To gain a more complete picture of responses to restoration across the sampled vertebrate community, we used joint species distribution models to simultaneously model multispecies responses (section 5.2). We also used continuous descriptors of the characteristics of sampled seismic lines (line of sight, line width, mounding height and line density), rather than the discrete restoration categories used in our previous models, and we incorporated annual trends. Across the sampled community, line characteristics explained relatively little variation in species detections, as compared to other factors like season, habitat, and unexplained site-level variation (captured through random effects). Nevertheless, line characteristics—particularly line of sight—explained some variation in detections for several species, notably white-tailed deer, wolf, moose, and sandhill crane (all > 5% variation explained). Residual co-occurrence patterns from these multispecies models confirmed that caribou were grouping separately from their predators and apparent competitors. Model-based predictions of shifts in community structure under fully restored conditions (i.e. line characteristics consistent with restored lines) suggested that caribou may experience less predation risk under restoration due to declines in habitat use by wolves and coyotes. Nevertheless, assessing changes in species interactions is challenging; we completed an initial evaluation of predator interactions in the Algar landscape (published in Tattersall et al. 2020a, https://doi.org/10.1002/ece3.6028; Appendix 6), but we recommend further work to more directly investigate shifts in species interactions following restoration.
To enable assessment of population-level responses, we evaluated models for estimating population density of focal species (section 7). Since these species do not have unique markings that allow identification of individuals, we developed unmarked spatial count (SC) and spatial partial identity (SPIM) models to estimate densities of caribou and black bear in Algar in different years, and in a different landscape within the Richardson caribou range. While additional research is needed to further evaluate and improve these models, our estimates are the first for these species and landscapes, and suggest a local increase in caribou density within the Algar survey area from 2016 to 2019. These estimates are consistent with the hypothesis that linear restoration and wolf population management are improving conditions for caribou in this landscape. We created a decision framework to guide further research and application in density estimation using models for unmarked and partially marked populations.
While our project was focused on monitoring wildlife responses to restoration, we also collected daily timelapse images that allowed us to characterize the local environment at camera trap stations. We developed an approach to measure vegetation phenology and productivity from camera trap images, demonstrating different patterns in the understory dynamics relative to overstory phenology captured by satellite remote sensing (section 8). Our phenology metrics indicated that plants on passively restored lines showed phenological patterns more similar to off-line undisturbed conditions. Phenology on actively restored lines was more variable, and, for some measures (e.g. length of growing season, date of senescence, productivity) was more similar to unrestored control and human-use lines than to off-line or naturally regenerating lines. We showed that wildlife activity was related to phenology at different scales, with a strong link between the occurrence of migratory sandhill cranes and the vegetative growing season. Weaker but significant links were documented between vegetation productivity (measured as greenness) and annual and weekly detections for both caribou and white-tailed deer, which we suggest is consistent with these herbivores tracking forage availability along seismic lines. Our phenology analyses established the ability of camera trapping to monitor the progress of habitat restoration by evaluating vegetation characteristics and its impacts on wildlife species.
In a final exploratory analysis, we compared wildlife community structure and behaviours of ungulate herbivores between the Algar study area and another camera trap survey area in the Richardson caribou range (section 9). We found marked differences in species detections and behavioural patterns between the landscapes, consistent with our hypotheses that the relatively more disturbed Algar landscape would have higher proportions of wolves and deer, and that ungulate prey would show more risk-averse behaviours where there were more wolves. We believe that such landscape-scale comparisons across standardized camera trap surveys represent an important tool for assessing the effectiveness of linear restoration programs and other caribou recovery actions.
This project has clearly demonstrated the utility of camera trap surveys for monitoring restoration effectiveness in boreal environments. The cameras facilitated the collection of longitudinal data on the effects of seismic line restoration on wildlife behaviour and community composition, plant phenology and productivity, and the links between them. We found some evidence for positive outcomes emerging from the Algar restoration program, including trends towards reductions in predator activity and abundance. However, our results highlight that predators and other species continue to heavily use restored and unrestored lines in this landscape, and that the short-term responses to restoration are likely insufficient to drive rapid recovery of caribou. We recommend that future efforts carefully consider these results and attempt more aggressive methods for linear restoration (e.g. line blocking), alongside other recovery measures (such as wolf control), and rigorously evaluate the outcomes within an adaptive management framework (see section 10 for detailed recommendations). The relatively weak short-term effects observed in Algar highlight the need for long-term monitoring of wildlife and vegetation responses to restoration efforts, and for landscape-scale comparisons between different restoration techniques and environmental contexts.
Related Paper in Remote Sensing in Ecology and Conservation
Related Paper in the Journal of Applied Ecology
# 15-ERPC-04 / 16-ERPC-02 / 17-ERPC-02 / 18-ERPC-01 / 19-ERPC-04