Where to begin? A flexible framework to prioritize caribou habitat restoration
Author contributions: MD, TH contributed to data collection, analyses, research design, and manuscript preparation; CB contributed to manuscript preparation; MC, KB, MK contributed to research design and manuscript preparation; ML contributed to data collection, analyses, and manuscript preparation; SB, RS contributed to research design and manuscript preparation.
Abstract
Habitat loss is a leading threat to many species at risk, and as such, the need for habitat restoration is widespread. In the boreal forests of Western Canada, habitat restoration is a key habitat management action needed to achieve self-sustaining populations of woodland caribou, a federally threatened species in decline. Hundreds of thousands of kilometers of linear features were created during the exploration or extraction of oil and gas that are no longer used, yet natural regeneration remains stagnated. Only a fraction of these linear features is restored each year, sparking the need for managers to prioritize efforts. We developed an algorithm to prioritize habitat restoration and demonstrate how it can be used to predict and monitor progress towards restoration goals. Our approach is based on the idea of maximizing the gain in unaltered caribou habitat per unit cost, while allowing for the inclusion of different goals, costs, and weighting criteria. Our algorithm ranked landscape units into five zones of restoration priority. The largest gain in unaltered habitat occurred following restoration of the highest priority zones, with diminishing returns as restoration proceeded. None of the caribou ranges reached habitat management targets when not considering restoration within energy project boundaries, even after all candidate linear features were restored. Our results highlight the need for ambitious, coordinated restoration, and the need for improved land-use planning to minimize alteration within caribou range. We demonstrate the flexibility of our algorithm by applying the framework to a case study in a mountain ecosystem.
Implications for Practice
- Through flexible algorithms that explicitly identify benefits, costs, and weightings, habitat restoration prioritization can guide efficient allocation of resources to improve conservation outcomes.
- Prioritization frameworks can also be used to predict success against restoration targets. Our results demonstrate that ambitious and collaborative efforts will be needed to meet restoration targets for woodland caribou in the boreal forests of Canada.
Introduction
Humans have had a vast impact on ecosystems worldwide. Land conversion of natural habitats and fragmentation poses an ongoing threat to biodiversity (Dobson et al. 1997; Tilman et al. 2017, 1994). Even in Canada, which is considered to be more intact than other jurisdictions (Watson et al. 2016; Coristine et al. 2019), habitat loss is the leading threat to many species at risk, and the risk has increased over time (Venter et al. 2006; Woo-Durand et al. 2020). Intact landscapes provide a range of options to protect and maintain biodiversity and ecosystem function. However, many landscapes are degraded to a point that threatens species persistence, and restoration is necessary to meet species-specific or ecosystem-based recovery goals (Possingham et al. 2015). Given that the pace of restoration may be limited by available resources, there is a need to prioritize where to start restoration based on ecological opportunities and economic costs (Joseph et al. 2009; Holl & Aide 2011).
Anthropogenic habitat alteration has been identified as a primary cause of the decline of Canada's woodland caribou (Rangifer tarandus), a species federally listed as threatened (COSEWIC 2002; Environment Canada 2012). Caribou live at low densities in boreal and montane forests spanning from the Atlantic to the Pacific coast. In spite of their broad distribution, caribou abundance is declining across most of their range (Festa-Bianchet et al. 2011). The conversion of mature forest stands to early seral vegetation, for example through forestry, creates forage for ungulates such as moose (Alces alces) and deer (Odocoileus spp.; Mumma et al. 2020), leading to increased predator abundance, and correspondingly higher predation on caribou (Serrouya et al. 2021). Furthermore, linear features associated with resource exploration and extraction, such as roads and seismic lines, facilitate the incursion of predators into caribou refugia (Davidson et al. 2020) and increase predator hunting efficiency (Dickie et al. 2017). Although the mechanisms resulting in caribou declines are complex, and undoubtedly exacerbated by climate-mediated effects on the boreal food web, anthropogenic habitat alteration has been identified as playing a key role (Johnson et al. 2020).
Caribou declines are highly pronounced in western Canada (McLoughlin et al. 2003; Hervieux et al. 2013). In these areas, petroleum exploration and forestry have created networks of linear features; exceeding 350,000 km across boreal caribou ranges, the majority of which are legacy seismic lines (Pattison et al. 2016). While some of these features continue to be actively used for resource extraction or recreation, many features are no longer in use (i.e., legacy features). Despite discontinued use, vegetation on legacy features tends to regenerate slowly or stagnate. Less than 10% of legacy seismic lines have advanced vegetation regeneration (Lee & Boutin 2006; Finnegan et al. 2018; Nagy-Reis et al. 2021). To support self-sustaining caribou populations, large unaltered tracts of land are required (Environment Canada 2012; Nagy-Reis et al. 2021). Thus, restoring linear features to return forest cover while simultaneously reducing predator hunting efficiency is a key action identified in caribou recovery strategies (Environment Canada 2012; Environment and Climate Change Canada [ECCC] 2020). Habitat alteration within existing and future energy development programs are subject to reclamation requirements through the Alberta Environmental Protection and Enhancement Act. However, many legacy features fall outside of these areas, and guidance is needed to support planning of collaborative restoration programs. With hundreds of thousands of kilometers of these legacy features across caribou ranges, and only a fraction being restored each year, managers must prioritize where to begin to make the best use of limited resources (Johnson et al. 2020; Nagy-Reis et al. 2021).
Systematic conservation planning frameworks (Margules & Pressey 2000), alongside multicriteria decision-making analyses (Adem Esmail & Geneletti 2018), provide data-driven processes in which clear and explicit criteria guide restoration planning and prioritization (Gilby et al. 2021; Martin et al. 2022). Prioritizing habitat restoration can occur at various scales of interest, from choosing between individual features within a specific area, to choosing between different subpopulations (Gilby et al. 2021). Previous linear feature restoration prioritization efforts have identified particular features or areas where restoration is needed using graph-theoretic approaches (Yemshanov et al. 2021), maximizing recovery potential using modeled vehicle traffic on seismic lines (Pigeon et al. 2016; Hornseth et al. 2018), and identifying areas where vegetation recovery is lacking (Van Rensen et al. 2015). We build on this foundation to create larger “bite-sized” regions in which restoration can maximize the ratio of gain in unaltered caribou habitat per unit cost within particular subpopulations, or even across subpopulations, while recognizing the needs of multiple organizations and the continued economic demands on the landscape (Bode et al. 2010; Smith et al. 2022).
Our focus is to develop a prioritization algorithm and demonstrate how it can be used to predict and monitor progress against Canada's federal recovery strategy's maximum disturbance target of 35% total altered habitat within each caribou range (Environment Canada 2012). There are six components of the framework: (1) identify landscape units in which to prioritize restoration; (2) calculate the benefit of restoration; (3) calculate the cost of restoration; (4) incorporate additional weighting criteria; (5) calculate the weighted gain per cost for each landscape unit; and (6) group units into ranked zones from highest to lowest weighted gain per cost. The main benefits of our approach are that it is tangible for managers to implement, and flexible enough to allow for the inclusion of different goals, costs, and weighting criteria. We demonstrate this flexibility by applying the framework first to boreal Alberta, then to an additional case study in a mountainous system. Although the context of this paper is restoring caribou habitat, the general framework has wide applicability for restoration on any working landscape, which has been identified as a global need (Wilson et al. 2006; Holl & Aide 2011).
Methods
Study Area
We developed the prioritization process for caribou ranges within the Oil Sands Region of Alberta, Canada. The Oil Sands Region covers 140,213 km2 of northern Alberta and overlaps seven caribou ranges (Fig. 1). The area has a linear feature density of more than 2 km/km2 (Alberta Biodiversity Monitoring Institute 2018a). We defined our study area as the extent of caribou ranges that predominantly fell within the Oil Sands Region (Richardson, West Side Athabasca River [WSAR], East Side Athabasca River [ESAR], Red Earth and Cold Lake; 69,309 km2). We excluded the Nipisi and Chinchaga ranges because most of these ranges are outside of the Oil Sands Region.
The study area is within the boreal forests of Alberta, with a mixture of upland forests interspersed with bog and fen peatland complexes, marshes, and swamps. There is limited topographic relief, with various small lakes and rivers. The climate is characteristic of the Boreal Plains Ecozone, with low precipitation (~450 mm/year). Ungulate species include moose, woodland caribou and white-tailed deer (Odocoileus virginianus). Predators of those ungulates include gray wolves (Canis lupus), coyote (Canis latrans), black bears (Ursus americanus) and lynx (Lynx canadensis). Other important prey includes beaver (Castor canadensis) and snowshoe hare (Lepus americanus).
The Prioritization Process
The goal of this work is to provide a tool for prioritizing restoration. We prioritized areas within each caribou range by developing criteria that quantified the increase in unaltered caribou habitat as a result of seismic line restoration (“gain in unaltered habitat”) relative to the cost of restoration, and included weighting criteria to reduce the potential of being altered again in the future and to increase the benefit to where caribou are known to be currently using habitat (Fig. 2).
All analyses were conducted using ESRI ArcGIS and R (R Core Team 2019). Data and code used in the prioritization process are available at https://github.com/MelanieDickie/RestorationPrioritization.
Identifying Landscape Units
First, we defined the landscape units across which restoration would be prioritized. We used publicly available townships to delineate landscape units, which are approximately 9.6 km × 9.6 km. These units represent an operationally appropriate scale whereby altered habitat in one area can be accessed and restored within a season and reflects the difficulties of moving people and equipment to remote worksites. While these landscape units were chosen based on an operational scale, restoration of multiple townships is aligned with the spatial needs of caribou for large patches of intact forest (Tracz et al., 2007). The study area included 894 townships. We included townships within caribou range buffered by 500 m to reflect edge effects of habitat alteration just outside of caribou range.
Calculating Gain in Unaltered Caribou Habitat Following Restoration
We defined habitat alteration as the footprint of anthropogenic land-use, buffered by 500 m, plus burned areas <40 years old, as per Canada's federal recovery strategy (Environment Canada 2011). We used the Alberta Biodiversity Monitoring Institute's 2018 Human Footprint Inventory (HFI; Alberta Biodiversity Monitoring Institute 2018b) to map human land-use. Low-impact seismic lines were excluded because they are inconsistently mapped and wolves do not select this class of linear feature, nor do they move faster on them than in unaltered habitats as they do conventional seismic lines (Dickie et al. 2017).
To calculate current habitat alteration, we merged the HFI with the provincial wildfire boundaries, aged <40 years old in 2018, to match the vintage of the HFI (Alberta Biodiversity Monitoring Institute 2018b), and calculated the percent area altered of each landscape unit. We estimated habitat alteration following restoration by removing all conventional seismic lines from the HFI and recalculating the percent altered area using the above definition. We necessarily assume that restoration treatments will successfully recover caribou habitat. Tests of restoration efficacy are in their infancy, though early tests suggest that restoration treatments can disrupt predator behavior (Dickie et al. 2021, 2022) and facilitate return forest cover (Filicetti et al. 2019). The spatial and temporal scales in which these results will translate into demographic responses by caribou are unknown. While our prioritization framework guides restoration planning in the near-term (within decades), the process of habitat recovery post-treatment will likely extend for decades.
Given the uncertainties in predicting future fire distribution and frequency, we assumed that burned areas were static (i.e., fires remained constant in the “current” and “future” calculations of habitat alteration). Restoration of other semi-permanent feature types may also be considered for restoration treatments while seismic line restoration is occurring. Other feature types include trails, wellpads, pipelines, transmission lines, industrial sites, and roads. We evaluated the effect of restoring additional types of human footprint, and the recovery of forestry cutblocks, but the impact of these feature types on the percentage altered was minimal (Table S1; Supplement S1). We therefore considered restoration of conventional seismic lines only.
Linear features within energy sector boundaries were not considered as candidates for restoration in this prioritization process. We assume that exploration and development will continue for decades within these boundaries and therefore assumed project areas to be fully altered, which is purposefully conservative and overestimates alteration. In reality, restoration will begin within many project boundaries by 2031 through sub-regional planning process and the Alberta Land Stewardship Act. Reclamation of habitat alteration in these areas is required separately under Alberta Environmental Protection and Enhancement Act approvals. We delineated energy sector project boundaries using the Alberta Energy Regulator's Scheme Mapper. We also engaged with Canada's Oil Sands Innovation Alliance to verify these data and updated boundaries where approvals existed. If a landscape unit partially overlapped a project boundary, we only included the portion of the landscape unit outside the boundary within the restoration prioritization process.
Calculating the Cost of Restoration
Next, we approximated the relative cost of restoration using the density of seismic lines to be restored (km/km2) in each landscape unit. This metric was used because the realized cost of restoration is highly variable depending on the treatment applied, habitat conditions, and logistical constraints. Furthermore, how these constraints influence the realized cost will vary over time as restoration learnings are developed and new infrastructure and technology is created to increase efficiencies. Cost estimates that reflect underlying landscape heterogeneity may improve estimates of gain per cost under current constraints, but may not reflect realities as these advancements are made. Conversely, under current constraints and as operational advancements are made the cost of restoration will largely reflect the amount of effort required; that is, the density of seismic lines requiring treatment. We divided the gain in unaltered habitat by the density of seismic lines in each landscape unit to provide a ratio or value of gain in unaltered habitat per unit cost, hereby termed “gain per cost.” Higher values indicate areas with a large increase in unaltered habitat for a comparatively lower cost.
Incorporating Additional Weighting Criteria
To direct restoration away from areas likely to be altered in the future, we weighted each landscape unit by the inverse of the estimated value of below-ground petroleum resources. We used the normalized Resource Valuation Layer (RVL) developed by the Canadian Association of Petroleum Producers to estimate the value of oil and gas deposits within each landscape unit (Canadian Association of Petroleum Producers 2016).
To direct restoration toward areas of high value for caribou, we weighted each landscape unit by the relative use by caribou. We estimated use from caribou telemetry locations provided by the Government of Alberta from Global Positioning System (GPS) collars. Collar deployment began in different years for each range, and for all ranges including collar data up to 2019. Deployment began in WSAR in 1998, followed by ESAR (2008), Richardson (2009), Red Earth (2011), and Cold Lake (2012). Although GPS data from WSAR extends over a longer monitoring period than the other ranges, many recently collared individuals (post-2010) use the same areas used by caribou in the earlier portion of the dataset (pre-2010). For each individual caribou, we developed a “utilization distribution” (i.e., the probability of use across an individual's home range) using the package “adehabitatHR” and using the reference method for calculating the smoothing parameter (Calenge 2006). To create a range-level index of caribou use, we averaged all individual utilization distributions in a range. We then calculated the average index of caribou use for each landscape unit in the study area.
Ranking Landscape Units
We grouped the weighted gain per cost into five priority zones with approximately equal numbers of landscape units within each zone. The choice of five zones was arbitrarily chosen to provide an operationally reasonable number of townships per zone per caribou range. Zones were ranked from 1 (highest) to 5 (lowest), such that the highest priority areas have a higher gain in unaltered habitat, lower costs, lower RVL, and higher caribou use. We prioritized landscape units into zones for each range, rather than across all ranges because the federal recovery strategy states that each population is to be recovered where feasible (Environment Canada 2012).
We estimated the cumulative percent of unaltered habitat as restoration progressed from Zone 1 through Zone 5 to evaluate the potential progress towards the federal maximum disturbance management threshold target of 35% altered habitat. The HFI maps human land-use at a much finer spatial resolution than the satellite imagery used to develop the 35% target. Therefore, habitat alteration estimates from HFI are higher than those that would be obtained using the ECCC 2015 anthropogenic disturbance layer (ECCC 2017a). To better place the gain in unaltered habitat following restoration in the context of the 35% maximum habitat alteration target, we developed a linear regression for each range to calibrate the two data sources (Table S2; Supplement S2). We present results using both the raw HFI values as well as the calibrated values using the relationship between the two dataset under current landscape conditions.
Results
We present the restoration priority zones in Figure 3, and the change in percent altered habitat as restoration proceeds cumulatively through each zone in Table 1. Habitat restoration within Zone 1 has the largest impact on percent area altered within each caribou range, ranging from a decrease of 7% in WSAR and ESAR, to a decrease of only 3% in Richardson. The reduction in percent of altered habitat (or gain in unaltered habitat) diminishes as restoration proceeds from the highest to lowest priority zones (Table 1). When assuming no restoration occurs within energy sector project boundaries, none of the ranges reach the 35% maximum habitat alteration target, even after restoring all five zones and calibrating to ECCC data. The WSAR caribou range is the closest to reaching the 35% maximum habitat alteration target following the restoration of all seismic lines and calibrating data to ECCC, declining from 67% altered habitat currently to 45% altered habitat (Table 1).
| Caribou range | Current % area altered | % Area altered post restoration | ||||
|---|---|---|---|---|---|---|
| Zone 1 | Zones 1–2 | Zones 1–3 | Zones 1–4 | Zones 1–5 | ||
| ABMI Human Footprint Inventory | ||||||
| Cold Lake | 93 | 87 | 82 | 81 | 80 | 80 |
| ESAR | 90 | 83 | 79 | 76 | 75 | 74 |
| Red Earth | 82 | 77 | 71 | 67 | 64 | 63 |
| Richardson | 91 | 88 | 87 | 87 | 87 | 87 |
| WSAR | 87 | 80 | 73 | 68 | 63 | 62 |
| ECCC Calibration | ||||||
| Cold Lake | 85 | 80 | 76 | 74 | 73 | 73 |
| ESAR | 80 | 73 | 69 | 66 | 65 | 64 |
| Red Earth | 63 | 59 | 55 | 52 | 50 | 49 |
| Richardson | 81 | 78 | 76 | 76 | 76 | 76 |
| WSAR | 67 | 61 | 55 | 50 | 46 | 45 |
Case Study: Applying the Prioritization Process to a Mountainous System
Like boreal woodland caribou, Southern Mountain Caribou also face threats from habitat alteration leading to increased predation via apparent competition which is exacerbated by linear features (Wittmer et al. 2005; Apps et al. 2013). Caribou in the mountain system are predominantly impacted by forest harvest resulting in the replacement of old-growth forests, highly selected for by caribou, with early successional forests which are avoided by caribou but selected by moose and their predators. Habitat restoration has therefore been similarly identified as an important action to achieve self-sustaining caribou populations in montane ecosystems (Environment Canada 2014). Restoring thousands of kilometers of unused forestry roads presents a similar need for prioritization to maximize the ratio of gain in unaltered caribou habitat per unit cost. However, the operational and ecological considerations in the mountainous system are different from those in the boreal system. For example, access to forestry roads often stretch along valley bottoms leading into high-elevation forests, impacting how logistical constraints and future alteration are considered in respect to restoration planning.
We adapted the boreal caribou prioritization model for caribou in the northern Columbia Mountains of British Columbia and present this modification as a case study for applying the prioritization process to different systems. This study area (6,637 km2) included much of the area occupied by the Columbia North, Columbia South, Frisby-Boulder caribou subpopulations and part of the area formerly occupied by the Central Rockies subpopulation. We engaged with local forestry tenure licensees operating in the study area to delineate feasible landscape units across which to prioritize for habitat restoration. Most landscape units followed existing watersheds and resource road networks ranging from 4.9 to 498 km2 (median = 86 km2). The variation is reflective of road network limitations and watershed size.
To calculate the gain in unaltered habitat following restoration, we estimated the percent habitat alteration defined as forest stands burned or harvested within the last 40 years, plus areas already approved for harvest. We only considered alteration in the last 40 years because older forests in the area have sufficiently regenerated to no longer support increased moose use (Serrouya et al. 2011; Mumma et al. 2020). We only considered forested areas as caribou habitat available for restoration, excluding unforested alpine meadows, rock, and ice. Unlike the federal recovery strategy for boreal woodland caribou (Environment Canada 2012), the federal recovery strategy for Southern Mountain Caribou (Environment Canada 2014) does not buffer anthropogenic habitat alteration. For this reason, and because in this area a horizontal 500-m buffer would span many valley bottoms and exclude most forested areas due to the topographical ruggedness of the landscape, we did not buffer altered habitat. Linear features considered to be non-restorable included all transmission lines, one paved highway, and three main forestry roads leading to helicopter-access winter ski resorts. As with the boreal process, we did not estimate the monetary cost of restoration, but instead used the density of restorable roads to index the cost of restoration.
To focus restoration in areas with the most immediate benefit to caribou, we increased the weighting of landscape units with a higher proportion of core caribou habitat and with higher known caribou use. Core caribou habitat was designated using spatial layers from the Government of British Columbia (Dodd & Carswell 2019). Caribou use was estimated using utilization distribution of GPS-collared caribou following the same process as for the boreal population. We also increased the weighting of landscape units where large proportions of the forest were already protected to build off of existing unaltered areas. Existing caribou habitat protection was obtained from Ungulate Winter Range boundaries from the Government Actions Regulation (Government of British Columbia 2005) designed to protect old-growth forests. We calculated the proportion of the landscape unit already protected for caribou. Finally, we decreased the weighting of landscape units with a higher proportion of habitat alteration preferred by moose to avoid obstructing hunter harvest in areas where moose are likely to be at higher densities, which would be counter-productive in a caribou recovery context (Serrouya et al. 2017). We defined moose habitat as forests younger than 20 years because this age class will continue to be productive, high-quality moose habitat (Mumma et al. 2020). Delaying restoration in areas with high-quality moose habitat would maintain moose hunting opportunities, potentially reducing local moose abundance until caribou habitat restoration occurs.
Discussion
Across many ecosystems habitat has been degraded beyond the point of protection, requiring active habitat restoration to restore ecosystem function (Suding 2011; Possingham et al. 2015). This pattern is the case for threatened and endangered woodland caribou: insufficient habitat has been protected to achieve self-sustaining populations, and yet habitat loss continues (Nagy-Reis et al. 2021). Decision makers are now faced with the daunting challenge of addressing restoration across broad areas that continue to be impacted by extensive networks of linear features. To address this challenge, a strategic approach to restoration is necessary (Joseph et al. 2009) because of the cost of habitat restoration, the extent of alteration, the many human interests on the land base, and the overlap between critical caribou habitat with valuable natural resources (Hebblewhite 2017). In this paper, we developed an adaptable framework building off of concepts of systematic conservation planning and multicriteria decision-making analyses to prioritize areas for restoration based on the premise of maximizing the ratio of gain in unaltered habitat per unit cost, with key modifiers such as caribou habitat use and future development potential.
Our algorithm provided ordinal groups that ranked landscape units into five zones to prioritize restoration. The largest gain in unaltered habitat occurred following restoration of the highest priority zone, with diminishing returns as restoration proceeded through the lowest zone. However, without restoration within project areas, we found that none of the ranges reached the federal recovery target of a maximum of 35% habitat alteration, even after restoring seismic lines within all five zones. We simulated the impact of seismic line restoration only because other semi-permanent feature types (e.g., wellpads and pipelines) had only minimal impacts on area altered within each range. However, taking advantage of mobilized equipment and personnel to restore additional features will likely improve the gain in unaltered habitat following restoration.
The WSAR and Red Earth caribou ranges were the only two ranges to drop below 50% altered. These two ranges are overlapped by relatively little wildfire, whereas significant areas of the Cold Lake, ESAR, and Richardson ranges have recently burned. Much of the remaining unburned habitat within Cold Lake and ESAR ranges are overlapped by large areas of energy project boundaries, which we assumed would remain fully altered given the recognition that successful restoration prioritization should account for multiple needs on the landscape (Bode et al. 2010; Smith et al. 2022). Our results, therefore, suggest that the restoration of legacy features will likely need to occur within active project boundaries, and that collaboration between organizations such as government and industry will be key for effective caribou habitat recovery. Likewise, collaborative programs that address restoration both within and outside project boundaries will likely increase operational efficiencies. Overall, our results highlight the need for ambitious progress towards effective habitat restoration, and the need to further improve land-use planning to minimize alteration within caribou range.
Broadly, the goal of managing anthropogenic habitat alteration is clear: reverse habitat loss by restoring human-altered habitat and reducing the rate of new alteration. But natural habitat alteration adds complexity, particularly for species that rely on old-growth or mature forests. Large portions of some caribou ranges, in particular Richardson, are burned. In these ranges, reaching a maximum of 35% habitat alteration when including wildfire is likely unachievable, particularly as burn size and frequency increase under predicted climate change scenarios (Weber & Flannigan 1997). The federal recovery strategy includes both anthropogenic and natural habitat alteration within the maximum disturbance management threshold of 35% (Environment Canada 2012), though does acknowledge the need for range-by-range flexibility (ECCC 2017b; Johnson et al. 2020). Though both fire and anthropogenic habitat alteration are thought to contribute to increased prey and predator abundances, recent evidence suggests burns in boreal wetlands may not create early seral forage for ungulate (DeMars et al. 2019). Furthermore, while caribou avoid burned areas, the link between use of burned areas and vital rates is unclear (Johnson et al. 2020; Konkolics et al. 2021). Fire may even facilitate the recovery of anthropogenic habitat alteration by resetting the successional trajectory to natural recovery (Filicetti & Nielsen 2018, 2020). Clarifying the relative contribution of these two types of habitat alteration to caribou declines will help to facilitate effective habitat management and planning.
For prioritization efforts to translate into effective restoration planning, it is imperative to incorporate the appropriate criteria for evaluating both costs and benefits (Joseph et al. 2009). The most appropriate criteria should specifically address identified threats, and are likely scale-dependent (Hobbs & Harris 2001; Wilson et al. 2007). Several caribou recovery documents recognize that prioritizing areas for restoration based on a “return on investment” is the most efficient approach to pursue (Ray 2014). It is also increasingly recognized that limited resources should be prioritized to be cost-effective and transparent (Bottrill et al. 2009; Schneider et al. 2010; Hebblewhite 2017; Cornwall 2018). We weighted the prioritization rankings to guide restoration away from areas that may be targeted for future development and towards areas of high caribou habitat value, combining economic and ecological objectives across the landscape. But at broader scales, managers may prioritize among specific caribou ranges. Such decisions could be based on caribou population size and trend, the degree of permanent land conversion that is virtually impossible to reverse (e.g., agriculture, cities, and highways), or the proximity to protected areas. At a finer scale, specific linear features may be prioritized due to known predator–prey dynamics and site limiting factors. Incorporating considerations at both broader and finer scales will enhance the utility of decision-support provided by our framework.
We incorporated caribou habitat use as a weighting criterion to focus restoration on areas that would most immediately benefit caribou. Unless overlapped by a project boundary, landscape units in Zone 1 capture all the highest caribou-use areas, as well as several areas with moderate caribou use, while lower priority landscape units are those that have near-zero caribou use. Caribou space-use was determined from a sample of GPS-collared female caribou from each range. Our approach assumes that collared individuals are representative of each range's population. Caribou fecal sample distributions obtained through independent surveys align with the distribution of GPS-collar locations (T. Hegel, personal communication). Although GPS collars were only deployed on females, this is the critical demographic to capture for conservation purposes, because caribou population growth is largely determined by recruitment and adult female survival (DeCesare et al. 2012). Increasing the priority of areas in which caribou have previously calved would also be ideal, though we were unable to incorporate calving here.
Our approach explicitly considered the energy and forestry sectors by directly engaging with these groups, however, there are crucial values that should be similarly considered when planning habitat restoration that we were unable to incorporate. Of utmost importance to enhancing success is meaningful engagement between western scientists and Indigenous communities, exemplified by the success of caribou recovery actions and habitat protection within the Klinse-za caribou subpopulation led by the West Moberly First Nations and Saulteau First Nations (Lamb et al. 2022). Similarly, engagement with non-Indigenous land users such as recreational users as well as trappers is needed to prevent potential land access conflicts (Hornseth et al. 2018). To ensure the highest benefit for caribou and people alike, and to ensure restoration treatment integrity is maintained, the timing and sequencing of restoration across the landscape will need to be considered.
In addition to incorporating the appropriate criteria, careful consideration of the method in which multiple criteria are combined is necessary and should match the goals of the prioritization process (Belton & Stewart 2002; Martin et al. 2022). We used two different approaches to combined criteria: multiplication for the boreal caribou case study and summation for the mountain caribou case study. Multiplication is suitable for applications when equal weightings are desired, but zeros in any criteria should result in a final score of zero regardless of the value of other criteria. For example, in the boreal process, multiplication relegated all landscape units with no documented caribou use or the highest likelihood of future development to the lowest priority zone, an appropriate outcome to our process. However, in the mountain caribou case study we incorporated multiple criteria where a disproportionate influence of zeros would not have been appropriate. For example, while we increased the priority of areas with higher proportions of core caribou habitat, it was deemed inappropriate for landscape units with no core caribou habitat (i.e., completely matrix caribou habitat) to be relegated to the lowest priority zone. Many alternate options are available to combine criteria, including order-based ranking and interactive non-linear approaches. Each option has pros and cons that should be carefully considered when choosing the appropriate combination processes to match needs (Belton & Stewart 2002). Sensitivity analyses can be used to evaluate the sensitivity of results to these decisions, particularly when the methodological choice is uncertain (Adem Esmail & Geneletti 2018). Sensitivity analyses can also evaluate which landscape units are most or least influenced by the methods used to combine criteria, further identifying areas that are consistently high or low priority.
For caribou and other species that require broad, intact landscapes to persist, restoration of legacy features may not be enough. Effective habitat management requires a net decline in habitat alteration, such that the rate of habitat gain must exceed the rate of habitat loss. The rate of restoration must outpace the rate of new habitat alteration. Beyond strategic and ambitious habitat restoration and technical innovation to increase the pace of restoration, increased habitat protection and technological innovation to decrease the rate of new habitat alteration will be important to achieving a net decline in habitat alteration.
Acknowledgments
The authors would like to thank Dr. Troy Hegel as well as the Canada's Oil Sands Innovation Alliance (COSIA) members, in particular, Jon Gareau and Lori Neufeld, for their collaboration on this project. The authors also thank Agathe Lebeau, Gwen Bridge and members of the North Columbia Restoration Prioritization group for support in developing the prioritization process in the Columbia region. The authors also thank Stephanie Andrews for analytical support. The authors thank Dr. Stephen Murphy and two anonymous reviewers for their constructive comments on a previous version of this manuscript. Funding for this research was provided by COSIA and Environment and Climtae Change Canada. The authors acknowledge that Michael Cody is employed by Cenovus, a member of COSIA. All coauthors agree that Cody's employment does not represent a conflict of interest.
