Scientific Assessment to Inform the Identification of Critical Habitat for Woodland Caribou, Boreal Population, in Canada - 2011 Update: Methodology

Previous Page Table of Contents Next Page

SARA S.2 defines critical habitat as “[…] the habitat that is necessary for the survival or recovery of a listed wildlife species […]”. Consideration of scale is fundamental to identifying the bio-physical attributes (i.e., habitat) required for the survival or recovery of boreal caribou (EC 2008). Caribou select habitat at multiple spatial scales to meet their life history requirements. At fine spatial scales, microclimate and food availability are important factors influencing caribou habitat selection. However, the primary limiting factor on boreal caribou populations is predation (Rettie and Messier 1998; Wittmer et al. 2005), associated with natural or human-induced landscape conditions that favour early seral stages and higher densities of alternative prey, resulting in increased risk of predation to caribou. Habitat conditions at the scale of boreal caribou ranges affect the demography of boreal caribou (e.g., survival and reproduction), which ultimately determines whether or not a population will survive. Therefore, a local population range was identified as the relevant spatial scale for the identification of critical habitat that provides the conditions required by boreal caribou.

The survival of boreal caribou local populations requires that both habitat and population conditions are conducive to overall stable or positive population growth and longer-term persistence. This state is referred to here as self-sustaining (see Definitions). If either habitat or population conditions are not favourable, the population will decline and eventually disappear in the absence of intervention. For example, a large population could disappear due to a recurring declining trend, whereas a small population could disappear due to stochastic events (e.g., severe winter). Self-sustaining local populations are required to improve the likelihood of maintaining boreal caribou in the wild.

Boreal caribou distribution in Canada spans seven ecozones and many more ecoregions (EC 2008). There is tremendous variation in ecological conditions across this distribution, to which boreal caribou populations exhibit variable local adaptations. Representing this variability with appropriate levels of redundancy is an essential consideration when identifying a distribution objective. Creating a patchy distribution is likely to elevate risk and promote continuation of overall range recession (up to 50% of the potential historical range is already no longer occupied). In addition, all local population or conservation units have been deemed biologically or technically feasible to recover. Therefore, the scope of the current science assessment is the current distribution of boreal caribou in Canada.

For the purpose of conducting the current assessment, critical habitat for boreal caribou was therefore defined as the resources and environmental conditions required for self-sustaining local populations, or groups of animals under similar local conditions, throughout their current distribution in Canada.

The present assessment was guided by similar principles that were established for the 2008 Scientific Review:

  1. Consider available published scientific information and seek multiple lines of evidence to support conclusions.
  2. Recognize the dynamic nature of boreal systems, and the resultant effects on boreal caribou habitat.
  3. Acknowledge and consider that both the physical and functional characteristics of habitat for this species operate at multiple spatial and temporal scales, including physical and functional properties.
  4. Recognize that variation in population structure, population and landscape condition, and state of knowledge may warrant finer scale approaches to refine the identification of critical habitat across the national distribution of this species.
  5. Apply precaution when evidence is insufficient to judge harm (precautionary principle).
  6. Recognize that ongoing research and monitoring, as part of an adaptive management approach, are key to reducing uncertainties over time, improving decision-making, and achieving management objectives.
  7. Recognize that critical habitat identification is a scientific process, with socio-economic considerations addressed in other phases of the overall SARA recovery planning process.

Similar to the 2008 Scientific Review, this assessment was designed to provide a scientific description and quantitative evaluation of critical habitat relative to the set of conditions (demographic and environmental) within each range. The framework and components developed in the 2008 Scientific Review were expanded to include additional scientific activities (described in Section 2.4), in order to augment the earlier assessment.

Figure 1 illustrates the overarching Critical Habitat Framework and its relationship to other stages of the recovery process. A general description of each step in the assessment process is provided below.

Step 1: Identification of the Current Distribution of Boreal Caribou in Canada

The current distribution of boreal caribou was used to define the geographic scope for the identification of critical habitat. The first step in the framework was to update the distribution based on the most recent and best available information.

The designation of the Canadian range for a species-at-risk ordinarily relies on the COSEWIC assessment process (COSEWIC 2010). Since there has been significant new information since the last COSEWIC assessment (COSEWIC 2002), an up-to-date distribution map (see section 2.4.1) was created as an interim product to conduct this critical habitat assessment.

Figure 1. Critical Habitat Framework for boreal caribou. Although other sources of information are considered to identify critical habitat (e.g., ATK), the top panel of this figure is focussed on the scientific activities to inform critical habitat identification.

Figure 1. A broad overview of the critical habitat framework for boreal caribou. The framework is composed of two parts. First, there is the scientific description of critical habitat which includes identifying the species current distribution, delineating local population ranges, determining the conditions required for self-sustaining local populations and last providing a scientific description of critical habitat. The science information feeds into the second part of the framework which is  the legal identification of and protection of critical habitat. This starts with the identification of the species’ critical habitat in the National Recovery Strategy, the identification of the recovery activities required to protect critical habitat for Action Planning and the implementation of the critical habitat recovery activities. The framework is embedded in adaptive management cycle where information gaps are identified as information needs, population responses to the implementation of recovery activities are monitored and both are used to inform and adapt scientific and management activities.

Step 2: Delineation of Boreal Caribou Ranges

A range is defined as the geographic area within which there is a high probability of occupancy by individuals of a local population, all of which are subjected to the same influences affecting vital rates over a defined time frame. Local populations may experience a limited exchange of individuals with other populations, such that the demography is affected mainly by local factors and not by immigration or emigration among groups. Boreal caribou ranges were identified as the appropriate units of analysis for ensuring the recovery and/or survival of boreal caribou (see section 2.1).

The 2008 Scientific Review noted that ranges were delineated inconsistently across jurisdictions. When sufficient information was available, the range assessment was conducted on recognized local populations. In other cases, management units delineated either mainly or partly by non-ecological considerations (e.g., administrative boundary, land management unit) were assessed. The report recognized the variation in the data and methods for range delineation as a potential source of uncertainty with respect to the assessment of the ability of boreal caribou ranges to maintain self-sustaining caribou populations (EC 2008).

The present report investigated the impact of using different delineation methods on range size, and a standardized delineation method was suggested. In addition, information on the type and quantity of data used to delineate ranges as provided by each jurisdiction was compiled, and a terminology was proposed to reflect the level of certainty with respect to range delineation (see section 2.4.2). However, the current assessment did not provide alternative range delineations using the reported methodology due to time constraints and awareness that parallel processes were being implemented by some jurisdictions to refine their range delineation. As such, the updated range delineation information provided by most jurisdictions at the time of implementing the Critical Habitat Framework was accepted as the best available knowledge.

Step 3: Identification of the Conditions Required for Self-Sustaining Local Populations

Animals use or extract resources (e.g., for food and shelter) from their environment for survival and reproduction. Population growth rate and persistence are related to population condition (trend and size), and are expected to deteriorate with the loss or changes in habitat that influence survival and reproduction. The 2008 Scientific Review acknowledged this link by using habitat condition (percent total disturbance) and population condition as a starting point for assessing the capacity for each range to support a self-sustaining caribou population.

In this update, an enhanced but analogous procedure was developed to increase the certainty in conclusions regarding the state of critical habitat in boreal caribou ranges in Canada. Some jurisdictions provided updated demographic data (trend, size, adult female survival, calf recruitment). These data were used to quantify the likelihood that boreal caribou populations were self-sustaining based on current population condition. However, data availability varied considerably among ranges, and population data were lacking entirely for some ranges.

The suite of analyses characterizing habitat condition was expanded to: 1) include additional bio-physical attributes influencing caribou survival and recruitment, such as the configuration of different habitat types; and 2) quantify the scale-dependent nature of caribou habitat by examining patterns of habitat selection at national versus regional scales.

Disturbance Mapping (section 2.4.3.1)

The 2008 Scientific Review identified anthropogenic and natural disturbances as significant predictors of habitat condition. The analysis of anthropogenic disturbance used the "national anthropogenic disturbance" database developed by Global Forest Watch Canada (Lee et al. 2006). Although the Global Forest Watch Canada data was a valuable contribution to the 2008 Scientific Review, it did not provide discrimination among different disturbance types and buffered all digitized human developments by 500 m.

In this update, new anthropogenic disturbance maps were created to investigate the relative impact of different disturbance types and their configuration on the assessment of boreal caribou ranges.

Habitat Selection (section 2.4.3.2)

A habitat selection analysis was conducted to identify additional bio-physical attributes influencing habitat condition, beyond the percentage of total disturbance. The analysis was conducted at both a national and ecological units (ecozone) scales to understand how variation in the availability of habitats across the boreal forest in Canada might influence patterns of habitat preference and avoidance.

Buffer Analysis (section 2.4.3.3)

In the current context, a "buffer" is referred to as an area assumed to be functionally unavailable to caribou due to its proximity to anthropogenic development. In the 2008 Scientific Review, the national anthropogenic disturbance dataset was developed with a 500 m buffer to each anthropogenic disturbance, but the nature of the data prevented any manipulation of the buffer. While recognizing this limitation, the 500 m buffer was considered a reasonable minimum at the time.

In the present update, the effects of 1) different buffer widths on the configuration of disturbance and 2) effects of landscape configuration and connectivity on caribou demography were examined.

Meta-Analysis of Boreal Caribou Population and Habitat Condition (section 2.4.3.4)

Understanding the relationship(s) between caribou population condition and the condition of the range is central to determining the amount of habitat required to support a self-sustaining population. In the 2008 Scientific Review, a meta-analysis was used to quantify the variation in calf recruitment across twenty-four (24) ranges in Canada as a function of total disturbance (fire and 500 m-buffered anthropogenic disturbance). This recruitment-disturbance relationship was the main tool for quantifying the capacity of a range to maintain a self-sustaining caribou population based on habitat condition.

In this update, the scope of that meta-analysis was broadened to incorporate updated disturbance mapping and results from a more precise habitat selection and buffer analysis to refine the characterization of habitat conditions within boreal caribou ranges, and to better explain the variability among local populations associated with the recruitment-disturbance relationship.

Current and Future Conditions (section 2.4.5)

In the 2008 Scientific Review, current range conditions were assessed using three lines of evidence (or indicators): population size, population trend, and total disturbance (%). Probabilities of persistence (then used to assess the state of "self-sustaining") were assigned to categorical states defined for each indicator informed by either expert opinion (trend), a population model (non-spatial PVA, population size), or the recruitment-disturbance relationship (total disturbance). The range-specific integrated probability of self-sustainability was derived from averaging the sum of indicator values for a range. In other words, it provided a static assessment of the capacity of a range to maintain a self-sustaining caribou population based on its current state.

The present update also evaluated the probability that the current conditions were sufficient to support self-sustaining caribou populations using a set of indicators: two of population growth, and one of persistence. However, the indicators were quantified using a generic population model and a probabilistic decision-analysis tool. The population modelling extended the results from the 2008 non-spatial PVA model (EC 2008).

Also, a habitat-dynamics model was developed to better understand how future changes in habitat conditions within a range might affect the sustainability of boreal caribou local populations. Habitat conditions were modeled based on, and limited to, the likelihood of future fires and natural forest recovery (i.e., regeneration) of disturbed habitats. The model was not designed to provide a full assessment of future conditions (i.e., it does not include future anthropogenic disturbance). Rather, this information can be used in combination with the persistence indicator to provide an indication of the level of active recovery likely to be required (in addition to passive recovery) for self-sustaining local populations, and in the interpretation of disturbance thresholds.

Step 4: Description of Critical Habitat

In the context of the present assessment, the scientific description of boreal caribou critical habitat for each range consists of the following four components: the delineation of the range; an integrated risk assessment of current capacity to maintain self-sustaining populations; information to support the identification of range-specific disturbance thresholds; and a description of the key bio-physical attributes within a range required by boreal caribou.

Integrated Risk Assessment (section 2.4.6.1)

A probabilistic decision-analysis tool (i.e., Bayesian Decision Network) was developed to
combine each available data input (population trend, population size, percentage total
disturbance) for a given range to assess the probabilities that current conditions within
boreal caribou ranges would support self-sustaining populations. An indicator-based
“lines of evidence” approach was used to evaluate two of the criteria related to self-sustaining populations (stable/positive population growth and persistence). This
approach was favored over the averaging of individual probabilities of population growth
10 and persistence (as per EC 2008) because: (i) the demographic and environmental factors that determine population growth rate and population persistence are related, although time lags may create a temporal mismatch between the respective factors, and; (ii) the quantity and type of information available for each range varies widely, and so
considering each type of data as a line of evidence enabled the application of a consistent set of decision rules for assessing the evidence. The results were used to identify the risk that the current habitat conditions of boreal caribou ranges would fail to maintain selfsustaining local populations.

The third criterion for self-sustaining populations, i.e., no active management, was assessed based on information available for each range. If a range was assessed as able to support a "self-sustaining" caribou population, based on population information criteria, but was known to be subject to management interventions, it was not considered to be "self-sustaining".

Range-specific Management Thresholds (section 2.4.6.2)

The probabilistic approach to the integrated risk assessment of critical habitat for recovery planning is complemented by the identification of similarly-derived probabilistic intervals of disturbance relative to current and projected caribou population state. This information can be used to support the establishment of risk-based management thresholds. While it falls beyond the scope of the present scientific assessment to recommend specific management thresholds, given the need to explicitly identity acceptable management risk, description of a methodology for deriving these is provided, along with examples of their potential application, and discussion of their interpretation relative to the criteria and indicators evaluated here.

Bio-Physical Attributes (section 2.4.6.3)

Bio-physical attributes are the habitat characteristics required by caribou to carry out the life processes necessary for survival and reproduction. The results from the habitat selection analyses (this report) and published reports were used to summarize key bio-physical attributes by ecozone.

Step 5: Identify Information Needs, Monitor and Adapt

Critical habitat for boreal caribou is an emergent property of dynamic boreal landscapes represented by a suite of conditions that are not fixed in either space or time. A robust research and monitoring program is essential to continually assess the identification and management of critical habitat for each local population or conservation unit, and adjust when needed. At a minimum, monitoring is required to ensure that the protection of critical habitat is effectively meeting specified recovery objectives for populations and distribution in the long term.

The process of adaptive management acknowledges and supports the adjustment of management actions in light of new knowledge. The adaptive management cycle is an essential component of the Critical Habitat Framework. Knowledge gaps and uncertainties are identified, evaluated, and reported as information needs, and addressed through management planning and implementation (see Section 4.0).

The implementation of the Scientific Description of Critical Habitat component of the Critical Habitat Framework (see Figure 1) is focussed on three main outcomes: 1) an integrated risk assessment of whether or not the current set of habitat and population conditions within a range were sufficient to support a self-sustaining caribou population; 2) a methodology to identify range-specific disturbance thresholds; 3) and a description of the bio-physical attributes of boreal caribou habitat. Figure 2 illustrates how the framework was expanded to achieve these outcomes and provide a description of critical habitat. The sections below provide a stepwise summary of the decision tools and analyses that were conducted to implement the framework.

Figure 2. Scientific description of boreal caribou critical habitat framework.

Figure 2. A more detailed description of the critical habitat framework for boreal caribou. As in Fig. 1, the framework begins with a scientific description of critical habitat which includes the identification of the species current distribution and delineating local population ranges. In this diagram, the step that involves determining the conditions for self-sustaining local population ranges has been elaborated. Range condition is assessed using information on current population condition which comes from available data on population size and population trend, and current habitat conditions which uses information on available habitat, levels of human disturbance and areas burned by natural fires. The information on habitat is used to quantify the effects of disturbance on caribou calf recruitment and input into a habitat-dynamics model predicting changes in future range condition. The assessment of current population condition, current habitat condition and future range condition are used to inform the scientific description of critical habitat which includes an integrated risk assessment of whether the range can support a self-sustaining local population (based on population and range condition), the identification of range-specific disturbance thresholds and a list of biophysical attributes necessary for the survival and recovery of boreal caribou. The scientific description of critical habitat is used to inform the legal identification of critical habitat by policy makers. Again the framework is coached in adaptive management, where information needs are identified, population responses are monitored and both are used to inform and adapt scientific and management activities.

The geographic scope of the assessment was defined by the current distribution, or extent of occurrence, of boreal caribou in Canada. Updated information from most of the jurisdictions on boreal caribou ranges (see Section 2.4.2) revealed some discrepancies with respect to the distribution information used in the context of the 2008 Scientific Review. The species distribution map was updated by increasing the distribution of boreal caribou to include all areas currently identified by the jurisdictions as boreal caribou ranges (Figure 3).

The two main areas where changes relative to the 2008 Scientific Review occurred were:

  1. Northwest Territories: a) western boundary moved eastward; and b) changes around Great Bear Lake; and
  2. Alberta: numerous changes to the distribution boundary in the province.

The updated distribution map confirms that boreal caribou are found in nine jurisdictions, extending from the Yukon Territory in the west, to Labrador in the east, and as far south as some islands in Lake Superior.

Figure 3. Distribution map of boreal caribou in Canada showing the current distribution of boreal caribou using updated information provided by jurisdictions. Note: Because of the lack of information on the historical distribution of boreal caribou in B.C. relative to the mountain ecotype of woodland caribou, the historical southern extent in that province is based on the boreal ecozones boundary.

Figure 3. Map illustrating the historical and current distribution of boreal caribou across the boreal ecozones in Canada.

National variation in range delineation

A survey of how ranges were being delineated across the boreal caribou distribution was conducted to assess the variation in the data and methods used across jurisdictions. An analysis was also developed to better understand the variation in type and quantity of data used to delineate ranges and the implications to the current assessment. Finally, a consistent approach to delineating ranges for this species was suggested to reduce the national variation in the future. This was accomplished by:

  1. requesting that jurisdictions provide detailed information on how caribou ranges were delineated to document data and methods currently applied;
  2. examining the impact of data availability on range size while controlling for other factors related to habitat condition (using data from the 2008 Scientific Review);
  3. developing a categorization of ranges along a continuum that reflects the level of certainty in range boundaries and highlights important biological considerations for each category; and
  4. classifying the updated delineation of boreal caribou ranges in Canada according to the categorization based on the type of information available and associated certainty in delineation.

The information provided by the jurisdictions indicated that data availability influenced the methods used for delineating boreal caribou ranges across Canada (Appendix 7.1). The analysis using data from the 2008 Scientific Review suggested that 65% of the variation in size of boreal caribou ranges was explained by data and methods used to delineate the ranges, and three surrogate measures of habitat quality including: the percentage of human disturbance on a range, the size of forest patches, and an inferred measure of forage availability, estimated from the cumulative fraction of photosynthetically active radiation (FPAR) (Table 1 in Appendix 7.1). The analysis revealed that high levels of human activity were associated with more discrete and isolated boreal caribou ranges and that data type had a significant impact on range size, after controlling for the effects of habitat quality.

The updated range boundaries for boreal caribou that were provided by jurisdictions were classified into three types reflecting the level of certainty in range boundaries: Conservation Units (low certainty), Improved Conservation Units (medium certainty), and Local Population (high certainty) (Figure 4). Suggested methods and considerations for developing a standardized approach to delineating each of the three types of ranges are discussed in Appendix 7.1.

Figure 4. Range delineation types developed to reflect variation in the data and methods used to delineate boreal caribou ranges across Canada and the level of certainty in the delineated boundaries.

Figure 4. Three types of boreal caribou ranges were identified along a continuum that reflects the level of certainty in delineated boundaries. Conservation Units fall at the lowest end of the certainty spectrum and represent areas delineated based on digital maps using information that is not necessarily specific to the area of interest. Next are Improved Conservation Units which are delineated using digital maps and short-term observational data on caribou specific to that area. Last, a Local population range is delineated using long-term observational data on habitat use and movement and has the highest level of certainty in delineated boundaries.

Revised national range delineation map

The boreal caribou range delineation map was updated using the best and most current information provided by jurisdictions. The map served as the basis for delimiting the spatial extent for the subsequent analyses of a range to maintain a self-sustaining population and ultimately the description of critical habitat.

Trans-boundary and large continuous ranges

Two special cases were highlighted as having important implications to the critical habitat analyses: trans-boundary ranges and very large continuous ranges. Few jurisdictions have coordinated efforts to harmonize information on trans-boundary ranges and range delineation often artificially stops at the political boundaries. The lack of joint monitoring and data-sharing between jurisdictions decreases the certainty of range delineations and subsequent critical habitat descriptions. For example, the cumulative disturbance across the trans-boundary range may exceed levels supporting population sustainability despite management efforts applied in either jurisdiction. Similarly, averaging habitat condition over a large, continuous area will mask spatial variation in disturbances, potentially resulting in range contraction where human development is concentrated (see Table 1 in Appendix 7.1).

Two aspects of the 2008 Scientific Review focused on describing the habitat conditions influencing the survival and recovery of boreal caribou in Canada. First, a meta-analysis concluded that the percentage total disturbance (fire and 500m buffered anthropogenic disturbances) negatively affected the rates of caribou recruitment (EC 2008). Second, boreal caribou habitat use was described across different ecozones in Canada.

In the present update, a number of additional analyses quantifying the relative impact of different disturbance and habitat types and their configuration on caribou demography were undertaken to improve the certainty of the assessment of whether ranges could maintain self-sustaining local populations based on habitat condition. These included:

  1. new digitized maps of anthropogenic disturbances and fires were created to facilitate analyses quantifying the impact of disturbance on caribou demography (Section 2.4.3.1);
  2. an analysis of caribou habitat selection was conducted using radio-collar locations provided by jurisdictions to augment available information on the relative importance of different habitat types to caribou. The analysis was conducted at several scales (entire boreal caribou distribution and stratified by ecozones) to better understand how regional context might influence the description of critical habitat (Section 2.4.3.2);
  3. new analyses were conducted to better understand how the spatial configuration of anthropogenic disturbance might influence caribou demography (Section 2.4.3.3); and
  4. an enhanced meta-analysis of the relationship between habitat condition and population condition (hereinafter referred to as the recruitment-disturbance relationship, Section 2.4.3.4).
2.4.3.1 Disturbance mapping

a) Anthropogenic

A method for locating and classifying anthropogenic development according to the disturbance type was developed and implemented to create a nationally consistent, repeatable geospatial dataset of unbuffered estimates of anthropogenic disturbances. This update also increased the temporal correspondence between the disturbance data and the demographic data used in subsequent analyses to quantify the effect of habitat quality and configuration on the demography of boreal caribou.

Anthropogenic disturbance was defined as any human-caused disturbance to the natural landscape that could be identified visually from Landsat imagery at a scale of 1:50,000. Disturbances were classified into two broad categories, linear and polygonal features, which were further broken down into eight sub-categories each (Table 1). For each anthropogenic feature type, a clear description was established to maintain consistency in identifying the various disturbances in the imagery by different interpreters. Although ancillary data were used to guide interpretation and feature labelling, features themselves were only digitized if they were clearly visible on Landsat imagery. This general rule set the baseline for developing more specific rules of interpretation and digitizing of the disturbance events.

Table 1. Categories of anthropogenic disturbance digitized to inform the implementation of the boreal caribou critical habitat description framework.
Linear Features Polygonal Features
Roads Cut areas
Power lines Mines
Railways Reservoirs
Seismic Lines Settlement
Pipelines Well Sites
Dams Agriculture
Airstrips Oil and gas1
Unknown Unknown2

1 Features associated with the oil and gas industry. This may include gas plant, batteries, pump station and compressor stations.
2 Areas believed to be anthropogenic disturbance, based on patterns and comparison to surrounding environment in the satellite imagery; however, the specific type of disturbance is unknown.

Two series of map products were produced. The first maps were to support the buffer analysis (see Section 2.4.3.3) and the meta-analysis of population and habitat condition (see Section 2.4.3.4). These data were collected from satellite imagery with dates that corresponded to the collection of demographic data for each local population (see Appendix 7.5), increasing the temporal correspondence between the disturbance and demographic data. In addition, current (2006-2010) mapping of each boreal caribou range as delineated by jurisdictions was performed to provide estimated areas of human disturbance required for the current and future range assessment (see Section 2.4.5). Only information on new anthropogenic disturbances was collected in areas that overlapped with the sample of ranges used in the meta-analysis (Section 2.4.3.4).

b) Fire

Estimates of fire used in the 2008 Scientific Review were calculated from the Canadian National Fire Database (CNFDB, maintained by the Canadian Forest Service (CFS)), augmented by additional coverage for Northwest Territories, that contained wildfires greater than 200 ha (CFS 2010, NWTCG 2010). A 50 year limit was used to identify areas disturbed by fire, and hence unsuitable for caribou, consistent with methodology applied by Sorenson et al. (2008).

For the present assessment, jurisdictional agencies were contacted to obtain the most complete and up-to-date information on fires. Information on fire within National Parks was provided from either Parks Canada, if available, or the CNFDB. The availability of fire data varied, in particular with respect to the first year of data collection, and the maximum number of years for which fire data was available across jurisdictions was 40 years. As a result, the fire data were standardized by using a 40 year limit to identify areas disturbed by fire (i.e., less than 40 yrs). Due to the small amount of land in the 40-50 year age class for fires (for areas where the information was available), the change from 50 years to 40 years post-fire resulted in only minor discrepancies in measures of area disturbed by fire between the 2008 Scientific Review and the present assessment.

2.4.3.2 Habitat selection

It is well accepted that boreal caribou habitat use can vary spatially in response to regional and local environmental conditions and habitat availability (see Appendix 7.3). The current ranges of boreal caribou in Canada spans nine ecozones (Figure 5). Ecozones represent areas with roughly the same climatic conditions, land features, and floral and faunal species. They provide a logical starting point for controlling for some of the regional variation in abiotic and biotic conditions experienced by caribou across their boreal distribution.

Figure 5. Location of current ranges of boreal caribou in relation to the ecozones in Canada.

Figure 5. A map showing the location of the 57 boreal caribou ranges relative to 15 ecozones in Canada.

Generally, resource selection (RSF) models are used to quantify a species' habitat use relative to its availability. RSFs were developFed to describe caribou habitat use at two broad-scales: a national model describing habitat use across the extent of the species' occurrence and ecozonal models describing habitat use across the different ecozones found in the boreal forest of Canada (Figure 6; Appendix 7.3). These scale-dependent analyses allowed for a better understanding of how the variability in habitat selection at different spatial scales might affect the description of critical habitat, by testing whether or not controlling for ecozonal variation could strengthen inferences about the biotic and abiotic conditions influencing caribou demography.

The RSF models were developed using animal location data from 581 radio-collared caribou distributed among 27 ranges, including 179,022 locations during 2000 to 2010 (Table 4.1 in Appendix 7.3). The caribou location data were provided by several jurisdictions. A nationally consistent digital database was developed to define the types of habitats available to boreal caribou across the country, including the identification of different types of forest, wetlands, disturbances (fires and roads), forage quality, slope, aspect and the roughness of the landscape (Table 5.1-5.2 in Appendix 7.3). The landcover data were derived from MODIS imagery, complemented with the Peatlands of Canada database. The fire data used was from a compilation generated by EC and acquired from individual provinces and territories along with data from the CNFD for the fires within National Parks. The fire data included burns from 1917 and 2010.

The RSF models corroborated many important habitat selection relationships of boreal caribou reported in the literature. For example, the models indicated that caribou consistently avoided areas with high road density and avoided recent burns (less than 40 years old), across all ecozones. In this context, avoidance is defined as a reduction in use compared to what would be expected based on availability. Previous research has shown that boreal caribou are associated with mature conifer stands and peatlands where terrestrial lichens are available for winter forage (Stuart-Smith et al. 1997, Neufeld 2006, O'Brien et al. 2006, Brown et al. 2007, Courtois et al. 2007). Thus, recent burns that destroy lichen and result in young seral stands are likely avoided by caribou across the boreal forest (Schaefer and Pruitt 1991, Vors et al. 2007, Sorensen et al. 2008). However, such results derived from coarse national or ecozonal datasets should be interpreted with further considerations of local environmental factors (i.e., frequency, size and severity of forest fires in a specific region). Fire regimes vary significantly across the national, and even, the ecozonal distribution of boreal caribou. Boreal caribou are adapted to the local environmental conditions defining each range. From a caribou habitat use perspective, a 40 year post - fire time period might be considered "old" in certain regions, while considered relatively "young" elsewhere. As such, regional idiosyncrasies in fire recovery time period should be investigated with more refined datasets prior to making decisions on the appropriate recovery time period relative to caribou habitat.

In general, the ecozone specific models were more discriminatory and had better predictive accuracy than the national model. A cross-validation analysis was performed using withheld caribou locations to assess the ability of the RSF models to successfully predict caribou habitat. The results indicated that, even at the national scale, high frequencies of the withheld caribou locations occurred within high use areas identified by the RSF models. There was, however, significant variation in ecozonal relationships that should be explored through more detailed analyses, based on more refined datasets.

Figure 6. a) National and b) ecozone specific resource selection functions (RSFs) for boreal woodland caribou across the extent of occurrence in Canada. The probability of selection is scaled between low (green) and high (red).

Figure 6. Graphic representation of the results from the national resource selection model for boreal caribou in Canada showing areas with a low probability of caribou use in green, intermediate caribou use in yellow/orange and high use in red.
Figure 6. Graphic representation of the results from the national resource selection model for boreal caribou in Canada showing areas with a low probability of caribou use in green, intermediate caribou use in yellow/orange and high use in red.

The relative ranking of habitat as predicted by the RSF models was used to identify high quality caribou habitat. High quality habitat was defined by the top three (3) quantiles of the predicted probability of occurrence of boreal caribou, and was incorporated into subsequent analyses examining the effects of habitat condition on caribou recruitment (see section 2.4.3.4).

2.4.3.3 Buffer analysis

The purpose of the buffer analysis was to quantify the ecological effects of human development on boreal caribou, relative to the extent by which human disturbances were buffered. It was also used to examine the impact of buffering on the configuration of anthropogenic disturbance and, in turn, the effect of configuration of disturbance in a range on caribou.

The effect of anthropogenic disturbances on boreal caribou was tested by comparing changes in the predictive power (R²) of a model describing variation in caribou recruitment as a function of percent total disturbance of the range, an aggregate measure of fire and buffered human disturbance, for 24 boreal caribou ranges across Canada (see Section 2.4.3.4 or Appendix 7.5 for details). The ten models tested differed only in the buffer radius applied equally to all types of anthropogenic development with the exception of reservoirs (see rationale for removing reservoirs from the disturbance footprint in Section 2.4.3.4).

The relationship between the disturbance-recruitment model’s predictive power and different buffer treatments on human disturbance was dome-shaped and was sub-divided into 3 zones: increasing, stable, and decreasing (Figure 7). The stable zone was defined as models with buffer treatments with an -value that was within 2.5% of the best model (1000 m buffer; Appendix 7.4). The 2.5 % threshold was chosen to approximate a one-tailed significance test of α = 0.025 with the tail of the distribution bound by the most extreme R²-value (i.e., best model). The most conservative buffer (500 m) within the stable zone (Figure 7) was selected to represent the effective area of anthropogenic disturbance.

Figure 7. R²-value of models describing recruitment as a function of percent total disturbance with different buffers applied to anthropogenic disturbance. The dashed line denotes the 500 m buffer selected to represent the effects of human disturbance. The three zones (increasing, stable, decreasing) represent trend in the relationship.

Figure 7. The effect of changing the buffer applied to anthropogenic disturbance on the predictive power of a model describing variation in caribou recruitment as a function of total, non-overlapping disturbance. The buffers tested ranged from 100 m to 4 km in width. The changes in predictive power (R2) are domed shaped (initially increase then decrease with increasing buffer on anthropogenic disturbance). The 500 m buffer was used to represent the ecological footprint of anthropogenic development on boreal caribou.

A sensitivity analysis indicated that the disturbance-recruitment relationship applied with a 500 m disturbance buffer width produced stable estimates of the effect of anthropogenic disturbance on caribou recruitment (see Appendix 7.4). Moreover, the 500 m width appeared to capture basic information about the effects of fragmentation or the spatial configuration of human disturbance on the landscape in addition to the effects of habitat loss. Only two of the six disturbance configuration metrics tested had a significant effect on caribou calf recruitment, after controlling for the percentage anthropogenic disturbance buffered by 500 m: edge density, a surrogate for quantifying the changes in the permeability of the landscape to predators, and the nearest-neighbour distance between disturbance patches, a surrogate measure of landscape connectivity. These two metrics of disturbance configuration were incorporated into the subsequent analysis to identify the relationship between population (recruitment) and habitat conditions (disturbance, see Section 2.4.3.4).

2.4.3.4 Meta-analysis of population and habitat condition

In addition to incorporating the results of the enhanced disturbance mapping, the scope of the meta-analysis of caribou calf recruitment in relation to disturbance was expanded to better explain and understand the influence of habitat quality (including the type and configuration of disturbance) on this relationship (see Appendix 7.5). The expanded version included the development of eleven (11) candidate models that quantified improvements in the disturbance mapping (M0 vs. M3; Table 2), tested the relative effects of different disturbance types (M1-8), incorporated the results of the effects of anthropogenic disturbance and the configuration of disturbance (M9), and evaluated the influence of undisturbed habitats (M10-12) including high quality caribou habitat (hqh) as derived from the habitat selection model (Model 12).

Table 2. Specification of candidate models for the national meta-analysis.
Model Predictor variables Description
M0 total_dist_2008 Total non-overlapping disturbance from 2008 (see EC 2008)
M1 anthro_2011 Anthropogenic disturbance (500m buffer; reservoirs removed)
M2 fire_2011 Fire proportion (unbuffered)
M3 total_dist_2011 Total non-overlapping disturbance (500m buffer; reservoirs removed)
M4 lnlinear_2011 Percent buffered linear disturbance (500m buffer)
M5 poly_2011 Polygonal anthropogenic disturbances (500m buffer; reservoirs removed)
M6 lnlinear + poly_2011 M4 + M5
M7 anthro + fire_excl_anthro_2011 M1 + fires exclusive of anthropogenic disturbances
M8 total_dist + fire_prop_dist_2011 M3 + fires as proportion of total disturbance
M9 total_dist + ln_nn_2011 M3 + area-weighted mean nearest neighbor distance (500m buffer)
M10 ifl_2011 Proportion intact forest landscape exclusive of anthropogenic disturbance
M11 ifl_nofire_2011 Proportion intact forest landscape exclusive of anthropogenic disturbance and fire
M12 total_dist + hqh_2011 M3 + proportion of high quality habitat

The top model (M3) explained 69% of the variation in calf recruitment across a sample of twenty-four (24) ranges based on the percent total disturbance (fire + 500 m buffered anthropogenic disturbance; Figure 8) on each range. This model was analogous to the top model used in the 2008 Scientific Review. However, the new disturbance maps, which allowed better temporal matching of demographic data with disturbance data, and exclusion of reservoirs from the disturbance estimates, resulted in a 12% gain in explanatory power over the 2008 model. Most of the negative effects of disturbance were attributed to human development (60% in isolation), while only 5% of the variation in recruitment could be attributed to fire alone (see Appendix 7.5). Nevertheless, their combined influence was greater than the sum of their individual contributions. Decomposing anthropogenic disturbance into linear and polygonal features did little to improve the predictive power of the recruitment model, but the negative effect of linear disturbance features was greater than the negative effect of polygonal disturbances (see Appendix 7.5).

Figure 8. Graph showing 50, 70 and 90 % prediction bands for the best univariate regression model (M3) of caribou recruitment and landscape disturbance.

Figure 8. Mean boreal caribou recruitment (calves/100 female cows) plotted versus the percent total, non-overlapping disturbance for 24 study areas in Canada. The relationship is negative with over 40 calves being recruited per 100 female cows at low disturbance levels (<5% disturbance) and fewer than 10 calves being recruited per 100 female cows when disturbance levels exceed 85%.

The disturbance-recruitment relationship from the meta-analysis was used to parameterize a model of habitat-based population growth (see Appendix 7.8) based on the percent total disturbance within each boreal caribou range, and this indicator was used for the integrated risk assessment (Section 2.4.6.1) and to derive the categories of risk for the disturbance thresholds (Section 2.4.6.2).

Jurisdictions were requested to share best available demographic data for the update assessment. Data availability and type varied widely between ranges (Appendix 7.1; Appendix 7.11).

2.4.5.1 Current condition

The assessment of current range conditions evaluated the probability that current range conditions were sufficient to support self-sustaining caribou populations. As per Section 2.1, a “self-sustaining” population is one that experiences stable or positive growth (trend) over the short term (first criterion), and is large enough (size) to persist over the long-term (second criterion) without active management intervention (third criterion), such as predator control. Table 3 describes the indicators developed to evaluate the three criteria, namely:

  1. the probability that caribou would experience stable or increasing population growth over the short-term, which was expressed as Pr (λ ≥ stable) over 20 years; and
  2. the probability that the population was large enough to avoid quasi-extinction, defined as a population with less than 10 reproductively active females over the longer term without the need for ongoing active management intervention (e.g., predator management or translocation from other populations), which was expressed as Pr (Nt ≥ Qext) over 50 years (see Appendix 7.8 for details of the two indicators). This indicator was used to assess the increased risk of extinction for small population sizes such that lower values of Pr (Nt ≥ Qext) indicate a higher risk of extinction and vice versa.

Two types of analyses were required to estimate the probabilities that ranges could maintain a self-sustaining local population: a non-spatial population viability analysis (using a generic population model; Appendix 7.6), and a probabilistic decision-analysis tool, or Bayesian Decision Network (BDN[2]; Appendix 7.8). Two different sources of information were used to define range-specific demographic parameters. The disturbance-recruitment relationship from the meta-analysis was used to parameterize the population model based on habitat condition (see below under Indicator from habitat condition), whereas the information provided by jurisdictions on population size and trend were used to define demographic parameters based on population condition (see below under Indicators from population condition). The self-sustainability indicators from habitat and population conditions were used to inform the integrated risk assessment, which is a component of the scientific description of critical habitat (Section 2.4.6).

Table 3. Indicators of self-sustainability used in the Integrated Risk Assessment to assess current conditions.
Indicator of self-sustainability Data used1 Description
1) Probability of population growth rate based on habitat condition, Pr (λ ≥ stable)habitat Population growth (λ) derived from the recruitment-disturbance relationship. The proportion of times the mean projected population growth over 20 years (yr) was either stable or increasing on average over that time interval Pr (λ ≥ stable), given a specified set of demographic estimates. The 20-yr time period corresponds to the IUCN criteria for evaluating rates of change and probabilities of population decline. Only extant populations, defined as populations greater than 10 animals at the end of the time interval, were used to calculate Pr (λ ≥ stable).
2) Probability of population growth rate based on population condition, Pr (λ ≥ stable)population Population information reported from jurisdictions.
3) Probability of population size exceeding quasi-extinction threshold,
Pr (Nt ≥ Qext)
Population size reported from jurisdictions. Estimates calculated assuming good demographic conditions (stable growth). The proportion of times a population remained extant over 50 yrs (Pr (Nt ≥ Qext) given a set of specified demographic estimates. Extant populations were defined as those with greater than 10 reproductively active females, based on IUCN criteria for assessing extinction risk (IUCN 2010).

1 Decision rules were developed for addressing the variability in data availability across ranges (see Appendix 7.8).

Population model

The generic population model (referred to as "ensemble model" in Appendix 7.6) was developed to create a large database of potential outcomes from which range-specific probabilities could be derived. The demographic variables used in the model (Table 4) were estimated from a review of the published literature on boreal caribou and cover the range of demographic values reported from the jurisdictions. The model projected annual changes in initial population (Nt = 1) over 20 and 50 years using all possible permutations of population parameters and assumed levels of error in each parameter (i.e., standard deviation, SD) to reflect different levels of annual stochastic variation. The projected changes in population size at each time (t) step (Nt = 1, Nt = 2,… Nt = 20) over 20 years were used to calculate the annual realized lambda (λ, proportional change in population size at t+1) for each combination of parameters (Appendix 7.6). These were used to define the range of lambda values for which the population showed decreasing, stable and increasing growth, on average over the 20 years given stochastic variation (Table 5) and to estimate the self-sustainability indicator of stable and increasing population growth (Pr (λ ≥ stable)). The upper and lower bound for defining a self-sustaining population was defined by λ=1.01 and λ=0.99, respectively. Many populations under a variety of demographic conditions are likely to be self-sustaining using this definition, particularly given assumed annual stochastic variation in population parameters and the effects of demographic stochasticity. However, some may not be self-sustaining (e.g. smaller populations considered over longer time frames), therefore care is required when interpreting whether one or more management responses are likely to improve the chances of a population being self-sustaining in the future. The projected changes in population size (Nt = 1, Nt = 2,… Nt = 50) over 50 years were used to calculate the self-sustainability indicator for assessing the increased risk of extinction for small population size (Pr (Nt ≥ Qext)).

Table 4. Ranges of parameters and their incremental step sizes used in factorial projections for the generic population modelling.
Parameter Range of Values # levels (increments)
N0 - initial population size (mature females) 20-10,000 18
Annual mean per capita recruitment - recruitment of females/female 0-0.3 13 (increments of 0.025)
Annual SD about recruitment 0.01-0.21 4 (increments of 0.07)
Annual potentially breeding female survival rate 0.70-1.0 16 (increments of 0.02)
Annual SD about survival 0.01-0.21 4 (increments of 0.07)
K 20 times N0 3 (values of 3, 6, and 20)
rmax 1.1-1.3 2 (1.1 and 1.3)
Table 5. Range of lambda values corresponding to each of the population trend categories determined through population simulations.
Population Category Range of λ
Decreasing <0.99
Stable 0.99-1.011
Increasing >1.01

1 above a λ = 0.99, populations have a high probability of remaining, on average, stable for long-periods, given stochastic variation (Appendix 7.6).

Indicator from habitat condition

One self-sustainability indicator was calculated from habitat condition: the probability that the local population would experience stable or increasing population growth (λ) over a 20-year period, i.e., Pr (λ ≥ stable)habitat, hereinafter referred to as the habitat indicator of population growth (Table 3). The percentage of total disturbance (fire and 500 m-buffered anthropogenic disturbance) within each range was used to estimate the recruitment rate (± SD) using the recruitment-disturbance relationship developed in section 2.4.3.4. The national average for adult survival (Sad = 0.85) was used in combination with the expected recruitment value and associated coefficients of variation for each range to calculate annual population growth used by the BDN to estimate range-specific Pr (λ ≥ stable)habitat. This indicator was calculated as an expected mean over all possible population sizes observed nationally to isolate the effects of population growth from the effects of population size (see section Indicators from population condition).

Indicators from population condition

Two self-sustainability indicators were calculated for each range based on population condition information: Pr (λ ≥ stable)population, hereinafter referred to as the population indicator of population growth and Pr (Nt ≥ Qext), hereinafter referred to as the indicator of quasi-extinction, a persistence measure used to assess the increased risk of extinction for small population size (Table 3).

The population indicator of population growth was estimated by the BDN for all ranges using the reported lambda values (λ ± SD) or population trend (decreasing, stable, or increasing; Table 5) reported by jurisdictions. A set of decision rules was developed to address the variability in partial data reported from jurisdictions for lambda. For example, the national average for adult survival (Sad = 0.85) was used in combination with observed recruitment values to estimate lambda if the reporting jurisdiction only provided recruitment estimates for a given range. As per the habitat indicator of population growth, the population indicator of population growth was calculated as an expected mean over all possible population sizes to reflect variability in expected outcomes based on observable population sizes at the national scale.

The indicator of quasi-extinction was estimated using population size for each range and assuming good demographic conditions (stable growth) to isolate the effect of population size from population growth. This provided an assessment of the increased risk of extinction for small populations due to stochastic events for the integrated risk assessment (see Section 2.4.6). This indicator was not calculated in the absence of population size data.

2.4.5.2 Future conditions

A semi-spatial habitat-dynamics model was developed (Appendix 7.7) to forecast future changes in the spatio-temporal patterns in habitat conditions and their impact on the assessment of a range’s ability to maintain a self-sustaining population. These projections of natural disturbance were used as input data for the estimation of the habitat indicator of population growth through time (see Table 3) which in turn provided information to support the consideration of range-specific disturbance thresholds. The three (3) steps required to estimate the indicator were:

1) The current state of habitat conditions within each range was characterized by information about fire regimes, anthropogenic disturbances, water bodies, and different habitat types identified by the Moderate Resolution Imaging Spectroradiometer (MODIS, a sensor on the Terra satellite). All features within the range were time stamped to track their change through time. Time since last fire was used to determine the age of forests following fires. Anthropogenic disturbances were assumed to have been created in the same year of the Landsat imagery that was used to digitize the disturbance (section 2.4.3.1). An expected stable or equilibrium age structure for the different forest types was created using long-term simulations (see Appendix 7.7).

2) The current range maps (from step 1) were projected 100 years into the future using the Spatially Explicit Landscape Event Simulator (SELES; Fall and Fall 2001) according to four broadly contrasting scenarios:

  1. Static conditions: Current conditions remain unchanged (i.e., there are no new anthropogenic or natural disturbances, ecological succession or recovery of disturbed habitat). While this is not intended as a realistic scenario, it is an analogue of the 2008 Scientific Review "static" analysis and extends it to allow estimation of demographic stochasticity in the habitat-based PVA model (Appendix 7.6).
  2. Recovery only: This projects the effect of passive (i.e., no active restoration) of disturbed areas over time on range condition (see Appendix 7.6 for rules of passive recovery). This allows for an assessment of the partial effect of reducing current levels of fire and anthropogenic disturbance on the range upon the population parameters irrespective of the confounding effects of other dynamics.
  3. Natural disturbances only: This scenario was used to estimate the partial effect of natural habitat dynamics (see Appendix 7.6 for rules of natural habitat dynamics), given current levels of anthropogenic disturbance upon the population parameters.
  4. Recovery + natural disturbance: This scenario was used to examine the combined effects of range dynamics assuming no further increase in anthropogenic disturbance upon the population parameters.

The spatial simulation maps were used to calculate annual estimates of the percent future total disturbance (fire and 500m buffered anthropogenic) for each of the four scenarios.

3) The future disturbance estimates were used to calculate an expected recruitment rate from the disturbance-recruitment relationship (see Section 2.4.3.4). Using the same approach as in Section 2.4.5.1, the expected recruitment was combined with a national mean adult survival rate (Sad = 0.85) and associated coefficients of variation for both to produce probability estimates of expected future growth rate (λ) of the population. The probability that the future range would experience stable or increasing growth Pr (λ ≥ stable) was estimated using the large database created by the Population Model and the BDN (Section 2.4.5.1) to produce range specific mean estimates of lambda (λ) for future levels of disturbance averaged over 0-20 years, 21-50 years and 51-100 years.

The results from the assessment of future conditions for each scenario and time interval are represented in the factsheets (Appendix 7.10) for each range as per Figure 9. Such information can be used to support the interpretation of the range-specific disturbance thresholds (see Section 2.4.6.2 and Appendix 7.9). However, they are not, and should not, be interpreted as projections of the actual condition of the range through time.

Figure 9. Probability that the population growth rate is either stable or positive (Pr (λ ≥ stable)) as a function of percent (%) total disturbance based on four (4) hypothetical habitat dynamic scenarios: (1) Current: static conditions; (2) Recovery Only: passive recovery of old disturbances; (3) Natural Disturbance Only (Nat. Dist. Only): new disturbances created by fire without passive recovery; and (4) Recovery + Nat. Dist.: the combined effects of new fires and passive recovery of old disturbances, averaged for three time intervals (1-20, 21-50, 51-100 yrs). Note: this example is for the Atikaki-Berens range.

Figure 9. figure legend should suffice. “Probability that the population growth rate is either stable or positive (Pr (λ ≥ stable)) as a function of percent (%) total disturbance based on four (4) hypothetical habitat dynamic scenarios: (1) Current: static conditions; (2) Recovery Only: passive recovery of old disturbances; (3) Natural Disturbance Only (Nat. Dist. Only): new disturbances created by fire without passive recovery; and (4) Recovery + Nat. Dist.: the combined effects of new fires and passive recovery of old disturbances, averaged for three time intervals (1-20, 21-50, 51-100 yrs).  Note: this example is for the Atikaki-Berens range.
2.4.6.1 Integrated Risk Assessment

"Risk" is defined here as the likelihood that a range can maintain a self-sustaining local population, and the uncertainty surrounding the indicators used in the assessment. The Integrated Risk Assessment for each local population had 2 main components (Tables 6 and 7):

  1. A statement about the probability that current range conditions, described in terms of habitat and population conditions, are sufficient to support a self-sustaining population (Table 6). This involved a lines and weight of evidence approach to integrate the three indicators of self-sustainability (see Section Decision rules for integration) and assign each range to one of five likelihood categories of self-sustainability. These categories represent a gradient of risk based on a modified version of the likelihood scale developed by the Intergovernmental Panel on Climate Change (IPCC 2005.).
  2. The uncertainty, or alternatively, the certainty in the Integrated Risk Assessment was assessed using two certainty measures: a statement of certainty that reflects the type of information used to estimate the indicators of self-sustainability (Table 7), and their consistency (described under Section Decision rules for Integration). A reported lambda measured over at least three years within the last 10 years was assigned a higher level of certainty in terms of the quality of information than a reported trend. A reported lambda measured over less than three years within the last 10 years was assigned the same level of certainty as a reported trend (Table 7).
Table 6. Likelihood scale for the Integrated Risk Assessment of current conditions.
Probability of the outcome (self-sustaining) Category of likelihood that the range condition can support a self-sustaining local population Self-sustainability outcome1
≥ 0.9 Very likely SS
< 0.9 to ≥ 0.6 Likely SS
< 0.6 to ≥ 0.4 About as likely as not NSS/SS
< 0.4 to ≥ 0.1 Unlikely NSS
< 0.1 Very unlikely NSS

1 SS: Self-Sustaining; NSS/SS: Not Self-Sustaining/Self-Sustaining; NSS: Not Self-Sustaining.

Table 7. Level of certainty associated with the availability of demographic data for a range. Data on habitat condition was not included because it was available for and standardized across all ranges.
Certainty statement of evidence based on availability of demographic data Demographic data available
Much evidence Reported lambda1 and population size
Considerable evidence Reported trend and population size
Some evidence Population size
Limited evidence No demographic data

1 Lambda must be measured over ≥3 years within the last 10 years. A reported lambda that did not satisfy this criterion was treated like a reported trend.

Decision Rules for Integration

A set of decision rules, illustrated in Figure 10, was established for weighting the individual assessments of self-sustainability based on the three (3) indicators and providing one integrated risk assessment. First, a self-sustainability outcome was determined for each indicator using the probability intervals in Table 6. The three individual outcomes were compared to determine if the three indicators of self-sustainability were in agreement (Step 1). If all three indicators suggested the same selfsustainability outcome (i.e., high agreement as defined in Table 8), the later was considered as the integrated risk assessment (Step 8). For this situation, the associated likelihood category was based on the indicator suggesting the lowest, more conservative, likelihood of self-sustainability.

Additional steps were taken to resolve disagreement among the indicators and identify the most plausible outcome given available range-specific information. In cases where the three indicators were in complete disagreement (i.e., low agreement as defined in Table 8), the precautionary principle was applied such that the outcome from the indicator suggesting the lowest probability of maintaining a self-sustaining population was assigned the most weight of evidence, and selected (Step 2) with its associated likelihood category as the integrated risk assessment outcome. For all other cases, Steps 3-7 of the decisions rules were implemented to identify the indicator with the most weight.

Conflicting assessment outcomes resulting from partial agreement (Table 8) between the habitat and population indicators of population growth (Step 4) was resolved by:

The precautionary principle (Guiding Principle 5) was applied in situations where the outcome from the indicator of population growth (outcome of Step 4a-b) disagreed with the outcome from the indicator of quasi-extinction used to flag the increased risk of extinction for small population sizes (Step 6). The integrated risk assessment was based on (i.e., more weight was attributed to) the indicator that produced the highest risk or lowest probability of meeting the goal of maintaining a self-sustaining population. This was synonymous with adjusting the integrated risk assessment to indicate cases where there might be an increased risk of extinction due to small population size.

As stated previously, in addition to the certainty in the data used (Table 7), the certainty in the integrated risk assessment was further defined through assessing the consistency in the information provided by each indicator for a given range. Consistency was evaluated based on the number of indicators that were in agreement with the indicator assigned the most weight of evidence in the integrated risk assessment (Table 8). The evaluation was conducted only on ranges with all 3 indicators of self-sustainability. As above, the approach used for the certainty statement represents a modified version of that developed by IPCC (2005).

Table 8. Level of certainty in the consistency of information.
Indicator of self-sustainability attributed the most weight of evidence Conditional consistency
criteria
Certainty statement of consistency of information
Pr (λ ≥ stable) Consistent with alternative Pr (λ ≥ stable) indicator High agreement
Pr (Nt ≥ Qext) Consistent with the two Pr (λ ≥ stable) indicators
Pr (λ ≥ stable) Consistent with
Pr (Nt ≥ Qext)
Partial agreement
Pr (Nt ≥ Qext) Consistent with one of the two Pr (λ ≥ stable) indicators
Pr (λ ≥ stable)
Pr (Nt ≥ Qext)
No consistency Low agreement

Figure 10. a) Decision rules applied to inform integrated risk assessment. b) Elaboration of rules used to resolve difference between the indicators of population growth.

Figure 10. a) decision rules to inform integrated risk assessment. Step 1: Are the three indicators of self-sustainability in agreement? If yes, then the integrated risk assessment is complete (step 8). If no, then go to step 2: are all three indicators of self-sustainability in disagreement? If yes, then go to step 7: apply the precautionary principle and select the indicator of self-sustainability associated with the highest level of risk (most conservative). This becomes the final integrated risk assessment (step 8). If no, go through steps 3 and 4 in yellow box to resolve differences. Step 3: are the indicators of population growth in agreement? If no, the go to step 4: apply decision rules to determine most reliable estimate of the indicator of population growth (described in more detail in figure 10b) and move to step 5. If yes (indicators of population growth are in agreement), then go directly to step 5: this becomes the final categorization for population growth. Step 6: is the final indicator of population growth in agreement with the indicator of quasi-extinction used to assess the increased risk of extinction for small population sizes? If no, the go to step 7: apply the precautionary principle and select the indicator of self-sustainability associated with the highest level of risk and use it for the final integrated risk assessment (step 8). If yes (indicator of population growth and quasi-extinction are in agreement), then move to step 8 and use them as the final integrated risk assessment. b) elaboration of decision rules to resolve differences between indicators (yellow box including step 3 and 4). Step 3: Are the indicators of population growth in agreement? If no, go to step 4a: Could the population indicator of population growth be the result of predator management? If yes, then use the habitat indicator of population growth as the final categorization indicator of population growth (step 5). If there is no predator management, then go to step 4b: is the population indicator of population growth based on lambda averaged over three years or more within the last ten years? If yes, then use the population indicator of population growth as the final categorization indicator of population growth for step 5. If no, then use the habitat indicator of population growth as the final categorization indicator of population growth for step 5. The decision rules following step 5 (steps 6-8) are described above.
Figure 10. a) decision rules to inform integrated risk assessment. Step 1: Are the three indicators of self-sustainability in agreement? If yes, then the integrated risk assessment is complete (step 8). If no, then go to step 2: are all three indicators of self-sustainability in disagreement? If yes, then go to step 7: apply the precautionary principle and select the indicator of self-sustainability associated with the highest level of risk (most conservative). This becomes the final integrated risk assessment (step 8). If no, go through steps 3 and 4 in yellow box to resolve differences. Step 3: are the indicators of population growth in agreement? If no, the go to step 4: apply decision rules to determine most reliable estimate of the indicator of population growth (described in more detail in figure 10b) and move to step 5. If yes (indicators of population growth are in agreement), then go directly to step 5: this becomes the final categorization for population growth. Step 6: is the final indicator of population growth in agreement with the indicator of quasi-extinction used to assess the increased risk of extinction for small population sizes? If no, the go to step 7: apply the precautionary principle and select the indicator of self-sustainability associated with the highest level of risk and use it for the final integrated risk assessment (step 8). If yes (indicator of population growth and quasi-extinction are in agreement), then move to step 8 and use them as the final integrated risk assessment. b) elaboration of decision rules to resolve differences between indicators (yellow box including step 3 and 4). Step 3: Are the indicators of population growth in agreement? If no, go to step 4a: Could the population indicator of population growth be the result of predator management? If yes, then use the habitat indicator of population growth as the final categorization indicator of population growth (step 5). If there is no predator management, then go to step 4b: is the population indicator of population growth based on lambda averaged over three years or more within the last ten years? If yes, then use the population indicator of population growth as the final categorization indicator of population growth for step 5. If no, then use the habitat indicator of population growth as the final categorization indicator of population growth for step 5. The decision rules following step 5 (steps 6-8) are described above.
2.4.6.2 Range-specific management thresholds

The integrated risk assessment is designed to evaluate a set of likelihoods related to the probability that current range conditions will support a self-sustaining population. Each likelihood statement is associated with a range of indicator values corresponding to different levels of confidence in the desired outcome; in this case, self-sustaining local populations (see Table 8). However, an assessment of current conditions does not answer the question: what must be conserved (for self-sustaining ranges) or recovered (for non-self-sustaining ranges) to achieve a desired level of certainty in meeting the recovery goal? In the context of the present assessment, this can be framed relative to risk and expressed through the identification of habitat-based management thresholds; specifically, disturbance thresholds. While it falls beyond the scope of a scientific assessment to determine management thresholds, information and methodologies to support this are presented here (see Appendix 7.9 for a full discussion of the approach).

Ecological thresholds can be identified when the relationship between an attribute of interest and the environmental driver of change, or stressor, is non-linear, suggesting an abrupt change across a small range of values (Groffman et al. 2006, Villard and Jonsson 2009, Samhouri et al. 2010). The existence of discrete ecological thresholds can inform management decisions, particularly when the threshold represents a clear boundary between two states that differ with respect to desired management outcome. However, when relationships between environmental stressors and ecological attributes are linear, or when transitions from one state to another are gradual, no ecological threshold can be defined, and consideration of management thresholds relies on the assignment of ecological risk along a continuum of conditions. In either case, establishment of management thresholds relies on management decisions regarding the level of acceptable risk. Such decisions can be informed by a probabilistic assessment of potential outcomes, relative to desired state, as represented here.

In the present assessment, range condition is a primary indicator related to the recovery criteria of stable or positive population growth, and is used here as the starting point for considering management thresholds related to critical habitat. Integration of the recruitment-disturbance relationship described in Section 2.4.3.4 with a mean annual adult female survival rate (Sad = 0.85), allows derivation of a lambda function for the habitat-based population growth indicator (Appendix 7.8). This relationship expresses the probability of observing a mean lambda over a 20-year period indicative of a stable or increasing population (λ ≥ stable), at varying levels of total range disturbance (Figure 11). While there are regions of greater certainty at the low and high ends of the disturbance gradient, there is no discrete threshold separating sustainable from unsustainable conditions. The likelihood of observing a non-declining population decreases, or risk of failure to achieve the recovery objective increases, with increasing levels of disturbance.

Figure 11. Probability of observing stable or positive growth (λ ≥ stable) of caribou populations over a 20-year period at varying levels of total range disturbance (fires ≤ 40 years + anthropogenic disturbances buffered by 500 m). Lambda (λ) was calculated using disturbance-specific recruitment values from the meta-analysis and a mean annual adult female survival rate of 0.85, consistent with other components of the critical habitat assessment (see Appendix 7.8). Certainty of outcome, ecological risk, and management scenarios are illustrated along a continuum of conditions.

Figure 11. Graph showing the decrease in probability of achieving stable or increasing population growth as a function of increasing disturbance. The latter is used to categorize the likelihood of achieving the recovery goal of a self-sustaining population or, conversely, the risk of not achieving the recovery goal to help inform management. For example, the probability of achieving stable or increasing population growth is high when disturbance levels are low. It is very likely that the goal of a self-sustaining population will be achieved. Thus, there is a low risk associated with not meeting the recovery goal. At higher levels of disturbance, the probability of stable or increasing population growth is low and it becomes very unlikely that the goal of a self-sustaining population will be achieved. The latter corresponds to a very high risk of not meeting the recovery goal and suggests that habitat restoration may be needed.

The disturbance values associated with the likelihood categories for the habitat-based population growth indicator are presented in Table 9. The intervals associated with each likelihood statement reflect a range of indicator values, consistent with a probabilistic representation of certainty in outcome. Within the integrated risk assessment, an assignment of “self-sustaining” was reached when available information suggested a range was more likely than not to support a self-sustaining population, based on a probability of sustained stable or positive growth ≥ 0.60 (Table 8; Figure 11). Similarly, when it was more unlikely than not that a range would support a self-sustaining population (probability of sustained stable or positive growth ≤ 0.40), the range received an assignment of a probable outcome of “not self-sustaining”. The assignments reflect a weight of evidence approach based on quantitative criteria. The high uncertainty in outcome associated with the intermediate likelihood category of “as likely as not” results in a joint assignment of probable outcomes.

Table 9. Intervals of total range disturbance associated with varying levels of certainty in outcome and risk relative to achieving the recovery objective of stable or positive population growth.
Probability of Sustained Stable or Positive Growth1 Likelihood of Desired Outcome Disturbance Interval2 Level of Risk
≥ 90% Very likely ≤ 10% Very low
< 90 to ≥ 60% Likely > 10-35% Low
< 60 to ≥ 40% As likely as not > 35-45% Moderate
< 40 to ≥ %10 Unlikely > 45-75% High
< 10% Very unlikely >75% Very high

1 Intervals adapted from IPCC 2005; time frame for assessing mean growth rate is 20 years.
2 See Figure 11.

The interpretation of risk has both objective and subjective components. The relative ranking of risk from very low to very high in Table 9 represents the continuum of conditions, and associated probabilities of outcomes, given a discrete ecological threshold has not been identified in the relationship of interest. This interpretation is thus objective, in that it represents a gradient in certainty of outcome based on the underlying ecological relationship. The assignment of risk to each likelihood category, and the likelihood intervals themselves, can also be interpreted as an expression of acceptance of varying levels of certainty in desired outcomes, and therefore subjective in nature. For example, acceptance of an “as likely as not” outcome as moderate risk from a management perspective is a value statement, and thus subject to change based on management objectives.

The present scientific assessment can speak to the likely outcomes and relative risk associated with current and future conditions, given a stated management objective, but it cannot determine the level of acceptable risk relative to realizing that objective. Specification of acceptable risk is necessary to establish management thresholds, because different levels of acceptable risk are expressed as different management thresholds. For example, if a “likely” outcome relative to the recovery objective of self-sustaining populations was considered acceptable from a management perspective, then the associated management threshold might be set at the upper end of the range of disturbance values associated with this likelihood category, with values below consistent with a desired state, and those above representing an undesirable or unacceptable state. A more or less conservative management approach would result in a lower or higher disturbance threshold. Because thresholds are typically used as a trigger for management actions, the use of incremental or tiered thresholds, representing a gradient of risk, can support graduated management responses along the recovery continuum.

Regardless of whether level of acceptable risk is expressed through a single or multiple thresholds, the associated disturbance intervals derived from Figure 11 reflect expected outcomes based on patterns evident at a national scale. They do not provide for range-specific assessments in the absence of additional information. Outcomes for individual populations will fall at different locations within these intervals, and may even fall outside them. Consistent with the integrated risk assessment, other indicators related to recovery criteria, also expressed relative to risk or likelihood of desired outcome, can be used to refine understanding of outcomes and interpretation of thresholds at a range-level (see Section 3.2 for the probability assignments associated with the two population indicators also used in the integrated risk assessment). The same decision rules applied in the integrated risk assessment would apply here (Figure 10, Section 2.4.6.1). Considerations of additional lines of evidence, such as projected extinction risk, based on the estimate of quasi-extinction derived from population size and trend (Table 6.8.1, Appendix 7.8), and potential future conditions, based on natural forest recovery of presently disturbed areas, and additional disturbance by fire (see Section 2.4.5.2), can also increase the overall certainty in outcome and consequently refine the interpretation of disturbance thresholds for individual populations relative to acceptable risk.

In general, the less information available, the less certainty there is in outcome. Multiple lines of evidence that suggest similar outcomes create greater certainty, as too does higher quality information. Certainty in outcome is the principal measure used to refine assessment of risk and inform the establishment and interpretation of disturbance thresholds for individual ranges. The steps necessary to support application of range-specific disturbance thresholds are illustrated in Figure 12, including a link to identification of appropriate management actions. In all cases where adjustment to interpretation of thresholds is considered, more information is required to make informed assessments of risk and certainty of outcomes.

Figure 12. Generalized approach to the assessment of risk and establishment and interpretation of disturbance-based management thresholds for boreal caribou.

Figure 12. Step 1: establish acceptable level of risk relative to certainty in desired outcome, and identify disturbance thresholds based on the relationship between disturbance and population growth. Step 2: Compare current level of range disturbance to identified thresholds values and assess risk relative to achieving the desired management outcome. Step 3: Consider additional assessment criteria and modify assignment of risk and threshold interpretation if certainty has increased or the probable outcome differs. Step 4: Evaluate projected future conditions for habitat and population, re-assess risk, and modify threshold interpretation if likely outcome differs. Step 5: Identify appropriate management actions based on final assessment of risk and associated threshold interpretation.

Further elaboration of the approaches described here is provided in Appendix 7.9. Examples of applications of these concepts are provided in the results section of the main document (see Section 3.3). The methodological framework presented provides a starting point for considering range-specific disturbance thresholds within a risk assessment framework. Their implementation requires determination of acceptable risk by managers.

2.4.6.3 Bio-physical attributes

Predation risk and forage availability are the primary factors that influence patterns of boreal caribou habitat use across Canada. In general, boreal caribou select mature and late seral-stage upland and lowland conifers and peatland complexes to spatially separate themselves from predators. These habitats often have an abundance of terrestrial and arboreal lichens, an important source of forage especially in winter. Grasses, sedges, herbaceous plants and lichens all become an important source of food from spring to fall when caribou broaden their diets. Similarly, caribou use habitats that facilitate escape from predators, such as shorelines and habitats with shallow snow, and many use islands for calving in spring. Caribou also show a general avoidance of shrub rich habitats, deciduous forests, and areas recently disturbed by fire and human activity (polygonal and linear disturbance). These habitats create favourable conditions for other ungulate species, which are the primary prey species for wolves and bears. In addition to creating sources of food for primary prey species, linear disturbance create travel corridors increasing the vulnerability of caribou to predators (James 1999; James and Stuart-Smith 2000; Nagy 2011).

The general characteristics of caribou habitat selection outlined above were described at the finer scale for seven ecozones and the five subregions of the Boreal Shield. The characteristics were compiled based on a literature review of boreal caribou habitat use in ecozones across their distribution in Canada (EC 2008), and from the results of the habitat selection analysis (see Section 2.4.3.2). No information was available for either the Southern Arctic ecozone or Taiga Cordillera ecozone (EC 2008). Within each ecozone, habitat selection was first described at a broad scale and then broken down according to the simple representation of the life history cycle of boreal caribou: calving, post-calving, the rut, winter and the habitats used as travel corridors. Habitats avoided by boreal caribou are described last. The latter framework is meant to facilitate interpretation with respect to describing the bio-physical attributes that allow caribou to carry out life processes necessary for species survival and recovery.

The interpretation of the habitat selection analysis requires the following considerations:


2 Bayesian decision networks (BDNs: also called probability networks) are statistical tools increasingly used in ecology and wildlife management to depict the influence of habitat or environmental predictor variables on ecological-response variables (Marcot et al. 2006).

Previous Page Table of Contents Next Page

Page details

Date modified: