Part 4 - Scientific Review for the Identification of Critical Habitat for Woodland Caribou, Boreal Population (Rangifer tarandus caribou) in Canada
6.4 Environmental Niche Analysis - Predicting potential occurrence of threatened boreal woodland caribou to support species recovery in Canada.
The Boreal Caribou Critical Habitat Science Review has pursued four complementary analytical approaches to re.ect the multi-scale, hierarchical interaction of species and their habitats; here we conducted an environmental niche analysis. We modeled the geographic extent of the environmental niche (fundamental and realized, e.g., abiotic and biotic) for boreal woodland caribou across its current extent of occurrence in Canada. While not directly incorporated into the prior analysis, the results presented here were expected to con. rm the current national distribution, and contribute to the local population management in the action planning stage. For example, re.ned and validated niche models could inform management on priority areas for habitat restoration where the current local extent is not large enough, as well as guide monitoring programs throughout the extent of occurrence as part of the adaptive management framework. The population and distribution objectives in the National Recovery Strategy (Environment Canada 2007) are to maintain existing local populations of boreal caribou that are self-sustaining and achieve growth of populations not currently self-sustaining, throughout the current extent of occurrence. Delineation and management of these local populations are key to the recovery of boreal caribou (Environment Canada 2007).
The geographic distribution of a species is a function of its ecology and evolutionary history, determined by diverse factors operating at different spatial scales, including climate (Case and Taper 2000, Soberon 2007). We assume that a species will be present at a given point where three conditions are met: a) abiotic conditions (such as climate) are favourable, b) biotic conditions (other species) allow species to maintain populations, and c) the region is accessible to dispersal from adjacent populations (Soberon and Peterson 2005, Soberon 2007). These three conditions describe a species niche, one of the fundamental theories in ecology of how organisms use their habitats. Niche theory suggests that . tness or habitat suitability is not monotonically related to conditions or resources, but instead decreases from either side of an optimal condition (Hirzel et al. 2002). A geographic area, with the appropriate set of abiotic factors, free of competition from biotic factors for a species in which the species may theoretically occur, may be regarded as the geographical expression of the fundamental niche (Hutchinson 1957). In contrast, an area where the abiotic conditions are favourable but we also consider biotic interactions, such as competition and predation, may be considered the geographical representation of the realized niche (Hutchinson 1957). A region that has the right set of biotic and abiotic factors and is accessible to the species (via dispersal) is the potential geographic distribution of the species (MacArthur 1967, Soberon 2007).
The recent availability of species occurrence data over large regions, for example from breeding bird surveys or large-scale wildlife surveys, combined with the availability of large-scale environmental climate and biotic data, has lead to an increase in approaches to model the distribution of species (Soberon 2007). Species distribution models are one type of empirical model relating spatial observations of an organism to environmental predictor variables, using a variety of statistical techniques, from logistic regression to more complex computation approaches (Guisan and Zimmerman 2000). Guisan and Thuiller (2005) suggested that environmental predictors for species distribution models should be chosen to capture the three main types of in.uences on species distribution: i) limiting factors or regulators, de.ned as factors controlling a species ecophysiology (e.g., temperature, water, soil), ii) disturbances (natural or human), and iii) resource availability, de.ned as all compounds that can be assimilated by organisms (e.g., energy and water). Spatial patterns in relationships between species and their environments vary with scale, often in a hierarchical manner (Johnson 1980, Pearson et al. 2004). Environmental niche models are conceptually similar to other species distribution models commonly employed in ecology (resource selection functions (Boyce and McDonald 1999), bioclimatic envelopes (Hijmans and Graham 2006) etc.), but niche models are explicitly linked to niche theory and usually address distribution across broad regional scales (Anderson et al. 2002). Environmental niche models reconstruct species’ ecological requirements (conditions or resources) and predict the geographical distribution of those requirements.
Ecological niche models (ENM) have been used to study issues in evolution (Peterson 2001), ecology (Anderson et al. 2002), and conservation (Peterson and Robins 2003). Their predictive models of species geographic distributions are important in a variety of conservation applications, such as conservation reserve design (Wilson et al. 2005), to predict the spread of invasive species (Peterson 2003), and to predict the effects of climatic change on species responses to future and past climates (Pearson and Dawson 2003, Hijmans and Graham 2006, Peterson et al. 2004). ENM models have been used to assess the distributional patterns of endangered species in many countries, including the United States (Godown and Peterson 2000), China (Chen and Peterson 2000), and eastern Mexico (Peterson et al. 2002). ENM models have also been used to incorporate multiple species and trophic interactions for example the implications for endangered spotted owls (Strix occidentalis) by invading barred owls (S. varia) facilitated by human disturbance in Washington and Oregon (Peterson and Robins 2003). Guisan et al. (2006) suggested that niche-based models may improve the sampling of rare species and Raxworthy et al. (2003) used ENM to target .eld surveys for under-studied reptiles and located previously undiscovered chameleon species in Madagascar.
As part of the science review process for boreal caribou, our goal was to support the identi.cation of critical habitat by employing environmental niche analysis to understand the pattern of occupancy in the current extent of occurrence. First, we examined the potential distribution (fundamental niche) as a function of climate and topography for two 30-year time periods: 1930 to 1960 and 1971 to 2000. Boreal caribou have experienced a range contraction at the southern limit of their distribution; therefore we hypothesized that the potential distribution of woodland caribou has shifted northwards between these two periods. This analysis may help determine contributions of climate change in limiting habitat use by caribou. Second, we predicted that the pattern of occupancy (realized niche) within the current extent of occurrence by biotic predictor variables, as derived from satellite imagery and other spatially explicit sources (e.g., Petorrelli et al. 2005). Woodland caribou declines are hypothesized to result from indirect effects of anthropogenic disturbance across their range (McLoughlin et al. 2003, Laliberte and Ripple 2004, Vors et al. 2007, Wittmer 2007). Proximally, human disturbance is thought to increase primary prey densities, and hence densities of predators, like wolves and black bears, that cause caribou populations to decline through apparent competition (Seip 1992, James and Stuart-Smith 2000). Therefore, we hypothesized that the realized niche would be constrained by these biotic interactions indexed through spatial measures of human disturbance across the boreal forest. Finally, we discussed the potential contributions of environmental niche modelling to various aspects of the ongoing, adaptive process of identifying critical habitat as identi.ed under the federal recovery plan for the species under the auspices of the Canadian Species at Risk Act.
The boreal caribou is a forest-dwelling sedentary ecotype of woodland caribou with an extent of occurrence over approximately 2.4 million km2, in eight provinces and territories, and occurring predominantly within .ve ecozones (Environment Canada 2007: Figure 1, Table 1). Often, boreal woodland caribou habitat is characterized as peatland complexes intermixed with mature to old pine, black spruce, and tamarack (e.g., Brown 2005) Forested peat complexes with abundant arboreal lichens and uplands dominated by mature conifers with dense ground lichens are typical of boreal caribou habitat, and are thought to provide for nutrient rich forage and as a refuge from higher predator densities associated with typical deer and moose habitat (e.g., deciduous/mixedwood, Thomas et al. 1996, McLoughlin 2003, 2005).
Geo-referenced boreal caribou observational location data were obtained from a variety of sources and consisted of various acquisition methods including: GPS (Global Positioning System) collar, ARGOS collar, VHF collar, aerial surveys, ground surveys, and incidental observations, ranging over time from the 1940’s to 2007. The database included over 1 million records of caribou observations. Two different datasets were used for niche modelling, to train and to validate the models, respectively. For the former, collared (GPS, ARGOS, VHF) data were used, whereas non-collared (surveys) data were held back for independent validation of outputs (Fielding and Bell 1997, Boyce et al. 2002).
To reduce spatial and temporal autocorrelation and minimize bias introduced by collar type, training data were limited to one location, per animal, per day by random selection where multiple daily acquisitions were captured (White and Garrott 1990). Because occurrence data frequently included location error, entries with uncertainty greater than 1 km were excluded for the study, regardless of acquisition method. The training dataset for current analyses consisted of over 217,000 points from collared animals, but the distribution of locations was
not uniform throughout the geographic range of boreal caribou (Table 1, Figure 1). Therefore, we strati.ed sampling to obtain datasets representative of the species-habitat variability in different ecozones across the extent of occurrence (Callaghan 2008). For modelling purposes, we produced ten subsets consisting of 10,000 points randomly selected from the 200,000 locations at the same ratio as the proportion of boreal caribou range represented by that ecozone (Boyce et al. 2002, Araujo and New 2006). Niche models are sensitive to sample size and to biases in the geographic distribution of the data (Peterson and Cohoon 1999, Stockwell and Peterson 2002). Statistical sampling designs outlined above have been suggested to limit these biases, while increasing model performance (Araujo and Guisan 2006). Although this balanced the coverage for ecozones, there were still considerable gaps in the geographic distribution of the occurrence data (Figure 1).
Appendix 6.4 - Table 1. Percent of Boreal caribou extent of occurrence in each ecozone and breakdown of collar locations used for training subsets input data.
Ecozone Boreal Plan Percent of Extent of Occurrence 13.5 Percent of .xes 21.4 Number of .xes 46561
Boreal Shield 41.1 43.5 94893
Hudson Plain 7.7 1.7 3809
Montane Cordillera 0.4 0.6 1207
Southern Arctic 2.2 0.1 134
Taiga Cordillera 0.1 0.0 29
Taiga Plain Taiga Shield 19.6 15.3 26.7 6.1 58115 13227
To predict the geographic extent of the boreal caribou environmental niche, we used abiotic and biotic variables including climate surfaces, topography, and biotic variables derived from satellite and existing vector data. Climate covariates were created using an interpolation technique based on thin-plate-smoothing splines (Hutchinson 1995). Biologically meaningful climate parameters (35 bioclimatic) were derived from monthly temperature and precipitation data that were averaged over two 30-year time periods: 1930 to 1960 and 1971 to 2000. Data were provided by the Canadian Forest Service at 30 arcseconds (~1 km) and 300 arcseconds (~10 km) resolutions (see McKenney et al. 2006). Potential variables were selected based on hypotheses developed from literature reviews of caribou and other northern ungulates (Table 2). Climatic variables have been shown to affect population dynamics in many large-bodied, northern ungulates through direct and indirect mechanisms at a variety of scales (Weladji et al. 2002). Indirect effects include for example, late winter precipitation and spring temperatures and precipitation on forage quality and its quantity in summer, and conditions of the summer range have shown associated effects on body size and reproductive success (Finstad et al. 2000). However, winter weather severity also has direct effects on population dynamics. Years with high snowfall may lead to increased winter calf mortality (Fancy and Whitten 1991), decreased body mass of calves (Cederlund et al. 1991) and lighter yearlings (Adams and Dale 1998). To reduce collinear predictor variables, we randomly sampled 10,000 grid cells from the entire country and derived Pearson’s correlation coef. cients for 35 bioclimatic parameters and elevation. We excluded variables that had a coef. cient of correlation >0.7 (Parra et al. 2004).
Precipitation in driest period
Total precipitation for 3 months prior to start
of growing season Growing degree days (gdd) above base temperature for 1st 6 weeks of growing season
Precipitation of coldest quarter
Gdd above base temperature 3 months prior to growing season Annual mean temperature Maximum temperature of warmest period Annual temperature range High summer/fall forage availability
– improved condition at breeding Early green-up – improved calf survival
Early green-up – improved calf survival
Food limitation caused by crusting or snow
depth Snowy late winters lead to improved summer forage
Range limit based on physiology Range limit based on physiology Range limit based on physiology
Digital elevation models (DEM) were derived from the Shuttle Radar Topography Mission (SRTM) data and obtained from the WorldClim website (www.worldclim.org) at 1-km and 10-km grid cell resolution.
To model realized niche, we attempted to capture attributes related to competition (e.g., resource availability and predation) that may restrict the occupied niche or environmental space. To account for forage resources we included: MODIS derived cumulative annual fraction of Photosynthetically Active Radiation (fPAR) (Coops et al. 2007, Huete et al. , Zhao et al. 2005), minimum annual fPAR (Coops et al. 2007), landcover (Latifovich, unpub), and peatland presence (Tarnocai 2005). The fPAR data were derived from a physically-based model that describes the propagation of light in plant canopies (Tian 2000) together with MODIS spectral bands. The cumulative annual fPAR re.ects the annual productivity of the site, whereas the minimum annual fPAR represents the minimum perennial cover of the site (Yang et al. 2006, Coops et al. 2007). Few studies incorporate information to account for predators or competitors directly in niche modelling, and those that have modelled the environmental niche of the predator or competitor and included them as a covariate (Peterson and Robbins 2003, Heikkinen et al. 2008). Few density data exist for the main predators of caribou across the boreal forest, yet predation by wolves and and black bears is the most frequently identi.ed limiting factor of caribou populations (Bergerud and Elliot 1989, Johnson et al. 2004). However, the principal driving factor changing predator distributions at the southern limit of caribou range is hypothesized to be anthropogenic disturbance. Modern commercial forestry creates new early seral forest stands which bene.t primary prey species, such as moose (Alces alces) and deer (Odocoileus spp), followed by wolves (Fuller 1981) resulting in increased predation rates on secondary prey such as caribou (Wittmer 2007). Human activities also include linear developments like roads, seismic exploration lines, pipelines, and utility corridors, all of which increase predation rates and ef. ciency of wolves preying on caribou (James and Stuart-Smith 2000, McKenzie 2006). Therefore, we approximated predation risk with: road density (calculated as the total distance of roads within 1-km pixel from the Updated Road Network (Geobase), the Statistics Canada Road Network (Statistics Canada) and the DMTI SpatialTM roads database GFWC 2007), disturbance (from GFWC anthropogenic footprint, GFWC 2007), mean forest patch size, number of forest patches, standard deviation forest patch size (derived from Earth Observation for Sustainable Development, calculated using (EOSD) gridded at 1 km, (Wulder et al. 2008).
Environmental Niche Modelling
Ecological niches of boreal caribou were modelled using Maximum Entropy (MaxEnt; Phillips et al. 2004, 2006). Maxent estimates the most uniform distribution (maximum entropy) of occurrence points across the study area, given the constraints that the expected value of each environmental variable under this estimated distribution matches its empirical average (Phillips et al. 2004, 2006). The raw output is a probability value (0-1) assigned to each map cell of the study area, which are then converted to a percentage of the cells with the highest probability value. This is termed the ‘cumulative’ output. Comparative studies using MaxEnt for species distribution modelling that used independent validation performance suggest that it is more accurate than other models (Elith et al. 2006, Hernandez et al. 2006) and does not require or incorporate known absences in the theoretical framework (Phillips et al. 2004). In MaxEnt, it is unnecessary to de.ne the occupancy threshold a priori. In fact, the spatially explicit continuous probability output may be one of the most relevant advantages of MaxEnt for Critical Habitat Identi.cation because it allows for the .ne distinction of habitat suitability in different areas (Kirk 2007). We examined the continuous cumulative output to determine the potential to distinguish a continuum of habitat suitability in different areas.
For intrinsic model evaluation, the area under the receiver-operating characteristic (ROC) curve (AUC) provides a single measure of model performance, independent of any particular choice of threshold (Fielding and Bell 1997). The ROC curve is obtained by plotting sensitivity (fraction of all positive instances that are classi.ed as positive or true positive rate) on the y-axis and 1-speci.city (fraction of all negative instances that are classi.ed as negative) for all possible thresholds. Since MaxEnt does not require or use absence data (negative), the program considers the problem of distinguishing presence from random, rather than presence from absence. Our ROC analysis used all the test localities as instances of presence and a sample of 10,000 random pixels drawn from the background as random instances (Phillips et al. 2006). A random prediction corresponds to an AUC of 0.5, the best discriminating model corresponds to an AUC of 1.0.
We produced environmental niche models for boreal caribou based on three independent environmental datasets to satisfy the objectives outlined above:
1) Potential distribution based on climate averages from 1971 to 2000 and elevation (current
fundamental niche). 2) Potential distribution based on climate averages from 1930 to 1960 and elevation (historic
fundamental niche). 3) Realized distribution based on biotic variables from recent satellite imagery (current
If observed range contractions by caribou resulted, at least in part, from climate change, then we expected that the current fundamental niche should differ most from the historical fundamental niche at more southerly reaches and/or regions where caribou were present historically. Further, if biotic interactions exacerbated by anthropogenic disturbance account for range contractions, then we expected that the current realized niche should be smaller than the current fundamental niche.
All scenarios used the dataset derived from collared animals for training the models. Ten random subsets were run individually with MaxEnt (v3.1) and the cumulative distribution pixel values were averaged over the ten runs to produce a .nal map.
Climate and Topography
The .nal models included the variables listed in Table 2. Mean AUC scores among sub-samples was 0.95, indicating that the model output was signi.cantly better than random. Figures 2 and 3 show the mean cumulative Maxent output for climate variables and topography niche models, from 1930 to 1960 and 1971 to 2000, respectively. Outputs showed that areas of highest probability in both maps correspond to areas where collar data were available to train the model (Figure 1). Similarly, areas with no training presence are not strongly predicted in either time period (e.g., areas in northern Manitoba). Visual inspection along the southern extent of the distribution suggested that the fundamental niche in Ontario and Quebec has not changed signi.cantly over the two time periods. In Alberta, however, the potential distribution may have receded northward. The earlier distribution map suggested presence to the disjointed Little Smokey population, whereas the map from the later time period did not (Figures 2 and 3).
Appendix 6.4 - Table 3. Biotic covariates used in environmental niche models.
Cumulative Annual fPAR
Minimal Annual fPAR
Mean forest patch size
Number of forest patches
Standard deviation of forest patch sizes
Covariates were screened for collinearity and variables included in the model are listed in Table 3. Mean AUC scores were 0.884 among the ten subsets. Figure 4 shows the cumulative MaxEnt output for the biotic models. Higher probabilities were associated with areas with high numbers of satellite collar .xes, but close examination of Alberta and British Columbia herds shows congruency with the ‘Local Population’ polygons in the National Recovery Strategy (Environment Canada 2007; Figure 5 a), where training data were not available. The model predicted a high probability of occurrence, consistent with the extent of occurrence across the range, with the exception of the distribution in northern Saskatchewan, northern NWT, and the northern part of the Quebec (Figures 4,5b).
Biotic ModelAverage Probability
0-1010 -2020 -3030 -4040 -5050 -6060 -7070 -8080 -9090 -100
Kilometers 075 150
Appendix 6.4 - Figure 5a. Biotic model output in western Canada.
Light blue lines are the local population polygons from the respective jurisdictions and the yellow dots indicate collar location used for
Biotic Model Average Probability
0-10 10 -20 20 -30
30 -40 40 -50 50 -60 60 -70 70 -80 80 -90
Kilometers 0 75 150 300
Appendix 6.4 - Figure 5b. Biotic model output in central Canada.Yellow dots indicate collar locations used for training data.
We found support for our .rst hypothesis in part of the country but not everywhere. For example, the fundamental niche, or potential distribution, of woodland caribou may have contracted marginally along its southern frontier in Alberta and Saskatchewan. Thus, some minor range contraction may have occurred in these regions owing to climate change in the past 30 years. In Ontario and Quebec, however, the fundamental niche has remained relatively constant and, based on mid-20th century climate data, does not extend to the southern extent of the Boreal Shield Ecozone, as is suggested by the historical distribution of woodland caribou. Our study design called for training datasets to be compiled using radio-tagged animals owing to the large datasets available and the wide geographic distribution across the extent of occurrence, but these data did not exist for the entire time period. Improved estimates of the historical fundamental niche may come with inclusion of other types of locational data (e.g., not telemetry) consistent with the period, which may include animals outside (south) of the current distribution. It is possible that the more southerly habitats comprised a different biophysical fundamental niche space that is not captured in current distribution of animals.
Our second hypothesis that the realized niche is smaller that the fundamental niche was supported in some parts of the country. In Ontario, for example, Figure 3 shows continuous areas of potential habitat for caribou as far south as the entire north shore of Lake Superior, including Pukaskwa National Park. Figures 4 and 5b revealed that areas of potential continuous occupancy that should otherwise be suitable are restricted to some 200-300 km north of the lake, consistent with the current extent of occurrence. A remnant population of boreal caribou exists in Pukaskwa National Park, likely because conditions suitable for their survival continue to persist along the lakeshore, inside the Park. However, forest management of the landscape between the Park and the current more northerly extent of occurrence has eliminated other suitable habitats (Vors et al. 2007). Our results also suggested that some patches of potential habitat exist in this latter area and that movement of individuals between the present continuous extent and Pukaskwa Park may be possible. Re.nement of these types of models may help to identify potential areas for connectivity and help determine priority areas for potential rehabilitation via landscape management.
Other studies have modelled population extirpations using niche models by combining climate variables and landcover data (Peterson et al. 2006). Climate, vegetation, and elevation datasets are often related (Hutchinson 1998). For example, in Canada, ‘greener’ areas get higher rainfall and also have higher temperatures (Ichii 2002). Elevation also shows a close relationship to temperature but the nature of this relationship is variable in space and time (Ichii 2002). Our analysis demonstrated correlation among many climate parameters used as predictor variables for caribou and the annual fraction of Photosythetically Active Radiation (fPAR) from MODIS, as expected. Inclusion of climate parameters (at 1-km resolution) in the ‘realized’ niche models effectively ‘washed out’ the precision of the predictions. In the climate surfaces, pixel values are interpolated from weather station data, whereas satellite-derived data are collected such that a systematic measurement is taken for each pixel. Based on consistent and recent coverage by remote sensing, 1-km biotic variables should re. ect spatial and temporal variation at a higher resolution than the climate data and satellite based models will be more representative of current distribution boundaries (Parra et al. 2004). Within the range of a species, satellite-based models should have less over-prediction (commission error), or higher speci.city, that is, higher probability of correctly predicting a cell as absent (Peterson et al. 2004, Parra et al. 2004). Further reduction of commission errors in the biotic models may come from exclusion of old locality records that re.ect available habitat at a previous time, but which may have been recently altered. Our training dataset was limited to point locations from the last 20 years to be consistent with timelines used in other areas of the document (Environment Canada 2007), whereas the biotic variables were more recent (last 5 years). Industrial activities that are probably deleterious to caribou populations have increased drastically in some areas over the last 20 years (McKenzie 2006). Restricting location data to be temporally consistent may improve performance of the satellite-based models.
Somewhat unique to our realized niche models was the inclusion of data (disturbance, road density, fragmentation parameters) to account for top-down or predator interactions in limiting species distributions. Hutchinson’s n-dimensional niche concept suggested that a species will occupy areas of the fundamental niche where the species is competitively dominant. However, interspeci.c competition also needs to be considered (Pulliam 2000). Evidence suggests that predation is a major factor in boreal caribou population dynamics and probability of persistence and thus should be considered when modelling caribou habitat occupancy (Sorenson 2008). Many recent satellite-based initiatives and worldwide efforts to maintain access to high quality space-based vegetation data ensure that the economic and timely availability of resource type information for modelling at broad geographic scales is secure (Yang et al. 2006). However deriving accurate and time-speci.c disturbance layers, such as linear feature density or other industrial activities at the scale required, is dif.cult and expensive. Improvement in the derivation and inferential capacity of these data and better relationships de.ning the spatial and temporal scale at which these top down predator interactions occur in caribou populations may improve the occupancy predictions.
A major limitation to any analysis, such as ours, is the geographic bias of locational data available to train the model (Peterson and Cahoon 1999, Johnson and Gillingham 2008, Phillips 2008). Our study design employed many protocols cited to improve model accuracy and reduce bias on model outcomes, including .ltering of GPS collar data (Rettie and McLoughlin, Friar et al. 2004), random subsetting and multiple model runs (Araujo and New 2006), and ecological strati.cation (Reese et al. 2005, Aroujo and Guisan 2006). Nonetheless, despite the large contributions of locational data from across the country, the extent of occurrence as outlined in Environment Canada (2007) is not completely sampled (Figure 1). The location of sampling areas highlights another important bias demonstrated, in theory and practice, to affect the outcome of niche modelling. Most studies have been done on caribou populations at the southern end of the range, while other studies have been conducted on low and/or declining populations (Environment Canada 2007). Niche theory and studies performed using environmental niche models suggest that to improve accuracy of predictions, known sink populations should not be included since this habitat may represent marginal niche space (resources and conditions) for viable populations (Pulliam 2000, Soberon 2007). Sample selection bias due to sampling effort (accessibility) has been shown to dramatically reduce the predictive performance of presence-only models, such as MaxEnt (Phillips 2008). Improved sampling design to represent the entire geographic distribution and attempting to capture the entire niche space of boreal caribou would improve overall model performance and value of the outputs (Jimenez-Valverde and Lobo 2006).
In summary, preliminary results using environmental niche models to study the distribution of boreal caribou at broad scales are important to support Critical Habitat Identi.cation. Species distribution models are increasingly used in conservation planning and management of rare or threatened species to understand the patterns and processes of occurrence on the landscape. The National Recovery Strategy delineates the extent of occurrence of boreal caribou and suggests that some portions of the shaded area (Figure 1) have higher probability of caribou occurrence than others (Environment Canada 2007). The strategy also considers local populations of boreal caribou to be the fundamental units of conservation and management for recovery and action planning. Further re.nement and more rigorous validation of the models presented here would contribute to understanding the areas of occupancy and local population ranges within the larger extent of occurrence. The vital rates required for management and recovery of boreal caribou are dif.cult to obtain because of the large areas that the animals occupy and the low densities at which they exist and because the forested areas that they occupy are dif.cult to survey with traditional aerial techniques (Environment Canada 2007). Spatial predictions from niche-based distribution models may be used to stratify sampling to increase ef.ciency. The new data can then be used to improve the original model and performed repeatedly. Such an adaptive process would re.ne predictions and support management and recovery of local populations at a regional scale. A large range of techniques now exists to predict species distributions, and various studies have demonstrated the predictive capability and accuracy with various types of species and input data availability (e.g. Elith et al. 2006). Presence-only models, such as MaxEnt, may be the most appropriate for rare or threatened species, and caribou in particular, because absences are not likely actual absences but false negatives. Future analyses will focus on model comparisons and reducing data bias to accurately predict boreal caribou occupancy across its extent of occurrence.
Adams L. G. and B.W. Dale. 1998. Reproductive Performance of Female Alaskan Caribou Journal of Wildlife Management 62:1184-1195
Anderson, R. P., M. Gómez-Laverde, and A. T. Peterson. 2002. Geographical distributions of spiny pocket mice in South America: Insights from predictive models. Global Ecology and Biogeography 11:131-141.
Araújo, M. B., and A. Guisan. 2006. Five (or so) challenges for species distribution modelling. Journal of Biogeography 33:1677-1688.
Araújo, M. B., and M. New. 2006. Ensemble forecasting of species distributions. Trends in Ecology & Evolution Volume 22:42-47
Bergerud A.T. and J.P. Elliott. 1989. Wolf predation in a multiple-ungulate system in northern British Columbia. Canadian Journal of Zoology 76:1551–1569
Boyce, M. S., P. R Vernier, S. E. Nielsen, & F. K. A. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157:281-300.
Boyce, M. S. and L. L. McDonald, 1999. Relating populations to habitats using resource selection functions. Trends in Ecology and Evolution 14:268-272.
Brown, G.S. 2005. Habitat selection by woodland caribou in managed boreal forest of northeastern Ontario. Ph. D. thesis. University of Guelph.
Callaghan, C. 2008. Habitat narrative. Boreal Caribou Critical Habitat Science Review. Appendix 4.3.
Case T.J. and M.L. Taper 2000. Interspeci.c Competition, Environmental Gradients, Gene Flow, and the Coevolution of Species’ Borders. American Naturalist 155:583-605
Cederlund G.N., H.K.G. Sand, A. Pehrson. 1991. Body Mass Dynamics of Moose Calves in Relation to Winter Severity. Journal of Wildlife Management 55:675-681
Chen, G.J. and A.T. Peterson. 2000. A New Technique For Predicting Distribution of Terrestrial Vertebrates Using Inferential Modeling. Zoological Research 21:231-237
Coops, N.C.,M.A. Wulder, D.C. Duro, T. Han and S. Berry. 2008. The development of a Canadian dynamic habitat index using multi-temporal satellite estimates of canopy light absorbance. Ecological Indicators 8:754-766
Elith, J., C.H. Graham, R.P. Anderson, M. Dudík, S. Ferrier, A.Guisan, R.J. Hijmans, F. Huettmann, J.R.Leathwick, A. Lehmann, J.Li, L.G. Lohmann, B.A. Loiselle, G. Manion,
C. Moritz, M. Nakamura, Y. Nakazawa, J.M. Overton, A.T. Peterson, S.J. Phillips, K. Richardson, R. Scachetti-Pereira, R.E. Schapire, J. Soberón, S. Williams, M.S. Wisz, &
N.E. Zimmermann. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129–151.
Environment Canada. 2007. Recovery strategy for the woodland caribou (Rangifer tarandus caribou), boreal population, in Canada. Draft, June 2007. Species at Risk Act Recovery Strategy Series. Ottawa: Environment Canada. v + 48 pp. plus appendices.
Fancy, S.C., and K.R. Whitten. 1991. Selection of calving sites by Porcupine Herd caribou. Canadian Journal of Zoology 69:1736-1743
Fielding, A. H. & J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation 24:38-49.
Finstad G.L., M. Berger, K. Lielland, A.K. Prichard. 2000. Climatic in.uence on forage quality, growth and reproduction of reindeer on the Seward Peninsula I: climate and forage quality. Rangifer Special Issue 12:144
Friar, J.L., S.E. Neilson, E.H. Merrill, S.R. Lele, M.S. Boyce, R.H.M. Munro, G.B. Stenhouse
H.L. Beyer. 2004. Removing GPS collar bias in habitat selection studies. Journal of Applied Ecology 41:201-212.
Fuller, T.K. and L.B. Keith. 1981. Woodland Caribou Population Dynamics in Northeastern Alberta. Journal of Wildlife Management. 45: 197-213.
GFWC. 2007. Recent Anthropogenic Changes within the Boreal Forests of Ontario and Their Potential Impacts on Woodland Caribou. http://www.globalforestwatch.ca
Godown, M. E., and A. T. Peterson. 2000. Preliminary distributional analysis of U.S. endangered bird species. Biodiversity and Conservation 9:1313-1322.
Guisan, A., O. Broennimann, R. Engler, N.G. Yoccoz, M. Vust, N.E. Zimmermann, and A.Lehmann. (2006) Using niche-based models to improve the sampling of rare species. Conservation Biology 20:501–511.
Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147-186.
Guisan, A. and W.Thuiller, 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993–1009.
Heikkinen R.K., M. Luoto, R. Virkkala, R.G. Pearson and J.H. Korber 2007. Biotic Interactions Improve Prediction of Boreal Bird Distributions at Macro-Scales. Global Ecology and Biogeography 16:754-763
Hernández, P.A., C.H. Graham, L.L. Master, D.L. Albert, 2006. The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29:773–785.
Hijmans R.J. & C.H. Graham. 2006. The Ability of Climate Envelope Models to Predict the Effect of Climate Change on Species Distributions. Global Change Biology 12: 2272-2281.
Hirzel, A. H., J. Hausser, D. Chessel, and N. Perrin. 2002. Ecological-niche factor analysis: How to compute habitat-suitability maps without absence data? Ecology 83:2027-2036.
Huete, A., K. Didan, T. Miura, E. P. Rodriguez, X. Gao, & L. G. Ferreira. 2002. Overview of the Radiometric and Biophysical Performance of the Modis Vegetation Indices. Remote Sensing of Environment 83:195-213.
Hutchinson, G. E. 1957. Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology 22:415-427.
Hutchinson M.F. 1995. Interpolation of mean rainfall using thin-plate smoothing splines. International Journal of Geographic Information Systems. 9:385-403.
Hutchinson, M.F 1998. Hutchinson, Interpolation of rainfall data with thin plate smoothing splines. II. Analysis of topographic dependence. Journal of Geographic Information Decision Analysis 2:168–185.
Ichii, K., A. Kawabata, Y. Yamaguchi. 2002. Global correlation analysis for NDVI and climatic variables and NDVI trends: 1982-1990 International Journal of Remote Sensing, 23:3873 – 3878.
James, A. R. C. & A. K. Stuart-Smith. 2000. Distribution of Caribou and Wolves in Relation to Linear Corridors. Journal of Wildlife Management 64:154-159.
Jimenez-Valverde A. and J. M. Lobo. 2006. The ghost of unbalanced species distribution data in geographical model predictions. Diversity and Distributions. 12:521-524.
Johnson D. H. 1980. The Comparison of Usage and Availability Measurements for Evaluating Resource Preference. Ecology 61:65-71.
Johnson C.J., D.R. Seip, M.S. Boyce. 2004. A quantitative approach to conservation planning: using resource selection functions to map the distribution of mountain caribou at multiple spatial scales Journal of Applied Ecology 41:238–251.
Johnson, C.J., and M.P. Gillingham. 2008. Sensitivity of species-distribution models to error, bias, and model design: An application to resource selection functions for woodland caribou. Ecological Modelling 213:143-155.
Kirk, D. 2007. Comparing empirical approaches to modelling species’ distributions and occurrence – relevance to critical habitat identi.cation. Unpublished Report to Parks Canada Agency.
Laliberte, A.S. and W.J. Ripple. 2004. Range contractions of North American carnivores and ungulates. BioScience 54:123-138.
MacArthur, R. H. 1967. The theory of the niche. Population biology and evolution, ed R. C. Lewontonin. pp. 159-176. Syracuse University Press, Syracuse, NY, USA.
McLoughlin, P.D., E. Dzus, B. Wynes, and S. Boutin. 2003. Declines in populations of woodland caribou. Journal of Wildlife Management. 67:755-761.
McLoughlin, P.D., J.S. Dunford, and S. Boutin. 2005. Relating predation mortality to broad-scale habitat selection. Journal of Animal Ecology 74:701-707.
McKenney D.W., J. H. Pedlar, P. Papadopola, M. F. Hutchinson 2006. The development of 1901–2000 historical monthly climate models for Canada and the United States. Agricultural and Forest Meteorology 138: 69-81.
McKenzie, H. 2006. Linear features impact predator-prey encounters: analysis with .rst passage time. University of Alberta.
Parra, J.L., C.C. Graham, J.F. Freile. 2004. Evaluating alternative data sets for ecological niche models of birds in the Andes. Ecography. 27:350-360.
Pearson R.G. & T.P. Dawson 2003. Predicting the impacts of climate change on the distribution of species: are bioclimatic envelope models useful? Global Ecology and Biogeography 12:361-371
Pearson R.G., T.P. Dawson and C Liu. 2004. Modelling species distributions in Britain: a hierarchical integration of climate and land-cover data. Ecography 27: 285-298.
Peterson, A.T., 2001. Predicting species’ geographic distributions based on ecological niche modeling. Condor 103:599–605.
Peterson A.T., M.A. Ortega-Huerta, J. Bartley, V. Sanchez-Cordero, J. Soberon, R.H. Buddemeier and D.R.B. Stockwell. 2002. Future projections for Mexican faunas under global climate change scenarios. Nature 416:626-629
Peterson, A.T. 2003. Projected climate change effects on Rocky Mountain and Great Plains birds: generalities of biodiversity consequences. Global Change Biology 9:647–655.
Peterson, A. T. and C. R. Robins. 2003. Using ecological niche modeling to predict Barred Owl invasions with implications for Spotted Owl conservation. Conservation Biology 17:1161–
Peterson, A.T., Cohoon, K.C.1999. Sensitivity of distributional prediction algorithms to geographic data completeness. Ecological Modelling 117:159–164.
Peterson A.T., E. Martinez-Meyer, C. Gonzalez-Salazar & P.W. Hall. 2004. Modeled Climate Change Effects on Distributions of Canadian Butter.y Species. Canadian Journal of Zoology 82:851-858
Peterson, A.T., V. Sánchez-Corderob, E. Martínez-Meyerb, A. G. Navarro-Sigüenzac, 2006. Tracking population extirpations via melding ecological niche modeling with land-cover information. Ecological Modelling 195: 229-236
Pettorelli, N., J. O.Vik, , A. Mysterud, , J.-M. Gaillard, C. J. Tucker, & N.-C. Stenseth, 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology and Evolution 20:503-510.
Phillips, S.J., M. Dudík, R.E. Schapire. 2004. A maximum entropy approach to species distribution modeling. Proceedings of the twenty-.rst international conference on Machine learning ACM International Conference Proceeding Series Vol. 69 p. 83.
Phillips, S. J., R. P. Anderson, and R. E. Schapire, 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231–259.
Phillips S.J. and M. Dudík. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31: 161-175
Pulliam, H. R. 2000. On the relationship between niche and distribution. Ecology Letters
Raxworthy C.J., E. Martinez-Meyer, N. Horning, R.A. Nussbaum, G.E. Schneider M.A. Ortega-Huerta, A. T. Peterson. 2003. Predicting distributions of known and unknown reptile species in Madagascar. Nature. 426:837-41
Reese, G.C., K.R. Wilson, J.A.Hoeting, C.H. Flather. 2005. Factors affecting species distribution predictions: A simulation modeling experiment. Ecological Applications 15:554
Rettie W.J. and P.D. McLoughlin. 1999. Overcoming radiotelemetry bias in habitat selection Studies Canadian Journal of Zoology. 77:1175–1184.
Seip, D. R. 1992. Factors limiting Woodland Caribou populations and their interrelationships with wolves and moose in southeastern British Columbia. Canadian Journal of Zoology 20:1494-1503.
Soberón, J., and A. T. Peterson. 2005. Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodiversity Informatics 2:1-10.
Soberon J. 2007. Grinnellian and Eltonian niches and geographic distributions of species. Ecology Letters 10:1115–1123
Sorensen T, McLoughlin PD, Hervieux D, Dzus E, Nolan J, Wynes B, Boutin S. 2008. Determining Sustainable Levels of Cumulative Effects for Boreal Caribou. Journal of Wildlife Management 72:900–905
141 APPENDIX 6.4
Stockwell, D.R.B. and A.T. Peterson, 2002. Controlling bias in biodiversity data. In: Scott, J.M., Heglund, P.J., Morrison, M.L. (Eds.), Predicting Species Occurrences: Issues of Scale and Accuracy. Island Press, Washington, DC, pp. 537–546.
Tarnocai, C., I.M. Kettles and B. Lacelle. 2005. Peatlands of Canada. Agriculture and Agri-Food Canada, Research Branch, Ottawa, scale 1:6 500 000.
Thomas, D. C., E. J. Edmonds, and W. K. Brown. 1996. The diet of Woodland Caribou populations in west-central Alberta. Rangifer Special Issue 9:337-342.
Thomas C.D., A. Cameron, R.E. Green, M. Bakkenes, L.J. Beaumont, Y.C. Collingham,
B.F.N. Erasmus, M.F. De Siqueira, A. Grainger, L. Hannah, L. Hughes, B.Huntley, A.S.Van Jaarsveld, G.F.Midgley, L.Miles, M.A.Ortega-Huerta, A.T.Peterson, S.L. Phillips and S.E.Williams. 2004. Extinction Risk From Climate Change. Nature 427:145-148
Thuiller, W., D.M. Richardson, P. Pysek, G.F. Midgley, G.O. Hughes, and M. Rouget, 2005. Niche-based modelling as a tool for predicting the risk of alien plant invasions at a global scale. Global Change Biology 11:2234–2250.
Tian Y, U. Zhang, Y. Knyazikhin, R.B. Myneni, J.M. Glassy, G. Dedieu, S.W. Running. 2000. Prototyping of MODIS LAI and FPAR algorithm with LASUR and LANDSAT data. IEEE Transactions on Geoscience and Remote Sensing. 38:2387-2401
Vors, L.S., J.A. Schaefer, B.A. Pond, A.R. Rodgers, B.A. Patterson. 2007. Woodland caribou extirpation and anthropogenic landscape change in Ontario. Journal of Wildlife Management. 71:1249-1256
R.B. Weladji, D.R. Klein, Ø. Holand, A. Mysterud 2002. Comparative response of Rangifer tarandus and other northern ungulates to climatic variability. Rangifer 22:33-50
White, G. C. and Garrott, R. A. 1990. Analysis of wildlife radio-tracking data. Academic Press, New York.
Wilson K.A. M.I. Westphal, H.P. Possingham, J. Elith. 2005. Sensitivity of Conservation Planning to Different Approaches to using Predicted Species Distribution Data. Biological Conservation 22:99-112.
Wittmer, H.U., B.N. McLellan, R. Serrouya, C.D. Apps. 2007. Changes in landscape composition in.uence the decline of a threatened woodland caribou population. Journal of Animal Ecology 76:568–579.
Wulder, M.A., J.C., White, T. Han, J.A. Cardille, T., Holland, N.C. Coops, and D. Grills. 2008. Landcover mapping of Canada’s forests: II. Forest fragmentation. Submitted to Canadian Journal of Remote Sensing.
Yang, W., B. Tan, D. Huang, M. Rautiainen, N. V. Shabanov, Y. Wang, J. L. Privette, K.F. Huemmrich, R. Fensholt, I. Sandholt, M. Weiss, D. E. Ahl, S. T. Gower, R. R. Nemani, Y., Knyazikhin, R. B. Myneni, 2006. MODIS Leaf Area Index Products: From Validation to Algorithm Improvement. IEEE Transactions on Geoscience and Remote Sensing 44:18851898
Zhao, M. S., F. A. Heinsch, R. R. Nemani, S. W. Running, 2005. Improvements of the MODIS Terrestrial Gross and Net Primary Production Global Data Set. Remote Sensing of Environment 95:164-176.
6.5 A National Meta-analysis of Boreal Caribou Demography and Range Disturbance
A key step in the critical habitat identi.cation process is determining attributes of a caribou range that support or compromise population persistence (e.g. the ability of the range to support a self-sustaining population). This meta-analysis compiled demographic data from boreal caribou populations across Canada to evaluate the hypothesized relationship between caribou population parameters and levels of anthropogenic and/or natural (. re) disturbance on caribou ranges. Results from this work provide quantitative guidelines for one of the three assessment criteria (range disturbance) used in the evaluation of local populations for critical habitat identi.cation.
Woodland caribou (Rangifer tarandus caribou) are designated a species-at-risk nationally, and in most provinces and territories within their range, due to broad-scale range recession and population declines, in large part associated with human settlement and disturbance (Bergerud 1974, Mallory and Hillis 1998, Schaefer 2003, Vors et al. 2007). This species is closely associated with late-successional coniferous forests and peatlands (Rettie and Messier 2000). These forests are a source of lichens, which comprise the bulk of woodland caribou diet – particularly in winter – but lichen availability is generally not considered a limiting factor (Schaefer and Pruitt 1991, Joly et al. 2003, Courtois et al. 2007). More importantly, these forests provide refugia from predators and other ungulates (Bergerud and Elliott 1986). Many woodland caribou populations are in decline, and the proximate cause is thought to be increased predation. Logging and other disturbances that increase the amount of early seral-stage forest promote higher densities of prey species such as moose (Alces alces) and white-tailed deer (Odocoileus virginianus), which support higher predator densities, especially wolves (Canis lupus) (Bergerud and Elliott 1986, Seip 1992, Stuart-Smith et al. 1997, Racey and Armstrong 2000, Wittmer et al. 2005, 2007). In addition, linear disturbances (e.g. roads, seismic lines) that accompany industrial development in the boreal forest facilitate greater predator mobility and hunting ef.ciency (James and Stuart-Smith 2000, Dyer et al. 2001, McLoughlin et al. 2003, James et al. 2004). Boreal caribou, an ecotype of woodland caribou, are declining throughout much of their North American range (McLoughlin et al. 2003). Given the increasing levels of industrial development in previously pristine areas, preventing or mitigating further population declines is increasingly the focus of management efforts.
In this study, a simple question is posed: is there a clear relationship between caribou demography and anthropogenic and/or natural (.re) disturbance levels on caribou ranges across the distribution of boreal caribou in Canada? We expected that adult survival, calf recruitment and overall population growth would be negatively related to changes that create favorable habitat for moose and deer, in keeping with the logic that increased primary prey increases predator density which contributes to caribou population decline. Caribou avoidance of industrial development (Bergerud 1974, Mallory and Hillis 1998, Dyer et al. 2001, Schaefer 2003) and recent burns (Schaefer and Pruitt 1991, Joly et al. 2003, Dunford et al. 2006) have been well documented; thus it is reasonable to postulate that these factors would negatively affect range condition with respect to the ability of an area to support a self-sustaining local population. Wittmer et al. (2007) found that the variation in adult female survival among 10 woodland caribou populations of the arboreal lichen-feeding ecotype was best explained by range condition. Further, in a review of 85 studies that examined impacts of human activity on caribou, Vistnes and Nellemann (2008) concluded that choice of spatial scale for examining impacts strongly in.uenced conclusions, recommending that accurate assessment required regional-scale studies. Finally, in a recent analysis of 6 boreal caribou populations in Alberta, Sorensen et al. (2008) demonstrated a negative relationship between range condition and population growth rate. Their 2 variable model, which included level of anthropogenic disturbance (%IND) and wild.re (%FIRE), explained 96% of the variation in caribou population growth rates. Hence, our selection of caribou range as the appropriate unit of analysis is justi.ed.
The Sorensen et al. (2008) regression model represents a signi.cant advance in our understanding of the effects of disturbance on caribou demography at the level of population ranges. However, the study was based on a small sample size and a limited range of values for anthropogenic disturbance (e.g. the minimum level of anthropogenic disturbance was 31.6%). As a result, while the data were suf.cient to demonstrate signi.cance in terms of a relationship between the dependent and independent variables, the model has limited scope for prediction beyond the geographic area and parameter space included in that study, and should be used cautiously within that region when predicting minimum levels at which negative effects on caribou population growth might occur. The objective here was to extend the Sorensen et al. (2008) analysis to populations of boreal caribou across Canada, in order to test whether the relationship documented was robust across a broader spectrum of range conditions, and guide evaluation of the ability of ranges to support self-sustaining populations. Original work on this study was initiated in 2006, as part of an independent effort to address this question. Augmentation and re.nement of this effort was undertaken in conjunction with the Environment Canada scienti.c review of critical habitat for boreal caribou.
Data collection – caribou
Researchers and management agencies were approached to supply demographic information on woodland caribou populations that had been studied for a minimum of two years (the smallest interval included in Sorensen et al. 2008), and for which adult female survival (as determined by radio-telemetry monitoring) and/or calf recruitment (late winter calf/cow population surveys) had been measured. The intent was to assemble data that exhibited a broad range of variation with respect to geography and degree of anthropogenic change to population ranges. A tabular data survey with instructions was circulated to potential contributors. Information on 25 boreal populations from 7 provinces and 1 territory was acquired (Figure 1). There was considerable variability in the intensity and duration of sampling, and availability of ancillary information.
Estimates of population condition
Of the 25 populations included in this study, data for assessing female survival and therefore estimating population growth rates were available for 15 (Table 1). Some populations had only a small number of female caribou collared and concomitant high variability in estimated survival. Therefore, to maximize the number of populations available for analysis, estimates of recruitment rates, which were available for all populations, were used as a surrogate of ‘population condition’. Bergerud and Elliot (1986, 1998) demonstrated that recruitment was directly related to population rate-of-growth in caribou, as well as in other ungulates. Furthermore, recruitment may be a better short-term indicator of population condition in rapidly changing landscapes than either female survival or population growth rate, given that calves are more susceptible to predation than adults, and high adult survival could initially mask the negative effects of landscape change.
To test the relationship between recruitment and population growth, and the appropriateness of using recruitment as the response variable to range condition in the regression analysis, data from the subset of populations for which recruitment and survival were available were used to estimate population rate of change (.) following Hatter and Bergerud (1991); see also McLoughlin et al. (2003) and Sorensen et al. (2008). However, because averages and not annual data were provided for some local populations, an arithmetic, rather than geometric mean (McLoughlin et al. 2003. Sorensen et al. 2008) was used to estimate average values for each parameter over the years of study included for each population (Table 1). Data for some populations were sub-sampled to be temporally consistent with available data on landscape change; in particular, to avoid inclusion of demographic data that potentially preceded the change. Also, some populations with long-term data exhibited trends suggesting that an average over the entire sampling interval was not representative of the current population condition. Where available, up to 4 years of most recent data, spanning a maximum sampling interval of 5 years, and with greatest temporal correspondence to the landscape change data, were used to estimate demographic parameters for analysis (Table 1). The 6 Alberta populations included in Sorensen et al. (2008) were also included in this study; however, the sampling intervals differed (1993-2001 vs. 2002-2006). Thus, it was possible to also evaluate the relationship between recruitment and population growth for a second subset of temporally non-overlapping data, based on Sorensen et al. (2008).
Delineation of population ranges
Range boundaries were provided by contributors for study populations, obtained from provincial or territorial sources for jurisdictionally-recognized population ranges, or generated from 100% minimum convex polygons (MCPs) of telemetry data provided by contributors. Delineation method is indicated in Table 1 and illustrated in Figure 1. Where a study population
corresponded closely to a jurisdictionally recognized range (e.g. =90% correspondence), the data were considered representative of the range, and the jurisdictional boundary was used for population delineation and characterization of range condition.
Characterization of range condition and model speci.cation
Following Sorensen et al. (2008), the relationship between recruitment and range condition was evaluated by comparing three candidate models. Model 1 considered the percent of the range area burned within the past 50 years of the most recent recruitment data for each population. Fire data from the Canadian Large Fire Database, augmented by additional coverage for the Northwest Territories, that contained wild.res >200 ha (NRCan 2008, GNWT 2008) were used. Model 2 considered the percent of the range area affected by anthropogenic disturbance, based on GIS layers obtained from Global Forest Watch Canada (GFWC). GFWC have compiled the only available, nationally-consistent coverage of anthropogenic disturbance across forested regions of Canada. All visible linear and polygonal anthropogenic disturbances were digitized from Landsat images from the period 1985–2003, and combined with additional coverage of roads, reservoirs and mines from databases spanning the period 2002-2006. Linear disturbances included roads, railroads, seismic lines, pipelines, and utility corridors; polygonal features included recently anthropogenically-converted areas such as settlements, populated industrial areas, croplands (both new and abandoned), reservoirs, cutblocks, and mining activity. All features in the database were buffered by 500 m to create a “zone of in.uence”, and merged to create a non-overlapping coverage of all anthropogenic disturbances. Detailed methodology is available from Lee et al. (2006). Sorensen et al. (2008) used a 250-m buffer when quantifying human disturbance. However, we did not have access to the raw data used in the GFWC analysis, so could not select an alternate or varying buffer width. Nevertheless, in a review of reindeer and caribou response to human activity from regional-scale landscape studies, Vistnes and Nellemann (2008) report reduced use by caribou of areas within 5 km of infrastructure and human activity, thus the 500-m buffer is not unreasonable. Lastly, Model 3 considers the combined effect of .re and anthropogenic disturbance, herein termed total disturbance.
Characterization of total disturbance and modeling procedure
Sorensen et al. (2008) used a 2 variable model to characterize total disturbance (%FIRE and %IND); however, they found a relatively high correlation between these 2 variables (Pearson correlation of 0.69) which tends to produce least-squares estimates that are exaggerated in absolute value (Montgomery et al. 2001). Multi-colinearity between these 2 variables could also in.uence parameterization because of the likely non-linear relationship between the proportion of area disturbed and the level of spatial overlap. Speci.cally, at low levels of disturbance the spatial overlap is likely to be low whereas the likelihood of overlap should increase at higher levels of disturbance. Visual inspection of the data revealed such a pattern. Therefore, to describe total disturbance when testing the hypothesis of primary interest (e.g. the combined effects of .re and anthropogenic disturbance), the merged mapped of non-overlapping disturbances was used to derive a single measure of total disturbance. This
Local Population Prov/Terr Years Available Sample # Years Years Used Range R S .
Red Wine NL 1981-1988, 1993-1997, 2001-2003 Y 3 2001-2003 J 45.4 n/a n/a
Mealy Mountain NL 1971, 1974-1975, 1977, 1981, 1985,1987, 1994, 2002, 2005 Y 2 2002, 2005 J 50.3 89.0 1.19
Lac Joseph NL 2000-2002, 2005, 2007, 2008 Y 4 2000-2002, 2005 J 34.3 n/a n/a
Val-d'Or QC 1987-1988, 1990-1991, 1995-2002,2004-2005 Y 4 2001-2002, 2004-2005 J 15.3 87.0 0.94
Manicouagan QC 1999-2001 N 3 1999-2001 J 50.5 75.0 1.00
Manouane QC 1999-2001 N 3 1999-2001 J 28.1 86.0 1.00
Pipmuacan QC 1999-2001 N 3 1999-2001 J 40.6 82.0 1.03
Charlevoix QC 2000-2001,2004-2006 Y 4 2001,2004-2006 J 35.0 n/a n/a
Jamesie QC 2002-2003 N 2 2002-2003 SA 27.4 n/a n/a
James Bay ON 1998-2000 N 3 1998-2000 SA 21.3 79.0 0.88
Pukaskwa ON 1973-1991, 1997, 1999, 2001 Y 3 1997, 1999, 2001 J 40.3 n/a n/a
Smoothstone-Wapawekka SK 1993-1995 N 3 1993-1995 SA 28.0 84.0 0.98
Caribou Mountain AB 1995-2007 Y 4 2003-2006 J 17.4 75.0 0.82
ESAR AB 1994-1997, 1999-2007 Y 4 2003-2006 J 13.4 86.6 0.93
Red Earth AB 1995-1997, 1999-2007 Y 4 2003-2006 J 13.6 81.9 0.88
WSAR AB 1994-2007 Y 4 2003-2006 J 20.9 84.2 0.94
Little Smoky AB 2000-2007 Y 4 2003-2006 J 12.3 82.2 0.88
Cold Lake AB 1999-2002, 2004-2007 Y 4 2002, 2004-2006 J 12.6 83.8 0.89
Chinchaga AB 2002-2007 Y 4 2003-2006 J 13.9 87.0 0.93
Snake-Sahtaneh BC 2004-2005 N 2 2004-2005 J 7.2 94.0 0.97
Cameron Hills NWT 2006-2008 N 3 2006-2008 SA 16.4 n/a n/a
Dehcho North NWT 2006-2008 N 3 2006-2008 SA 20.7 n/a n/a
Dehcho South NWT 2006-2008 N 3 2006-2008 SA 32.3 n/a n/a
GSA South NWT 2004-200
- Date modified: