SEARCH

SEARCH BY CITATION

Keywords:

  • critical habitat;
  • habitat suitability index models;
  • local ecological knowledge;
  • Rangifer tarandus;
  • recovery planning;
  • resource selection functions;
  • Species at Risk Act;
  • traditional ecological knowledge

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

Negotiating the complexities of wildlife management increasingly requires new approaches, especially where data may be limited. A robust combination of traditional ecological knowledge (TEK) and western science has the potential to improve management decisions and enhance the validity of ecological inferences. We examined the strengths and weaknesses of predicting woodland caribou (Rangifer tarandus caribou) habitat selection with resource selection functions (RSF) based on western science and TEK-based models within the territory of the Taku River Tlingit First Nation of northern British Columbia. We developed seasonal RSF models with data from 10 global positioning system collared caribou. We generated TEK-based habitat suitability index models from interviews with Taku River Tlingit members. We tested the ability of both habitat models to spatially predict the occurrence of collared caribou locations. To portray differences between the models, we statistically and visually compared the spatial predictions of TEK and RSF modeling approaches using Kappa statistics and k-fold cross validation. Kappa statistics of habitat ranks from the models showed substantial agreement during summer (K = 0.649) and fair agreement during winter (K = 0.337). We found that both TEK and RSF models predicted independent caribou locations (Spearman's rank correlations from k-fold cross-validation ranged from 0.612 to 0.997). Differences in model performance were a result of RSF models predicting more relatively high quality habitat than TEK models. Given the widespread declines of woodland caribou across the boreal forest of Canada, and the requirement of the Canadian Species at Risk Act to incorporate both traditional and western science approaches into recovery planning, our results demonstrate that TEK-based habitat models can effectively inform recovery planning for this imperiled species. © 2013 The Wildlife Society.

The value of using traditional ecological knowledge (TEK) in contemporary wildlife management, policy, and conservation is growing (Schmidt and Stricker 2010, Huntington 2011). People who spend significant time on the land have intimate knowledge of wildlife distributions, behavior, movements, body conditions, and habitat relationships (Lyver and Łutsël K'é Dene First Nation 2005, Kendrick and Manseau 2008). TEK often emerges over long temporal scales and includes population-level inferences about wildlife that can offer complementary insights to shorter-term and smaller-scale scientific studies (Moller et al. 2004, Rist et al. 2010). For example, TEK has been successful in providing novel insights into species ecology (Ramstad et al. 2007), identifying baseline conditions (Nichols et al. 2004), describing historical species distributions (Ferguson and Messier 2000, Santomauro et al. 2012), and resolving land management issues (Sandström et al. 2003). Recent studies demonstrate that when TEK is brought into play early in a wildlife management process, the combination with scientific data can lead to more efficient and effective wildlife management decisions (Kendrick and Manseau 2008, O'Flaherty et al. 2008). Finding collaborative opportunities to include TEK in modern wildlife management is a priority for management agencies, conservation organizations, and indigenous people. This is especially true in northern North America where management is often explicitly required to include TEK.

TEK is often defined as an understanding of the environment that comes from a historical continuity in resource use in a particular place (Berkes 1999). In this context, the term traditional implies knowledge handed down through generations in the form of oral history, in contrast to local ecological knowledge (LEK) that is based on direct experiences (Anadón et al. 2009). However, because LEK is regularly embedded within cultural practices, distinguishing LEK from TEK can be difficult, especially when referring to knowledge of indigenous peoples. Thus, TEK and LEK can both represent a significant source of high quality information about ecological relationships. Though fundamental differences between TEK, LEK, and western science exist, all are empirical knowledge systems established through observation and experience (Berkes et al. 2000). All have the ability to test predictions, interpret results within a cultural framework, and adjust expectations when presented with new data (Davis and Ruddle 2010). As a result, TEK and western science often have related questions, predictions, and goals in relation to the management of natural resources.

The conservation and recovery of woodland caribou (Rangifer tarandus caribou) is an endangered species priority that directly intersects with the needs, interests, and knowledge of indigenous people throughout northern North America (Vors and Boyce 2009, Festa-Bianchet et al. 2011). Across Canada, boreal and southern mountain woodland caribou are listed as threatened under the Canadian Species at Risk Act (SARA) because of habitat loss associated with oil, gas, mining, and forestry extraction (Wittmer et al. 2005a, Apps and McLellan 2006, Schaefer and Mahoney 2007). Human development has also altered predator-prey relationships causing widespread population declines (Vors and Boyce 2009) and recently, extirpation of some herds (Wittmer et al. 2005b, Hebblewhite et al. 2010). Recovery planning for woodland caribou is challenged by the spatial scales associated with identifying critical habitat at a national level, factious politics of endangered species regulation, and the legal requirement under SARA to explicitly include TEK in the recovery process (Government of Canada 2002). Prominence of TEK in woodland caribou recovery planning is perhaps unprecedented among world endangered species legislation, reflecting the importance of this species to indigenous people and the legal treaty rights to the lands in which caribou exist. Although potentially beneficial as a conservation strategy, few examples demonstrate how to use TEK to guide endangered species recovery.

Endangered species legislation prioritizes the conservation of critical habitat, a legal concept that can be challenging to define (Mooers et al. 2010). The first step is to understand habitat selection by relating the species selection of resource units to environmental variables (Boyce et al. 2002). Resource selection functions (RSF) use a statistical approach to measure habitat selection by examining use or avoidance of a resource relative to its availability (Manly et al. 2002). However, RSFs are limited by the spatio-temporal scales of animal location data, which are often collected for only a few animals (relative to the entire population) and over a short-time frame. Habitat suitability index (HSI) models, alternatively, predict habitat quality based on expert opinion (U.S. Fish and Wildlife Service 1981). Thus, the basis of HSI modeling is consistent with TEK and LEK. Using TEK and LEK to develop HSI models is particularly useful when statistical limitations result from nonexistent, incomplete, or biased empirical data (Johnson and Gillingham 2004, Doswald et al. 2007, Perera et al. 2012). In an era of inadequate funding and time for animal-based field studies, the value of TEK should be recognized, particularly in areas where intensive scientific studies are unrealistic. A challenge to the effective use of TEK has been how to explicitly use TEK and ensure that it meets rigorous quality standards. For example, managers must examine how well models that are informed by TEK (e.g., HSI models) predict animal locations, especially for endangered or threatened species (Nielsen et al. 2003, Doswald et al. 2007).

The goal of our study was to examine how alternative modeling approaches and information sources can be used in conjunction to inform endangered species recovery. Our first objective was to develop and test independent TEK and western science woodland caribou habitat models. Second, we statistically and visually compared the spatial predictions of TEK and western science models using k-fold cross-validation and spatial Kappa statistics. We conclude by making recommendations for the use of TEK in wildlife management and endangered species recovery planning.

STUDY AREA

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

We focused on an 11,594-km2 home range area of the Atlin northern mountain woodland caribou herd that intersects the 48,000-km2 traditional territory of the Taku River Tlingit First Nation (TRTFN) in northwest British Columbia (Fig. 1). Mountain ranges with high peaks, broad plateaus, and wide valleys characterize the study area. Elevations range from 660 m to 2,000 m. Lower elevation boreal forests included open lodgepole pine (Pinus contorta latifolia), subalpine fir (Abies lasiocarpa), and white spruce (Picea glauca). Mid-elevations transition into krummholz where willow (Salix spp.) and dwarf birch (Betula glandulosa) dominate. Alpine habitats consist of herbs, bryophytes, sedges (Carex spp.), and altai fescue (Festuca altaica) tundra. Trembling aspen (Populus tremuloides), black cottonwood (Populus balsamifera trichocarpa), and willow dominate valley bottoms. More details of the study area can be found in Polfus et al. (2011).

image

Figure 1. The home range area of the Atlin northern mountain woodland caribou herd that intersects the traditional territory of the Taku River Tlingit First Nation in North America on the border of the Yukon Territory and British Columbia, Canada.

Download figure to PowerPoint

The northern mountain woodland caribou ecotype occurs throughout the Yukon Territory, Northwest Territories, and northwestern British Columbia and includes the traditional territory boundaries of 33 First Nations (Environment Canada 2012a). Northern mountain woodland caribou were listed under SARA as a species of special concern in 2004 in response to population declines resulting from habitat loss and fragmentation, human-induced changes to predator–prey dynamics, overharvest, and proliferation of human development (Environment Canada 2012a). Historically, tens of thousands of Tlingit maintained camps and villages from  Tlèn (Atlin Lake) to T'àkhú (the Taku River) near Juneau, Alaska (McClellan 1981). Woodland caribou have always been a culturally important source of meat and other animal products for the TRTFN. In the early 1990s, concerns for population declines of the Atlin, Carcross-Squanga, and Ibex caribou herds caused many First Nation hunters to reduce or eliminate their harvest of caribou (Farnell 2009). Recent aerial surveys indicated that the Carcross-Squanga and Ibex herds have increasing population sizes with sustainable recruitment, whereas the Atlin herd consistently exhibits stable or decreasing recruitment indices indicative of population declines (Taku River Tlingit First Nation and British Columbia 2010).

METHODS

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

The Ministry of Water, Land, and Air Protection of British Columbia monitored caribou in the Atlin herd with global positioning system (GPS) collars (GPS 2000, LOTEK, Aurora, Ontario, Canada) between January 2000 and January 2002 and very high frequency (VHF) collars from December 1999 to March 2003 as part of an ungulate monitoring program (Polfus 2010). Caribou were captured by helicopter net-gun and physically restrained during collaring (Resource Inventory Committee 1998; procedures were part of a study plan approved by the Ministry of Water, Land, and Air Protection). Locations were recorded by GPS collars every 4 hr. Fix-rates were >90%; therefore, we assumed no significant habitat-induced fix-rate bias (Frair et al. 2010). Seasonal VHF locations (which were used only for model validation) were collected from fixed-wing aircraft 1–2 times/month. The GPS locations used for RSF modeling had locational accuracy of approximately 31 m (e.g., Hebblewhite 2006). We did not explicitly account for telemetry error because a primary focus of the study was on evaluating caribou response to human activity, and continuous covariates such as distance to human activity are known to be less affected by positional error than categorical covariates (Montgomery et al. 2011).

Model Variables

We included resource variables previously identified as influential predictors of northern woodland caribou habitat selection. Northern woodland caribou forage on terrestrial lichen in winter in forested and alpine windswept areas (Johnson et al. 2000, Gustine and Parker 2008, Farnell 2009). We used a 13-class land cover classification that was derived from Landsat TM satellite imagery (30-m resolution, minimum map unit was 900 m2; Polfus 2010). We extracted elevation (m), slope (degrees), aspect (degrees), and hillshade from the Terrain Resource Information Management digital elevation model (30-m resolution) using ArcGIS 9.3.1 (Environmental Systems Research Institute, Redlands, CA). High hillshade values represent western slopes with high sun exposure, low values are indicative of shaded slopes. Snow affects caribou resource selection (Johnson et al. 2000); thus, we generated percent snow cover by averaging all 8-day composites of maximum snow extent maps in both winter and summer separately at 500-m resolution from Moderate Resolution Imaging Spectroradiometer (MODIS) satellites (Hall et al. 2000). To represent alpine areas where residual snow patches were likely to persist, we divided the number of days snow occupied a MODIS cell by the number of days in the May–June period to generate spatial models of percent snow cover for May and June. The RSF models also incorporated an index of primary productivity that was produced by averaging all 16-day MODIS composites of the normalized difference vegetation index (NDVI) in winter and summer at a 250-m resolution (Pettorelli et al. 2005). Caribou avoid recent burns because fire destroys slow growing lichen (Gustine and Parker 2008). Though fires are rare in the study area, we extracted burned lodgepole (stands that had burned since 1950) from the lodgepole pine land cover class with data from provincial fire history. More information about derivation of the habitat covariates can be found in Polfus et al. (2011).

Resource Selection Functions

We developed RSF models with a use-availability design (Manly et al. 2002) by comparing covariates at caribou GPS locations to random available locations within pooled 99% fixed-kernel seasonal ranges (for detailed methods, see Polfus et al. 2011). We generated available points using a 1:1 ratio of used to available locations. We developed models at the second-order (landscape) scale (Johnson 1980) during winter (15 Nov–15 May) and summer (16 May–14 Nov). We defined seasons based on elevational shifts in caribou locations (sensu DeCesare et al. 2012). We evaluated habitat selection with generalized linear mixed-models with a random intercept for each animal to account for sample size and autocorrelation (Gillies et al. 2006, Fieberg et al. 2010). We included human covariates in the models as a cumulative zone of influence (ZOI; Gunn et al. 2011). The distance the ZOI extended around human covariates varied by development type and season, and represented the area around developments avoided by caribou (Polfus et al. 2011). We spatially mapped the RSFs in ArcGIS 9.3.1 and divided the probability of selection into 10 equal-ranked bins (based on the number of pixels) from 1 (low quality habitat) to 10 (high quality habitat).

TEK Habitat Suitability Index Models

The TRTFN and Round River Conservation Studies staff conducted interviews with 8 TRTFN members in the winter and spring of 2000 and 2001 to develop a regional conservation area design as part of landuse planning, and to develop HSI models for caribou. Interviews were conducted with permission from and in long-term collaboration with the TRTFN. All interviews were conducted in English. Participants were regarded as expert hunters, gatherers, and community elders and were selected by band members from the approximately 100 TRTFN members that reside in the Atlin area (Department of Indian Affairs and Northern Development 2001). All interviewees had >40 years of experience living, hunting, and subsisting in the study area. The knowledge represented information likely obtained by individuals through a combination of direct experience, shared information, and childhood education and tutelage by elders and parents. Thus, cultural learning, stories, and teachings (TEK) were inextricably intertwined with information collected over a lifetime of living and working on the land (LEK).

Questions about cultural practices and knowledge specific to numerous animal species were used to guide individual semi-directive interviews (Huntington 1998, Lyver and Łutsël K'é Dene First Nation 2005). Interview length depended on the knowledge of the participant, varying from an hour to several days. Questions about seasonal use and food resources for key wildlife species were explored during interviews. Participants drew important areas and animal locations on maps. All interviews were voice recorded and later transcribed. The TRTFN retained copies of all the recordings and hold intellectual property rights to the research materials. We analyzed information from interviews on caribou habitat requirements (i.e., land cover types, habitat associations, seasonal foraging strategies, and other resources that were associated with caribou during different times of the year) and summarized this information.

The process used to develop the TEK-HSI models is similar to other HSI modeling efforts (Johnson and Gillingham 2004) that include expert data (literature or expert opinion) and is interpretive and highly contextual. The HSI values represent the relative value of different environmental features (such as land cover type, availability of water, etc.) to caribou habitat use. We linked the habitat descriptions described in interviews with the same spatial resource covariates that we used to generate the RSF models (Table 1). We subsequently used these variables to create rule-based HSI models for summer (Jun–Nov) and winter (Dec–May) to match seasonal periods of the RSFs. Habitat quality was based both on the number of respondents (0–8) and the consistency of interviewee responses relative to each potential habitat characteristic. We scaled the HSI models from 10 (highest value) to 1 (lowest value) to match the RSF scale. We gave variables associated with a high number of respondents reporting similar observations the highest ranks (rank column in Tables 2 and 3). For example, if ≥75% of participants were in agreement about caribou associations with a specific seasonal habitat type, that variable received the highest rank (Table 2; e.g., low elevation lodgepole pine forests received a 9). If <75% of participants reported caribou associations with a habitat variable that variable received a lower rank (Table 2; e.g., low elevation river valleys received a 5). When <25% of participants identified a habitat variable, it received one of the lowest ranks (Table 2; e.g., low elevation lakes received a 2). We built models by applying the values (rank columns in Tables 2 and 3) to the combinations of ecological conditions for land cover type, elevation, and aspect in ArcGIS 9.3.1. In this way, we built models additively by overlaying all values as raster layers in ArcGIS 9.3.1 until all information (from all respondents) was included in one map layer. We could not incorporate human developments into the prediction of habitat quality in the HSI models because the original interviews had been designed to gather baseline habitat relationships in the absence of human impacts.

Table 1. Seasonal caribou (β) coefficients and standard errors for covariates from generalized linear mixed models with a random intercept predicting relative selection at the second-order scale (Johnson 1980) for the Atlin herd of northern mountain woodland caribou in northern British Columbia, 2000–2002
CovariateSummerWinter
Selectivity βSE95% CISelectivity βSE95% CI
  • a

    Lodgepole pine.

  • b

    Normalized difference vegetation index.

  • c

    Zone of influence.

LPa-lichen−0.7330.1465−1.020−0.4450.5690.06240.4470.691
Mixed conifer−0.8570.0920−1.037−0.676    
Krummholz0.3290.11310.1070.550−0.9190.1399−1.193−0.645
Burn LP    −0.8660.1684−1.196−0.536
Spruce-fir    0.2320.06250.1100.355
Low valley willow    0.6870.09370.5030.870
Alpine shrub0.4950.10310.2930.697    
Alpine tundra0.5960.11170.3770.815−0.6990.1634−1.019−0.379
Rock0.2980.13880.0260.570−1.6590.6140−2.862−0.456
Water−3.1980.3123−3.810−2.586−0.8270.1519−1.125−0.529
Elevation0.0120.00120.0100.0140.0170.00120.0150.019
Elevation2−4.44E−064.540E−07−5.33E−06−3.55E−06−7.23E−065.640E−07−8.34E−06−6.13E−06
Slope0.0370.00780.0220.053−0.0500.0034−0.057−0.044
Slope2−0.0020.0002−0.003−0.002    
Hillshade0.0040.00060.0020.0050.0060.00090.0050.008
NDVIb summer−2.71E−041.660E−05−3.04E−04−2.39E−040.0030.00030.0030.004
NDVI summer2    −2.83E−072.310E−08−3.29E−07−2.38E−07
Percent snow winter8.2120.37537.488.959.5520.65758.26410.841
Percent snow winter2    −7.6550.4531−8.543−6.766
Human ZOIc summer−0.4780.0608−0.598−0.359    
Human ZOI winter    −0.9540.0739−1.099−0.809
Constant−14.9900.6986−16.360−13.621−22.7951.1244−24.999−20.591
Table 2. Winter resource covariates used to generate a habitat suitability index model for the Atlin herd of northern mountain woodland caribou from interviews with members of the Taku River Tlingit First Nation in northern British Columbia, Canada, 2000–2002. Rank corresponds to the habitat value in the model
Interview descriptionLand cover typeElevation (m)AspectRank
  • a

    Lodgepole pine.

  • b

    A value of 1 is added to the current value of the land cover types listed when they occur <1 km from a lake.

Low elevation lakesLake<1,150All2
High in mountainsAlpine tundra>1,150All2
Open, high elevation windswept slopesAlpine tundra, rock, snow>1,15090°–180°3
Low elevation forestSpruce-fir, mixed conifer, mixedwood<1,150All4
Eat caribou moss (lichen)LPa-lichen>1,150All5
Low elevation river valleysAlpine shrub, low valley willow<1,150All5
Low elevation forest near LP forestSpruce-fir, mixed conifer, mixedwood <500 m from LP-lichen<1,150All7
Low elevation LP forestLP-lichen<1,150All9
Lakes as escape terrainLow elevation forests (LP-lichen, spruce-fir, mixed conifer, mixedwood <1,150 m) and low elevation valleys (alpine shrub, low valley willow <1,150 m) <1 km from a lake AllAdd 1b
Table 3. Summer resource covariates used to generate a habitat suitability index model for the Atlin herd of northern mountain woodland caribou from interviews with members of the Taku River Tlingit First Nation in northern British Columbia, Canada, 2000–2002. Rank corresponds to the habitat value in the model
Interview descriptionLand cover typeElevation (m)AspectRank
  • a

    Lodgepole pine.

  • b

    A value of 2 is added to the current value of the land cover types listed in this row when the land cover types fall above 1,150 m and between the aspects of 315°–135°.

  • c

    Moderate resolution imaging spectroradiometer satellite.

Below treeline, wide-rangingLPa-lichen<1,150All2
Mountain sides and slopes, wide rangingKrummholz, low valley willow, alpine shrub<1,150All3
Mountain sides and slopes, eat grass and lichenAlpine tundra<1,150All4
Snow to escape insectsSnow<1,150All4
Below treeline, mountain sides and slopes, wide rangingLow valley willow, LP-lichen, spruce-fir, mixed conifer, mixedwood>1,150AllAdd 1
High in mountains, graze on grass and other vegetationAlpine tundra, alpine shrub, krummholz, rock, snow>1,150AllAdd 3
North facing slopes to escape insects on snow patchesAlpine tundra, alpine shrub, krummholz, rock, snow>1,150315°–135°Add 2b
Use last of snow to escape insectsAlpine tundra, alpine shrub, krummholz, rock, snow in area with >50% snow cover May and June (MODISc snow cover data)AllAllAdd 1

Model Evaluations and Comparisons

The utility of habitat models, whether they are RSF or HSI, depends on their ability to predict animal locations (Roloff and Kernohan 1999, Wiens et al. 2008). We used independent VHF data to assess if the RSF and TEK models predicted caribou use. In addition, we also used the GPS data (that was used to build the RSF models) to assess if the TEK models predicted caribou use. For each evaluation, we calculated the number of locations that occurred within the 10 habitat rank classes normalized by the area of that class (Boyce et al. 2002, Johnson and Gillingham 2005). We used Spearman's rank correlation (rs) to test the correlation between predicted habitat quality and the frequency of caribou locations, with the expectation that a greater frequency of locations would occur in high quality habitat (Boyce et al. 2002). To visually examine where spatial discrepancies occurred between models, we subtracted the spatial predictions of the TEK models from the RSF models and mapped the difference between ranks in each cell. To statistically compare spatial predictions of habitat quality between the RSF and TEK models, we simplified the number of habitat ranks from 10 to 3 which represented high, medium, and low quality habitat to simplify decision making for managers (Johnson and Gillingham 2005). We used a weighted Kappa statistic to evaluate the agreement in the distribution of the 10,000 random points in habitat quality ranks (Monserud and Leemans 1992). A Kappa index value of 1 indicates perfect agreement, whereas a value of 0 indicates that the observed agreement was approximately equal to what would be expected by chance (Monserud and Leemans 1992).

RESULTS

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

GPS collars were placed on 8 female and 2 male caribou and VHF collars were placed on a separate sample of 13 female and 4 male caribou. We obtained 16,270 GPS locations and 661 VHF locations. Mean group size of the Atlin herd was 14 individuals, though congregations of up to 200 animals occur (K. Diemert, British Columbia Ministry of Forests, Land, and Natural Resource Operations, unpublished report and data). We assumed that a sample of 27 collared caribou captured in separate groups represented the Atlin herd because caribou are social. Further, beta coefficients for resource selection stabilized and standard errors reached an asymptote when we included 8–10 caribou in the models (see Figs. S1 and S2, available online at www.onlinelibrary.wiley.com). We pooled males and females as resource use did not differ between the sexes (χ2 test, winter P = 0.455, summer P = 0.816; Polfus et al. 2011).

Resource Selection Function Models

At the landscape scale, caribou showed significant avoidance of summer and winter ZOIs (Polfus et al. 2011). In winter, caribou selected intermediate elevations (1,179 m) and lichen-lodgepole pine complexes, spruce-fir forests, and low elevation willow (Table 1). Caribou avoided krummholz, alpine environments, steep slopes, burned lodgepole pine, rock, and water in winter. In summer, caribou habitat selection shifted to higher elevations (1,363 m) and caribou displayed strong selection for krummholz, alpine shrub and tundra, rock, slopes with greater sun exposure, and areas that had high percent snow cover in winter. Caribou selection was negatively associated with lodgepole pine, mixed conifer forests, water, and steep slopes (Table 1).

TEK Habitat Suitability Index Models

The 8 TRTFN members were generally in agreement about caribou habitat use (i.e., individuals did not contradict each other). The TRTFN members indicated that in winter caribou used low elevation forests, especially mature lodgepole pine with lichen groundcover (Table 2). They also indicated that caribou used low elevation valleys in river bottoms and open windswept slopes in alpine areas depending on snow conditions. Low elevation lakes were also identified as predator escape terrain and were thought to be used by caribou in winter as mineral licks (Table 2). Participants reported that in summer caribou used predominately high elevation alpine environments during the entire period and could often be found on remnant snow patches to escape insects. They also indicated that caribou were wide-ranging and used mountainsides and slopes where they foraged on grass, willow, and lichen (Table 3).

Model Comparison and Evaluation

Correlations between the summer and winter RSF models and frequency of withheld caribou locations were high (rs = 0.994 and 0.997, respectively). The TEK-based HSI models also had high predictive performance when evaluated with caribou location data. Output from the summer TEK model was highly correlated with the frequency of caribou GPS locations (rs = 0.910), though the correlation was lesser for VHF locations (rs = 0.612). Output from the winter TEK model was more highly correlated with frequency of VHF caribou locations (rs = 0.806) compared to GPS locations (rs = 0.750).

Visual inspection of the differences between the 10 habitat class RSF and TEK maps indicated that most spatial discrepancies were a result of the RSF models predicting more high quality habitat than the TEK models (Fig. 2). In winter, discrepancies were most apparent on north and west slopes, on small lakes, and within a large historical burn in the northern part of the study area (see Fig. S3, available online at www.onlinelibrary.wiley.com). The winter TEK model predicted more high quality habitat in and around the town of Atlin than the RSF model. In summer, areas of low elevation were given higher rank by the RSF than by the TEK model (see Fig. S4, available online at www.onlinelibrary.wiley.com). Reducing the number of rank classes from 10 to 3 standardized the amount of area in each habitat class; high (30%), medium (30%), and low (40%) and facilitated statistical comparison between the 2 models. The Kappa value for the summer TEK and RSF models was 0.649 (SE = 0.008) and indicated substantial spatial agreement (0.61–0.80), whereas the Kappa value for the winter TEK and RSF model was 0.337 (SE = 0.008), which represented fair agreement (0.21–0.40).

image

Figure 2. Second-order resource selection function (RSF) maps generated with global positioning system collar locations (left column) and habitat suitability index (HSI) model maps generated with the traditional ecological knowledge (TEK) of the Taku River Tlingit First Nation (TRTFN, right column) for summer (a) and winter (b) habitat use of the Atlin herd of northern woodland caribou, northern British Columbia, Canada, 2000–2002. The relative probability of selection in the RSF models is scaled between low (green) and high (red).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

Our study is the first to quantitatively compare TEK-based caribou habitat models with habitat models developed using western science approaches. The high predictive ability of the models and correlations between HSI and RSF model outputs demonstrates that TEK can be an effective tool for wildlife management and recovery planning. Specifically, both modeling approaches predicted that caribou selected low elevation lodgepole pine forests in winter, which is typical of northern mountain populations (Poole et al. 2000, Florkiewicz et al. 2007, Gustine and Parker 2008). During summer, both models predicted caribou use of alpine habitats, which is likely a result of selection for high quality forage and relief from insect harassment (Ion and Kershaw 1989). Our results support the growing recognition of TEK approaches in conservation initiatives around the world (Berkes 2004).

Comparisons between different modeling approaches can reveal assumptions about modeling processes, potential biases, and differences in scales of inference. For example, Johnson and Gillingham (2005) found that an HSI model of caribou habitat in British Columbia was a poor predictor of caribou distribution when compared to RSF and species niche models. However, the HSI model was not developed with local expert knowledge, which may have limited performance. In our case, despite high overall agreement between approaches, we found several informative differences between TEK and RSF models. First, the RSF models predicted a greater area of habitat in RSF ranks 8–10 (winter 3,475 km2) than occurred in the corresponding habitat ranks of the TEK models (winter 1,705 km2) because the habitat classes in the TEK models did not have equal areas. Thus, to facilitate statistical comparison, we reduced the number of habitat classes so the areas of the resulting 3 classes (high, medium, and low) were similar between the RSF and TEK models (i.e., high quality habitat in the TEK models included ranks 7–10; 3,242 km2). Further, RSF models allow remote sensing covariates such as slope, aspect, hillshade, and NDVI to be simultaneously incorporated in the model; all of which are challenging to match with TEK because people do not often discuss this type of variable. Thus, the RSF's ability to incorporate remote sensing covariates likely resulted in some of the spatial discrepancies between models (see Fig. S3, available online at www.onlinelibrary.wiley.com). In summer, the TEK and RSF models had relatively high spatial agreement. Most discrepancies occurred in valley bottoms, which were often considered medium quality habitat in the RSF models but low quality habitat in the TEK-HSI models because of an elevational cut-off used to indicate caribou preferred alpine habitats.

TEK can also provide unique perspectives and suggest new avenues for scientific exploration (Ramstad et al. 2007). For example, local resident knowledge of jaguar (Panthera onca) occurrence in Nicaragua overlapped with a GIS-based least-cost corridor analysis by only 33%, a result that affirmed the importance of examining alternate data sources and the value of local knowledge in guiding conservation planning (Zeller et al. 2011). In our study, respondents described lakes as escape terrain in winter. Caribou are also highly visible on frozen lakes and commonly dig through the snow to lick the ice and feed on vegetation in muskrat (Ondatra zibethicus) push-ups (J. Williams, M. Connor, TRTFN, personal communication). However, the RSF models indicated that caribou avoided lakes in winter. Resource selection analyses may emphasize behaviors with high residency times in a land cover type, and underestimate importance of land cover types used for movement (Cleveland et al. 2012). Additionally, in the RSF models, caribou avoidance of large lakes may have masked selection for small lakes because all lakes were classified as a single land cover type. Caribou likely differentiated between large and small lakes. Frozen lakes provide an example where TEK suggests the need for additional information gathering around a potentially important but otherwise overlooked habitat type.

We also found habitat quality differences in the winter TEK and RSF models within a large burn that was considered low quality based on TEK but high quality in the RSF models. Few caribou locations occurred within the burn. The TEK captured a habitat relationship not quantified in the RSF model but one that likely has importance in understanding caribou habitat quality and distribution. Rather than use differences between models to determine which is more correct, managers must acknowledge that neither the RSF nor TEK models are necessarily better than the other and that objective comparison can increase insights into both methodological approaches (Ramstad et al. 2007).

Several studies have attributed long temporal but relatively small spatial scales to ecological information collected from TEK (Usher 2000, Moller et al. 2004, Fraser et al. 2006, Rist et al. 2010). In the Atlin area, our model comparisons indicate that predictions from the TEK models most closely resembled caribou habitat selection at the scale of the herd home range (Polfus 2010). Methods used to collect TEK may be adapted to compliment the scale of western scientific studies (Ramstad et al. 2007). For example, Gagnon and Berteaux (2009) reported that in Nunavut, Canada, TEK regarding arctic fox (Vulpes lagopus) populations provided information at a larger spatial scale than current scientific data (by describing an area up to 23,000 km2), but that the scale of TEK on the distribution of molting locations and migrations of snow goose (Chen caerulescens) was similar to western scientific documentation. In Switzerland, lynx (Lynx lynx) habitat models developed with the local knowledge from game wardens performed better than models developed with knowledge from international lynx experts when validated against information from the local area (Doswald et al. 2007). However, the model derived from lynx experts was more transferable to other parts of Switzerland. New statistical approaches like scale-integrated RSFs, may allow managers to synthesize results across scales to predict species occurrence across a variety of extents (DeCesare et al. 2012).

The TEK HSI models did not include direct reference to the influence of human development on caribou habitat selection which contributed to some discrepancies. We found the winter TEK models predicted higher quality habitat closer to human disturbances than the RSF models. For these reasons, TEK may be useful for predicting the habitat quality in areas that are currently avoided by caribou because of human developments. Thus, TEK might be used to locate suitable sites for habitat restoration (Patthey et al. 2008), guide dynamic estimates of habitat quality (Nielsen et al. 2010), or produce population abundance targets (Serrouya et al. 2011). Because interview structure is flexible, managers may be able to design interview questions to include information about historical distributions that could aid in the identification of underlying habitat quality in areas that are avoided at present. Information about potential habitats that are currently unused is difficult to reveal with telemetry data. Knowledge of potential habitats could be especially important in areas of recent extirpation or in places that have experienced recent increases in human development.

Governments around the world are recognizing the importance of indigenous rights in wildlife management. Recently, the TRTFN and British Columbia signed a joint land use plan (Taku River Tlingit First Nation and Province of British Columbia 2011) that includes explicit direction for the management of land use activities, cumulative effects, and habitat loss that may affect the Atlin caribou herd. Managers must find new ways to incorporate TEK in natural resource management (Colchester 2004). Merely testing TEK with western science (Gilchrist et al. 2005, Anadón et al. 2009) fails to appreciate that western scientific approaches are not necessarily closer to truth than other approaches (Gratani et al. 2011). Instead, collaborative comparisons provide an avenue towards a more complete understanding of ecosystems and can lead to more effective management decisions. Using TEK requires respecting cultural differences, developing meaningful collaborations, and ensuring that knowledge is not taken out of context, misinterpreted, or misused (Usher 2000, Huntington 2011). Though obstacles may impede the translation of ideas and concepts between worldviews, our example clearly shows TEK-based habitat models have the potential to provide useful habitat identification and increase the ability of indigenous people to influence decisions that affect their community, culture, and lifestyle (Berkes 2004).

MANAGEMENT IMPLICATIONS

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

Our work demonstrates that TEK can be used by wildlife management organizations in response to the ongoing challenges of woodland caribou conservation and recovery planning across Canada (Festa-Bianchet et al. 2011). Recent court rulings upholding First Nations concerns and legal rights regarding boreal caribou declines emphasize the importance of increasing First Nation involvement in conservation strategies. Although the collection and incorporation of TEK in recovery planning is required under SARA, political controversy, inadequate guidelines, and weak policy direction have hindered the inclusion of TEK. Current approaches by the Canadian Government to collect TEK to inform SARA recovery planning for woodland caribou could benefit from adopting a collaborative community-based approach. Traditional knowledge holders may also gain confidence in co-management settings through a process of collaborative validation (Gratani et al. 2011). Quantitative TEK-based caribou habitat models, such as ours, are especially important in areas where western scientific data, such as telemetry data, are limited. For example, collecting western scientific data for the 57 local boreal woodland caribou populations that span 1.5 million km2 will be difficult (Environment Canada 2012b). Managers need to accept, however, that although collaborative approaches can ultimately lead to more successful conservation decision making, they require a substantial time and financial commitment to be successful. Finally, our efforts demonstrate that TEK-based habitat models not only facilitate critical habitat identification, but offer an approach that can meaningfully incorporate knowledge held by First Nation people into a more complete understanding of caribou ecology and, ultimately, recovery planning.

ACKNOWLEDGMENTS

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

We thank the Taku River Tlingit First Nation elders and community members who shared their knowledge to benefit this work including: J. Williams, B. Jack, A. Williams, T. Jack, H. Carlick, R. Carlick, G. Thorlakson, R. Johnson, J. Jack, P. Tizya, and the late D. Jack. In addition, project oversight for the TRTFN was provided by S. Carlick, N. Gordon, and M. Connor. Funding was provided by the Habitat Stewardship Program for Species at Risk of Environment Canada, the Taku River Tlingit First Nation, Round River Conservation Studies, NASA grant # NNX11AO47G, and the University of Montana. The partnership between the TRTFN and Round River Conservation Studies facilitated funding proposals, project management, reporting, and the development of project partnerships. We thank Round River staff who helped with field work and conducted interviews, especially D. Milek and A. Noson. R. Tingey provided valuable GIS support. University of Calgary collaborators G. McDermid and A. McLane assisted with developing the land cover classification. The manuscript benefited from the revisions of M. Mitchell, P. Krausman, C. Polfus, M. Festa-Bianchet, D. Visscher, and G. Roloff. Animal location data were provided by K. Diemert, N. Maclean, and the British Columbia Ministry of Water, Land, and Air Protection.

LITERATURE CITED

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information
  • Anadón, J. D., A. Giménez, R. Ballestar, and I. Pérez. 2009. Evaluation of local ecological knowledge as a method for collecting extensive data on animal abundance. Conservation Biology 23:617625.
  • Apps, C. D., and B. N. McLellan. 2006. Factors influencing the dispersion and fragmentation of endangered mountain caribou populations. Biological Conservation 130:8497.
  • Berkes, F. 1999. Sacred ecology, traditional ecological knowledge and resource management. Taylor and Francis, Philadelphia, Pennsylvania, USA.
  • Berkes, F. 2004. Rethinking community-based conservation. Conservation Biology 18:621630.
  • Berkes, F., J. Colding, and C. Folke. 2000. Rediscovery of traditional ecological knowledge as adaptive management. Ecological Applications 10:12511262.
  • Boyce, M. S., P. R. Vernier, S. E. Nielsen, and F. K. A. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modelling 157:281300.
  • Cleveland, S. M., M. Hebblewhite, M. Thompson, and R. Henderson. 2012. Linking elk movement and resource selection to hunting pressure in a heterogeneous landscape. Wildlife Society Bulletin 36:658668.
  • Colchester, M. 2004. Conservation policy and indigenous peoples. Environmental Science & Policy 7:145153.
  • Davis, A., and K. Ruddle. 2010. Constructing confidence: rational skepticism and systematic enquiry in local ecological knowledge research. Ecological Applications 20:880894.
  • DeCesare, N. J., M. Hebblewhite, F. Schmiegelow, D. Hervieux, G. J. McDermid, L. Neufeld, M. Bradley, J. Whittington, K. G. Smith, L. E. Morgantini, M. Wheatley, and M. Musiani. 2012. Transcending scale dependence in identifying habitat with resource selection functions. Ecological Applications 22:10681083.
  • Department of Indian Affairs and Northern Development. 2001. Registered Indian population by sex and residence 2000. Ottawa, Canada: <http://epub.sub.uni-hamburg.de/epub/volltexte/2010/2880/pdf/rip00_e.pdf>. Accessed 26 Mar 2013.
  • Doswald, N., F. Zimmermann, and U. Breitenmoser. 2007. Testing expert groups for a habitat suitability model for the lynx Lynx lynx in the Swiss Alps. Wildlife Biology 13:430446.
  • Environment Canada. 2012a. Management plan for the northern mountain population of woodland caribou (Rangifer tarandus caribou) in Canada. Species at Risk Act Management Plan Series. Environment Canada, Ottawa, Ontario, Canada.
  • Environment Canada. 2012b. Recovery strategy for the woodland caribou (Rangifer tarandus caribou), boreal population, in Canada. Species at Risk Act Recovery Strategy Series. Environment Canada, Ottawa, Ontario, Canada.
  • Farnell, R. 2009. Three decades of caribou recovery programs in Yukon: a paradigm shift in wildlife management. Department of Environment, Government of Yukon, Whitehorse, Yukon Territory, Canada.
  • Ferguson, M. A. D., and F. Messier. 2000. Mass emigration of arctic tundra caribou from a traditional winter range: population dynamics and physical condition. Journal of Wildlife Management 64:168178.
  • Festa-Bianchet, M., J. C. Ray, S. Boutin, S. Côté, and A. Gunn. 2011. Conservation of caribou (Rangifer tarandus) in Canada: an uncertain future. Canadian Journal of Zoology 89:419434.
  • Fieberg, J., J. Matthiopoulos, M. Hebblewhite, M. S. Boyce, and J. L. Frair. 2010. Correlation and studies of habitat selection: problem, red herring or opportunity? Philosophical Transactions of the Royal Society B-Biological Sciences 365:22332244.
  • Florkiewicz, R., R. Maraj, T. Hegel, and M. Waterreus. 2007. The effects of human land use on the winter habitat of the recovering Carcross woodland caribou herd in suburban Yukon Territory, Canada. Rangifer Special Issue 17:181197.
  • Frair, J. L., J. Fieberg, M. Hebblewhite, F. Cagnacci, N. J. DeCesare, and L. Pedrotti. 2010. Resolving issues of imprecise and habitat-biased locations in ecological analyses using GPS telemetry data. Philosophical Transactions of the Royal Society B-Biological Sciences 365:21872200.
  • Fraser, D. J., T. Coon, M. R. Prince, R. Dion, and L. Bernatchez. 2006. Integrating traditional and evolutionary knowledge in biodiversity conservation: a population level case study. Ecology and Society 11(2) <http://www.ecologyandsociety.org/vol11/iss2/art4/>. Accessed 13 Nov 2013.
  • Gagnon, C. A., and D. Berteaux. 2009. Integrating traditional ecological knowledge and ecological science: a question of scale. Ecology and Society 14(2) <http://www.ecologyandsociety.org/vol14/iss2/art19/>. Accessed 13 Nov 2013.
  • Gilchrist, G., M. Mallory, and F. Merkel. 2005. Can local ecological knowledge contribute to wildlife management? Case studies of migratory birds. Ecology and Society 10(1) <http://www.ecologyandsociety.org/vol10/iss1/art20/>. Accessed 13 Nov 2013.
  • Gillies, C. S., M. Hebblewhite, S. E. Nielsen, M. A. Krawchuk, C. L. Aldridge, J. L. Frair, D. J. Saher, C. E. Stevens, and C. L. Jerde. 2006. Application of random effects to the study of resource selection by animals. Journal of Animal Ecology 75:887898.
  • Government of Canada. 2002. Species at Risk Act: an act respecting the protection of wildlife species at risk in Canada. Government of Canada, Ottawa, Canada.
  • Gratani, M., J. R. A. Butler, F. Royee, P. Valentine, D. Burrows, W. I. Canendo, and A. S. Anderson. 2011. Is validation of indigenous ecological knowledge a disrespectful process? A case study of traditional fishing poisons and invasive fish management from the wet tropics, Australia. Ecology and Society 16(3) <http://www.ecologyandsociety.org/vol16/iss3/art25/>. Accessed 13 Nov 2013.
  • Gunn, A., C. J. Johnson, J. S. Nishi, C. J. Daniel, D. E. Russell, M. Carlson, and J. Z. Adamczewski. 2011. Understanding the cumulative effects of human activities on barren-ground caribou. Pages 113134 in P. R. Krausman, and L. K. Harris, editors. Cumulative effects in wildlife management: impact mitigation. CRC Press, Boca Raton, Florida, USA.
  • Gustine, D. D., and K. L. Parker. 2008. Variation in the seasonal selection of resources by woodland caribou in northern British Columbia. Canadian Journal of Zoology 86:812825.
  • Hall, D. K., G. A. Riggs, and V. V. Salomonson. 2000. MODIS/Terra snow cover 8-day L3. Global 500m Grid V03, February 2000 to February 2002. National Snow and Ice Data Center, Digital Media, Boulder, Colorado, USA.
  • Hebblewhite, M. 2006. Linking predation risk and forage to ungulate population dynamics. Dissertation, University of Alberta, Edmonton, Canada.
  • Hebblewhite, M., C. White, and M. Musiani. 2010. Revisiting extinction in national parks: mountain caribou in Banff. Conservation Biology 24:341344.
  • Huntington, H. P. 1998. Observations on the utility of the semi-directive interview for documenting traditional ecological knowledge. Arctic 51:237242.
  • Huntington, H. P. 2011. Arctic science: the local perspective. Nature 478:182183.
  • Ion, P. G., and G. P. Kershaw. 1989. The selection of snowpatches as relief habitat by woodland caribou (Rangifer tarandus caribou). Macmillan Pass, Selwyn/Mackenzie, Mountains, N.W.T., Canada. Arctic and Alpine Research 21: 203-211.
  • Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:6571.
  • Johnson, C. J., and M. P. Gillingham. 2004. Mapping uncertainty: sensitivity of wildlife habitat ratings to expert opinion. Journal of Applied Ecology 41:10321041.
  • Johnson, C. J., and M. P. Gillingham. 2005. An evaluation of mapped species distribution models used for conservation planning. Environmental Conservation 32:117128.
  • Johnson, C. J., K. L. Parker, and D. C. Heard. 2000. Feeding site selection by woodland caribou in north-central British Columbia. Rangifer Special Issue 12:159172.
  • Kendrick, A., and M. Manseau. 2008. Representing traditional knowledge: resource management and Inuit knowledge of barren-ground caribou. Society and Natural Resources 21:404418.
  • Lyver. P. O., and Łutsël K'é Dene First Nation. 2005. Monitoring barren-ground caribou body condition with Denésôåiné traditional knowledge. Arctic 58:4454.
  • Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. P. Erickson. 2002. Resource selection by animals: statistical design and analysis for field studies. Second edition. Kluwer Academic Publishers, Dordrecht, the Netherlands.
  • McClellan, C. 1981. Inland tlingit. Pages 469480 in J. Helm, editor. Handbook of North American Indians. Vol. 6, Subarctic. Smithsonian Institution, Washington, D.C., USA.
  • Moller, H., F. Berkes, P. O. Lyver, and M. Kislalioglu. 2004. Combining science and traditional ecological knowledge: monitoring populations for co-management. Ecology and Society 9(3) <http://www.ecologyandsociety.org/vol9/iss3/art2/>. Accessed 13 Nov 2013.
  • Monserud, R. A., and R. Leemans. 1992. Comparing global vegetation maps with the Kappa statistic. Ecological Modelling 62:275293.
  • Montgomery, R. A., G. J. Roloff, and J. M. Ver Hoef. 2011. Implications of ignoring telemetry error on inference in wildlife resource use models. Journal of Wildlife Management 75:702708.
  • Mooers, A. O., D. F. Doak, C. S. Findlay, D. M. Green, C. Grouios, L. L. Manne, A. Rashvand, M. A. Rudd, and J. Whitton. 2010. Science, policy, and species at risk in Canada. BioScience 60:843849.
  • Nichols, T., F. Burkes, D. Jolly, and N. B. Snow, the community of Sachs Harbour. 2004. Climate change and sea ice: local observations from the Canadian western arctic. Arctic 57:6879.
  • Nielsen, S. E., M. S. Boyce, G. B. Stenhouse, and R. H. M. Munro. 2003. Development and testing of phenologically driven grizzly bear habitat models. Ecoscience 10:110.
  • Nielsen, S. E., G. McDermid, G. B. Stenhouse, and M. S. Boyce. 2010. Dynamic wildlife habitat models: seasonal foods and mortality risk predict occupancy-abundance and habitat selection in grizzly bears. Biological Conservation 143:16231634.
  • O'Flaherty, R. M., I. J. Davidson-Hunt, and M. Manseau. 2008. Indigenous knowledge and values in planning for sustainable forestry: Pikangikum First Nation and the Whitefeather Forest Initiative. Ecology and Society 13(1) <http://www.ecologyandsociety.org/vol13/iss1/art6/>. Accessed 13 Nov 2013.
  • Patthey, P., S. Wirthner, N. Signorell, and R. Arlettaz. 2008. Impact of outdoor winter sports on the abundance of a key indicator species of alpine ecosystems. Journal of Applied Ecology 45:17041711.
  • Perera, A. H., C. A. Drew, and C. J. Johnson. 2012. Experts, expert knowledge, and their roles in landscape ecological applications. Pages 110 in A. H. Perera, C. A. Drew, and C. J. Johnson, editors. Expert knowledge and its application in landscape ecology. Springer, New York, New York, USA.
  • Pettorelli, N., J. O. Vik, A. Mysterud, J. M. Gaillard, C. J. Tucker, and N. C. Stenseth. 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution 20:503510.
  • Polfus, J. L. 2010. Assessing cumulative human impacts on northern woodland caribou with traditional ecological knowledge and resource selection functions. Thesis, University of Montana, Missoula, USA.
  • Polfus, J. L., M. Hebblewhite, and K. Heinemeyer. 2011. Identifying indirect habitat loss and avoidance of human infrastructure by northern mountain woodland caribou. Biological Conservation 144:26372646.
  • Poole, K. G., D. C. Heard, and G. Mowat. 2000. Habitat use by woodland caribou near Takla Lake in central British Columbia. Canadian Journal of Zoology 78:15521561.
  • Ramstad, K. M., N. J. Nelson, G. Paine, D. Beech, A. Paul, P. Paul, F. W. Allendorf, and C. H. Daugherty. 2007. Species and cultural conservation in New Zealand: Maori traditional ecological knowledge of tuatara. Conservation Biology 21:455464.
  • Resource Inventory Committee. 1998. Wildlife radio telemetry, standards for components of British Columbia's biodiversity no 5. Ministry of Environment, Lands and Parks, Victoria, British Columbia, Canada.
  • Rist, L., R. U. Shaanker, E. J. Milner-Gulland, and J. Ghazoul. 2010. The use of traditional ecological knowledge in forest management: an example from India. Ecology and Society 15(1) <http://www.ecologyandsociety.org/vol15/iss1/art3/>. Accessed 13 Nov 2013.
  • Roloff, G. J., and B. J. Kernohan. 1999. Evaluating reliability of habitat suitability index models. Wildlife Society Bulletin 27:973985.
  • Sandström, P., T. G. Pahlén, L. Edenius, H. Tømmervik, O. Hagner, L. Hemberg, H. Olsson, K. Baer, T. Stenlund, L. G. Brandt, and M. Egberth. 2003. Conflict resolution by participatory management: remote sensing and GIS as tools for communicating land-use needs for reindeer herding in northern Sweden. Ambio 32:557567.
  • Santomauro, D., C. J. Johnson, and G. Fondahl. 2012. Historical-ecological evaluation of the long-term distribution of woodland caribou and moose in central British Columbia. Ecosphere 3:119.
  • Schaefer, J. A., and S. P. Mahoney. 2007. Effects of progressive clearcut logging on Newfoundland caribou. Journal of Wildlife Management 71:17531757.
  • Schmidt, P. M., and H. Stricker. 2010. What tradition teaches. Wildlife Professional 4(4):4044.
  • Serrouya, R., B. N. McLellan, S. Boutin, D. R. Seip, and S. E. Nielsen. 2011. Developing a population target for an overabundant ungulate for ecosystem restoration. Journal of Applied Ecology 48:935942.
  • Taku River Tlingit First Nation, and British Columbia. 2010. Interim collaborative harvest management plans for Atlin caribou, Atlin East sheep and moose and lower Taku grizzly bear. Report of the Framework Agreement for Shared, Decision Making Respecting Land Use and Wildlife Management. Integrated Land Management Bureau, Smithers, British Columbia, Canada.
  • Taku River Tlingit First Nation, and British Columbia. 2011. Wóoshtin wudidaa: Atlin Taku Land Use Plan. Taku River Tlingit First Nation Land and, Resources Department, Atlin, British Columbia, Canada.
  • U.S. Fish and Wildlife Service. 1981. Standards for the development of habitat suitability index models. U.S. Fish and Wildlife Service, ESM 103, Department of the Interior, Washington, D.C., USA.
  • Usher, P. J. 2000. Traditional ecological knowledge in environmental assessment and management. Arctic 53:183193.
  • Vors, L. S., and M. S. Boyce. 2009. Global declines of caribou and reindeer. Global Change Biology 15:26262633.
  • Wiens, T. S., B. C. Dale, M. S. Boyce, and G. P. Kershaw. 2008. Three way k-fold cross-validation of resource selection functions. Ecological Modelling 212:244255.
  • Wittmer, H. U., B. N. McLellan, D. R. Seip, J. A. Young, T. A. Kinley, G. S. Watts, and D. Hamilton. 2005a. Population dynamics of the endangered mountain ecotype of woodland caribou (Rangifer tarandus caribou) in British Columbia, Canada. Canadian Journal of Zoology 83:407418.
  • Wittmer, H. U., A. R. E. Sinclair, and B. N. McLellan. 2005b. The role of predation in the decline and extirpation of woodland caribou. Oecologia 144:257267.
  • Zeller, K. A., S. Nijhawan, R. Salom-Pérez, S. H. Potosme, and J. E. Hines. 2011. Integrating occupancy modeling and interview data for corridor identification: a case study for jaguars in Nicaragua. Biological Conservation 144:892901.

Supporting Information

  1. Top of page
  2. ABSTRACT
  3. STUDY AREA
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. MANAGEMENT IMPLICATIONS
  8. ACKNOWLEDGMENTS
  9. LITERATURE CITED
  10. Supporting Information

Additional supporting information may be found in the online version of this article.

FilenameFormatSizeDescription
jwmg643-sm-0001-SupData-S1.docx1171K

Table S1. Example of how different groupings of caribou were selected to test the effect of sample size on RSF model predictions. For example, in the first set (1 caribou in model), locations from each caribou (C1–C5 and C22–C26) are used to produce the model; in the second set (2 caribou in model), 2 individual caribou are randomly selected to produce the model and this is replicated 10 times, etc. This is repeated for sets of 3–10 caribou models.

Figure S1. Coefficients for trees in RSF models that include 1–10 caribou. The beta coefficient estimate stabilizes around 6 caribou included in the model.

Figure S2. Coefficients for water in RSF models that include 1–10 caribou. The beta coefficient estimate stabilizes around 5 caribou included in the model.

Figure S3. Spatial discrepancies between rankings of winter resource selection function (RSF) generated with spatial information from caribou locations and rankings of winter habitat suitability index model generated with the traditional ecological (TEK) knowledge of the Taku River Tlingit First Nation in northern British Columbia, Canada. Warm colors indicate areas where the ranking of habitat quality predicted by the TEK model was higher than the ranking of the RSF model. Cool colors indicate places where the RSF predicted higher habitat quality than the TEK model. The numbers represent the difference in habitat rankings (1–10) where only discrepancies in rankings of greater than 5 are colored.

Figure S4. Spatial discrepancies between rankings of summer resource selection function (RSF) generated with spatial information from caribou locations and rankings of winter habitat suitability index model generated with the traditional ecological (TEK) knowledge of the Taku River Tlingit First Nation in northern British Columbia, Canada. Warm colors indicate areas where the ranking of habitat quality predicted by the TEK model was higher than the ranking of the RSF model. Cool colors indicate places where the RSF predicted higher habitat quality than the TEK model. The numbers represent the difference in habitat rankings (1–10) where only discrepancies in rankings of greater than 5 are colored.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.