Have geographical influences and changing abundance led to sub-popula- tion structure in the Ahiak caribou herd, Nunavut, Canada?

We examined the premise that changing abundance and environmental conditions influence the seasonal dispersion and distribution of migratory tundra caribou (Rangifer tarandus groenlandicus). The Ahiak herd’s (north-central Nunavut, Canada) calving shifted from dispersed on islands to gregarious calving on the mainland coast. As abundance further increased, the calving ground elongated east and west such that we proposed a longitudinal climate gradient. As well, the calving ground’s east and west ends are different distances from the tree-line, which dips south closer to Hudson Bay. We proposed that whether caribou winter on the tundra or within boreal forest and the different climate across the long calving ground could contribute to differential survival and productivity such that sub-population structure would result. At the scale of the individual cows (identified through satellite-collars), we did not find inter-annual spatial fidelity to either the western or eastern parts of the calving ground. At the population scale (aerial surveys of calving distribution), we also did not find discontinuities in calving distribution. The spatial association of individual cows during calving compared with their association during the rut was inconsistent among years, but overall, cows that calve together, rut together. At this time and with the available evidence, we could not infer sub-population structure from shifts in dispersion and distribution as influenced by geography and changes in abundance for the Ahiak herd.


Introduction
The general premise for this paper is that physical geography (landforms and climate) and abundance influence migratory tundra caribou's (Rangifer tarandus groenlandicus) seasonal and annual distribution (Bergerud et al., 2008). Seasonal ranges contract or expand with changes in abundance within the constraints imposed by how physical geography influences distribution.
The Ahiak herd (previously named as the Queen Maud Gulf herd; Gunn et al., 2000) seasonally ranges mostly in north central Nunavut, Canada (Fig. 1). We first describe how abundance and physical geography influence the herd's calving and winter distribution. Sec-ondly, we examined whether those influences contribute toward sub-population structure within the Ahiak herd. Wells & Richmond (1995) defined a subpopulation as "an arbitrary spatially-delimited subset of individuals from within a population".
The geographical influences are firstly, the configuration of Queen Maud Gulf 's coast line which stretches in a relatively straight east-west direction for 300 km and has 100s of islands and islets. Compared to other caribou calving grounds, the Ahiak herd's annual calving grounds (sensu Russell et al., 2002) are unusually long and narrow (~340 km x ~75 km).
The second geographical influence is how the extent of the low-lying coastal plains of the northeast Nunavut mainland affects the position of the tree-line and in turn, caribou winter and pre-calving migration distribution. Within mainland Canada, from west to east toward Hudson Bay, the low elevation arctic coastal plain widens out and consequently the tree-line (Timoney et al., 1992) dips south and widens (Fig. 2). The tree-line is more accurately described as a forest-tundra biome (Timoney et al., 1992;Payette et al., 2001) lying between the southern limit (<0.1% cover) of upland tundra and the northern limit (<0.1% cover) of trees >3-4 m tall. From Mackenzie Delta to north of Yellowknife, the forest-tundra biome is 60 -150 km wide compared to 230-340 km for the zone from the Dubawnt River to central Manitoba-Keewatin (Timoney et al., 1992). The southeast extension of the tundra west of Hudson Bay increases the likelihood for some Ahiak caribou to be wintering on the tundra.
In the paper's second part, we examine how the geographical influences (long straight coast line and position of the tree-line) could lead to sub-population structure. A longitudinal environmental gradient in the timing of snowmelt or plant green-up across the east-west length of the calving ground could influence early calf survival. If individual cows had fidelity between years to either the western or eastern calving ground and the environmental gradient is marked; sub-population structure could develop if the association of cows on the calving grounds is similar to the pattern of association during the rut. Additionally, if individual cows wintering on the tundra or within the forests are more likely to calve at the east or west end of the elongated coastal calving ground, then sub-population structure could develop or be accentuated in response to different environmental conditions during winter and pre-calving migration. Cows wintering within foresttundra and lichen woodland communities of the boreal forest (Payette et al., 2001), versus those wintering on the tundra, will face different ecological conditions during winter and pre-calving migration, which could affect their reproductive success and survival.
would depend on inter-annual fidelity by individual breeding females within the calving ground. While annual fidelity of individual cows to calving grounds is well documented, fidelity to specific sites has not been reported (for example, Fancy & Whitten, 1991). Our premise was that the length of the calving ground would be sufficient for cows to possibly exhibit annual fidelity to part of the calving ground.
Our objective is to review evidence for use patterns of the calving grounds at the individual scale (based on satellite-collared cows) and at the population-scale (using information from aerial surveys of calving distribution). Our hypothesis is that the Ahiak calving grounds are comprised of two or more calving sub-populations.
By focusing on a relatively detailed description of space use by the Ahiak herd, we aim to contribute to better understand the spatial dynamics of migratory caribou. The Ahiak herd is unusual in its long narrow calving ground and if it contributes to sub-population, the information will be useful to management. We need to be aware of similarities and differences between herds especially in the context of longerterm changes in space use as herds change in abundance.

Methods
We compiled information from government reports (Table 1) to derive descriptive statistics and spatial analyses for telemetry (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) and aerial survey data (2006-10) (Government of Northwest Territories, Department of Environment and Natural Resources). We also used 11 calving and rut locations from 2008-10 from a cooperative project for monitoring and baseline studies, Agnico-Eagle Mines Limited, AREVA Ltd and Government of Nunavut (M. Campbell, pers. comm.;Gebauer et al., 2011). We tested four predictions that followed from our negative hypothesis: 1) Distribution of calving cows would be discontinuous across the calving ground and individual breeding females would show interannual fidelity to discrete areas within annual calving grounds observed in consecutive years.
2) Based on the sample of collared cows, 5% or less of the collared breeding females would switch calving locations in subsequent years between the western, central and eastern areas of an annual calving ground. (We selected 5% based on annual rates of switches of breeding females).
3) Based on aerial surveys of the Ahiak calving grounds, 2006-10, there would be an observable pattern of discontinuity along the eastwest axis in the distribution of breeding females and newborn calves (measured at a grid scale interval of 10 km). 4) Cows that calved together on an annual calving ground (at the scale of the east or west halves of the calving ground), will not solely associate together during the subsequent fall rut. Thus, within a year, collared breeding females that calved within i) the western portion of the calving ground, or ii) the combined central and eastern portions would not be associated with other cows that calved within the same area of the coastal calving ground during the subsequent fall rut (temporal index is 20 October) at a rate greater than would be expected by chance (i.e., >50% probability).

Distribution and abundance of caribou calving in the Queen Maud Gulf
To describe caribou calving along the coast of Queen Maud Gulf relative to changes in abundance, we compiled earlier accounts (Gavin, 1945;Appendix G in Gunn et al., 2000), precalving surveys from 1983and 1993(Heard et al., 1987Buckland et al., 2000), and systematic aerial calving ground surveys in 1986, 1996, and 2006-2010 which were strip transect surveys designed to map the distribution and estimate densities of caribou (Gunn et al., 2000;D. Johnson, unpubl. data). We also include  (Calef & Hubert, 2002;Rescan, 2011). The 2006-10 aerial systematic surveys used transects aligned north-south and on-transect sightings were recorded as sex age classes and allocated to 10 km segments to quantify both dispersion and distribution of caribou on the calving grounds (D. Johnson, unpubl. data; Poole et al., 2013). To compare the 2006 to 2010 systematic calving distribution (D. Johnson, unpubl. data), we created 90% fixed kernel polygons on the un-weighted distribution of breeding females each year and examined overlap between sequential calving grounds; where percent overlap in polygons = (Common area in year i and year i+1)/((∑ area year i and year i+1)/2).
Climate gradients across the calving ground Russell et al. (2013) used a retroactive spatial climate database to create climate descriptions for the seasonal ranges of circumpolar migratory tundra caribou. The database consisted of 22 "caribou-relevant" climate indicators that covered the period 1979-2011. The actual derivations of these 22 variables are presented in Russell et al. (2013). We used this database to compare climate in the three Ahiak calving ground blocks (see next section for a description of the blocks). We tested for differences in climate variables and decadal trends among the blocks using an Analysis of Variance and Duncan's-Waller multiple range tests (SAS, version 9.10). Seasons presented in the analysis were: calving (June), summer (July and August), autumn (September -November), winter (December -March) and spring (April -May).

Queen Maud Gulf calving and rut associations
To examine the relationship between calving and rut associations, we used both individual and population-scale information. The individual scale information was from 74 adult cows fitted with conventional satellite or GPS collars and we used those data to describe the dispersion of individuals and their association. The various projects fitting satellite collars on adult female caribou had different objectives and, with the exception of 2007, all caribou were collared on winter range (Gunn et al., 2000;Gunn & D'Hont, 2002;Gunn et al., 2013). In July 2007, the collaring was conducted on the post-calving range (Johnson & Fleck, 2009). We acquired the data from the Government of Northwest Territories through their Wildlife Management Information System (WMIS). We accessed the calving and rutting locations of caribou cows collared near Baker Lake in 2008 and 2009 from Gebauer et al. (2011; available on the Nunavut Impact Review Board's Public Registry; accessed December 2011). We compiled a database using collared cows with at least one calving location on Queen Maud Gulf coastal calving ground 1996-2010. We also included those cows with a history of calving on the traditional Beverly calving grounds (south of Garry Lakes) but which had shifted to the Queen Maud Gulf coastal calving ground (Nagy et al., 2011;Gunn et al., 2012). Calving date was based on examination of daily movement rates coupled with spatial movements in GIS, such that calving was determined from a rapid drop in daily movement rate (generally to <2 km/day) and localization, followed by a 5-10 day period of reduced movement (Fancy & Whitten, 1991). The rut and winter distributions were indexed from collar locations on 20 October and 1 February each year, respectively (Gunn et al., 2000).
For the population-scale analyses, Poole & Nishi (unpubl. data) had divided the calving ground into three blocks based on a qualitative assessment of caribou density observed during systematic aerial surveys from 2006-10 (GNWT ENR WMIS): Ahiak West (AHW) -from the western boundary of calving as far east as the Armark River (roughly the eastern boundary of the high or medium density zones in the past); Ahiak East (AHE) -from Armark River east to just east of McNaughton Lake; and Ahiak Adelaide (AHAD) -Adelaide Peninsula to the western shore of Chantrey Inlet (Fig. 2). The east-west width of the blocks varied: AHW was 235 km wide; AWE 120 km and AHAD 85 km. We used these three blocks to compile descriptive statistics on the consecutive calving locations to describe inter-annual patterns of calving location fidelity for individual calving cows and distances to rut and winter locations. We tabulated the distances between consecutive annual calving locations to describe the frequency of distance classes between consecutive calving year i to year i + 1. For some analyses, we combined the blocks as two calving associations (see below).
To compare whether cows that calved together were also associated during the rut based on the locations of collared cows, we combined AHE with AHAD to increase sample sizes. This gave us two sets of calving locations which we termed calving associations, the western and central/eastern. First we described the dispersion of the collared cows during the rut relative to each other. For each year, we calculated for the two calving associations and the combined rut locations the mean (average X and Y coordinate) centroid and the mean distances (and standard deviations) from the centroids to the individual caribou locations; the distances be-tween centroids for the calving associations and combined rut locations were also calculated. Distances calculations were performed using in-house developed programs that use MapInfo Professional libraries (Pitney Bowes Software, Troy, NY).
Our second approach to compare whether cows that calved together were associated during the rut was to determine whether the overlap in the area used by the two rutting associations was due to chance or were in disagreement. To describe the area used, we calculated minimum convex polygons (MCP) buffered with one standard deviation for the rut locations based on the associations of individual cows during calving. Mapping was done in MapInfo Professional software. We compared the two rut MCPs to test if their differences are due to 'chance' or disagreement using the Kappa Index of Agreement (KIA) (http:// www.spatialanalysisonline.com/output/). We used Idrisi GIS software (Clark Labs, Worcester, MA) to calculate the KIA for all pairings of the calving associations (by year) and combined rut locations and MCPs of the western and central/eastern calving associations. To judge the strength of the agreement, we used the classification by Landis & Kock (1977) (< 0 -poor; 0 to 0.2 -slight; 0.21 to 0.4 -fair; 0.41 to 0.6 -moderate; 0.61 to 0.8 -substantial; 0.81 to 1.0 -almost perfect).
For the spatial analysis of rut association, we removed three outliers as their rut distribution was highly distant (>400 km) from the other collared cows. We had no way of knowing the reproductive status, physical condition or health of these caribou which are possible reasons for the geographic isolation of single cows (or to the extent they were with other uncollared cows).

Distribution and abundance of caribou calving in the Queen Maud Gulf
Caribou numbers were low in the Queen Maud Gulf coastal area and Adelaide Peninsula in the 1920s and increasing by the 1960s although population estimates were not obtained (Gavin, 1945;Banfield, 1950;Kelsall, 1968;Pelly, 2000;Gunn et al., 2000). The rate of increase between the 1950s and 1980s was qualitative until the first systematic estimates in 1986 and 1996, which estimated that the herd increased 2-3 fold. The mean density of caribou on survey in 1996 was almost three times higher than in 1986 ( veys indicated no apparent trend in the density of caribou on the calving ground ( Table 2). The lower densities in June 2008 may have reflected an unusual year as pregnancy rates were low and not all cows may have reached the calving ground or were later than the timing of the survey (D. Johnson, unpubl. data). Gavin (1945) describes caribou calving dispersed on the numerous islands and islets along the shallow coast of Queen Maud Gulf. In June 1985 and 1986, cows and newborn calves were seen on the coastal islands (Gunn et al., 2000;A. Gunn, unpubl. field notes). However, gregarious calving was observed on the adjacent mainland coast and during a systematic transect survey in June 1986, calving density was measured and distribution mapped with the western boundary of calving cows being the Simpson River (Gunn et al., 2000).
By June 1995, the distribution of cows and calves extended to west of the Perry River and north of MacAlpine Lake (Gunn, 1996). The June 1995 survey overlapped a narrow survey area south of Kent Peninsula where baseline information on caribou distribution was collect-ed from 1994 to 2002 for the Hope Bay mine project (Calef & Hubert, 2002). In June 1996, east-west systematic transects east of Bathurst Inlet mapped calving distribution east to the coast of Chantrey Inlet (Gunn et al., 2000). The southern boundary of the calving ground was not delineated. Compared to 1986, the 1996 calving distribution had extended east and west of the 1986 distribution and was continuous across the width of the Queen Maud Gulf coastal area from Chantrey Inlet west to the Hope Bay area east of Bathurst Inlet (Gunn et al., 1997;Gunn et al., 2000) (Fig. 3).
Additionally, the basis for assuming that the 1986 calving ground had extended west was Inuit observations of pre-calving cows heading east of Bathurst Inlet (Gunn et al., 2000). The capture sites east of Bathurst Inlet and movements of the collared caribou in April-June 1996 also supported an eastward pre-calving migration toward coastal Queen Maud Gulf. The calving locations for both 1996 and 1997 revealed use of the length of coastal calving grounds between the Ellice and Simpson rivers (Gunn et al., 2000). Between 1996 and2006, Fig. 3. The approximate extent of the Ahiak herd's calving distribution in June 1986 and 1996 (Gunn et al., 2000), and 1996-97 calving locations from collared cows. the western boundary of calving distribution contracted from approximately the area just east of the Hope Bay-Spyder Lake corridor in 1996 to the vicinity of the Ellice River in 2006 (Gunn et al., 2000;D. Johnson, unpubl. data). The comparison of the 2006-10 calving distributions along the Queen Maud Gulf showed an average 82% overlap (range 81-84%) during 2006-10, with changes among years primarily at eastern and western ends rather than changes in width (Fig. 4).

Climate gradients across the calving ground
There was a pattern in both winter and spring snow depth among blocks. In winter the AHE block had consistently highest snow 40.5 ± 2.1 (SE) cm and AHW consistently lowest snow depths (34.6 ± 2.0 cm). For the spring period leading up to calving the AHE block had significantly deeper snow than AHW (41.9 ± 2.7 cm versus 34.2 ± 2.2 cm, respectively) in every decade and higher than AHAD (35.3 ± 1.9 cm) in the 1990s. In the 2000s decade, this difference between AHE and AHW extended into the calving period (Table 3). Snow density in spring was consistently higher in AHE than AHW in all decades, significantly so in the 1980s and 1990s (0.272 ± 0.007 gm•cm-3 versus 0.254 ± 0.008 gm•cm-3, respectively).
If we consider decadal differences (Table 4), spring snow depths were consistently higher in the 1990s while the 1980s had the lowest spring snow depths. Spring snow depths were significantly higher in the 1990s compared to the 2000s in AHAD (40.6 ± 2.6 cm versus 35.3 ± 1.9 cm, respectively) and AHE (45.9 ± 2.7 cm versus 38.1 ± 2.0 cm, respectively). For all three blocks, drought index was higher in the 1990s compared to the other 2 decades during the calving (4.6 ± 0.91 versus 2.5 ± 0.46, respectively) and summer (19.3 ± 5.5 versus 9.7 ± 2.4, respectively) seasons. We did find differences in summer precipitation in the AHW and AHE blocks. In the AHE the mean daily precipitation in summer was significantly higher in the 1980s compared to the 2000s (46.6 ± 6.4 cumulative mm. month versus 36.6 ± 3.6, respectively) while the AHW had significantly higher summer precipitation in the 1980s compared to both the 1990s and the 2000s (46.3 ± 6.5 versus 35.1 ± 4.6, respectively). Further during the calving period in AHW the 2000s were the wettest decade, significantly higher than the 1990s and the 1980s (46.6 ± 6.4 ver-sus 36.6 ± 3.6, respectively). Although there was a decadal and spatial pattern in respect to growing degree days (GDD) to July 15th, the relatively high annual variation meant that differences between decades and among blocks were not significant. Comparing decades, the 1990s experienced the highest GDD, while the 2000s were the coolest. Among blocks the AHW was consistently warmer (higher GDD) in every decade while the AHE had the lowest GDD in all decades. Mean GDD to July 15th were 388 ± 94.5 (AHW), 356 ± 76.9 (AHAD) and 345 ± 80.2 (AHE).  excluded six cows (8.1% -6/74 cows; 5.7% -6/105 collar-years) which calved on the Ahiak calving ground and switched to either the north east mainland (NEM) or Bathurst (BA) calving ground (1 cow AHW -BA -BA; one cow AHE -AHE -NEM -AHAD -AHE; two cows AHAD -NEM and one cow NEM -AHE). However, we included the eight cows whose initial known calving ground was the traditional Beverly ground and which subsequently calved on the Ahiak calving ground for at least 1 year. Agnico-Eagle Mines Limited and Government of Nunavut (M. Campbell, pers. comm.) deployed nine collars near Baker Lake, Nunavut in May 2008 and together with support from AREVA Ltd., deployed 21 collars on the tundra north of Baker Lake were in November 2009 (Gebauer et al., 2011). Four collared cows calved on Adelaide Peninsula in 2008; two in 2009 and five in 2010 calved south of Adelaide Peninsula (Gebauer et al., 2011). The other cows calved northeast of Chantrey Inlet within previously recorded calving distributions (Gunn & Fournier, 2000a;. We had only 1 year of calving locations for 18 cows on the Ahiak herd's calving ground and 15 of those (83%) calved within AHW. Of the 49 cows with 2 or more years of calving locations, 27 cows (55%) did not switch among the three blocks covering the length of the Ahiak calving ground and 24 (89%) of those cows calved in AHW. Conversely, 22 of 49 cows (45%) with more than 1 year's calving locations switched among blocks. Of these 22 cows that switched blocks, 13 of 14 (93%) cows with their initial calving location on the Ahiak calving ground calved 1 year in AHW then shifted to AHE or AHAD. Most shifts were west to east but two  cows did shift from AHE or AHAD to AHW. Eight cows which initially calved on the traditional Beverly calving ground switched to the Ahiak calving ground: five of those cows with 2 or more years calving locations on the Ahiak calving ground switched among the blocks.

Queen Maud Gulf calving and rut associations
To test whether individual breeding females would show inter-annual fidelity to discrete areas within annual calving grounds observed in consecutive years, we examined the frequency of distance classes between consecutive calving year 1 to year 2; 72 paired calving locations. We restricted the period to 2006-10 as annual sample sizes were largest. Most (82%) were 25-150 km apart and the 10% tail end of the frequencies were separated by >150 km, which suggests use of the extent of the calving ground (Fig. 5).

Frequency of distance classes between consecutive calving
Frequency calving years

25-km categories
Rut distribution relative to calving associations We predicted that cows that calved together on an annual calving ground (at the scale of the east or west halves of the calving ground), will not solely associate together during the subsequent fall rut. Thus, within a year, collared breeding females that calved within i) the western portion of the calving ground, or ii) the combined central/eastern portions would not be associated with other cows that calved within the same area of the coastal calving ground during the subsequent fall rut (temporal index is 20 October) at a rate greater than would be expected by chance (i.e., >50% probability). Our results were that the degree of dispersion of the collared cows during the rut (20 October) as measured by the mean distance of individual locations relative to the centroid for the two calving associations did not differ annually between the two calving associations (western: x = 118 ± 8.8 SE; central/eastern: x = 136 ± 14.3; t = 1.98, df = 108, P = 0.29; Table 6; Fig. 6). In 2007 the dispersion was less than the other years (individual locations were more tightly clustered around the centroid) for both calving associations. In 2010, the individual locations were more clustered for cows from the western calving association).
The distribution of the rut varied between years based on the geographic location and respective position of the centroids of western and central/eastern calving associations to all rut locations (Fig. 6). The distribution suggested that the centroids of western calving associations during the rut are shifted to the southwest and central/eastern calving associations to the northeast (centres of activity ranging 117-276 km apart), except in 2007 when the cows from both the east and west calving associations had least dispersion (centroids ~32 km) and were furthest to the southwest (Fig. 6, Table 7). The straight-line distances between the centroids of the rut distribution based on the two calving associations varied among years (33 -276 km). The greatest distance between the centroids was in 2008 (276 km) which was likely an atypical year as pregnancy rates were low during 2007-08 winter (D. Johnson, unpubl. data  Between years, cows from the two calving associations overlapped to varying degrees during the rut (2006-10; Fig. 7, Table 7). In 2007 and 2009, the agreement between the two calving associations and rut overlap was high to moderate suggesting that cows which calved together were together (in closer association) during the rut. However, the agreement was slight in 2008 and fair in 2010 (Table 7).
Calving distribution relative to the tree-line To describe the pre-calving migration distance travelled across tundra relative to tree-line, we measured the distance from calving locations to the closest point on both the northern and southern edges of forest-tundra biome (Timoney et al., 1992;Payette et al., 2001). The mean distances differed significantly among calving blocks (Table 8; PROC GLM, F > 43.0, df = 2,146, P < 0.0001; Duncan's multiple range test, P < 0.05). For both the distance to northern and southern edges of the forest-tundra biome (Table 8), caribou calving in the western part of the Ahiak calving ground (AHW) were closer than caribou from the eastern portion (AHE), which was closer than those calving locations on the Adelaide Peninsula (AHAD). Most distances to the northern edge of tree-line were to the tongue of the forest-tundra biome that follows the Thelon River north to Aberdeen Lake (Fig. 1). Caribou that calved on the Adelaide Peninsula would have travelled an average 31% greater distance if they had wintered within the boreal forest compared with caribou that calving in the AHW calving block.
Winter distribution relative to the tree-line Caribou wintered mostly on the tundra and forest-tundra biome. In total, 57% of February locations were on tundra; 24 % in forest-tundra biome, and 18% of locations were in the boreal forest (south of forest-tundra biome). The percentages changed with the destination of precalving migration as 52%, 64% and 84% of the cows calving in AHW, AHE, and AHAD, respectively, wintered on the tundra. For caribou wintering in the boreal forest, 21%; 14% and 0% of the caribou returned to calve in AHW,   1997-2010, to the north (N) and south (S) of forest -tundra biome (Timoney et al,. 1992, Payette et al., 2001. AHE, and AHAD, respectively. The pre-calving migration straight-line distance (distance between winter location and calving) differed among calving blocks (PROC GLM, F = 3.17, df = 2.95, P = 0.047) and was similar (Duncan's multiple range test, P < 0.05) for AHW (525 ± 17.4 km; n = 75) and AHE (513 ± 60 km; n = 14), but significantly shorter for AHAD (397 ± 32.7 km; n = 11) (Table 9). Individual cows did not show high fidelity to wintering within tundra, tree-line, or taiga. Nearly three quarters (73%) of cows with ≥2 years of wintering locations switched among blocks (n = 33). Similarly, 68% (n = 47) sequential pairs of wintering areas switched between years.
Population-scale: Calving distribution based on 2006-10 calving ground surveys To test our prediction that the distribution of calving cows would be discontinuous across the calving ground, we used the 2006-10 aerial systematic surveys with transects spaced at 10 or 20 km intervals (D. Johnson, unpubl. data; Poole et al., 2013). We examined for a pattern of discontinuity along the east-west axis in the distribution of breeding females and newborn calves (measured at a grid scale interval of 10 km).The distribution data do not reveal any obvious spatial breaks in the distribution of cows and calves along the east-west axis of the coastal calving area. There were no north-south tran- sects with no breeding cows and/or calves, or only bulls and juveniles (Fig. 8).

Discussion
Changing abundance and geography (the long relatively straight coastline) influenced the dispersion and distribution of the Ahiak herd's calving. When abundance was low, caribou calved dispersed across the many islands along the Queen Maud Gulf coast. As abundance increased by the mid-1980s through mid-1990s, calving shifted to gregarious calving on a long narrow coastal calving ground. The main elongation of the calving ground was between 1986 and 1995 -a period when densities on the calving ground increased three-fold. The tree-line shifts south as width of the Hudson Bay coastal plain increases, and its orientation relative to the elongated calving ground increased the likelihood that cows calving on the eastern part had wintered on the tundra. However, individual cows annually varied as to whether they wintered in the boreal forest or tundra. We did not find support for our four predictions about the use patterns of the calving grounds at the individual scale (based on satellite-collared cows) and at the populationscale (using information from aerial surveys of calving distribution). Thus our hypothesis (the Ahiak calving grounds are comprised of two or more calving sub-populations) was not supported. We had projected that the climate gradients across the elongated calving ground and its location relative to whether cows were more likely to winter on the tundra or boreal forest would affect survival and productivity and lead to sub-population structure.
At the population scale, we found no evidence that the distribution of calving cows or non-breeders was discontinuous in any one year although there were variations in density. For example, Poole et al. (2013) applied an In-verse Distance Weighted (IDW) interpolation to map the spatial pattern of density clumping using 2007 (Fig. 9).
The areas of higher density of breeding cows were connected by areas with lower density of cows and calves which coincides with the area of deeper snow in winter and spring. This pattern of spatial variation in the high densities is similar to, for example the Central Arctic herd, which has two areas of higher calving densities linked by areas of lower density and the cows and calves mingle during the other seasons of the year (Arthur et al., 2009).
At the individual scale of satellite-collared cows, while most collared cows used the western third of the calving ground, almost half the individual cows shifted their annual calving locations within the length of the calving ground. Satellite-collared cows annually varied in the degree to which they associated during the rut and there was no clear-cut pattern of cows calving in one half of the calving ground remaining as a consistent (inter-annual) association during the rut. The cows mainly wintered on  the tundra and within the forest-tundra biome, which partly reflects the geographical extent of the tundra. Individual cows varied among years as to whether they wintered in the forest-tundra biome or above tree-line on the tundra. We used three methods to estimate calving relative to rut associations. The overlap of the individual rut locations have a stronger pattern of clustering of the calving associations during the rut than the distances between the centers of the rut distribution. This is because MCPs do not consider the dispersion of constituent caribou locations that lie within them (Nilsen et al., 2008), while the mean distances that form the centroids are sensitive to the dispersion of the individual locations. The dispersion of individual locations likely has both process (biological) and measurement error (sample size). Inter-annual comparisons are limited because the membership of the calving associations var-ied among years (i.e., the scale of calving spatial fidelity was not at the block scale). The overall results suggest that for 2006-10, cows that calve together also rut together. Low sample size and uncertainty about how the collared cows represent the seasonal distribution of all caribou at finer geographical scales are possible limitations of our conclusions.
The level of information available on the Ahiak herd over time varied, with the least amount of information for the earlier years. The frequency of aerial surveys of the coastal calving area to describe the calving distribution and abundance of breeding females was uneven over time. Similarly, the annual sample of collared cows occurred after the time when abundance was increasing  and was low during 2001-06, which weakens inferences that can be drawn from the telemetry data. We also acknowledge that we have described the caribou dispersion and distribution over a 25 year period when abundance of the Ahiak and neighbouring herds changed.
Changing abundance in neighbouring herds may play a role in the likelihood of sub-population structure. An extreme decline in the neighbouring Beverly herd (J. Adamczewski, unpubl. data) may have caused a loss of gregarious calving on the traditional Beverly range, and subsequent movement of cows to the Ahiak herd although documentation of this Fig.9. Distribution of breeding female caribou observed during the Ahiak survey flown in June 2007. Relative density conducted using Inverse Distance Weighting (IDW) mapping interpolation using a 10 x 10 km pixel, exponent of 2, and a 15 km search radius (Poole et al., 2013: Fig. 23). through satellite-collared cows only started in 2007 (Gunn et al., 2012). Alternatively, Nagy et al. (2011: Fig. 3, p. 234) proposed that the Queen Maud Gulf (i.e., Ahiak) herd's long narrow calving ground could be the consequence of adjacent calving by two different sub-populations: the Beverly herd shifting from its traditional calving ground in the mid-1990s to the western part of Queen Maud Gulf and the caribou calving on the eastern Queen Maud Gulf coast in the 1980s. However, we could not infer sub-population structure for the coastal calving area based on the currently available evidence.
Another possibility for both a historical influence on the Ahiak calving ground and future sub-population structure is the caribou calving on Adelaide Peninsula. Until about the 1920s and 1930s, caribou migrated across Adelaide Peninsula in May and crossed to King William Island where they calved and summered before returning to Adelaide Peninsula in the fall (summarized in Appendix G of Gunn et al., 2000). In June 1976, a few cow-calf pairs were counted south of Adelaide Peninsula (although the flight-lines did not extend further west; Fischer et al., 1977). It is speculative whether the 1976 calving distribution shifted northwest (to the area mapped as calving in 1986), northeast of Chantrey Inlet, or disappeared. The 1976 calving area was surveyed in 1986 and no caribou were seen (Gunn & Fournier, 2000a). The possibility that calving shifted northeast of Chantrey Inlet is suggested from observations of calving caribou recorded in 1975, 1985, 1986, 1989and 1991(Gunn & Fournier, 2000a and from seasonal movements of satellite collared cows in 1991-93 (Gunn & Fournier, 2000b). The four satellite-collared cows in 1991-93 showed that caribou wintering south of the Boothia Peninsula toward Baker Lake would calve northeast of Chantrey Inlet (and not Adelaide Peninsula) (Gunn & Fournier, 2000b).
The caribou which calved on Adelaide Pen-insula mostly wintered on the tundra and reduced the length of their pre-calving migration by just over 100 km relative to the caribou which calved west of Adelaide Peninsula; caribou calving west of Adelaide Peninsula were more likely to winter within the closer forest and forest-tundra biome. We suggest that the shorter pre-calving migration distance for caribou that calved on Adelaide Peninsula may be from an energetic trade-off between the costs of foraging on the tundra, travel, and predation risk (Couturier et al., 2010). Currently, the high proportion of cows shifting calving locations along the length of the calving ground among years and the switching of cows wintering on the tundra or taiga/foresttundra biome suggests a high degree of plasticity -the ability to change biology or behavior to respond to changes in the environment -in individual behaviour within a population (for example, Couturier et al., 2010). However, variance in inter-annual selection of calving and wintering areas by individual adult females may be reduced if changing environmental conditions increase energetic costs and reduce fitness of cows and calves. While we could not infer population sub-structure for the Ahiak herd from the currently available evidence, it is not a future impossibility given the Ahiak herd's particular geographical attributes -the long narrow calving ground and the southeast tilt and width of the tree-line.