Attributes of Woodland Caribou Migration Habitat in West-Central British Columbia

W e examined sites used by 73 caribou located by radio-tracking during spring migration through a forested travel corridor. The physical and botanical characteristics o f these sites were described to find what features (if any) distinguished them from the forest at large. Raised and open aspect, sparse tree cover, free-draining soils, and a simple flora w i th abundant lichens were features c o m m o n to all the sites, but were lacking i n the denser forest around. Scores for these attributes were ordinated by multidimensional scaling o f similarities and differences from site to site. Separate searings for (i) terrain features, (ii) tree cover attributes, and (iii) floristic content each yielded a single cluster o f points w i t h few outliers and no particular pattern or trend. The inference is that the sites conformed to a single type w i th l imited variation. A profile o f the distinguishing characteristics was compiled and then applied to transects through the general migration area by persons unfamiliar w i t h it beforehand. Sites w h i c h matched the profile proved easy to identify, even though they comprised only a small proport ion o f the area. Sites w i th high scores for the most distinctive attributes had more caribou at the time o f radio-tracking. K e y words : habitat descriptors, o r d i n a t i o n , l i chens , r a d i o t r a c k i n g , forest management , Rangifer tarandus Rangifer, Special Issue No. 9,355-364 Introduction T h e T w e e d s m u i r E n t i a k o w o o d l a n d c a r i b o u h e r d contains about 500 animals ( C i c h o w s k i & B a n n e r , 1993). U n t i l recent ly , these w e r e some o f the least s tud ied c a r i b o u i n B r i t i s h C o l u m b i a . I n the m i d 1980s, u s ing r ad io t r ack ing , M a r s h a l l (1986) def ined the range o f the h e r d and its seasonal pat tern o f use. T h e animals s u m m e r i n n o r t h e r n T w e e d s m u i r Pa rk , a n d w i n t e r i n the E n t i a k o drainage south-east o f T e t a c h u c k L a k e (F ig . 1). T h e Che la s l i e R i v e r d ra i nage serves as a c o n n e c t i n g travel route , and i n certa in years part o f the h e r d m a y w i n t e r w i t h i n that route , rather than c o n t i n u i n g o n to the E n t i a k o (Steventon , 1994). C i c h o w s k i (1989) descr ibed habitat preferences o n the E n t i a k o range, bu t site se lect ion d u r i n g m i g r a t i o n was no t w e l l unders tood . A s l o g g i n g and o ther l and uses beg in to encroach o n the range o f this herd , the need to k n o w m o r e about habitat usage was b e c o m i n g urgent. W e e x a m i n e d a sample o f sites used b y the h e r d d u r i n g spr ing m i g r a t i o n i n 1993, and c o m p a r e d the i r phys ica l and bo t an i ca l characteristics. T h e a i m was to find w h a t features, i f any, w e r e shared amongst these sites, a n d d i s t inguished t h e m f r o m o ther parts o f the t ravel c o r r i d o r . T h e s tudy was d o n e as part o f a forest m a n a g e m e n t p l a n for the area.


Introduction
The Tweedsmuir-Entiako woodland caribou herd contains about 500 animals (Cichowski & Banner, 1993).Until recently, these were some of the least studied caribou in British Columbia.In the mid-1980s, using radio-tracking, Marshall (1986) defined the range of the herd and its seasonal pattern of use.
The animals summer in northern Tweedsmuir Park, and winter in the Entiako drainage south-east of Tetachuck Lake (Fig. 1).The Chelaslie River drainage serves as a connecting travel route, and in certain years part of the herd may winter within that route, rather than continuing on to the Entiako (Steventon, 1994).Cichowski (1989) described habitat preferences on the Entiako range, but site selection during migration was not well understood.As logging and other land uses begin to encroach on the range of this herd, the need to know more about habitat usage was becoming urgent.
We examined a sample of sites used by the herd during spring migration in 1993, and compared their physical and botanical characteristics.The aim was to find what features, if any, were shared amongst these sites, and distinguished them from other parts of the travel corridor.The study was done as part of a forest management plan for the area.

Study area
The Chelaslie corridor (61 000 ha) is a gently sloping plateau with undulating terrain, from 850 m elevati- Others which were screened by trees may well have been missed, however, and so this total was a minimum.
When migration ended, the 16 sites were plotted onto forest cover maps and located on the ground by compass.Each was described by terrain, soil, and floristic criteria, with special attention given to the terrestrial lichen flora.Physical and floristic descriptors (Tables 1 and 3) were from the list by Luttmerding et al. (1990) et al. (1992).

Data analysis (sample sites)
The data in Tables 1-3 (for all tables, see Appendix) were converted to standardized scores expressing the similarity or difference in each attribute, site by site.Scores were compiled separately for the three types of attribute (soil and terrain, tree cover, and flora).For soil and terrain: a pair of sites with the same attribute scored 1, or zero if the two sites differed.Summing these scores over all attributes gave the net difference in soil and terrain for that sitepair.Tree attributes were scored the same way as for soil and terrain.For the rest of the flora, the species list for all 16 sites combined was compared to the lists for each site-pair.A species which was present at both sites scored 1. Likewise a species absent from both sites.A species which was present at one but absent at the other scored zero.These scores were then summed to express the net difference in flora, in the same way as for soil/terrain and trees.
For each of Tables 1-3, these steps generated a matrix of net differences for all possible site-pairings (120 pairs).
The three sets of differences were ordinated by monotonic multidimensional scaling (MDS) for indications of clustering or segregation into sitegroups, or evidence of a linear or other gradational trend.Scaling was done first in two dimensions and then in three (where each dimension represents a different combination of variables in the datasethere, the matrix of net-difference scores among the 16 sites).Dimensions beyond 3 or 4 seldom account for much of the variation in a dataset (Wilkinson, 1990).In the present case, adding the third dimension explained only 5% more of the variation in scores.Therefore, only the first two dimensions are plotted in Figs 2-4.

Transect strips
Attributes which best defined the 16 sample sites were then checked for consistency and occurrence in the area at large.Ten transects (Fig. 1) totalling 65.4 km were run through places having a mixture of merchantable and non-merchantable timber types, so that value as timber and value to caribou could be compared.When a site was found which fit the profile of the sample sites, its size, tree density, crown closure, and lichen cover and moss cover

Number of caribou at the 16 sample sites
The number of caribou seen per site during the radio-tracking period ranged from 1 (5 sites) to 20 (one site).Numbers per site were compared with site attributes singly and in combination, under the assumption that numbers at a site were a measure of caribou use of that site, and that level of use was a reflection of site suitability -specifically, the state of the forest cover and lichen supply.The analysis was therefore confined to six variables: (1) lichen abundance (all spp.), (2) percentage Cladina spp., (3) stem density, (4) tree size (merchantable volume), (5) crown closure, and (6) percentage slash.These were compared by step-down multiple regression.
There was no a priori reason to reduce these variables in any particular sequence, and so they were deleted by size of effect in simple regression, largest first.As well as to find whether caribou numbers did vary with these attributes, the aim was see which of them had most potential for ranking sites in value to caribou.

Sample sites
Each of the 16 sample sites differed distinctly from its surrounds.This was apparent as soon as the site was located on the ground.Their similarity to each other was obvious as well.Raised and open aspect, coarse free-draining soils, sparse tree cover, little or no large woody debris, and a simple ground flora with abundant lichens were characteristics shared by all.Most of the sites were small and narrow (less than 100 m in width), but some extended for more than 1 km atop eskers and glaciofluvial terraces next to sedge meadows and willow swamps.Tables 1-3 show the attributes of each site.Together, these attributes contrasted sharply with the denser forest around.Indeed: just the combined presence of juniper, Arctostaphylos uva-ursi, Shepherdia canadensis, and a Cladina-dominant lichen flora beneath a sparse cover of lodgepole pine, is sufficient to distinguish these sites from all other types within the SBSdk biogeoclimatic subzone (DeLong et ai, 1993).
The results of multidimensional scaling confirmed the similarity of the sites.Separate searings for (i) terrain features, (ii) tree cover attributes, and (iii) ground flora each yielded a single cluster of points with few outliers and no linear or other systematic trend (Figs.2-4).The inference is that the 16 sites conformed to a single type with limited variation.A composite profile of the sites is shown in Table 4.

Numbers of caribou at the sites
Total lichen cover, percentage Cladina, tree density, canopy closure, merchantable volume, and debris cover, together explained nearly 60% of the difference in caribou numbers from site to site (Table 5).degrees of freedom available, the index was not statistically significant.Individually, however, crown closure and merchantable volume had a quite substantial effect: when these two factors were dropped from the index, R 2 fell by more than half.Stem density and lichen cover appeared to have only a small effect, and slash cover none at all, even though caribou tend to avoid dense timber clogged with slash, and lichens are their dietary mainstay.However, all but one of the 16 sites had stem densities less than 2000 per ha, and none of the sites had much slash (Table 1).Lichen cover varies with crown closure and stem density (next Section).

Transect strips
The 10 transects through the wider study area identified 64 sites which fit the general profile in Table 4.
Every transect had at least 2 such sites; the most was Index Fig. 5. Index of site attributes / versus number of caribou seen at the sites.
11 (Table 6).All but one of the 64 sites were in forest cover composed of mature semi-open lodgepole pine, mostly less than 2000 trees per ha -the density used by the BC Forest Service to define a stand of timber which was 'passable' for travel by this caribou herd (Ministry of Forests, 1992).All but three sites had densities less than 1500 trees per ha, and ninety percent had a canopy closure of 25% or less.
Lichen ground coverage ranged from 5% to 50%, with most sites having 5-25%.Moss cover (mostly Plemvzium schreberi) varied inversely with this (Kendall X = -0.57).The sparser the timber and the more open its canopy overhead, the higher was the ground-coverage by lichens underneath (Kendall X = -0.36 and -0.35 respectively) and the lower was the Rangifer, Special Issue No. 9,1996 cover by moss (all X values significant at p <0.001).
Lichen:moss ratios showed no relationship to site size, and so the sites were merely summed rather than weighted by size when calculating the extent of 'lichen habitat' on the transect strips.
The sum of the widths of the 64 sites where they intercepted the transects was 4700 m.This was 7.2% of the total transect length.Most of the sites were roughly 2.5 times longer than wide; hence the sum of the lengths was at least 11 600 m (17.7%, ignoring very elongated sites whose total length was unknown (pursuing these sites to their full length would have required long deviations from the transect route)).By simple division, the mean of these two percentages was 12.5.

Associations with other habitat types
Eleven of the 16 radio-sample sites were within 200 m of treeless lakeshores, meadows, or swamps.Most of these openings were crossed by game-trails with caribou tracks or droppings.Some trails led from one dry lichen site, through a nearby meadow, to another lichen site nearby.Along the 10 transect strips, sites with lichens tended to occur in scattered groups interspersed with low-density timber and meadows or swamps.The pattern as a whole appeared to consist of series or clusters of feeding sites linked by trails through semi-open timber and treeless ground.

Discussion
The 16 sample sites resembled each other and differed distinctly from their surrounds.When being approached through the surrounding timber, most of the sites were readily visible some distance ahead as openings in the forest with little or no understorey and a carpet of lichens on the ground.
Assuming that their extent on the transects reflected proportion by area, sites of this type comprised at least 12.5% of the migration corridorprobably more, since the unknown length of very long sites was ignored.Thus, any one of the 23 radio-collared caribou had at least a 12.5% chance of finding such a site at random.However, the chances of all 23 doing so were less than 1 in 2 21 (ie, 0.125 23 ).This degree of consistency was not likely accidental.(Four sites had more than one radio-collared caribou.If these are treated as just one location per site, the likelihood is still only 1 in 3 14 that caribou found all 16 sites by chance (ie, 0.125 16 )).
State of the tree cover was better than lichen cover at predicting the number of caribou seen at a site.However, the estimates of lichen cover were rather crude, were correlated with the tree cover, and were made after the caribou had already grazed the site.Also, the caribou seen at a site were not necessarily all of the caribou present at the time.
Thus, the coefficients in Fig. 5 do not necessarily express the actual relationship between caribou numbers and site attributes.Nonetheless, a trend was apparent, and was strong enough to suggest that the degree of site usage does vary with differences in these attributes.(The presence of some caribou at a site might have attracted others to it.However, if 'herd behavior' instead of site attributes determined the size of the group, Fig. 5 should not have shown a trend).
The floristic features in Table 3 resemble the 'dry lichen' and mixed 'lichen/moss' vegetation types which Cichowski & Banner (1993) described as important to this caribou herd on the winter range.Among the sites studied by us, lichen cover and moss cover alternated with each other in a gradational way rather than discretely, and the sites as a whole suggested a continuum of differences rather than a set of disjunct types (Fig. 4).Level of use by caribou appears to vary likewise.
After this study had ended, Steventon (1994) plotted the full set of 153 radio-locations onto forest cover maps which rated each part of the area in timber-production terms.He found that the percentage of locations on sites rated 'low' or 'poor' exceeded the extent of these sites on the area.The opposite held for sites rated 'medium' or 'good'.The apparent preference of the caribou for less-productive forest may lower the risk from logging within their range.
On poor dry sites with approximately 2000 stems per ha or less, the timber becomes sparse enough for lichens to occur, but is usually too low in volume for logging to be worthwhile.
Appendix: Tables 1-6 Table 1.Terrain and soil attributes of the sample sites.
duff depth (cm) 1 3.    Rangifer, Special Issue No. 9,1996 Fig. 1.Project area adjoining the northeast portion of Tweedsmuir Park, west-central British Columbia, showing the general spring migration route of the caribou (arrowed band).Redrawn fromMarshall (1986).Circles show the approximate location of 16 habitat sampling sites; zig-zag lines represent 10 habitat survey transects.
were scored in classes of 1 to 4, corresponding to quartiles of the range of values shown by the 16 sample sites.One of us (BM) surveyed 2 of the transects and supervised the other 8, which were done by three people with no previous knowledge of the sample sites or the rest of the area.This allowed us to test the criteria in Table 4 for their ability to indicate habitat sites without needing prior experience or data on caribou themselves.The surveys also allowed us to roughly estimate the extent and distribution of such sites throughout the migration corridor.

Fig. 2 .
Fig. 2. Multidimensional scaling (MDS) of soil and terrain attributes at the 16 sample sites.Dimensions 1 and 2 represent combinations of attribute scores which account for the largest and the next largest amount of variation among the sites.

b
The reduction in explained variance when the attribute was deleted during step-down multiple regression.*p<0.05.

Table 2 .
Tree attributes at the sample sites.Size (diameter) of debris, or height elevated above ground surface (whichever was greater).
d M midslope; U upper slope; C crest.e B L loam; S sand; C clay; Si silt.hH high; M moderate; L low.' X xeromor; H hemimor. J Depth to first mineralized horizon.kLogs, dead branches etc.1

Table 3 .
Plant species content (percentage cover) of the sample sites.
' Species represented by >1 individual at a site.Species occuring as a few scattered individuals were given a score of 1.

Table 4 .
Key-feature summary of the sample sites.

moraine, esker, or mounded blanket till
Hummocky, undulating, ridge or benchSloping or raised, giving an outlook over surroundsVariable, but most often SE, S, SW Boldface type denotes the most characteristic attributes.Number of species with significant ground cover (1% or more).Other definitions: see Tables 1-3 and text. *

Table 5 .
Percentage of variance explained by the site attributes in Fig.5.
" The variance explained by each attribute in simple regression with caribou numbers.