Feeding site selection by woodland caribou in north-central Brit ish Columbia

We examined the foraging habits of the northern woodland caribou ecotype {Rangifer tarandus caribou) at the scale of the individual feeding site. Field data were collected in north-central British Columbia over two winters (Dec 1996-Apr 1998). We trailed caribou and measured vegetation characteristics (species composition and percent cover), snow conditions (depth, density, and hardness), and canopy closure at terrestrial and arboreal feeding sites, and at random sites where feeding had not occurred. Logistic regression was used to determine the attributes of feeding sites that were important to predicting fine scale habitat selection in forested and alpine areas. In the forest, caribou selected feeding sites that had a greater percent cover of Cladina mitis and Cladonia spp, lower snow depths, and a lower percentage of debris and moss. Biomass of Bryoria spp. at the 1-2 m stratum above the snow significantly contributed to predicting what trees caribou chose as arboreal feeding sites. In the alpine, caribou selected feeding sites with a greater percent cover of Cladina mitis, Cladina rangiferina, Cetraria cucullata, Cetraria nivalis, Thamnolia spp., and Stereocaulon alpinum as well as lower snow depths.


Introduction
The habitat requirements of the northern woodland caribou ecotype of British Columbia are largely unknown (Harrison & Surgenor, 1996).This ecotype has been the subject of few studies, but is known to inhabit areas of low to moderate snow depths in low elevation forests, and to forage primarily on terrestrial lichens during winter (Hatler, 1986;Cichowski, 1993;Lance & Mills, 1996;Wood, 1996).Most caribou research in British Columbia has focused on the mountain caribou ecotype which spends little time in low elevation areas during the winter, but forages instead on arboreal lichens at high elevations (Servheen & Lyon, 1989;Terry, 1994).
Further understanding of the life history strategies of the northern woodland caribou ecotype is important in view of increasing demands for timber in the province.Wintering populations of this ecotype use low elevation forests that are valued for commercial wood products (Cichowski, 1993;Wood, 1996).Consequently, they are likely to be negatively affected by habitat alteration, fragmentation, and increased road access.
Rangifer, Special Issue No. 12, 2000 As part of a larger research project to define the processes that affect the movements and distribution of northern woodland caribou across the landscape, we investigated the influence of forage species, abundance, and accessibility on the selection of individual feeding sites during winter.
Specifically, we examined: 1. the influence of snow depth, density, and hardness as well as vegetation composition and abundance on the selection of terrestrial feeding sites at small spatial scales in forested and alpine habitats; and 2. the influence of lichen biomass on the selection of arboreal feeding sites.

Study Area
The group of caribou chosen for this study is known as the Wolverine herd (Heard & Vagt, 1998), and ranges throughout a 5100-km 2 area, approximately 250 km northwest of Prince George, British Columbia (Fig. 1).Terrain varies, from valley bot- krummholz form (MacKinnon et al, 1990;DeLong etal., 1993).

Materials and methods
Field investigations occurred at two to three week intervals between December and April, 1996-1997and 1997-1998 Using a random number table, a maximum of 12 sites were randomly selected on the transect for measurement: 3 sites where there had not been terrestrial feeding, 3 trees where there were no signs of arboreal feeding, and, if present, 3 crateting sites and 3 arboreal feeding sites (Fig. 2).
For statistical analyses, measures at feeding and non-feeding sites wete pooled across ttansects.To minimise the likelihood of recording the behaviour of the same animal more than once (i.e., pseudoreplication; Hurlbert, 1984), we limited the number of samples to not exceed the observed or, where animals were not sighted, the average number of caribou typically occurring within a group during the winter (H=9; Wood, 1996 Rangifer, Special Issue No. 12, 2000 unpubl.).Furthermore, because we wanted to sample all collared animals and visit as many geographically unique locations as possible, we restricted the maximum number of 100-m transects sampled at one location to 3, regardless of the number of animals observed.To further reduce the effects of spatial autocorrelation and allow an opportunity for changes in behaviour across space, and presumably time, successive transects were separated by a distance of 100 m.Therefore, at a location, we sampled a maximum of 9 terrestrial and 9 arboreal feeding sites and the 18 associated random sites across 3 transects.
In the alpine, safety concerns and the aggregated distribution of the feeding sites required us to use a 50 X 50-m quadrat rather than a 100-m segment of track.All craters in the quadrat were counted, and we randomly selected 3 to 6 craters for measurements, depending on time and weather constraints.
The corresponding non-feeding sites were located at a random compass bearing and random number of paces (1-20 paces) from rhe sampled craters, regardless of the quadrat boundaries.For statistical analyses, measures at feeding and non-feeding sites were pooled across quadrats.
At all tetrestrial feeding and random sites, snow depth was measured to the nearest 0.5 cm, and the penetrability (i.e., hardness) of the upper layer was esrimated with an instrument of out own design which was similar to the Rammsonde penetrometer.
A British Columbia Ministry of Environment, Lands, and Parks (1981) Snow Survey Sampling Kit was used to measure snow density by inserting a cylinder of known volume vertically into the snow, recording the depth minus the soil plug, and weighing the contents.Because the scale used to measure the mass of the cored snow is insensitive at low snow depths, density could not be reliably calculated for alpine sites.For cratered sites, the least disturbed edges were used for sampling.Following snow measurements, the snow was cleared and the percent cover of ground vegetation was assessed with a 0.5 m X 0.5-m point frame consisting of 16 vertical pins (Bookhout, 1994).Lichen and moss were identified to species, genus or morphological group, depending on ease and reliability of field classification.Species that occurred at fewer than 10 sample sites were pooled with the next most similar species or genus group, or were excluded.Percent cover of evergreen dwarf shrubs, grasses (Poaceae), and sedges were also recorded.However, with the exception of grass at alpine sites, there was no evi-dence of grazing on those plant types, so they were excluded from the analysis.At forested terrestrial sites, a moosehorn coverscope (Moosehorn Coverscopes, Medford, Oregon, USA) was used to assess percent canopy closure by raking one measure directly above each sampled site.
At each arboreal feeding and associated random site, a lichen clump (Bryoria spp.) with a predetermined oven-dried weight was used as a Standard Lichen Unit to visually estimate arboreal lichen biomass (Antifeau, 1987;Stevenson & Enns, 1993).
The number of similar Units that occurred within the reach of a typical caribou (1-2 m above the snow) was counted and multiplied by the mass of the Standard Lichen Unit to obtain total biomass within the 1-2-m stratum.Tree species and diameter at breast height were also recorded.
We used multiple logistic regression analyses to estimate the influence of percent cover of vegetation, snow conditions, and canopy closure on the selection of terrestrial feeding sites by caribou in foresred and alpine areas.To assess the selection of atboreal feeding sites, we tested a simple logistic regression model, consisting of foraged versus random trees as the dependent variable and grams of arboreal lichen in the 1-2 m stratum as the independent variable.
For the multiple logistic regression models (terrestrial forest and alpine), the Wald backward eliminarion procedure (SPSS Version 8.0) was used to idenrify the most parsimonious model for describing site selection of cratering locations (Menard, 1995).As recommended by Bendel & Afifi (1977), the a of 0.05 was relaxed to 0.15 during the backward elimination procedures to reduce the likelihood of excluding important variables.We used Pearson correlation values and tolerance scores with a collinearity threshold of 0.20 (Menard, 1995) to diagnose the presence of multicollinearity amongst the independent variables.Collineariry is the product of two or more highly correlared variables.It is an indication of redundancy within the statistical model and can lead to inflated error terms and in extreme cases render matrix inversion unstable (Tabachnick & Fidell, 1996).Although logistic regression is robust to most multivariate assumptions, data and model screening procedures were employed as recommended by Menard (1995) and Tabachnick & Fidell (1996); procedures were reported only if model validity was threatened.
For both terrestrial and arboreal feeding sites, we used the proportional reduction in the Jj statistic

162
(R 2 l) to indicate how much the inclusion of each significant explanatory variable improved model fit; the higher the value, the better the measured variables explain the differences between selected and random sites (i.e., analogous to the linear regression R 2 ) (Hosmer & Lemeshow, 1989).Odds ratios were used to interpret the effect of each explanatory variable on the response variable and are more intuitive than the regression coefficient when discussing the relative strength of each explanatory variable.
Univariate logistic function plots were used to graphically present the relationships between statistically significant vegetation, debris, and snow variables and the predicted probability of a caribou selecting a feeding site (Tabachnick & Fidell, 1996).
To provide a relative measure of the availability of forage species, we used Bonferroni corrected 95% confidence intervals to test differences in mean percent cover of lichen, mosses, grass, and debris between feeding and random sites, and among species (Neter et al., 1990).The relationship between tree diameter at breast height and amount of arboreal lichen was investigated with a simple linear regression equation.An (X of 0.05 was used for all tests of statistical significance.

Results
Over the two winters we examined caribou feeding sites along 85 forest transects and 23 alpine quadrats (Fig. 1).We sampled 461 terrestrial (206 feeding, 255 random) and 356 arboreal (102 feed- Rangifer, Special Issue No. 12, 2000 Table 1.Lichen and moss species and groups identified at terrestrial feeding and random sites; classification is based on ease and reliability of field identification, and frequency of occurrence in north-central British Columbia (Dec 1996-Apr 1998).

Feeding Sites in Forest Locations
Average snow depths at cratered sites ranged from 23-97 cm and at random sites from 27-102 cm.
Average snow hardness at cratered and random sites ranged from 0.27-3.19g/cm 2 and 0.25-4.2g/cm 2 of the between feeding site variation (Table 2).Snow depth, percent cover of debris, C. mitis, Cladonia spp., and the two moss classes significantly contributed to the statistical differentiation of cratered and random sites (Fig. 4).Cladonia spp.
had the highest odds ratio at +4.3% and the greatest influence on the selection of cratering sites by caribou (Table 2, Fig. 4).Snow depth had the least influence on selection of a feeding site; in this case, the odds ratio implies that a 1 cm increase in snow depth will reduce the likelihood that a caribou will crater by 2% (Table 2, Fig. 4).
Although tolerance scores for each variable in the schreberi (r=-0.370).This indicates that there may not be a direct relationship between site selection and the presence or absence of these moss and lichen species.
When choosing to browse arboreal lichen, caribou selected those trees with a greater biomass of Bryoria spp.than found in randomly available trees.
On average, selected trees had 4.9 g (±0.74) in the 1-2 m stratum versus 2.3 g (±0.24) for random trees.Pinus contorta was the dominant tree species at both selected (81%) and random sites (90%).The simple logistic regression model indicated that the amount of Bryoria spp. was a meaningful predictor of what trees caribou chose to browse (Table 2, Fig. 5).The model accounted for only a small amount of the variation berween feeding and random sires (R 2 L=0.039); however, 72.2% of the cases were correctly classified as feeding or random sites.The odds ratio indicated that a 1 g increase in the amount of Bryoria spp.would increase the likelihood of a caribou foraging by 9.9%.There was a significant, but weak linear relationship between tree diameter and arboreal lichen abundance (F= 17.495, df=250, P<0.001,7v 2 =0.066).

Feeding Sites in Alpine Locations
Average snow depth per quadrat ranged from 3-37 and 0-69 cm, and snow hardness between 0.54¬ 28.89 and 0-30.38 g/cm 2 for cratered and random sites, respectively.Percent cover of lichen classes was typically greater at cratered sites, but not sig-  Cladina rangiferina demonstrating the largest differences (Fig. 6).Debris was the only variable to illustrate a significant difference in percent cover, being more prominent at random (mean=37.3%±3.30)than cratered sites (mean=20.0%±1.99).
The multiple logistic regression model used to describe site selection of terrestrial feeding sites in the alpine accounted for 31% of the between site variation, and correctly classified 76.5% of the cratered and random sites (Table 3).Statistically
Furthermore, the lack of comparison control sites has frequently resulted in the analysis of forage availability as opposed to selection by the animals.
We attempted to improve upon these studies by investigating the influence of lichen species in combination with the limiting effects of snow on the fine scale selection of feeding sites in forested and alpine areas.

Selection of Feeding Sites by Caribou
Using data collected over two years across a broad geographic area, we developed statistically significant models to predict the selection by woodland caribou of terrestrial and arboreal feeding sites in forested locations, and terrestrial feeding sites in alpine areas.All three of the models had relatively low explanatory power (iv 2 L) indicating that the independent variables (i.e., ground cover and snow condition) captured only a small proportion of the differences between selected and random sites.We believe that this is a consequence of four sources of Second, although we are confident in our ability to identify feeding sites, it is possible that some sites were incorrectly classified.Caribou may have cratered but not fed at certain terrestrial sites, or trees may have been incorrectly classified as browsed when they were not.Sampling error also may have been introduced by classifying our random sites as non-selected sites when in actuality caribou did not make a choice, but passed by that location for reasons not directly related to a foraging decision (e.g., satiation, minor disturbance).Furthermore, because caribou remain in an area for some period of time, our random samples may contain a proportion of sites that would have been cratered at a later date.
To reduce this source of error, we should have chosen random sites where it could be confirmed that a caribou had made a decision not to crater, such as unexcavated sniffing holes (e.g., Helle, 1984).
Because snow conditions often made the identification of sniffing sites difficult, this approach was abandoned in favour of sampling random locations along the tracks.
Third, we assumed that the lichens remaining at a sampled feeding site were representative of the pte-cratering lichen cover, although the foraging and digging actions of caribou may have resulted in our underestimating the percent cover of lichen at feeding sites.To minimise this bias, we chose percent cover, as opposed to biomass, as our measure of relative lichen availability.Caribou rarely cropped the entire lichen thalus, thus using a point ftame with 6.25% increments we were able to accurately and precisely measure percent cover by species at feeding sites.
Rangifer, Special Issue No. 12, 2000 Fourth, selection strategies of the caribou may have changed during or between winters, confounding the importance of individual variables.For example, nutritional requirements may vaty over time or abundance of lichen species may vary spatially, resulting in tempotally variable selection patterns.This, and the sources of error listed above did not invalidate our results, but rather forced us to test a more conservative model (which may have decreased the likelihood of obtaining significant differences).

Influence of Vegetation on Feeding Site Selection
Numerous conclusions, in some cases contradictory, have been reported by researchers using field studies We also recognise that there may be inter-population variability, but feel that our results placed in the cor text of other works add to the understanding of the similarities and plasticity in foraging habits of these animals.
Our data indicate that northern woodland caribou select cratering sites based on the percent cover of several lichen species.In most cases our results agree with other studies.For example, C. mitis is commonly reported as being preferred or selected by caribou and reindeer (Helle & Saastamoinen, 1979;Helle, 1984;Lance & Mills, 1996) Through our conclusions we do not infer causal relationships between feeding site selection and the importance of individual lichen and moss species.
We emphasise this caveat because of the high correlations between several of the significant lichen and moss species.For example, where the model shows a strong effect for lichen and mosses at forested sites, caribou may be selecting for lichens or may be avoiding mosses; the statistical importance of one may be the product of the presence or absence of the other.Pleurozium schreberi may be an important discriminating variable only because it occurs where C. mitis and Cladonia spp.are not found, not because caribou avoid sites where it is found.High negative correlations likely occur because these species of moss and lichen have distinct light and moisture requirements and, therefore, grow in different locations (Robinson et al, 1989;Ahti & Oksanen, 1990).
Interpretation of our results is complicated by the inconsistencies in selected lichen species across forested and alpine sites.Most notably, C. rangiferina and S. alpinum, which were important discriminating variables at alpine sites, were not selected, even though available, by caribou at forested sites.
Our results from the forested sites agree with most of the above cited studies that have shown that these species, especially Stereocaulon spp., are relatively less palatable.This discrepancy suggests that depending on location, forest or alpine, animals may have differenr foraging strategies.
We observed that the majority of the lichens found in forested areas appeared more vigorous and occurred in greater abundance than those in the alpine (Figs. 3 & 6;C. J. Johnson,unpubl.).
Furthermore, at alpine sites clumps of lichen were more unevenly distributed, being separared by bare areas of rock or debris, as reflecred by the high negative correlation between debris and S. alpinum.
Caribou in the less productive alpine areas may be 168 less selective, taking advantage of those sites with the greatest amount of lichen regardless of palatability.The use of a larger number of species and less palatable yet more prevalent lichens, such as S.
alpinum, may be an adaptation to a less productive environment where foraging decisions are based largely on availability.This is consistent with the hypothesis of Bergerud & Nolan (1970) that caribou are adaptive and flexible in the forage species they select.
In our study area, woodland caribou in the forest fed on both terrestrial and arboreal lichens; although, based on feeding site frequency, it appeared that cratering is the predominant activity (C.J. Johnson, unpubl.).Comparable findings were reported for our study animals by Wood (1996) and for other woodland caribou populations (Cichowski, 1993).Selection of arboreal lichen may increase following some threshold in accessibility or availability of terrestrial lichen (Bergerud, 1974;Sulkava & Helle, 1975;Helle & Saastamoinen, 1979;Helle, 1984;Vandal & Barrette, 1985).
Our study animals selected trees, principally P.
contorta, that supported the greatest biomass of arboreal lichen.Across the transects we sampled, which occurred mainly in P. contorta or mixed P. contorta -P.glauca x P. engelmannii stands, the predominant epiphyte was Bryoria spp. with only trace amounts of Alectoria sarmentosa.Bryoria spp.has been reported as a highly palatable food type (Danell et al, 1994) and studies of the mountain caribou ecotype have revealed preference for this lichen group over other alectorioid species (Rominger & Robbins, 1996).The lack of a strong linear relationship between amount of lichen within the 1-2 m stratum and tree diameter suggests that lichen growth and the selection of arboreal feeding sites is related to factors orher than tree size.

Influence of Snow Conditions and Canopy Closure on Site Selection
Although caribou are well adapted to deep snow environments (Telfer & Kelsall, 1984), snow can hinder both the accessibility and detection of forage.Previous studies identified the threshold depth for cratering by caribou and reindeer to range from 50-80 cm (Formozov, 1946;Pruitt, 1959;Stardom, 1975 including hardness and ice layers (Formozov, 1946;Skogland, 1978 1988;Brown & Theberge, 1990).Bergerud & Nolan (1970) concluded that Newfoundland caribou could not smell terrestrial lichens under snow exceeding 25 cm in depth, but Helle (1984) reported that reindeer in Finland detected lichens through a snow thickness of 91 cm.
Over our two-year study period, the maximum crater depths we observed were 97 and 50 cm for forested and alpine sites, respectively.
Canopy closure increases snow interception and correspondingly reduces snow depth and the effort necessary to expose lichens (Schaefer, 1996).Across the range of the Wolverine herd, canopy closure did not affect the selection of cratering sites.In contrast, Cichowski (1993) and Lance & Mills (1996) found that cratering occurred most often in forested areas with more open canopies.In both cases, however, there was an interaction with the presence of terrestrial lichen suggesting that open canopy stands were more productive.Our analysis used a moosehorn coverscope as opposed to a visual estimate of canopy closure (Cichowski, 1993;Lance & Mills, 1996).The latter estimates closure of a much larger portion of the canopy (i.e., scale of the stand) than the coverscope (i.e., scale of the feeding site).
This likely accounts for rhe differences between our results and other studies.
If a caribou attempted to forage optimally by expending as little energy as possible when cratering, then selection of sites with shallower, softer, and less dense snow would be expected as long as the additional search time did not exceed the cost of finding more accessible lichens (Fancy & White, 1985).In agreement with this premise, LaPerriere & Lent (1977) found snow depths and hardness to be less in feeding areas relative to adjacent uncratered areas.At the individual feeding sites we surveyed, caribou appeared to partially meet these criteria by selecting locations to crater where snow depths were shallower than random sites.The greatest effect, as indicated by the odds ratio and univariate logistic plots (Fig. 4, 7), was in the alpine where because of uneven topography and drifting snow, we observed snow depths to be much more variable.
Neither snow hardness nor density appeared to influence crater site selection.In other studies, Frid Cichowski (1993) found that crater sites had greater snow depths, but reduced penetrability when compared to random sites.Duquette (1988) studying the Porcupine herd, reported that snow depths were deeper along migration trails than within adjacent feeding areas, and snow hardness did not differ between the two areas.

Management Implications
Our research suggests that particular scale-specific habitat characteristics may be important to manage Our results describe selection of foraging sites by caribou at one explicitly defined scale, the individual feeding site.However, the relationship between an organism and its environment is often complicated by multiscale influences.Factors from both finer and broader scales may act in unison to elicit responses that may not be detected by measurements designed to record responses at one particular scale.To accommodate the recording and understanding of these interactions, a multiscale hierarchical approach should be pursued (Senft et ah, 1987;Kotliar & Wiens, 1990;Wiens et ah, 1993).
This study was designed to measure just one of many scales that may be televant to how caribou perceive and respond to their environment (Johnson, 1980).The results and conclusions must, therefore, be viewed within the context of other scale-sensitive influences on movement and distribution across the landscape (e.g., large scale distribution of snow, habitat patch configuration, predation risk) which are necessary considerations when managing the winter range of woodland caribou (Cumming, 1992).We are currently investigating the affects of those influences on the foraging behaviour, movements, and distribution of woodland caribou at stand and landscape scales.

Fig. 1 .
Fig. 1.Forest transects and alpine quadrats located across the winter range of the Wolverine herd (Dec 1996-Apr 1998).
. After locating recent tracks of GPS-collared caribou or groups of non-collared caribou in the forest by air or ground survey, we assessed the immediate area for signs of foraging behaviour.Terrestrial feeding sites were charac-Rangifer, Special Issue No. 12, 2000 terised by meandering tracks, craters, and/or sniffing holes.Arboreal feeding sires were characterised by trampling, broken twigs, and fallen arboreal lichen at the base of trees.If some sign of foraging behaviour was present, we selected a random starting point in the snow along the caribou tracks greater than or equal to 20 paces from any ecotone border.Following this, we placed a measuring tape along a section of track that traversed a relatively homogenous stand or vegetation type, and all terrestrial (craters) and arboreal feeding sites found on a 100-m segment of track (transect) were counted.
Fig. 2. Schematic representation of the sampling design used along a 100-m segment of recent caribou tracks in the snow.

Fig. 3 .
Fig. 3. Percent ground cover of lichens at random (» = 255) and cratered (n = 2Q6) sites in forested locations.Vertical lines represent a half width of a Bonferroni-corrected 95% confidence interval and asterisks designate statistically significant differences between corresponding sites.

Fig. 4 .
Fig. 4. Predicted probability of caribou cratering at terrestrial forest sites relative to the percent cover of vegetation or debris (measured in units of 6.25% cover) and snow depth (cm).Symbols illustrate the range of data collected in this study.
Fig. 5. Predicted probability of caribou choosing an arboreal feeding site relative to the grams of Bryoria spp.within the 1-2 m stratum above the snow.Symbols illustrate the range of data collected in this study.

Fig. 6 .
Fig. 6.Percent ground cover of lichens at random (72 = 66) and cratered (« = 70) sites in alpine locations.Vertical lines represent a half width of a Bonferroni-corrected 95% confidence interval and asterisks designate statistically significant differences between corresponding sites.

(
1998) found no effect of snow depth or penetrability on crater site selection, but attributed this to the relatively low snow depths of his study area (mean=31.5cm, standard deviation = 5.8).
for, or consider during an assessment of the winter range of the northern woodland caribou of British Columbia.Forested areas should be managed to contain terrestrial lichen mats with a high percent cover of C. mitis, Cladonia spp., and a high biomass of arboreal lichen (Bryoria spp.).Cladina mitts, C. rangiferina, C. cucullata, C. nivalis, S. alpinum, and Thamnolia spp.are important species that should be considered when assessing and managing alpine areas.Because snow may limit access to forage, and restrict use to specific areas of the range, snow depths should be considered in conjunction with the availability of lichens when assessing the suitability and availability of caribou winter range.

Table 2 .
Summary of multiple logistic regression model derived using the Wald backward elimination procedure for terrestrial and arboreal feeding sites in forested locations in north-central British Columbia(Dec 1996-Apr   1998).TERRESTRIAL FEEDING SITES («=460; model % 2 = 128.576,df=6,P<0.001)Percentcover of all of the lichen species was greater at cratered sites, but non-overlapping confidence intervals revealed differences only for Cladina mitis and Cladonia spp.(Fig.3).At cratered sites C.(±2.19) for cratered and random sites, respectively.Canopy closure ranged from an average of 27.1% (±1.85) at cratered sites to 28.8% (±1.61) at random sites.

Table 3
. Summary of multiple logistic regression model derived using the Wald backward elimination procedure for terrestrial feeding sites in alpine locations in north-central British Columbia(Dec 1996-Apr 1998).RangiSer, Special Issue No. 12, 2000 Frid (1998)Frid (1998)in the southern Yukon is the most comparable to ours in method and species designation.He reported that the probability of a woodland caribou digging a crater increased as the percent cover of Cladonia Danell et al. (1994))iments have concluded that caribou (R. t. caribou) preferred a mixrure of C. stellaris, C. mitis, lowed by Stereocaulon, Cetraria and Peltigera spp.; the proportions of these species, however, may have been more related to availability than to selection(Russell et al, 1993).Danell et al. (1994)assigned high preference rankings to Cladina arbuscula, which is morphologically indistinguishable from C. mitis, C. rangiferina, and S. paschale and a low ranking to P. schreberi.