Does seasonal variation in forage quality influence the potential for resource competition between muskoxen and Peary caribou on Banks Island ?

Interand intra-annual variation in forage quality may influence population dynamics of Peary caribou and muskoxen on Banks Island. From 1993 to 1998 we collected 300 composite samples of sedge (Carex aquatilis and Carex spp.), willow (Salix arctica), legume (Oxytropis spp. and Astragalus spp.), and avens (Dryas integrifolia). Samples were collected in mid-June (start of the growing season), mid-July (peak of the growing season), mid-late August (senescence), and early (November), mid(February), and late(April/May) winter. We analysed forages for percent digestibility (in vitro acid-pepsin dry matter digestibility), crude protein (CP), fibre, lignin, and energy content. There was significant inter-annual variation in levels of lignin, fibre, and energy, and significant intra-annual (seasonal) vari¬ ation for all quality measures and forages, which reflected the strong difference in quality between summer and win¬ ter. We discuss the relationship between forage quality and seasonal diet composition of Peary caribou and muskoxen, and the potential implications for the reduced Peary caribou and high muskoxen populations.

Sedges (Carex aquatilis and Carex spp.), arctic willow (Salix arctica), legumes (Oxytropis spp.and Astragalus spp.), and avens (Dryas integrifolia) collec¬ tively represent >65% of the monthly diet of Peary caribou and a90% of the monthly diet of muskox-en on Banks Island (Larter & Nagy 1997;unpubl. data).In this paper we report seasonal and annual (1993)(1994)(1995)(1996)(1997)(1998) changes in crude protein, digestibility, lignin, fibre, and energy content of forages that are important dietary items of Peary caribou and muskoxen on Banks Island, and discuss whether patterns of forage utilization by caribou and muskoxen are consistent with Hofmann's (1989;2000) prediction that muskoxen should use fibrous forages of relatively low quality whereas caribou should select easily digestible high quality forage.Population changes of these herbivores on Bank's Island may be related, in part, to variation in avail¬ ability and quality of the main forage species.

Study area
Banks Island is the most western island in the Canadian arctic archipelago and covers approxi¬ mately 70 000 km 2 .The climate is arctic maritime along coastal areas, tending toward arctic desert conditions inland (Zoltai et al., 1980).Mean monthly temperatures are below 0 o C from Sept¬ ember through May, and mean minimum daily temperatures range from -30 to -40 o C from Dec¬ ember to March.Snow cover persists into June.The depth of snow cover varies, being greatly affected by wind, and is deepest in low-lying areas.Larter & Nagy (2000) provide a more detailed account of snow conditions in various habitats.Summers are short and cool, with mean maximum daily temper¬ atures ranging from 5 to 10 o C from June through August.Annual mean precipitation is 90 mm (Zoltai et al., 1980).Sachs Harbour (125 inhabi¬ tants) is the only permanent settlement on the Island.
Habitat descriptions were adapted from Kevan (1974), Wilkinson et al. (1976), andFerguson (1991).There are 4 major terrestrial habitats: wet sedge meadow (WSM), upland barren (UB), hum¬ mock tundra (HT), and stony barren (SB).WSM are generally level hydric and hygric lowlands char¬ acterized by Carex aquatilis, Eriophorum scheuchzeri, and Dupontia fisheri; vegetative cover is nearly 100% except for standing water.UB are well drained sites found on the upper and middle parts of slopes.Vegetative cover is 20-50% and is domi¬ nated by Dryas integrifolia and Salix arctica.HT is found on moderately steep slopes and is character¬ ized by individual hummocks, which are vegetated primarily by dwarf shrubs (D. integrifolia, S. arctica, and Cassiope tetragona); vegetative cover is 35-50%.SB has a coarse gravelly substrate and is sparsely vegetated (<10%).This habitat is found on wind blown areas, ridges, and gravel and sand bars.A more detailed description of the flora of Banks Island can be found in Wilkinson et al. (1976), Porsild &Cody (1980), andZoltai et al. (1980).

Sample collection/preparation
During summer we collected vegetation samples at the start of the growing season (13-21 June), peak of the growing season (16-22 July), and senescence (18-27 August).We collected sedge (Carex aquatilis and Carex spp.) from WSM and UB res¬ pectively, willow (Salix arctica) and legumes (Oxytropis spp.and Astragalus spp.) from UB and HT, and avens (Dryas integrifolia) from UB.Samples were collected from two different sites.
During winter we collected samples in early-(7¬ 18 November), mid-(12-26 February), and latewinter (20 April to 2 May) from WSM (C.aquatilis) and UB (legumes and Dryas).Sedge was collected over 5 winters (1993-94 to 1997-98) whereas legumes and Dryas were collected for the latter 3 winters (1995-96 to 1997-98) after it became apparent that these forages were an impor¬ tant component of the winter diet of Peary caribou (Larter & Nagy, 1997).Samples were collected from the same general area in both sites each year.
Forage samples were composed of numerous individual plants, including flowers if present, clipped at ground level (>25 g wet weight), except for willow samples.Willow samples were parti¬ tioned into leaf and stem components, and leaves were plucked from numerous individual plants.Current year's growth of willow stems was clipped from numerous individual plants during mid-July and late-August only because stem growth had rarely been initiated by mid-June.
Samples collected in summer were stored in brown paper bags and allowed to air dry in the field prior to being transported to the laboratory in Inuvik.Sedge samples were separated into their green live matter and dead components.All other samples were considered to be current year's live growth.Samples collected in winter remained frozen until transported to the laboratory in Inuvik where they were thawed at room temperature for 24 h.Winter samples were not separated into live and dead components.All samples were dried at 60 Rangifer, 22 (2), 2002 o C for 48 h, and ground through a 1 mm screen with a centrifugal mill.
Subsamples of all forage samples (<10 g dry weight) were analyzed at the Animal Science Department, University of British Columbia to determine their dry matter, nitrogen, energy, lignin, and fibre content.We determined percent digestibility at the Inuvik laboratory.

Forage quality analyses
Dry matter content was determined for all samples.Duplicate samples (n = 8) were analyzed to deter¬ mine the accuracy of the measurement (99.8%).All analyses were calculated on a dry weight basis.We determined percent nitrogen concentrations for all samples by micro-Kjeldahl (Nelson & Sommers, 1973).Each sample was run once, and duplicate samples were run to determine the accuracy of the measurement (96.9%; n = 48).We calculated per¬ cent crude protein (CP) content by the standard conversion (6.25 x percent nitrogen).We deter¬ mined percent digestibility for all samples except one sample of live Carex spp.Percent digestibility was determined by in vitro acid-pepsin digestibili¬ ty following Tilley & Terry (1963) and Spalinger (1980).Larter (1992;1997) found this simple method provided an index of forage fibre content comparable to that of the more complicated aciddetergent fibre technique (Van Soest, 1967).We used the mean percent digested for the statistical analysis (n = 4 or 5 separate runs).High digestibil¬ ity values indicate low fibre content and vice versa (Larter, 1992).
We determined percent lignin content by aciddetergent lignin (ADL) (Van Soest, 1963) and per¬ cent fibre content by acid-detergent fibre (ADF) (Van Soest, 1967).We determined lignin and fibre content for Salix, legume, Dryas, live summer C. aquatilis, and winter C. aquatilis samples only.Samples were run once through each analysis.Duplicate samples were run to determine the accu¬ racy of the measurements for acid-detergent lignin (77.7%; n = 29) and acid-detergent fibre (95.1%; n = 67).
We used bomb calorimetry (LECO AC-300 Automated Bomb Calorimeter) to determine ener¬ gy content (cal/g converted to kJ/g by multiplying by 4.184/1000).We determined energy content for all forage samples collected from June 1993 to April 1996, except for 2 for which we lacked ade¬ quate material.From June 1996 to May 1998 we determined energy content for Dryas, legume, live summer C. aquatilis, and winter C. aquatilis sam¬ ples only.Each sample was analyzed once, however duplicate samples (n = 18) indicated that the accu¬ racy of the measurement was high (99.3%).

Statistical analyses
For the purposes of statistical analysis we pooled forage quality measures across both sample areas, based on the rationale provided by Larter & Nagy (2001a), and partitioned sampling time into three summer (June, July, August) and three winter (November, February, April) periods.For each quality measure (CP, digestibility, ADL, ADF, energy) we used a three-way ANOVA to test for significant main effects (forage type, sample period [season], and year), and all interaction terms (SPSS, version 10.0.7, 2000).All analyses were based on Type III Sum of Squares and all forage quality measures were log-transformed prior to analysis in order to pass Levene's test of equality of variances.A full factorial (unbalanced) model was run for CP, digestibility and energy, however we did not include Carex spp. in the analysis for ADL and ADF because of missing data.Scheffe's test was applied in post-hoc analyses assuming a value for a = 0.05.The relationship between different quality measures was analyzed using correlation analysis (Pearson coefficient) and discriminant function analysis (SPSS, 2000).

Results
Inter-annual variation There were significant differences between years for acid-detergent lignin (ADL), acid-detergent fibre (ADF) and energy, but not crude protein (CP) or digestibility (Table 1).Post-hoc tests indicated that significant year effects could be partly accounted for by differences between the earlier and later years of sampling (Table 2), although this pattern was not entirely consistent.We do not know why ADL values in 1995 were so low, but this pattern was observed for all forages except the legumes.

Intra-annual (seasonal) variation
Strong seasonal effects were apparent for all quality measures.The main effect of Season and Forage* Season interactions were significant in all cases (Table 1), and reflected a strong difference between summer and winter.Within the summer or winter sampling periods respectively there were also dif¬ ferences over several months for most quality meas¬ ures, however the magnitude of these differences was much less than the differences between sum¬ mer and winter (Figs. 1 and 2)..8.4.5.1.9.7.0 3 .8 .6 .6 . 1 .

Fibre
Fibre content was lowest in June and July (Fig. 1).Levels increased in August but these values were similar to winter periods (Figs. 1 and 2).Post-hoc tests indicated that fibre levels were highest in Dryas and Salix stem, and then lower and very similar in all other forages (Scheffe test, a = 0.05; Figs. 1 and 2).Fibre levels were significantly lower in 1996 and 1997, compared to 1993 -1995.During winter fibre levels were lowest in C. aquatilis and highest in Dryas (Fig. 2).

Energy
Energy content ranged from 17-20 kJ/g for all for-148 ages, and tended to be lower in the last two years of the study compared to the first three years (Scheffe test, a = 0.05).Salix leaves and stems, C. aquatilis and Carex spp.had the highest energy levels, fol¬ lowed by Dryas and finally legumes (Scheffe test, a= 0.05; Figs. 1 and 2).However, there were sig¬ nificant seasonal differences as well (Table 1).During winter C. aquatilis had the highest and Dryas the lowest energy content (Fig. 2).order to determine the general relationship betwe¬ en quality measures for all forages (Table 3).Dur¬ ing summer, most measures were significantly cor¬ related (negative or positive), but there were fewer significant correlations during winter.Patterns of correlations were consistent between seasons, with the exception of the relationship between ADL and energy, which was positive in summer and negative in winter.Differences between forage species con¬ tributed to the overall correlation between meas¬ ures, but these correlation relationships were also consistent within forage species.Discriminant function analysis was used to build a predictive model of forage classification based on the five quality measures for five forages.This pro¬ cedure generated a set of discriminant functions based on linear combinations of the predictor vari¬ ables (quality measures) that provided the best dis¬ crimination between the forages.The canonical correlation for a discriminant function is the square root of the ratio of the between-groups sum of squares to the total sum of squares, and is the pro¬ portion of the total variability explained by differ¬ ences between groups.All five quality measures were entered simultaneously, and the pooled within-groups covariance matrix was used to classify each case.

Correlation between quality measures and classification °ff°r ages
In winter, all three forage species were clearly discriminated (Fig. 3a).Function 1 explained 54.1% of the variance (canonical correlation 0.910), while Function 2 explained 45.9% of the variance (canonical correlation 0.969).The strongest correlations with Function 1 were CP (0.657) and digestibility (0.340), while the strongest correlations with Function 2 were ADL (0.617), ADF (0.308) and energy (-0.210).The relationship between quality measures on each axis is consistent with the results of the pair-wise corre¬ lation matrix (Table 3).
In summer there was more overlap among the five forages (Fig. 3b).Function 1 explained 72.8% of the variance (canonical correlation 0.910), and was most strongly correlated with digestibility ( ¬ 0.652).Function 2 explained 23.5% of the variance (canonical correlation 0.780), and was most strong¬ ly correlated with ADF (0.737).The remaining quality measures (ADL, energy and CP) explained little of the variation (< 4%).C. aquatilis and legumes were most clearly discriminated, while Dryas and Salix stem had the greatest overlap.

Discussion
We observed significant inter-annual variation in the quality of key forages in the diet of caribou and muskoxen on Banks Island, similar to patterns reported by Larter & Nagy (2001a).Although there are relatively consistent inter-seasonal chang¬ es in quality measures for fibre, energy, and lignin content, the absolute levels vary substantially, thereby affecting the quality of forages available for herbivores in different years.Specific inter-annual differences are likely related to variability in pre¬ cipitation.For example, the lower fibre content in forages during summer 1996 and 1997 is likely related to moisture, as these summers were wetter than previous ones.Summer 1994 was drier than other years and if drier conditions cause increases in energy content this might explain higher levels in \ 1994 (Larter & Nagy, 2001a).We have no explana¬ tion for why lignin content was lower in 1995-96 than other years.
As anticipated all measures of forage quality also had strong seasonal components in both summer and winter.While the quality of individual forages clearly changed seasonally, a focus on one measure of forage quality at a time may not provide a com¬ plete picture of the overall quality of a particular type of forage.When we examined the correlation between forage quality measures on the most important forages for caribou and muskoxen strong patterns were observed (Table 3).In general, forage species are discriminated by digestibility in sum¬ mer, while crude protein and lignin were most important in winter (Fig. 3).An understanding of variation in forage quality among species is impor¬ tant because the basis for forage selection is differ¬ ent for caribou and muskoxen.
On Banks Island, both caribou and muskoxen forage extensively on willow leaves during June and July when willow leaves are highly digestible, have a high CP and are low in fibre and lignin con¬ tent.For intermediate feeders (IM) like caribou this is expected, however this would not be expected for grazers (GR) like muskoxen.Staaland & Thing (1991) reported that muskoxen in Greenland uti¬ lized Salix arctica in May and July; S. arctica was rich in hemicellulose during this time.This selec¬ tivity for S. arctica was believed to be in response to ruminal mucosal enlargement, which occurs dra¬ matically between May and June and permits rapid and maximal absorption of nutrients (Staaland & Thing 1991).In July, legumes make up a substan¬ tial portion of the diet of muskoxen on Banks Island (Oakes et al., 1992;Larter & Nagy, 1997).Crude protein content has declined in legumes somewhat since June, but legume biomass (N.Larter, unpubl.)and digestibility are greatest in July.Hofmann (2000) acknowledges that muskoxen are seasonally selective and possibly their heavy use of high quality legume and Salix on Banks Island during summer is in response to ruminal mucosal enlargement.
Sedge makes up a substantial portion of the muskox diet during August (~85%) and the winter months November to April (mean 68%, range 34-83%) (Larter & Nagy, 1997;unpubl.).During winter, sedge was the least fibrous of the three for¬ ages we examined, however, it was also the least digestible and had low CP (Fig. 2).The biomass of sedge is far greater than legumes and Dryas (N.Larter & J. Nagy, unpubl.),which make it an ideal forage for GR species such as muskoxen.Interest¬ ingly, willow also makes up a substantial part of the muskox diet in September (51%), October (75%), January (51%), March (36%) and May (48%) (Larter & Nagy, 1997;unpubl.).There is no leaf growth at these times of the year therefore we assume willow stem is being consumed.Visual analysis of rumen contents confirm the presence of willow stems (N.Larter & J. Nagy, unpubl.).While we lack data on the quality of willow stems during winter, it is probable that we can assume high lignin, high fibre, and higher energy content.Whether the high energy content of willow stems mitigates the high lignin content for GR-type her¬ bivores is unknown.Increased willow in the muskox diet in late winter may be a response to high animal density and reduced per capita sedge availability during this period.Differences in the proportion of sedge in the summer diet of muskoxen may also have been density related (Larter & Nagy, 2001b).Deciduous shrubs are often an important part of caribou summer diet, so a reduc¬ tion in their abundance from browsing prior to the growing season (by muskoxen) may have important consequences for caribou population dynamics (Ouellet et al., 1994).
Peary caribou on Banks Island generally selected forages with similarly low lignin, moderate crude protein, and high digestibility during summer, as predicted for IM-type herbivores.Surprisingly, legumes only made up small proportions of their diet in June and August (<15%), and were virtual¬ ly absent in their July diet when digestibility of legumes was the highest of all forages (Fig. 1).

Rangifer, 22 (2), 2002
Reindeer on South Georgia have a limited number of grasses available during summer and they gener¬ ally selected grass species with low lignin, moder¬ ate protein, and high digestibility (Mathiesen & Utsi, 2000).
Unlike most caribou/reindeer populations, the winter diet of Banks Island Peary caribou includes negligible lichen because of extremely low biomass (Larter & Nagy, 1997).During winter, legumes and Dryas make up as substantial portion of the caribou diet (50-80%).Legumes are the most digestible and have the highest CP in winter while Dryas have the highest fibre and lignin content (Fig. 2); available biomass of Dryas is greater than legumes (N.Larter, unpubl.).Caribou, as an IM, would be expected to select legumes.During winter, metabolizable energy becomes a more dom¬ inant component sought in forage than crude protein (White et al., 1981;Klein, 1990) and legumes have a higher energy content and are more digestible than Dryas.In winters with noticeably less snow cover the proportion of legume in the winter diet was higher and the proportion of Dryas lower (N.Larter & J. Nagy, unpubl.).Reindeer are better adapted to surviving periods of starvation (Aagnes, 1998) and although they cannot adjust to roughage with high fibre content they can better adjust to fibrous forage (Hofmann, 2000;Mathi¬ esen et al., 2000a).This might explain the dyna¬ mics of legumes and Dryas in the winter diet of Banks Island caribou.
The high degree of diet overlap observed on Banks Island (Larter & Nagy, 1997) is not entirely consistent with results from studies on other near¬ by islands in the western Canadian arctic.On southeastern Victoria Island, Staaland et al. (1997) concluded that although caribou and muskoxen coexist in close proximity, they appear primarily adapted to different diets and foraging strategies.The predominantly graminoid diet of muskoxen and more varied, browse-dominated diet of caribou should reduce the likelihood of competition.A similar conclusion was reached by Thomas & Edmonds (1984) for caribou and muskoxen on Melville Island.These observations are also consis¬ tent with Hofmann's (1989;2000) characterization of muskoxen as grazers, and caribou as intermedi¬ ate feeders, however the lower level of diet overlap on these islands may reflect differences in densities of animals and relative availability of forage com¬ pared to the situation on Banks Island.
Overall, our analysis of forage quality, combined with earlier studies of caribou and muskoxen diets on Banks Island (Larter & Nagy, 1997;2001b), suggest that large muskoxen populations could be competing for the preferred forage of caribou during critical times of the year.While both species have some flexibility in their abilities to utilize a range of forages, there are distinct anatomical constraints, particularly for caribou (Hofmann, 2000).Although a number of studies have been critical of Hofmann's classification scheme (e.g.Robbins et al., 1995;Van Soest, 1996), the available data appears to be supportive and indicate that this scheme is useful in assessing requirements of various species (Staaland et al., 1997;Hofmann, 2000;Mathiesen et al., 2000b).On Banks Island, the potential for caribou numbers to increase may be constrained by the availability of suitable forage in the presence of muskoxen.

Table 2 .
Results of Scheffe post-hoc analysis of interannual (year) effects of 3-way ANOVA of dif¬ ferences in crude protein (CP), acid-detergent lignin (ADL), acid-detergent fibre (ADF), digestibility, and energy content of six forages from Banks Island.Values are means for sum¬ mer sampling periods only, and values that are not_significantly different between years are marked in common (underline, bold, italics).
Post-hoc tests indicated that the highest CP values were shared by Salix leaves and legumes, fol¬ lowed by Salix stem and C. aquatilis, then Dryas and finally Carex spp.(Scheffe test, a = 0.05).During winter legumes had the highest CP (Fig.2), and CP of C. aquatilis was higher during the last two winters of the study compared to the earlier years. 1 and 2).Lignin levels were lowest in 1995-96.During winter lignin levels were substantially lower in legumes and C. aquatilis than in Dryas (Fig.2).