Caching decisions have been studied for many species, but large-scale variation of selective preferences due to environmental heterogeneity has rarely been examined. We investigated large-scale patterns of selective caching behavior in the American pika (Ochotona princeps), a non-hibernating generalist herbivore that caches vegetation in haypiles to serve as winter food. At 13 sites throughout the southern Rocky Mountains, we identified the three most common available and cached plant species and analyzed them for dietary quality. Selectivity at each site was measured as the difference between average quality of the most common cached and available vegetation. Pikas consistently cached materials of higher quality than the most common available vegetation. Selectivity for high nitrogen corresponded to quality of available plants and site summer temperature, whereas selectivity for high water content was negatively correlated with elevation. Our results indicate that pikas cache a variety of plant materials while using different selectivity cues, illustrating the complexity of pika caching behavior across a range of environmental conditions and forage values. Future studies on caching behavior should take into account differential selectivity due to environmental heterogeneity.
Introduction
Arctic and alpine ecosystems present unique challenges for species due to extreme seasonality and unpredictable resource availability. Species that dwell in these environments are forced to adopt strategies to survive extensive food shortages. Whereas some animals use physiological strategies to cope with periods of low food availability, others must rely on behavioral adaptations (Craig et al., 1979; Katz, 1974). Food caching is one behavioral strategy that has evolved to ensure the availability of food during energetically stressful periods (Smith and Reichman, 1984). When selecting food items for storage, caching species often take into account energy content, nutritional quality, and perishability of vegetation to ensure that the cache will sustain its gatherer through times of need (Gendron and Reichman, 1995; Hadj-Chikh et al., 1996; Lewis, 1982). Despite a rich literature on caching behavior, caching decisions have rarely been tied to environmental heterogeneity across space. Large-scale variation of selective preferences is poorly understood, particularly as it relates to topographical and climatic features. Given that current global environmental changes are predicted to drastically alter plant community composition in alpine systems (Guisan and Theurillat, 2000), this information gap must be closed. This study begins to examine patterns of selective foraging behavior across space in the American pika (Ochotona princeps).
The American pika, a lagomorph endemic to western North America, is a generalist herbivore that caches vegetation throughout summer and early autumn in haypiles. These caches are stored in rocky crevices in talus (broken rock) slopes, the pika's preferred habitat. Pikas do not hibernate, so they collect vegetation to serve as a winter food supply when harsh environmental conditions prevent foraging (Conner, 1983). Pikas are an ideal study species for investigating selectivity, as each individual maintains a separate territory and thus has a restricted subset of vegetation available in the form of the plants adjacent to or within the talus forming that territory (Huntly et al., 1986; Roach et al., 2001). We chose this species to explore large-scale patterns of selectivity because pikas exhibit limited dispersal between sites, minimizing lateral behavioral transfer across our study area. In addition, forage quality or quantity has proven important to pika population persistence throughout the American West (Wilkening et al., 2011, Erb, unpublished data).
Pikas vary in the forage species they consume and cache across their range and are considered generalists (Smith and Weston, 1990). However, at particular sites, pikas have been documented to preferentially harvest some plants extensively while ignoring other available plant species (Krear, 1965; Millar and Zwickel, 1972). To date, hypotheses on which vegetative characteristics influence forage selection are diverse and largely site-specific. Decades of research suggest alternative forage preferences based on plant size and morphology (Huntly et al., 1986), spatial distribution of plants (Roach et al., 2001), nutrition (West, 1980), presence of secondary compounds (Dearing, 1996), and novelty (Ivins, 1984). Previous studies are conflicting in the variables found to influence selection and the degree to which they do so. The diversity of selective caching explanations in this species indicates that pika decisions are complex and may vary across space and time.
Incongruity regarding pika forage preferences may reflect variation in climatic and topographic conditions among previous studies, each limited to one or two adjacent sites. Environmental heterogeneity might impact both the degree of selectivity employed by pika populations and the preference for different plant characteristics at each site. Our objective was to provide an evaluation of selective caching behavior by pikas throughout the southern Rocky Mountains and, as such, to explore how site-specific variables may influence caching preferences.
Here we perform large-scale analyses of pika caching behavior as it relates to site-specific environmental conditions. Pika populations each experience a different combination of food choices based on environmental controls on plant communities. Complex interactions among climate, topography, soils, and biotic communities in mountainous environments yield heterogeneous landscapes with great variation in vegetation composition (Cairns and Malanson, 1998). Because topographical and climatic features are often related to plant composition and nutrient content (Cairns and Malanson, 1998; Körner, 1998), they may influence the foraging behavior and degree of selectivity exhibited by different pika populations. In addition, climate is likely to alter foraging behavior by restricting access to forage in harsh conditions and shifting the nutritional inputs that are limited in a given population. To test the relationship between caching selectivity and environmental variables, we investigated the following questions:
Q1 : Do pika populations exhibit preferential caching behavior based on dietary elements across a large geographic scale?
Q2: Does the degree of selectivity depend on the quality of available vegetation at a site?
Q3: Can the degree and nature of selectivity patterns be explained by broad measures of environmental variation?
Methods
STUDY AREA
Thirteen sites known to be occupied by pikas both currently and historically (pre-1980) were selected via random sample stratified by latitude and elevation from a larger sample of 69 such sites in the southern Rockies. The 13 study sites ranged in latitude from the southernmost tip of the distribution of O. princeps at Nambe Lake, New Mexico (35°79′N), to the Snowy Range of the Medicine Bow Mountains, Wyoming (41°35′N; Fig. 1). Elevations ranged from 3036 m at Trap Lake, Colorado, to 3717 m at Nambe Lake, New Mexico. The order of site visits was designed to minimize phenological impacts on plant quality. Visits were timed to allow for an estimated three-day delay in plant development per 100 m in elevation or one degree latitude (Schwartz and Caprio, 2003; Bowman et al., 1999). Based on these calculations, all sites were visited at approximately the same phenological stage (±3 days), with southern, lower-elevation sites visited early (in July) and northern, high-elevation sites visited later in the season (August).
DATA COLLECTION
Haypiles were detected by walking transects across the entire talus patch, perpendicular to the dominant aspect of the habitat. The first transect was set 10 m above the lowest elevation of the talus within the site. Additional transects were surveyed upslope at 60 m ground-level intervals as long as talus was present. If any one transect would intersect more than 200 m of contiguous talus, only 200 m of the transect was surveyed. In this case, the full transect consisted of three sub-transects: two sub-transects of 50 m, each extending inward from the edges of the talus patch, plus one sub-transect of 100 m, positioned in the center of the talus patch. This subset was chosen to best represent vegetative characteristics of the site, as most vegetation exists near the talus edge.
Two observers walked along each transect, one looking for haypiles up-slope and one looking down-slope. Detected fresh haypiles were dissected to determine plant composition, and only green vegetation was identified to ensure that the vegetation analyzed had been gathered in the current year. All accessible haypile vegetation was extracted and sorted by species. The three most common plant species present in each haypile were identified, and their relative volumes in the haypile were estimated based on visible biomass. Graminoid species were lumped into a single category (Dearing, 1996), and all other taxa were identified to species using Weber's Rocky Mountain Flora (1976). It is standard practice for alpine graminoids to be lumped in haypile analyses because it is assumed that they have similar nutritional content from the perspective of the pika (Dearing, 1996; Gliwicz et al., 2006). All detected fresh haypiles were surveyed, and all vegetation was returned to its original haypile following dissection. We determined the three most commonly cached species at each site by counting occurrences across haypiles.
Vegetation surveys were conducted adjacent to three randomly selected haypiles within each site. Our sampling framework was designed to capture available vegetation most accessible to pikas, which often build haypiles along talus edges. We used the point intercept method (PIM) to estimate individual species proportions of available biomass. Point counts from the PIM are highly correlated to biomass in many ecosystems, including montane meadow plant communities (Bråthen and Hagberg, 2004; Jonasson, 1988). Three 50-m vegetation transects were sampled around each haypile, each running perpendicular to the dominant aspect at the haypile, as in Wilkening et al. (2011): the first transect was centered on the haypile and extended 25 m horizontally to each side, and two parallel transects were positioned 15 m up-slope and 15 m down-slope from the same haypile. A vertical marker was positioned at every meter along each transect, and plant species touching this marker were identified and recorded along all nine 50 m transects sampled within each site. Total counts were made of each taxon's occurrence in the vegetation survey, and the three most abundant available plants were calculated from data averaged across all transects at each site. Plants were identified to species for all taxa but graminoids, which, as in haypile surveys, were treated as a single taxon.
At each site, samples of the three most common plant species in haypiles and the three most common available plant species were collected from growing vegetation. For the remainder of this paper, “common” available and cached species will refer to these fresh samples collected representing common available and hayed species. Collection from intact plants was standardized by cutting all specimens at ground level and in full sun to eliminate the influence of dew on initial weights and minimize environmental variation caused by differences in light availability (Bowman et al., 1999). In addition, multiple source plants of each species were selected from the area within 30 m of a pika haypile at each site. Samples from intact plants were collected in paper bags and immediately weighed to the nearest 0.25 g with a 60 g spring scale (PESOLA AG, Rebmattli 19, CH-6340 Baar, Switzerland). Paper bags were weighed prior to the addition of the samples, and this weight was subtracted from the total mass to determine the wet weight of each sample.
VEGETATION ANALYSIS
Of potential plant characteristics that could influence selectivity, we focused on indicators of plant quality, examining nitrogen content (as an approximate measure of protein; AOAC, 1965) and water content. Plant samples were thoroughly air-dried in a mesh bag at the University of Colorado and then dried in an oven for 24 h at 60 °C. Samples were weighed immediately after drying to assess water content. Water content was computed as the total percent weight lost during drying (weight at harvest vs. weight following oven drying). Using liquid nitrogen and a mortar and pestle, 0.5 g of each sample was finely ground. Adding liquid nitrogen does not change the chemical composition of the sample, but facilitates the grinding process (Torregrossa and Dearing, 2009). We packaged 3 to 5 mg of each sample in 5 × 9 mm tin capsules and run though a Costech ECS 4010 elemental CHNSO analyzer (Costech Analytical Technologies, Valencia, California) to determine percent nitrogen.
STATISTICAL ANALYSIS
As an estimate of the overall contribution of a plant species to the vegetation available at site k, the relative proportion of each focal species was determined by:
where ni is the number of detections of species i along the vegetation transects and N is the total number of individual plants detected along the transects.
The proportion of species i at each site k relative to all cached material observed in haypiles was characterized by:
where hj = size of haypile j and pi = estimated proportion of species i of the three most common plants cached in haypile j. The relative size of each haypile was estimated in units of loose (observed) dry volume, each unit being 10 × 10 × 30 cm (C. Ray, personal communication).
Nitrogen content (Yk), and water content (Zk), at site k of available and cached vegetation in the environment (m = 1) or in haypiles (m = 2) were characterized by the following equations:
where Sm = [Aik, Cik], N = nitrogen (% weight), d = dry weight (in mg), and w = wet weight (in mg).
The selectivity index for each variable at a given site was calculated as:
where δA = [Yk1, Z,k1] and δC = [Yk2, Zk2] depending on variable of interest. Therefore, the selectivity index at a site for a given dietary variable is the residual value between the cached to available vegetation quality ratio and 1, the expected ratio given no selectivity. The relative magnitude of the selectivity index changed with each variable, depending on variance within a sample set. A positive selectivity index represents selective caching for higher values of the given nutritional factor than are found in available vegetation. A negative selectivity index represents selective caching for lower values of the given nutritional factor than are found in available vegetation. Zero is the expected value given no selectivity.
Because we did not expect the ratio of cached to available vegetation quality to be distributed normally, all analyses were conducted using the log transformation of the selectivity index:
To address our first question, if pikas exhibit selective caching behavior across sites, we used paired t-tests to determine if nutritional composition of common plants found in haypiles was different from common available vegetation. To address our second question, if degree of selectivity depends on vegetation quality, we tested if the slope of the relationship between cached and available vegetation was significantly different from 1 by calculating the slope confidence interval:
where β1 is the slope of the relationship, tα/2 is the t value of the coefficient β1 at significance level α, and SE(β1) is the standard error of β1 (Montgomery et al., 2012). We calculated confidence intervals at α values of 0.05 and 0.01 for each vegetative characteristic.
To investigate our third question, if patterns of selectivity are related to environmental factors, we tested if selectivity was related to elevation, mean summer temperature, and total annual precipitation through linear regression. We chose mean summer temperature as the most appropriate measure of temperature because chronic heat stress is thought to limit foraging time in pikas (MacArthur and Wang, 1974). Mean summer temperature was calculated as the mean temperature recorded in the months of June, July, and August by HOBO U-series data loggers (Onset U10-003) placed in pika haypiles. We chose total mean annual precipitation as the best precipitation measure, because winter and spring snow runoff and summer rains all contribute to summer plant growth. Annual precip- itation was obtained from PRISM Climate Group (2010). Only univariate models were tested due to sample size constraints.
Results
Of the 13 sites analyzed, 133 haypiles were dissected, and 60 plant samples were collected and analyzed. In no cases were the identities of the top three available and cached taxa identical. Graminoids were among the most common available and cached species at all but one site, comprising the most abundant available vegetation at 92% of sites and the most abundant cached vegetation at 69% of sites. Forb composition was highly heterogeneous across our study area; only 25% of the available forb species recorded were common at multiple sites. The lists of common available and cached species and values of nutritional quality at each site are presented in the Appendix.
PREFERENTIAL CACHING BEHAVIOR AMONG SITES
We found a positive relationship between common available and cached vegetation for nitrogen content (R2 = 0.388, P = 0.023; Fig. 2, part a) and water content (R2 = 0.757, P = 0.0001; Fig. 2, part b). Cached vegetation at a site was significantly higher in quality than environmental vegetation in terms of nitrogen content (t = -3.21, d.f. = 12, P = 0.008) and water content (t = -5.20, d.f. = 12, P = 0.0002), resulting in positive selectivity indices for each vegetative characteristic at most sites (Table 1).
DEGREE OF SELECTIVITY AND AVAILABLE VEGETATION QUALITY
The slopes of the relationship between quality of cached and available vegetation for nitrogen content (β = 0.521, P < 0.05, 95% CI 0.134 – 0.909) was significantly less than 1, suggesting that selectivity for higher quality plants decreased as the quality of available plants increased. The slope of the relationship between water content in common cached vegetation and available vegetation was not significantly less than 1 (β = 0.877, P > 0.05, 95% CI 0.583 – 1.171), indicating that as water content in available vegetation decreased, pikas did not compensate by significantly increasing selectivity.
PATTERNS OF SELECTIVITY AND ENVIRONMENTAL VARIATION
Average summer temperature had a positive relationship with selectivity for nitrogen content (R2 = 0.433, P = 0.020; Fig. 3, part a). The selectivity index for nitrogen content did not show a strong relationship with elevation or precipitation (P > 0.05). There was a negative relationship between elevation and the selectivity index for high water content (R2 = 0.366, P = 0.029; Fig. 3, part b). We did not find a relationship between percent water content in common available vegetation with elevation (P > 0.05); therefore, available water content cannot explain the variability we observed in selectivity along an elevational gradient. The selectivity index for water content did not show a strong relationship with temperature or precipitation (P > 0.05).
FIGURE 2.
Relationship between the quality of the three most common cached plants and the quality of the three most common available plants for: (a) % nitrogen content; and (b) % water content. The solid line is the best-fit linear regression line. The dashed line is the 1-to-1 relationship expected given no selective haying behavior. Each point represents the weighted average plant quality at an individual site.
Discussion
In this study, pikas cached a large variety of plants across a vast elevational and latitudinal range. Graminoids were consistently among the most common available plants and almost always found in haypiles. While it is common practice to group graminoids (Dearing, 1996), we acknowledge that this practice may partially explain the prevalence of graminoids as one of the three most common available and cached plants in our survey. Almost all cached forbs at each site were different from those cached at other sites. Pikas were exposed to a large diversity of plant communities, represented by the variation in common plants present in the environment between sites. Within the plant characteristics we analyzed, pikas selected higher quality foods than were common in the surrounding environment in almost all cases (Fig. 2). Pikas therefore exhibited both inter-site generalist behavior and intra-site selective caching behavior.
TABLE 1
Selectivity indices (I) of nutritional factors by site and environmental variables. A positive value for I represents selective caching for higher values of the given nutritional factor than found in available vegetation; a negative value for I represents selective caching for lower values of the given nutritional factor than found in available vegetation. Zero is the expected value given no selectivity. Sites are listed alphabetically.
Pika populations that were provided higher quality available vegetation (with respect to nitrogen content) showed a lower degree of selectivity than populations with low quality available vegetation. This result would be expected for long-term foraging optimization (Craig et al., 1979; Katz, 1974), as pikas living in areas with predominantly low-N (and therefore low-protein) plants are more likely to supplement their cache with higher energy foods. Because of the energetic costs of haying, pikas with adequately nutritious foods readily available would not be expected to expend excess energy seeking out more nutritious and perhaps less common plants (Huntly et al., 1986).
Along with the quality of available vegetation, environmental factors also explain variation in pika selectivity within our study area. Pikas exposed to hotter average summer temperatures demonstrated higher selectivity for nitrogen content. Given pikas' tendency to decrease daytime activity in high ambient temperatures (MacArthur and Wang, 1974; Smith, 1974), pikas may cache less forage at hotter sites. Our results suggest that some pikas are able to compensate for decreased forage quantity at hot sites by selecting vegetation of higher nutritional quality. This finding is consistent with results from both the Great Basin (Wilkening et al., 2011) and the southern Rockies (Erb, unpublished data), indicating that pika populations are more resilient to higher summer temperatures when presented with higher environmental forb availability and diversity.
FIGURE 3.
Selectivity for plants in relation to environmental variables for: (a) mean summer temperature and selectivity index for % nitrogen; and (b) elevation and selectivity index for % water content.
Our results indicate that pikas respond differently in their selectivity for high water content. Rather than compensating for low water availability in plants, pikas were only found to selectively cache plants with high water content as elevation declined. Because water content of commonly available vegetation did not correlate with elevation, an external factor must drive this relationship. We propose that longer snow seasons and increased availability of free water at high elevations may explain this phenomenon. Low elevation sites often experience later initial snowfall and are characterized by dry late autumns in the southern Rocky Mountains. Pikas at lower elevation sites may therefore have greater need for lush forage in the autumn than do their high elevation neighbors. High elevation sites in the alpine also generally have more free water due to a higher occurrence of rock-ice features and persistent snowfields (Millar and Westfall, 2008), which act as reservoirs (Clow et al., 2003) and provide a steady influx of free water throughout the year in the form of under-talus streams (Millar and Westfall, 2010). Pikas have been reported to drink from water sources, although the frequency and importance of this behavior are unknown (C. Ray, personal communication, J. Varner, personal communication). We propose that pikas may engage in this behavior more often under the talus than is observed during surface activities, due to subsurface accumulation of water. If this is the case, a decrease in selectivity for high water content in vegetation at high elevations could be explained by higher levels of available free water for drinking.
We found that pikas responded differently to environmental conditions in selective caching behavior on the measures of food quality investigated. Selectivity for high nitrogen was related both to the quality of vegetation available and an external environmental condition, summer temperature. However, preference for high water content appeared to correlate only to an external environmental condition, elevation. Selection for plant qualities therefore appears to be contingent on different, site-specific cues and complicates our understanding of how these animals choose materials for their caches.
Acknowledgement of differential selectivity by pikas based on environmental factors may reconcile previous contradictions among studies on pikas foraging selectivity. Although the available literature on pika caching preferences is largely restricted to singlesite studies, all studies combined constitute a substantially large and diverse study area when each local site is considered as part of a larger system. Our study begins to describe possible external influences on preferences for caching materials, but further research is needed to uncover the complexities that dictate patterns of selectivity across the species' range. Future studies on caching preferences of pikas should be carried out on large spatial and temporal scales and should examine a variety of plant characteristics, including nutrition, morphology, and phenolic content, in order to obtain a comprehensive understanding of the mechanisms of selectivity in this geographically widespread caching mammal.
This study provides a framework to investigate patterns of selectivity in other wide-ranging taxa, particularly those living in similarly extreme environments. Although we were unable to assess all potentially important environmental and vegetative characteristics in our system, we believe this study provides an important steppingstone in understanding how species interact with the variable environments arctic and alpine systems present. As climate change continues to alter biological systems, there is a growing need to understand how species vary their behaviors in different climatic conditions. A large-scale approach to studying foraging behavior provides an opportunity to better understand adaptive capacity and resiliency in species that may be exposed to further environmental variability in the decades and centuries to come.
Acknowledgments
This project was funded by the University of Colorado at Boulder and the Indian Peaks Wilderness Alliance. We thank two anonymous reviewers for their constructive feedback on our manuscript. Sincere thanks to C. Ray and R. Guralnick for their invaluable contribution in providing statistical, editorial, and intellectual guidance. We thank to B. Bowman and N. Barger for providing laboratory facilities and A. Darrouzet-Nardi and S. Castle for assisting in plant sample processing and chemical analyses. We thank A. Smith and J. Benson for their editorial contributions. Finally, thanks to A. Stecker and N. Kleist for assisting in data collection.
References Cited
Appendices
Appendix
List of sites and top three available and cached plants with associated nutritional values. % WC = percent water content and %N = percent nitrogen. Species 1, 2, and 3 are in order of abundance of either cached or available vegetation at a site. Sites are listed alphabetically.
Cached vegetation.
Available vegetation.
