Seasonal specialization in diet of the Humboldt marten (Martes caurina humboldtensis) in California and the importance of prey size

Abstract

Carnivorous mammals < 25 kg typically prey on species < 50% of their body mass but can choose prey whose energy value varies from small proportions of their daily needs to exceeding them. We hypothesized that for carnivores < 25 kg specializing in vertebrate prey, prey sizes closest to meeting daily energy needs would be most frequently depredated. We tested this hypothesis by reconstructing the diet of Humboldt martens using 528 scats and calculating the proportion of metabolizable energy (PME) that each prey taxon contributed to the diet. Overall, mammals dominated the diet (PME = 72%), followed by birds (PME = 22%), with berries, insects, and reptiles contributing < 10% PME. Sciurids comprised the largest proportion of all prey, representing 42% of overall PME, ranging from 29% (spring) to 51% (summer). While > 37 prey taxa were identified in the annual diet, only 11 contributed > 5% PME in any single season and the 4 dominant prey taxa in any single season represented 59–64% of that season’s PME. Medium-sized prey (85–225 g) composed 55–66% PME from summer through winter and 2.6 to 8.4 times PME compared to small (< 40 g) and large (> 250 g) prey during these 3 seasons, respectively. When PME for the most frequently consumed individual medium-sized prey (e.g., chipmunks) declined seasonally, martens switched to alternative medium-sized prey (2.8- and 2.5-fold increases in medium-sized birds and flying squirrels, respectively), increased use of large prey (> 8-fold increase), but changed use of small prey least. The annual importance of medium-sized prey, and seasonal shifts to similar-sized or larger prey during winter-spring seasons, both support our hypothesis that the most frequently depredated prey in the diet of Humboldt martens have body sizes closest to meeting daily energy needs.

Determining the importance of food items in the diet can provide insights into a species’ ecology, habitat relationships, and predicting how natural and anthropogenic habitat changes will affect food availability. The relationships between the size of the predator and their prey are central in structuring trophic linkages in food webs (Cohen et al. 1993). Within the Carnivora, species that feed primarily on vertebrates exhibit selection for sizes of prey that are a function of their own mass (Sinclair et al. 2003; Owen-Smith and Mills 2008). These predator–prey size relationships are not consistent among all species of carnivores, with large carnivores (> 25 kg) taking greater ranges of prey sizes and frequently killing prey much heavier than their own mass. Smaller carnivores (< 25 kg) typically prey on species within a much narrower range of body mass and typically do not include species > 50% of their own mass (Carbone et al. 1999; Sinclair et al. 2003).

For most carnivores < 25 kg, the daily prey biomass needs are typically < 50% of their body mass (McNab 1989). Prey that weigh < 50% of the body mass of their predators can range from meeting little of their daily energy needs, when they are small, to exceeding daily energy when they are large. If a carnivore does not have specific nutritional requirements obtainable only from selected prey species, then prey selection should directly relate to the energy available from each possible prey type (Powell 1979). The amount of energy available from a prey species is closely correlated with its body size (e.g., Powell 1981); the size of prey items, therefore, is an important characteristic of a predator’s diet. Moreover, daily foraging strategies should be focused on encountering energetically profitable prey. Predators should therefore theoretically choose home ranges including the quantity and quality of resources sufficient to support the density and diversity of prey species with body sizes that meet year-round energetic needs.

Among the Carnivora, mustelids have some of the highest energy needs, due to the combination of their energetically inefficient long, thin shape (Brown and Lasiewski 1972), and limited ability to store excess energy as fat (Buskirk and Harlow 1989). For North American martens (Martes americana and M. caurina—Dawson and Cook 2012), estimated field metabolic rates suggest they require daily prey biomass equivalent to 15–25% of their body mass to meet their energetic needs throughout the year (More 1978; Gilbert et al. 2009). While some small carnivores can meet their energy needs by consuming hyper-abundant small food resources such as fruits and invertebrates (Muñoz‐Garcia and Williams 2005), North American martens prey primarily on vertebrates (Martin 1994). For martens, which range in mass from 700 to 1,000 g, the amount of prey biomass equivalent to 15–25% of their body mass would range from 105 to 250 g. Therefore, if martens selectively kill prey with body masses that closely approximate their daily energy needs, species within this body size range should be their most utilized prey.

Carnivores may have optimal prey sizes that balance their energetic needs with foraging costs; however, finding and killing optimally sized prey is never guaranteed. What is killed by a predator is a function of opportunity, availability, and accessibility, which have seasonal, spatial, and random components. If carnivores are operating as optimal foragers, they should select the most profitable prey at a given time, and should only use alternative, less profitable prey when abundances of the most profitable prey are low (Oksanen et al. 2001). However, if a carnivore is selecting for specific body size of prey, over time we expect the distribution of prey sizes to center on a size that optimizes the energy gained when compared to the energy required to search for, kill, and eat it. Alternatively, if a generalist foraging strategy is employed, the distribution of prey sizes should simply reflect their availability. In the end, the importance of body size for prey selection by carnivores can be evaluated by comparing the use of prey size classes to their availabilities and by identifying how the diet shifts in response to seasonal changes in the availability of prey.

We investigated the role of prey size classes and seasonal availability on the diet of the Humboldt marten (Martes caurina humboldtensis), a small carnivore (~ 1 kg) inhabiting the low-elevation coastal forests of California and Oregon. The Humboldt marten is a subspecies of the Pacific marten and a subject of conservation efforts (Zielinski et al. 2001; Slauson et al. 2007). Much of the literature on North American martens describes them as a habitat specialist (see review in Thompson et al. 2012) but a dietary generalist (see review in Martin 1994), which appears to have detracted attention from identifying how key prey populations are linked to the habitats that martens select. Martens prey predominantly on vertebrates, and previous diet studies have suggested that small (< 40 g; Myodes—Martin 1994) and large (> 250 g; Lepus americanus—Cumberland et al. 2001) prey taxa are major components of North American marten diets. Methods used in these prior studies have not been consistent, however, and lack of clarity on the relative energetic importance of specific prey sizes in the diet of martens remains to be addressed.

Directly observing the feeding habits of martens is not feasible due to their secretive nature and low densities. However, diet can be assessed indirectly by scat analysis. Reconstruction of the diet from scat is challenging due to several inherent biases. First, field sampling of scats can be unrepresentative of both true diet composition and prey selection. Second, small prey are overrepresented in the diet because their higher surface area to volume ratio makes their remains in scats more frequently detected than the remains from larger prey (Weaver 1993; Cumberland et al. 2001). This is why “frequency of occurrence (FO),” the most commonly reported measure of importance of a prey item in the diet, overrepresents smaller prey (Klare et al. 2011). Third, FO also misrepresents the caloric value of prey, in that small prey have proportionally fewer calories and more indigestible matter due to their higher surface area to volume ratios. Fourth, sample size issues are of concern for studies on species that consume a large number of taxa, although some guidance has been provided by Trites and Joy (2005).

We can minimize the impact of some of the biases by seeking alternatives to the use of FO. In this paper, we describe the diet using new models we developed to interpret the remains in scats in terms of the proportion of metabolizable energy (PME) that each prey item represents. Metabolizable energy (ME) is the most proximate measure of the energy value of each prey item to an energy-maximizing predator and PME is the least biased representation of each prey item in the diet (Klare et al. 2011). We demonstrate the superiority of PME (see Supplementary Data SD1) and then use PME to produce the first description of the diet of the Humboldt marten, including the effects of season, sex, and age on diet. We evaluate the hypothesis that prey taxa with body sizes that most closely approximate the daily energetic needs of the Humboldt marten should be utilized most in their overall and seasonal diet to optimize foraging efficiency. If true, we predict that prey with body sizes closest to the marten’s daily energetic needs will contribute substantially more to the overall annual diet than smaller or larger prey. We also predict that if the availability of prey taxa in the preferred size range decreases seasonally, martens will shift to alternative prey of similar or larger body sizes. In this prediction, we identify that prey size should by the primary motivating factor in prey choice, followed by its abundance within the preferred body size range. If our hypothesis is rejected, we predict that the overall contribution of prey body sizes will not differ substantially and that if the availability of prey in the preferred size range changes seasonally, martens will shift to alternative prey of smaller body size but which tend to have greater abundance. In this prediction, we identify that prey abundance, irrespective of body size, should be the primary motivating factor in prey choice.

Materials and Methods

Study area

The 700-km² study area is located in southern Del Norte and northern Humboldt counties in coastal northwestern California (41°30′00″N, 123°45′00″W; Fig. 1). It occupies portions of the Klamath-Siskiyou and Northern California Coastal Forest ecoregions (Ricketts et al. 1999). The study area ranges from 500 to 1,300 m in elevation and is located 8 to 40 km inland from the ocean. The climate is an inland expression of the maritime regime, characterized by moderate temperatures and a Mediterranean climate typified by a distinct wet period in the winter and a dry period in the summer. Precipitation comes largely as rain, totaling between 200 and 300 cm annually. Snow occurs sporadically and rarely persists below 900 m elevation. Fog occurs on the western edge of the study area and further interior along major stream drainages, providing a source of moisture during the summer when there is typically very little rain.

The combination of moderate temperatures, high annual precipitation, and summer fog supports dense and continuous tree cover throughout most of the study area and dense shrub cover in mesic sites. Douglas-fir (Pseudotsuga menziesii) and tanoak (Lithocarpus densiflora) forest types are most common, with redwood (Sequoia sempervirens) types becoming more prevalent on the western edge. Additionally, areas with serpentine soils occur in the study area, and this soil type fosters several structurally and compositionally distinct forest types and supports a rich diversity of plant species (Kruckeberg 1984). In these soil types, low levels of essential nutrients and high concentrations of detrimental elements offer a harsh growing environment for plants (Jenny 1980), resulting in open and rocky sites with slow-growing woody plants and stunted trees (Jimerson et al. 1995). A dense understory of shrubs is characteristic of Humboldt marten habitat, whether it is associated with an overstory of mature trees or more open serpentine-influenced areas (Slauson et al. 2007). The study population occurs predominantly on unmanaged forest habitat (< 20% harvested) on the Six Rivers National Forest (hereafter, core area), although the western edge of the population occurs in predominantly managed forest habitat (> 90% harvested; hereafter edge area) previously owned by a private timber company but now owned and managed by the Yurok tribe (Fig. 1).

Field collection of scats

From 2000 to 2009 and from 2013 to 2014, we collected scats opportunistically during all 4 seasons of the year. From 2000 to 2009, scats were collected in the ~500-km² “core” study area and from 2013 to 2014, scats were collected in a ~200-km² “edge” study area (Fig. 1). We collected scats from various sources including: in live traps from captured animals in collaboration with another project (Slauson et al. 2007; K. M. Slauson, pers. obs.), at baited track and camera stations, from scat detector dogs trained to detect only marten scats (Long et al. 2008), and at den or rest structures discovered by tracking radiocollared individuals. Scats containing significant amounts of trap bait were excluded from analysis. We also excluded remains that were unlikely to be directly consumed for the purpose of acquiring energy, such as insects of the family Formicidae indirectly consumed with fruit, from grooming activities (e.g., marten hairs, ticks), and plant and soil debris (e.g., rocks, bark, conifer needles) commonly attached to scat exteriors. Scats were placed in plastic bags and frozen or placed in vials with desiccant prior to processing. Our scat collection methods permitted identification of the sex and age of the marten for those scats collected from known individuals in live traps, rest sites of known individuals, or collected on track plates where sex was identified by track measurement (Slauson et al. 2008).

Identification of prey remains

Identifying prey remains required the separation and cleaning of hard and soft remains. We placed each scat in a tight mesh nylon stocking and soaked it for > 12 h in a warm water bath. We gently manipulated the remains to break up any solid pieces and then rinsed them using warm tap water over a 500-μm soil sieve. The remaining scat elements were then put into a petri dish for wet sorting. While wet, we sorted the scat elements into their component groups (i.e., hair, bone, feathers, scales, insect remains, plant material, soil) and placed them in a food dehydrator under the lowest temperature setting for > 24 h. Once dry, we placed scat component groups into glass vials for identification and storage.

We assigned prey remains to the most discriminating taxonomic level possible. We compared the characteristics of scat component groups to reference collection materials for hair, scales (reptiles), feathers, and skeletons (mammals). We also used published and unpublished keys to identify guard hairs (e.g., Mayer 1952; Adorjan and Kolenosky 1969; Moore et al. 1974), and field guides (e.g., Sibley 2000) to identify feathers. Identifications of mammals typically occurred using skeletal elements, usually teeth and claws, and guard-hair characteristics. We identified a subset of birds on the basis of the size of flight feather shafts and overall feather color, and beak and claw characteristics. In most cases, birds could not be identified to species, but we classified them to one of the following body size classes: small < 40 g, medium 40–250 g, and large > 250 g. These size classifications were based on comparison of the sizes of remains present, either flight feather shafts or bones, to reference materials. Exoskeletal remains of insects (or their nest materials for Hymenoptera) were used to identify most insects to order; however, social insects were typically identifiable to genus and species by comparison to photographic keys (e.g., Haggard and Haggard 2006). We compared seeds and fruits of plants to reference collections made from the study area region and by comparison with published keys (e.g., Jepson 1993).

Development of the marten biomass calculation models

Before we could calculate prey remains in terms of ME, we first had to estimate the biomass of prey represented by their remains in scats. To do so, we re-analyzed data from the feeding trials of Zielinski (1981), who developed correction factors to relate the number of scats produced per individual prey item to the biomass of prey consumed using an adult domestic European ferret (Mustela furo). European ferrets, which are bred for docility and are similar in size (10–15% larger) and shape to a marten, were considered a suitable substitute for developing the scat–biomass relationship. Our goal was to determine how many scats were produced per unit weight of each potential prey species. Prey species fed to the ferret were selected to represent the diversity of weights that had been reported in marten diets, including small (< 40 g) prey species: deer mice (Peromyscus maniculatus; 17.0–17.7 g), passerine birds (15.5–21.9 g), voles (Microtus spp.; 35.9–37.9 g); and medium (50–150 g) prey species: chipmunks (Eutamias spp.; 50.4–53.2 g), northern flying squirrels (Glaucomys sabrinus; 117.8–140.6 g), Douglas squirrel (Tamiasciurus douglasii; 189.8–248.9 g). Feeding trials were conducted over a 4-month period, with 4 to 43 trials conducted per prey species (Zielinski 1981). During each trial, individual carcasses of each prey type were weighed and offered to the ferret 1 at a time for 4 days. The prey remains were weighed daily and replaced until they had been eaten completely or ignored by the ferret for 2 days. Scats were collected daily and each trial produced an observation of fresh mass of prey consumed per gram of remains (dry mass) in scats for the trial species. This ratio (grams of prey/scat) was regressed against prey weight, resulting in a significant correlation (r² = 0.89, P = 0.0001) described by the relationship:

where y is the fresh mass of prey consumed (g) per scat produced (dry weight) and x is the live body weight of prey items.

For birds, we only had access to feeding trial data for small passerines (n = 26) with mean weights from 15.5 to 21.9 g. Therefore, we used the difference between the predicted grams of prey for similar-sized mammals to develop a correction factor (predicted mammal grams of prey/scat − observed bird grams of prey/scat = mean of −0.59 g) that was added to the mammal biomass model to produce the following bird biomass model:

For reptiles and insects, no feeding trial data were available. Thus, for reptiles, we assumed a similar reduction in grams of prey per scat as birds, due to their high surface to volume ratio and high volume of indigestible matter (scales) and applied the bird model. For insects, we developed 2 biomass models; the first was for cases when single individuals are entirely consumed and the second when martens feed on the nests of social wasps and bees. For the former, we assumed individual insects were passed in single scats and estimated the number of individuals in a scat using structures representing single individuals (e.g., heads) and multiplied this by their average species weight:

Insect species weights were provided by Haggard and Haggard (2006).

For the cases where martens feed on nests of social wasps and bees, and scats were composed of adult and larvae remains, we assumed the same conversion rate of fresh biomass of adults and larvae to wet scat mass as for birds in the ferret feeding trials. Because we were uncertain if martens consumed the entire nests of social wasp and bees, we used the proportion of wet scat mass to estimate the amount of nest biomass consumed:

For berries, similar to insects, entire individual fruits are consumed and we used randomly selected scats composed of ≥ 95% of a single species of berry to estimate the average number of berries consumed to produce a single scat (medium-sized berries [e.g., salal] $x¯$ = 11.3 berries/scat, SE = 1.8, n = 7; small-sized berries [e.g., Vaccinium spp.] $x¯$ = 21.6 berries/scat, SE = 2.7, n = 9). We then collected and weighed 10 berries from each of the most frequently consumed fruits to estimate the average berry weight (medium-sized berries [salal] $x¯$ = 0.63 g, SE = 0.05, n = 10; small-sized fruits [Vaccinium spp.] $x¯$ = 0.21 g, SE = 0.02, n = 10). Finally, we calculated berry size class-specific biomass estimates for the mass of small (4.5 g berries/scat) and medium (7.1 g berries/scat) berries typically producing a single scat and used these values to calculate the overall biomass of berries in the following equation:

Biomass to ME conversion model

For all vertebrate prey items that represented > 5% of the biomass consumed, we estimated the amount of ME (kcal) available in the total amount of each prey item consumed. For vertebrate prey, we first used the published estimates of ME available from feeding trials for martens (Thompson 1986) and fishers (Powell 1981) to develop the following regression equation describing the significant (r² = 0.94, P = 0.007) positive relationship between increasing prey body mass or meal size and increasing amount of ME:

where y is the amount of ME (kcal/g) available from a vertebrate prey item and X is the live body mass (g) of that prey item. Then, to estimate the total amount of ME (kcal) obtained from the biomass consumed of each prey item, we used the following equation:

For berries, no feeding trials were available for mustelids similar to martens, so we used feeding trials on gray foxes (Urocyon cinereoargenteus—Ball and Golightly 1992) and their metabolizable energy estimate of 0.18 kcal/g of Himalayan blackberries (Rubus armeniacus) in the following equation:

For comparing ME among prey items, we calculated the PME for each taxon by dividing its estimated ME by the total amount of ME from all food items combined.

Summary of overall marten diet and statistical comparisons of diet by groups and prey size

Our primary metric for describing the marten diet was in terms of PME, which is the least biased method and reflects the fundamental currency of consumption. However, we also include a comparison of the importance of dietary components as ranked by PME versus how they were represented by the proportion of biomass (PB) and FO (Supplementary Data SD1).

We used the sample size thresholds identified by Trites and Joy (2005) to determine which comparisons had sufficient sample sizes to detect moderate effect sizes (≥ 0.30) with α = 0.05 and statistical power (1 − β) ≥ 0.80. They found that at least 59 scats were needed to distinguish prey species representing > 5% FO in the overall diet but that at least 94 scats were necessary to make statistically valid comparisons between scat sample groups (e.g., season, sex). We first compared sex and age classes to determine if data from scats from martens of different sexes or ages were sufficiently similar to justify pooling them, and to justify pooling these groups with scats in which the sex and age of the marten that produced them were unknown. We then evaluated the influence of season and prey body size. To limit the number of prey categories used in each comparison, we only used the observed counts of species occurrence for prey species representing > 5% of PME. We used standard contingency table tests to identify significant differences and used Fisher’s exact tests in program R (version 3.2.1; www.r-project.org). We compared PME for prey body sizes using these same methods. Our methods adhered to the American Society of Mammalogists guidelines for use of live animals in research (Sikes et al. 2016).

To evaluate whether prey use was selective, proportions of used prey were compared to their proportional availabilities. We did not estimate the availabilities of prey items in our study area but instead used estimates of density per hectare from the published literature. Abundance estimates for each taxon of small mammal included were converted to estimates of density per hectare and averaged from all studies within the study region that were conducted in similar vegetation types. We then calculated the relative availabilities for the most frequently consumed small- and medium-sized mammals in order to calculate their expected ratio:

Density estimates were then converted to the ME per hectare for each prey body size by summing the ME estimates for all mammal species in each size class. The expected medium- to small-mammal PME ratio was calculated using the equation above and compared to the observed PME ratio to provide a relative measure of prey selection.

Results

A total of 577 scats were collected, 49 of which included significant amounts of trap bait and were excluded from analysis resulting in a final sample of 528 scats. Seasonally, 277 (53%) were collected in fall, 53 (10%) in winter, 41 (8%) in spring, 151 (29%) in summer, and 6 (1%) were missing collection dates and could not be assigned to a season. A total of 82% of the scats were collected in either summer or fall. Non-bait scats (n = 528) were collected during the course of livetrapping and processing (35%; n = 187), surveys using track plates and remote cameras (11%; n = 60), finding resting sites of radiomarked individuals (45%; n = 237), and using scat detector dogs (8%; n = 44). A total of 332 (63%) of the scats could be identified to the sex of the depositor (male n = 194 [58%], female n = 138 [42%]) from a total of 52 individuals (27 males, 25 females). The age class of the marten depositing the scat was known for 235 (53.8%) of the scats.

Overall diet

The diet of the Humboldt marten included 5 classes of prey, in decreasing rank order in terms of PME: mammals (72.2%), birds (21.9%), reptiles (7.0%), insects (PB = 5.3%), and plants (2.6%; Table 1). Together, mammals and birds represented 94.1% of all ME in the diet (Table 1; Fig. 2). Sixteen taxa of mammals were found in the diet, representing 4 orders; however, Rodentia accounted for 85.2% of identifiable mammals and PME was dominated by 2 families, Sciuridae (42.2%) and Cricetidae (20.6%; Table 1). Of the 69 instances where bird remains were identifiable, 57.9% were Passeriformes, with Steller’s jay (Cyanocitta stelleri) being the most frequently identifiable bird species, and medium-sized birds contributing the most PME in the overall diet (Table 1). The most frequently occurring reptile remains were from alligator lizards (Elgaria spp.) followed by snakes (Thamnophis spp.); however, each represented < 5% PME. Overall, plant matter contributed little to overall PME (2.6%) and was nearly exclusively berries (98.5%). The berries consumed were predominately ericaceous species, in decreasing rank order of frequency: salal, evergreen huckleberry (Vaccinium ovatum), California red huckleberry (V. parviflora), and manzanita (Arctostaphylos spp.; Table 1). Identifiable insect remains were from 3 orders, Hymenoptera, Coleoptera, and Orthoptera. Only scats containing Hymenopterans, including social wasps (Vespula sp.), bald-faced hornets (Dolichovespula maculata), and European honey bees (Apis mellifera), were composed primarily of insect remains (26% of all insect scats), and these scats included remains of adults, larvae, and nest materials. Many small Coleopterans appeared to be consumed indirectly and often co-occurred with berries or insectivorous prey.

Sex- and age-specific diet comparisons

A total of 194 scats were from males and 138 were from females. However, scat samples were predominantly from the fall season (70.5% males, 73.9% females) and sample sizes for other seasons were insufficient (range = 9–23) to accurately characterize sex-specific seasonal diets (Trites and Joy 2005). Therefore, we pooled scats across all seasons for sex-specific comparisons. Because the differences between the sexes were indistinguishable (Fisher’s exact test; P = 0.19, with 10 prey categories), we pooled scats from both sexes with scats of unknown sex for subsequent analyses.

We collected a total of 168 scats (118 male, 50 female) from adult (> 2 years) and 127 scats (50 male, 77 female) from subadult (< 2 years) martens; sufficient sample sizes for age class comparisons but not for age class within sex comparisons. Scats were predominantly collected during the fall (80.0% adult, 70.5% subadult) with between 1.3% to 11.6% of the remainder of scat samples distributed among all other seasons for both age classes. The diets of the 2 age classes were indistinguishable (Fisher’s exact test; P = 0.47, with 11 prey categories) and, therefore, we pooled scats from martens of known age classes with scats of unknown age class for subsequent diet analyses. In sum, the lack of significant differences for sex and age classes resulted in pooling all scats by season for seasonal and body size analyses.

Seasonal dietary shifts

A total of 53, 41, 151, and 277 scats were analyzed from winter, spring, summer, and fall, respectively. We pooled winter and spring (n = 94) for analysis to meet minimum sample size needs, but sample size was only sufficient (Trites and Joy 2005) for independent analyses of summer (n = 151) and fall (n = 277). Mammals were the predominant prey group across the seasons and fluctuations in the proportion of mammals in the diet were inversely and proportionately equal to fluctuations of birds and reptiles in the diet (Fig. 3). The seasonal peaks in berry consumption reflect their availability, with individual species showing greatest use following their fruiting phenology. Red huckleberries typically ripen earliest (K. M. Slauson, pers. obs.) and showed an earlier peak in consumption than later ripening fruits such as salal and evergreen huckleberry. Insect consumption peaked in summer, when hives likely have their greatest amount of calories, in the form of queen larvae in hives of wasps and hornets and peak honey stores in non-native honeybee hives. Reptiles had their largest proportion in the diet in spring (12.2%) and summer (9.7%), when mammals were consumed least (Fig. 3).

Eleven taxonomic groups represented > 5% of PME during 1 or more season (Fig. 4). The majority (59–64%) of the PME in any season, however, was composed of only 4 taxon groups (chipmunks, medium-sized birds, red-backed voles, and either Douglas squirrels [summer-fall], Humboldt’s flying squirrels [winter], or large birds [spring]). However, the frequency of these dominant 4 prey taxa varied least from summer to fall (Fisher’s exact test; P = 0.10) and greatest from fall to winter-spring (Fisher’s exact test; P = 0.04) and from winter-spring to summer (Fisher’s exact test; P = 0.006). Overall, sciurids (85–225 g) were the largest component of the diet, comprising 42.2% of the diet annually. Their representation in the diet was highest in the summer and fall (50–51%) and somewhat lower, but substantial, in the winter and spring (29–34%; Table 1; Fig. 4). Chipmunks (85 g) were the most commonly depredated sciurid and contributed 22.7–33.9% of PME during spring through fall (Fig. 4). Medium-sized birds (~108 g) were the second-largest component of the diet, comprising 22.7% of the winter diet and 15.8% of the summer diet. Red-backed voles (25 g) represented a modest but consistent (6.1–15.1%) contribution to the diet across all seasons. Seasonally important species included flying squirrels (120 g; 12.1% in winter), and large birds (> 200 g; 17.3% in winter; Table 1; Fig. 4).

Prey body sizes

Medium-sized vertebrate prey (body masses of 85–225 g; Fig. 4) collectively comprised the greatest PME (59.1%), representing 2.6 times the PME in the diet than small (< 40 g) vertebrate prey (22.9%) and 8.4 times the PME than large (> 250 g) vertebrate prey (7.0%). However, the relative proportions of prey body sizes in the diet were not consistent across seasons, with the greatest differences in prey body sizes between fall and winter-spring (Fisher’s exact test; P = 0.002) and winter-spring and summer (Fisher’s exact test; P = 0.0009) and no significant differences from summer to fall (Fisher’s exact test; P = 0.23). Seasonally, medium-sized prey represented the greatest PME during summer (65.9%), more than half the proportion in fall (55.0%) and winter (58.6%), and the lowest proportion in spring (37.6%; Figs. 3 and 4). The representation of small prey varied least across the seasons, ranging from a low of 18.7% in summer to a high of 27.8% in spring. Large prey showed the greatest fluctuations across the seasons, averaging 2.4% in summer and fall, but increasing > 8-fold to 20.2% in winter and spring (Figs. 3 and 4). In response to a 2.5-fold decline in the proportion of chipmunks in the diet from fall to winter, most likely due the reduced activity of chipmunks in the winter, martens showed similar rates of increase in both medium-sized birds (2.8-fold increase) and Humboldt’s flying squirrels (2.5-fold increase; Fig. 4).

Ninety-five percent confidence intervals for the literature-based estimates of prey densities for the most frequent mammal prey in the diet were 9.5 to 20.8 deer mice/ha, 0.8 to 16.2 red-backed voles/ha, 2.1 to 6.1 chipmunks/ha, 0.4 to 2.4 flying squirrels/ha, and 0.08 to 0.94 Douglas squirrels/ha (Table 2). Converting these taxa-specific density ranges to ME yielded estimates of 181 to 768 kcal/ha of small-sized mammal prey versus 310 to 1,271 kcal/ha of medium-sized mammal prey per hectare and expected PME ratios of 1.71 to 1.65 kcal of medium prey to small prey per hectare (Table 2). Observed PME for these small mammals was 17.2% versus 58.4% for medium-sized mammals and yielded an observed PME ratio of small- to medium-sized prey of 3.4. This observed PME ratio was approximately double (1.99 times greater) that of the expected PME ratio derived from the taxa-specific density estimates and suggests that martens are exhibiting selection for vertebrate prey body sizes.

Discussion

The diet of the Humboldt marten was dominated by vertebrate prey, a finding consistent with regional studies of marten diets conducted in the inland mountains of California (Zielinski et al. 1983; Zielinski and Duncan 2004) and a range-wide study for both North American marten species (Martin 1994). The importance of small, non-vertebrate prey, such as berries and insects, has been emphasized in other studies that used FO methods (Martin 1994), but FO overrepresents the smallest food items compared to biomass (Cumberland et al. 2001) and PME (Supplementary Data SD1) analysis methods. This bias may make the smallest vertebrate species, as well as insects and berries, appear to comprise a greater part of the diet than they actually do (Supplementary Data SD1). Nonetheless, we acknowledge that insects and berries may be important for their nutritional or water content (Westoby 1978), reminding us that representing some food items only on the basis of their estimated energy value may also have its shortcomings.

Our conclusion that mammals—and in particular sciurids and cricetids—are dominant in the diet of the Humboldt marten is also consistent with previous regional studies of Pacific martens (Zielinski et al. 1983; Zielinski and Duncan 2004). Ground-dwelling sciurids (e.g., ground squirrels and chipmunks) predominate in the summer-fall diet and arboreal sciurids (Douglas and flying squirrels) in the winter-spring diet, consistent with the seasonal shift between these rodent groups reported by Zielinski et al. (1983). In contrast to a number of studies reviewed by Martin (1994) who reported that voles were a “major diet item throughout the range” of both North American marten species, we found that voles were never the dominant prey item in any season, and reached their highest relative proportion in spring when prey resources appeared most limited. Thus, in our study location, we did not find support for the preeminence of voles as a key marten prey item. The importance of red-backed and other voles in the diets of martens elsewhere appears to be related to variation in prey–habitat relationships and the diversity of prey communities. For example, Microtus were a major prey species in study areas that included forest-meadow mosaics (Zielinski et al. 1983), whereas there are virtually no wet meadows in our study area. Furthermore, Myodes has often been identified as a key prey item for martens; however, studies doing so have either been carried out in places where there is a lower diversity of alternative prey, such as in more northern latitudes (e.g., Buskirk and McDonald 1984; Thompson and Colgan 1987) or on islands (e.g., Bateman 1986).

In addition to the paucity of voles in the diet, the Humboldt marten appears unique in respect to the proportion of birds in its diet. Based on FO (used here only for comparison purposes), the frequency of birds we observed is the second highest reported (26% versus 29%, Quick 1955, reviewed in Martin 1994). Furthermore, when evaluated by the PME estimator, birds had the highest seasonal importance for the spring diet of Humboldt martens compared to other prey taxa. This is likely a reflection of the greater year-round abundance of birds in environments like our study area, characterized by temperate coastal forests with mild temperatures that remain largely snow-free during winter. The relatively mild winter conditions in coastal forests support wintering populations of several passerine species (e.g., varied thrushes and American robins) that are the most frequently consumed by martens in winter-spring, a phenomenon that is consistent with the importance of birds from other diet studies of North American martens in temperate areas of their range (e.g., Nagorsen et al. 1989).

The diet of the Humboldt marten was notably diverse; the > 37 taxa we discovered is the most reported for any diet study of North American martens (Martin 1994). This diversity is consistent with the trend in higher diet diversity in lower-latitude ecosystems with more diverse prey assemblages (Fleming 1973). This diversity is most conspicuous when considering the number of prey types consumed over the course of the year. However, this annual diversity masks the fact that the majority of PME in the diet is comprised of relatively few prey species, particularly when examined by season. Throughout the year, only 4 taxa represented the majority (59–64%) of the PME in any season. Although some shifts in important prey species occurred between seasons, the 4 taxa representing the majority of PME annually included only 6 total taxonomic groups, including 1 small prey (red-backed vole), 4 medium-sized prey (chipmunks, medium-sized birds, Douglas squirrel, and Humboldt’s flying squirrel), and 1 large-sized prey (large-sized birds). During summer and fall, Siskiyou chipmunks (T. siskiyou) were the most dominant individual prey taxon, representing 23–34% of the PME, respectively. Chipmunks are typically the most abundant species of the most frequently consumed mammals (Carey 1992; Rosenburg and Anthony 1993; Hayes et al. 1995), suggesting martens may be seasonally selecting the most abundant species within the medium-sized taxon group. When ground-dwelling sciurids become less available while they are hibernating in winter and early spring (Kenagy and Barnes 1988; Rosenburg and Anthony 1993), martens appeared to shift to larger-bodied (> 108 g) and presumably lower density taxa, primarily medium to large birds (e.g., Rusch et al. 2000) and flying squirrels (Table 2). This seasonal shift in mammal prey use, from ground-dwelling to arboreal sciurids from summer-fall to winter-spring, is consistent with another study of martens in California (Zielinski et al. 1983).

During the majority of the year, medium-sized (85–225 g) vertebrate prey were the dominant contributor to PME in the diet of the Humboldt marten, comprising greater than twice the contribution from small prey and more than 8 times that of larger prey. This is consistent with our first prediction that prey with body sizes that most closely approximate the daily energetic needs of martens will comprise the greatest part of their diet. This may be because a prey item in this size range will, on average, provide from 77% to 110% of the daily energy needs for a marten (More 1978; Gilbert et al. 2009). However, prey use was not consistent throughout the seasons and our second prediction specified that if the availability of important prey taxa changes seasonally, martens will shift to alternative prey of similar or larger body sizes but lower abundances. We discovered this shift from fall to the pooled winter-spring seasons when the availability of chipmunks declined due to hibernation and the marten exhibited an almost 3-fold increase in alternative medium-sized prey (flying squirrels and birds) and a > 8-fold increase in large-bodied, but presumably lower density avian prey. Even though small mammal prey typically occur at higher densities than medium and large prey, their use was consistently low (18–27% in any season) and there was no noticeable uptick in the use of either mice or voles when the chipmunks and squirrels became less available in the winter. Collectively, our findings support the hypothesis that prey taxa with body sizes that most closely meet or exceed the daily energetic requirements of Humboldt martens are the prevalent prey taxa in their diets. This suggests that although martens can be considered dietary generalists, due to the variety of prey in their annual diet, they also can be viewed as specializing within any season on a few profitable prey with body sizes that most closely meet their daily energetic needs.

The selection and relative importance of prey for carnivores is influenced by a number of characteristics other than prey size. Availability, detectability, and handling times vary among the prey items in the diet of the Humboldt marten. We reviewed relative densities from the published literature (Table 2) and discovered that martens preyed upon—as indexed by PME—medium-sized versus small-sized mammals at rates twice as high as would be expected based on our literature-based index of their relative availabilities alone. This suggests that body size may be a more important factor in prey selection than availability. However, availability does appear to play a role in the selection of prey among species in the medium body size category, both within and between seasons. During any single season martens preyed on the most abundant single taxa of sciurid (Tamias from summer through fall) and when the abundance of Tamias seasonally declined due to hibernation, martens exhibited the largest magnitude shift to the next most abundant sciurid, Glaucomys. This shift from a diurnal to nocturnal sciurid also requires a shift in foraging activity time from fall to winter, a phenomenon observed in marten populations elsewhere in California (Zielinski et al. 1983). In spring, when medium-sized prey are consumed least and presumed to be least available, martens are faced with the decision of switching to more-abundant small prey or less-abundant larger prey. Despite the added time and increased energy required to find and to acquire less-abundant larger versus smaller prey, Humboldt martens increased their consumption of larger-bodied prey in spring. Collectively, these results suggest that Humboldt martens select prey hierarchically, with selection for optimal prey body size operating first, then selection for the taxa most available during each season of the year. The importance of body size and abundance in prey selection has been demonstrated in numerous species of large carnivores (e.g., Karanth and Sunquist 1995; Sinclair et al. 2003; Owen-smith and Mills 2008). Here, we provide evidence that body size also affects prey selection for a smaller carnivore. Furthermore, our results suggest that the relationship of prey size to predator size in smaller carnivores, such as the marten, may more closely relate to meeting shorter-term, possibly daily, energy needs than for larger carnivores that are capable of killing and defending very large prey that meet energy needs over longer periods (e.g., many days).

Supplementary Data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.—Details of the comparisons for the 3 methods for representing prey importance in the diet, including FO, PB, and PME.

Acknowledgments

We would like to thank N. Klass and N. Buckler for assisting with laboratory processing and identification of prey remains from scats. We thank M. Delheimer, C. McNamara, P. Tweedy, B. Barry, B. Marckmann, and other field crew members for the field collection of scats. N. Duncan provided training in scat analysis methods.

Literature Cited