Ethics Statement

In Canada, black bears are not considered as a species at risk according to the Committee on the Status of Endangered Species in Canada (COSEWIC). We thus captured, collared and released 21 individuals in strict accordance with the recommendations of the Canadian Council on Animal Care. Both captures and manipulations of study animals were approved by the Animal Welfare Committee of the Ministère des Forêts, de la Faune et des Parcs du Québec (hereafter referred to as MFFP; certificates #CPA-FAUNE 2011–30 and 2012–17). Captures were conducted on public lands, under the supervision of the Québec government (i.e., MFFP), so no specific permissions were required.

Study area

The study area was located north of Saguenay (Québec, Canada) and covered approximately 6300 km², centered on Portneuf Lake (48°42’–49°17’N, 70°03’–70°42’W). Forests are transitional between the spruce–moss domain and the balsam fir Abies balsamea (Linnaeus) Mill.–white birch Betula papyrifera Marshall domain. Historical logging activities resulted in the harvest of ~35% of the forest area. Dense stands of deciduous trees, including willow Salix spp. Linnaeus and trembling aspen Populus tremuloides Mill., regenerate after logging. There is a high diversity of berry producing shrubs, the most common being blueberry, serviceberry Amelanchier Mill. spp., bunchberry Cornus Canadensis Linnaeus, skunk currant Ribes glandulosum Grauer, raspberry Rubus ideaus Linnaeus, wild and bristly sarsaparilla Aralia nudicaulis Linnaeus and Aralia hispida Vent., and chokecherry Prunus virginiana Linnaeus. Animal prey items available to black bear in the study area include beaver Castor canadensis Kuhl, moose Alces americanus Clinton, caribou Rangifer tarandus caribou Gmelin calves, snowshoe hare Lepus americanus Erxleben, ruffed grouse Bonasa umbellus Linnaeus, spruce grouse Falcipennis canadensis Linnaeus, as well as Canada goose Branta canadensis Linnaeus, which often nest in the area. Anthropogenic infrastructure is abundant and evenly distributed with a cabins density of 0.32 cabin/km² and a forest road density of 1.8 km/km². The mean annual temperature ranges between -2.5 and 0°C and annual precipitation fluctuates between 1000 and 1300 mm, of which 30 to 35% falls as snow [29]. The elevation ranges between 300 and 800 m with low rolling hills.

Telemetry survey and sampling protocol

Black bears were captured in June and July of 2011 and 2012 using a padded foot snare or a culvert trap, then immobilized with a Telazol-Ketamin-Xylazine (5:4:1) mix. Adults (females > 54 kg and males > 68 kg, based on data from previous captures in the same area) were equipped with a GPS/Argos collar (TGW-4583H-2, Telonics, AZ, USA) programmed to attempt recording a location every two hours. Individual weight, length, premolar teeth (for age determination), and presence of young (visible cub, lactating cues, or information collected during den visits in winter of 2012 and 2013) were collected to build an individual set of intrinsic characteristics. Collars were programmed to send GPS locations from the 7 previous days once a week via Argos satellites at a rate of four collars per day. This schedule allowed us to investigate known locations of four different bears daily while ensuring that surveyed locations were not too old. We randomly selected two or four locations per bear each week to search for feces, depending on the week; because collars sometimes failed to connect with the Argos satellites, we temporarily increased sampling efforts on individuals that connected successfully. Feces found at each location (within a 10-m radius) were collected in a plastic bag and frozen until lab analysis. When a collared female was with her cubs, we distinguished adult and cub feces by their size and diameter. Cub feces were only used for general comparison purpose. Because black bear are primarily solitary or with cubs, except during the reproduction season (June 10^th to July 9^th), and then only for few days [30], we assume that collected feces came from collared bears.

Laboratory analyses

Frozen feces were warmed at room temperature for half a day before analyses and then gently mixed until homogeneous. A subsample of ~100 g was weighed and then washed through sieves (1-mm, 0.5-mm and 0.1-mm meshes) and the proportion of total remains present in the 1-mm and 0.5-mm sieves was estimated. The 0.1-mm sieve was used only to capture small and rare items that were not retained in the 0.5-mm sieve, or to help in fragment identification. Items were identified to the species level for ingested plants and mammals, while all bird species (mainly waterfowls, passerines and grouses species) were grouped together. Insects were classified as ants (adults and larvae) or yellow-jackets and wasps Vespidae spp. Anthropogenic food included pieces of plastic bag, corn, hen feathers, and other elements usually not found in nature. The detritus category included all non-food items, such as woody debris eaten while consuming ants, conifer needles, moss, small rocks, and other debris eaten by bears. Remains of each sieve were spread on hundred-checkered plates and proportions of each item category in the total volume (i.e., the number of 1-% squares filled compared to other remains) was assessed. The total contribution of each item or group of items in a particular scat was corrected according to their contribution in each sieve and the sieve proportion of the total remains retained. To account for differences in digestibility of food sources, we used correction factors, developed for black bear [31] and grizzly bear (Ursus arctos horribilis Ord), a closely related species [4]. These factors are based on the percent contribution of each item, yielding percent dry matter biomass ingested. Population means, by season and by item, were therefore calculated and presented with their standard deviation.

Habitat use analyses

Previous studies estimated the average transit time of food items in the digestive tracts of black bear and brown bear (Ursus arctos Linnaeus), a closely related species, to range from ~ 4 to 16 hours [18,32,33]. We used the twelve locations previously recorded before the feces collection sites to characterize habitat use patterns 24 hours prior to the estimated feces deposition time. This allowed the inclusion of potential variability in gastrointestinal transit time depending on food type. The best time interval was subsequently assessed statistically (see statistical analyses section). We accounted for habitats used between GPS locations by considering habitat features contained within an ellipse encompassing two consecutive points. To do so, we used the package {adehabitatHR} [34] in R 2.15.3 [35] and performed Brownian bridges [36] using only two consecutive locations at a time. Different parameters had to be set a priori to control ellipse extent and shape. The first parameter (sig1) is related to animal speed and path tortuosity, and represents the probability that the animal diverges from the direct path (Euclidian distance), considering the distance and the time between successive locations. We selected this parameter with the liker function in {adehabitatHR} [36] using the annual individual location dataset to obtain an individual parameter value (mean = 4.98, min = 3.01, max = 6.76). The second parameter (sig2) represents the standard deviation of the distance from recorded GPS location to real animal location. As we filtered locations based on positional dilution of precision (PDOP < 10, SE of location error = 0.15, [37]), we set sig2 at 5; following a priori tests, we noted that this parameter had little influence on Brownian bridge shape and dimension unless it varied greatly (i.e., over what is commonly observed in GPS location precision standard deviation). The output was a raster with individual pixel values representing different probabilities of use by the bear between the two successive GPS locations. We transformed rasters into smoothed polygons using the getverticeshr function by including only pixels with a probability of use > 75% (as calculated by Brownian bridges) to obtain ellipses (see S1 Fig).

We used 1: 20,000 land cover maps provided by the MFFP, which are updated each year with new natural and anthropogenic disturbance polygons (e.g., forest fires, cutblocks, windthrows). Minimum mapping unit size was 4 ha for forested polygons and 2 ha for nonforested areas (e.g., water bodies). We classified forest stands into categories (Lake, Swamp, Conifer, Cut (0–5), Cut (6–20), Regeneration, Open; see Table 1 for description) relevant for bear ecology based on studies of their habitat selection in Québec [22,38]. Finally, the proportion of each habitat type and the density of anthropogenic structures (Table 1) were calculated within each ellipse using ArcGIS 10.0 [39].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of variables and associated measurement units.

https://doi.org/10.1371/journal.pone.0129857.t001

Statistical analyses

We frequently collected multiple feces (~45% of sites with > 1 scat, = 1.80, SD = 1.24) at a single GPS location, especially at bear resting sites (i.e., where bears stay for 4 to 8 h). Ignoring the exact moment each feces was deposited can bias our analyses, but we took that into account by selecting the first location recorded in a 20-m buffer around the feces location (i.e., considering a ~10-m precision of GPS location) as the departure point for a 24-h backward ellipses delineation. These ellipses were used to characterize the habitat where bears potentially fed instead of where they rested. As we did not have information on the exact time each feces was dropped, we analyzed feces separately and then calculated the mean food items dry matter biomass in feces for each site.

We used the constrained correspondence analysis (CCA, library {vegan} [40]) to relate fecal content to individual characteristics and habitat used during foraging. To account for repeated measures on individuals (i.e., sites), we included individual as strata in ANOVAs determining axis significance. We applied the same procedure for the method envfit in library {vegan}, a permutation test (n = 999) performed to establish length, direction and significance of environmental vectors. This allowed the permutation of observations within the specified stratum in order to control for variance induced by intra-individual variation and unequal sampling.

We first related feces content to individual characteristics (i.e., sex, age, reproductive status, weight and body condition index) and grouped individuals based on shared traits. We then compared feces content of individuals sharing similar diets, as defined by the first CCA, with habitat use while foraging. We previously defined the time interval, among different possibilities in the 24 h range (S1 Table), which best explained the feces content. This was done using the proportion of variance explained by the CCA comparing the matrix of visited habitats (i.e., the proportion of cover type and density of roads and rivers in selected ellipses) with the matrix of food items identified in the feces of all individuals. We previously removed food items found in less than three feces to account for only relevant food sources. We also removed variables that were correlated (r > 0.6) and made sure that the variance inflation factor remained < 10 [41].

A population-level diet was also estimated for adult and for cubs by pooling all sites (mean dry mass of ingested food items ± SD) and separating them by season (spring; May 15^th–June 14^th, summer; June 15^th–July 31^st and fall; August 1^st–September 14^th). Seasons were based on a sharp and clear shift in diet observed in our samples and to local plant phenology (unpublished data). We did not split diets by season for the individual-level analyses because of limited sample size.

All statistical analyses were carried out using R 2.15.3 and R 3.1.2 [35]. Data are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.75r70 [42].

Diet at the population level

Traditionally, fecal analyses were used to determine the average diet composition of a population without assigning feces to a unique individual, except in studies using DNA analyses on epithelial cells (e.g. [3,15]) and recent studies using GPS locations (e.g. [20,45,46]). To make our results comparable to previous studies we performed our analyses at both the population and individual/group levels.

At the population level, seasonal diet estimates were similar to observations made in the boreal forest by other research teams (e.g. [22,46]), especially for plants and insects. However, the diversity and importance of mammal prey species we documented seemed relatively uncommon for black bears. For example, we found beaver remains in feces during spring and fall (2.7% and 6.8% of the dry matter biomass found in feces, respectively), although beaver has seldom been identified as a primary prey item in terms of both the availability and ingestion of prey items (but see [47] for an isolated bear population on an island in Lake Superior). Similarly, snowshoe hare, particularly abundant in our study area during the survey, was a common food item in bear diet, especially in spring and summer (8.1% and 7.7% of the dry matter biomass found in feces, respectively). This could be explained by the high availability of leverets during these periods [48]. The importance of hare also exceeded what has typically been observed in North American black bear diets (e.g. [13,22]). In contrast, moose and caribou calves were almost absent from our samples even though these ungulate prey are common in the study area. Moose calf hairs were only found in one cub scat and we cannot ascertain whether the calf was preyed on or scavanged. The diet of cubs, which includes more digestible food items than the diet of adults, could be an artifact of the onset of weaning after the period of leaves palatability [30,31].

Individual variation in diet

We found variation in diet composition between individuals and these differences were related to sex and reproductive status, but not to age or body condition, as shown by the CCA. The abundance of the main food sources (i.e., ants, blueberry, grass, and willow) in scats did not differ between bears. Some food items were nonetheless found in higher abundance in feces of bears of same sex and reproductive status. For example, hare and vespidae were consumed more frequently by females with cubs-of-the-year, whereas young poplar leaves and smilacina berries of the previous year were more often observed in feces of females accompanied by yearlings. We believe that such contrasted patterns could be related to different habitat use and physiological requirements, although we did not measure individual energetic balance. Indeed, lactating females have high energetic demands related to cub nursing, especially in proteins [26]. Small mammal prey species could therefore be of great importance, especially prey living in closed and dense forest stands such as hares [49]. Incidentally, these habitats are also selected by females with cubs in our study area, probably for protection against predators and infanticidal males [25].

Linking fecal content to habitat use

As some groups of individuals seemed to share similar and distinct diets, we looked further to see if these patterns were translated to relationships between ingested food items and habitats used during foraging bouts. Bear foraging activity was probably not limited to the habitats visited between successive locations, and bears were not eating only during the time interval considered for a foraging bout (10–16 hours). Nonetheless, it remains highly likely that the majority of food items found in feces originated from those habitats and that the Brownian bridge ellipses encompassed most of the habitats that were really used by bears, regardless of the inherent inclusion of unused habitats. This is supported by multiple significant relationships that we highlighted between food sources and particular habitat types using a limited number of replicates and a relatively short time interval (10–16 h).

Our results suggest that the link between fecal contents and habitat use varied between groups, with consistent behavioral differences between sex and reproductive status. While lone bears of both sexes used closed roads and old cuts (6–20 years old) to feed on ants and birds, females with yearlings and females with cubs selected similar habitat types and focused their foraging activity on blueberry and on bunchberry, beaver, and creeping snowberry, respectively. This is a possible example of differential use of similar habitat types. Although difficult to ascertain and strongly dependant on sample size, habitat use delineation, and food availability (not measured in this study), such patterns could result from intraspecific competition for food [30], discrepancy in nutritional needs [26], or displacement of females with young by lone bears to prevent aggression and cannibalism on offsprings [25]. Temporal segregation could also explain such differences. Indeed, blueberries, an abundant fall food source, were consumed by females with yearlings in these habitats. Lone bears, on the other hand, fed on ants, which is mostly a summer resource (Fig 1). Although both groups foraged in these habitat types, it seems that they did not use them during the same season and sought different food sources.

Analyses of habitat use and diet content for females with cubs-of-the-year showed lower explained variation than for other groups (Fig 2B, 2C and 2D), a counterintuitive result, especially for a group that undergoes strong selective pressure [26,50]. This could highlight the inconsistensies in foraging behavior patterns within this group of females. Although they share similar behavioral constraints (e.g., cubs protection and nursing [23–27]), the availability of habitat types within their seasonal home range may vary, increase the intra-group variance and mask common strategies between females.

Methodological advantages and cautionary comments

We used a traditional approach by analyzing scat contents and went a step further by linking diet composition to timing of feeding and fine-scale habitat use patterns obtained from GPS/Argos telemetry. Nonetheless, decreasing location intervals and use of inertial navigation systems coupled with GPS [17] could fill the remaining gap between fixes and yield a continuous path for a better habitat use delineation. We chose the interval that best explained the relationship between feces content and habitat used, in accordance with literature on black and brown bears. However, feeding captive black bears with similar diet than their wild counterparts [18] could enhance the reliability of our data by identifying specific and more precise time intervals (although captivity could also influence transit time). Moreover, increasing sample size would be a critical feature, both in terms of individuals considered and of feces collected by individual, especially if the objective is to study inter-individual variation in foraging strategies.

We recognize that other techniques (e.g., isotopic ratios, GPS collars with video camera) could partially overcome these limitations, but note that considerable uncertainty still surrounds these methods. Scat DNA analyses are alternate ways to relate feces to a specific animal and assess diet diversity. The former implies a recovery of the feces shortly after excretion to prevent DNA damage of intestine epithelial cells found on the surface of feces [51] and yields little information on individual characteristics besides sex. The latter (diet analysis based on DNA metabarcoding) is a reliable but expensive way to detect all ingested food items, provided that appropriate DNA markers are used. The performance of these markers to assess the relative importance of each item in biomass ingested is however still debated, especially for diverse diets [52].