Diet and Macronutrient Optimization in Wild Ursids: A Comparison of Grizzly Bears with Sympatric and Allopatric Black Bears

When fed ad libitum, ursids can maximize mass gain by selecting mixed diets wherein protein provides 17 ± 4% of digestible energy, relative to carbohydrates or lipids. In the wild, this ability is likely constrained by seasonal food availability, limits of intake rate as body size increases, and competition. By visiting locations of 37 individuals during 274 bear-days, we documented foods consumed by grizzly (Ursus arctos) and black bears (Ursus americanus) in Grand Teton National Park during 2004–2006. Based on published nutritional data, we estimated foods and macronutrients as percentages of daily energy intake. Using principal components and cluster analyses, we identified 14 daily diet types. Only 4 diets, accounting for 21% of days, provided protein levels within the optimal range. Nine diets (75% of days) led to over-consumption of protein, and 1 diet (3% of days) led to under-consumption. Highest protein levels were associated with animal matter (i.e., insects, vertebrates), which accounted for 46–47% of daily energy for both species. As predicted: 1) daily diets dominated by high-energy vertebrates were positively associated with grizzly bears and mean percent protein intake was positively associated with body mass; 2) diets dominated by low-protein fruits were positively associated with smaller-bodied black bears; and 3) mean protein was highest during spring, when high-energy plant foods were scarce, however it was also higher than optimal during summer and fall. Contrary to our prediction: 4) allopatric black bears did not exhibit food selection for high-energy foods similar to grizzly bears. Although optimal gain of body mass was typically constrained, bears usually opted for the energetically superior trade-off of consuming high-energy, high-protein foods. Given protein digestion efficiency similar to obligate carnivores, this choice likely supported mass gain, consistent with studies showing monthly increases in percent body fat among bears in this region.


S1 Appendix. Field methods and quantification of relative daily intake of foods by grizzly and black bears, based on site visits of GPS-monitored individuals, Grand Teton National Park, Wyoming, USA, 2004-2006
For each bear-day, we visited all successful GPS locations collected for a 24-hour period, except for those bears with fix intervals of <1 hour, for which we visited at least one GPS fix each hour. Site visits typically occurred within a few days to 2 weeks after the GPS bear locations, with an average of 7 days. Within 20 m of each location, we searched for scats and evidence of feeding activity (e.g., grazing, excavations, carcass remains).
Approximate age of scats was evaluated by condition and moisture content. Scats that appeared too recent or too old, relative to the GPS date, were not included in analyses. Within our raw sample, number of scats collected per bear-day ranged from 0 to 24, but the count of 24 was an outlier. This bear-day was associated with a female with yearling offspring. Whereas cub scats were usually distinguishable from their mother's and were not included in analyses, scats of yearling offspring were not always distinguishable from their mother's. Therefore, we reviewed the data and found that only one grizzly bear female in our sample was accompanied by yearling offspring. This family group accounted for 4 bear-days: 2 days with lower counts of scats (n = 6), 1 day with a high count (n = 12) and the other day with the extreme high count (n = 24). For the two days with higher counts, we sub-sampled the scats (selecting 33% of scats for each trial), evaluated the resulting difference in estimated scat contents, and found little difference. This suggested that, if some portion of these scats were left by yearling offspring, they were consuming the same foods as their mother. However, we did not wish to overestimate the contribution of the foods in scats, due to an excessive number of collected scats. Therefore, for these two days with high scat counts that were associated with yearling family groups, we combined contents of all scats collected at each site and treated them as a single scat. This reduced the scat counts from 12 and 24 to 5 and 9, respectively. Our final scat sample ranged from 0 to 13 scat/day, with a mean of 2.6 (Fig. a). These numbers are consistent with Roth (1976), who documented defecation rates of brown bears in captivity and reported a range of 0 to11 scats/day, with seasonal means ranging from 2 to 7.
Our methods required that we combine observations of feeding activity with scats to estimate relative consumption of different foods and food types within a bear-day. All observations of feeding sign at GPS bear locations were subjectively classified according to three levels of intensity: 1) Light consumption -when only a few bites of a food were likely consumed, exemplified by grazing on a few individuals plants; fruit gleaned from a few branches; an isolated excavation for insects; and feeding on an old previously-consumed carcass.
2) Moderate consumption -when many bites were likely consumed, exemplified by observation of a measurable area of grazing; disturbance of several shrubs, trees, or logs for feeding; and feeding on a small vertebrate.
3) Heavy consumption -when a large volume of a food was likely consumed, exemplified by disturbance of a large proportion of the plot area for grazing, excavation, or other feeding; and feeding on a large, fresh vertebrate.
Sometimes multiple, usually consecutive fixes were obtained at the same site.
Observations of feeding sign were assigned to the first fix only, unless bears were observed to feed on a large, fresh ungulate carcass or were observed to feed at a large whitebark pine cache (with intensity values of 3). In this case, additional scat equivalents for this feeding sign were assigned to later fixes, but only if the time between locations (i.e., the time the bear stayed at the site) exceeded 3 hours.
We used intensity scores to assign a "scat equivalent" percentage to each observed food.
For each food type, we calculated the proportions of 1, 2, and 3 intensity scores recorded from feeding sign. To estimate the appropriate scat equivalent percentage values, we utilized the distribution of percent dietary (i.e., corrected) volumes of foods observed, by food type, within our entire scat sample (Table a). For example, 69% of the 899 observations of above-ground feeding sign were assigned an intensity score of 1. Correspondingly, the lowest 69% of scat volumes (1492 of 2162) for above-ground vegetation ranged from <1% to 29%, with a median of 9%. Using this process, we estimated the median observed percent volumes corresponding to the three intensity values for each food type. Summarizing across food types, the mean, median and mode values were all equal, therefore we used these values for our assignment of scat equivalent percentages (i.e., %%, 30%, and 75%).
To evaluate the potential influence of these assigned values on our results, we substituted alternative values and summarized food consumption for each bear-day. First, we substituted the minimum values among the food types (i.e., 2%, 13%, and 75%). Secondly, we substituted the maximum values among the food types (i.e., 9%, 51%, and 100%).
Percentages of daily digestible energy provided by each of the 8 food types observed on a given bear-day differed by -18% to 18% comparing assigned values to alternative minimum values, and differed by -17% to 20% comparing assigned values to alternative maximum values.
The median difference ranged from -1% to 2%, with most equal to 0. All distributions were centered near 0, indicating that estimates were unbiased relative to the assigned scat equivalent values (Fig. b, for example). Similarly, estimates of the percentages of daily digestible energy provided by each of the 3 macronutrients differed by -11% to 12% comparing assigned values to alternative minimum and maximum values. Median differences were all equal to 0, thus all distributions were centered at 0, indicating that estimates were unbiased relative to the assigned scat equivalent values (Fig. c).