Dietary plasticity and the importance of salmon to brown bear (Ursus arctos) body size and condition in a low Arctic ecosystem

Ecological flexibility within animal populations can allow for variation in resource use and foraging decisions. We estimated brown bear (Ursus arctos) diet composition in Gates of the Arctic National Park and Preserve, Alaska from 2013 to 2015 to evaluate how variation in foraging behavior influences body condition and size. We used stable carbon (δ13C) and nitrogen (δ15N) isotope analyses of sectioned brown bear hair samples to evaluate assimilated diet. We then developed a set of a priori linear models to evaluate differences in the diet composition of brown bears (n = 80) in relation to body fat (%) and mass. The proportion of meat (salmon [Oncorhynchus keta] and terrestrial meat combined) in the diet from July through late September varied between male and female bears, with males (x-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\stackrel{-}{x}$$\end{document} = 62%, SD = 30) assimilating significantly more meat than females (x-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\stackrel{-}{x}$$\end{document} = 40%, SD = 29). Most of the meat consumed came from marine-derived resources for males (53% of the total diet or 86% of the meat) and females (31% of the total diet or 77% of the meat). As we found the range of observed diets was unrelated to physiological outcomes (i.e., percentage body fat), we suggest that ecological flexibility within populations may provide an adaptive advantage by allowing individuals to reduce competition with conspecifics by foraging on alternate food resources. Identifying variable foraging behaviors within a population can allow for a better understanding of complex behaviors and, ultimately, lead to more informed management decisions related to habitat use, development, and harvest.


Introduction
Ecological flexibility within animal populations can facilitate diverse behaviors among individuals (Gordon 1991), including variation in resource use and foraging decisions (Bolnick et al. 2003;Svanback and Bolnick 2007). The niche variation hypothesis (NVH; Van Valen 1965) posits that populations with greater niche breadth should display higher levels of among-individual phenotypic and behavioral variation and that behaviorally flexible individuals should have an adaptive advantage. Variation in resource use within a population can allow individuals to meet their energetic needs by exploiting underutilized resources, thereby reducing competition among conspecifics while maintaining a similar level of fitness ). Foraging plasticity within a species has been well documented (Bolnick et al. 2003;Araújo et al. 2011;Edwards et al. 2011;Lafferty et al. 2015;Mangipane et al. 2018) and demonstrates how individuals within a population can range from generalist to highly specialized.

3
Brown bears (Ursus arctos) are generalist omnivores (Hilderbrand et al. 1996) that persist in diverse habitats Belant et al. 2010;Hilderbrand et al. 2018a) with a variety of available food resources. Under varying environmental conditions across the species' range, brown bear populations have diets that range from those dominated by vegetation (McLellan et al. 1995) to diets that are high in animal protein, such as salmon (Oncorhynchus spp., Hilderbrand et al. 1999) or ungulates (Persson et al. 2001). Within populations, brown bear can also display variable foraging strategies with diets that range from highly mixed (i.e., meat and vegetation) to exceedingly specialized (Lafferty et al. 2015;Costello et al. 2016;Mangipane et al. 2018). Despite variation in diet, individuals within a given population can achieve similar body fat levels (Mangipane et al. 2018). The ability to achieve similar physiological outcomes (i.e., body condition) regardless of diet demonstrates the ability for brown bears to display high levels of behavioral plasticity (Van Daele et al. 2012) as seen through foraging behavior.
Differences in foraging strategies among individuals within a population may be based on body size and daily energy requirements (Robbins et al. 2007). Studies of captive brown bears suggest that mixed diets of 24 ± 5% meat and 76 ± 5% fruit are the most biochemically efficient for body mass and fat gain in small-bodied bears (Felicetti et al. 2003;Robbins et al. 2007). However, larger-bodied bears have higher energy requirements due to their increased size, which makes foraging on spatially dispersed berries less energetically efficient (Welch et al. 1997). Therefore, large bears often default to an energy-maximizing strategy in which they consume large amounts of highly nutritious meat Costello et al. 2016;Mangipane et al. 2018). In addition to the nutritional benefits small-bodied bears receive from foraging on biochemically optimal mixed diets, they may also benefit by reducing competition and the associated risks posed by larger individuals. For example, female brown bears in southeast Alaska consumed less salmon than sympatric larger males, presumably to reduce risk of infanticide (Ben-David et al. 2004). Therefore, alternate foraging strategies can provide nutritional benefits as well as benefits related to risk reduction, and these advantages ultimately provide support for brown bear behavior following the NVH (Van Valen 1965;Lafferty et al. 2015;Mangipane et al. 2018).
Our objective was to estimate brown bear diet composition in and around Gates of the Arctic National Park and Preserve in north central Alaska, and to evaluate how variation in foraging behavior is related to body size and condition (i.e., percentage body fat). In line with the NVH, we predicted that all individuals would achieve similar levels of body fat despite variable diets. Additionally, we predicted that due to higher daily energy requirements (Robbins et al. 2007), larger-bodied bears would use an energy-maximizing strategy ) and consume high levels of meat. Alternatively, if the NVH did not hold and variability in individual diets results in differing body fat percentages, smaller bears could have higher body fat percentages due to their ability to forage closer to a biochemically optimal diet (Welch et al. 1997;Robbins et al. 2007).

Study area
Gates of the Arctic National Park and Preserve is a remote and undeveloped park located in the Brooks Range of northern Alaska ( Fig. 1; 68° N, 154° W). The 34,287 km 2 landscape is characterized by spruce forests (Picea spp.), lowland riparian areas, and high alpine terrain. Average annual temperature is − 5 °C; however, summer temperatures can reach above 20 °C (Wilson et al. 2014). Snow cover is present in the study area from October through May. Brown bear food resources in Gates of the Arctic are considered to be more limited than in more southerly coastal systems due to a short growing season and low ungulate density (Gasaway et al. 1992). Terrestrial meat resources available to bears include moose (Alces alces), caribou (Rangifer tarandus), Dall's sheep (Ovis dalli), and Arctic ground squirrels (Urocitellus parryii). Berries (Vaccinium spp., Empetrum nigrum, Shepherdia canadensis) and herbaceous vegetation are also available on a seasonal basis. Although not previously thought to be a significant food resource for bears in Gates of the Arctic (Sorum et al. 2019), chum salmon (Oncorhynchus keta) have been found to occur in the system in relatively low abundance from mid-July to early September (Alaska Department of Fish and Game Anadromous Waters Catalog https ://www.adfg.alask a.gov/sf/SARR/AWC/Catal og).

Animal capture and sample collection
In spring (April-June) 2014-2016, we captured brown bears via aerial darting from helicopters (Taylor et al. 1989). Bears were anesthetized with a 10-12 mg/kg mixture of tiletamine hydrochloride and zolazepam hydrochloride (Telazol; Fort Dodge Laboratories, Inc., Fort Dodge, IA; Taylor et al. 1989). At the time of capture, we determined sex, estimated age based on tooth wear (Hilderbrand et al. 2018b), and weighed individuals using an electronic scale (± 0.5 kg; MSI-7200, Measurement Systems International, Seattle, WA). We then used bioelectrical impedance analysis to estimate percentage body fat (Farley and Robbins 1994;Hilderbrand et al. 1998). Individuals with body fat percentages below 3% (n = 4) were likely the result of measurement errors, due to the fact that survival with less than 3% body 1 3 fat is unlikely (J. Teisberg, unpublished data). Therefore, we rounded body fat values up to 3% for these individuals. Additionally, we removed one individual whose body fat was 50.4%, also likely due to a measurement error. At the time of capture we also collected guard hair samples from between the front shoulders to estimate assimilated diet during the previous fall (Hobson et al. 2000). All hair samples were placed in individually labeled paper envelopes and stored in a − 20 °C freezer until analyzed.

Evaluating diet composition
We used stable carbon (δ 13 C) and nitrogen (δ 15 N) isotope analyses of the hair samples to evaluate assimilated diet, with each bear-year combination identified as an individual sampling unit. Samples were analyzed by the Environment and Natural Resources Institute Stable Isotope Laboratory at the University of Alaska, Anchorage (https ://www.uaa. alask a.edu/enri/labs/sils) and followed the methods and reporting outlined by Rogers et al. (2015); Mangipane et al. (2018). To assess the influence of diet on body condition, we used sectioned guard hair samples (from hair collected in spring just after den emergence) to evaluate assimilated diet during the time period when fat gain is most important for bears (i.e., hyperphagia; Cattet et al. 2002). Hair growth in bears occurs from early May through October ). Based on a 1.5 cm/month growth rate (Felicetti et al. 2004) and the estimated assimilation rate for dietary changes to be detectable in hair (isotopic equilibrium is reached at 30-40 days; Oelze 2016), we used the 4 cm of hair closest to the root for analysis to represent the time period from July through late September of the previous year. Additionally, we adjusted δ 13 C values of hair samples collected in 2015 and 2014 by -0.022‰ and -0.044‰, respectively, to account for the Suess effect, which is the global decrease in δ 13 C content of atmospheric carbon dioxide, as a consequence of humans burning fossil fuels during the past 150 years (Tans et al. 1979;Hopkins and Ferguson 2012). We report isotopic values in delta (δ) notation such that δ 13 C or δ 15 N = [(R sample / R standard ) − 1] × 1000, where R sample and R standard are the 13 C/ 12 C, or 15 N/ 14 N, ratios of the sample and standard, respectively. The standards are PeeDee Belemnite limestone for carbon and atmospheric N 2 for nitrogen.
To determine the proportional contribution of major foods to brown bear diets, we a priori identified three isotopically distinct and important food categories: vegetation, terrestrial meat, and salmon. Although location-specific isotope data for food items in our study area would have been preferred, these data were unavailable at the time of this study. Therefore, we used existing isotopic values reported  (Table 1). We included generalized plant values and terrestrial meat values for Arctic and sub-Arctic tundra habitats that were previously reported by Mowat and Heard (2006). Additionally, we included values for Chum salmon in Alaska previously reported in a meta-analysis by Johnson and Schindler (2009). Using δ 13 C and δ 15 N values for each bear as well as δ 13 C and δ 15 N values, standard deviations, and trophic discrimination values for the major food categories (Mowat and Heard 2006;Johnson and Schindler 2009), we confirmed suitable isotopic mixing space based on visual inspection of the isotopic bi-plot (Fig. 2). We then estimated the proportional contribution of each food category to the diet of brown bears using the Bayesian isotopic mixing model MixSIAR (version 3.1.10; Stock and Semmens 2013). We modeled sex as a fixed effect and included a random effect for individual as well as a term for process error (Moore and Semmens 2008). We fit our model with the Markov Chain Monte Carlo (MCMC) method using uninformative priors to estimate posterior distributions for each food category (Carlin and Siddhartha 1995). We used Gibbs sampling and applied 160,000 chain lengths, a burn-in of 10,000, and thinning of 10 across three chains with JAGS (Plummer 2013). We confirmed model convergence by visually inspecting trace plots and with the Gelman-Rubin diagnostic ( R < 1.05 indicating convergence; Gelman and Rubin 1992), subsequently reporting the mean posterior proportional contributions of each food category (± 95% credible intervals) to the diet of individual bears.

Models and model selection
To avoid including correlated covariates in the same model, we tested continuous covariates using the Pearson productmoment correlation coefficient (r), assuming multicollinearity existed if |r|> 0.70 (Dormann et al. 2013). We then developed a set of models to evaluate the influence of sex and proportion of total meat in the diet (salmon and terrestrial meat combined) on body mass and body fat. Initially, we included age in the model set; however, it was not retained in the final model because it did not improve model performance (i.e., did not improve model fit and added an additional variable) and was not significant in any of the models it was included in. Additionally, the random effect of individual did not improve model fit, and due to only having three years of data, we did not include it in the analysis (Bolker et al. 2009). We then chose the highest ranked model based on small-sample Akaike's Information Criterion (AIC c ; Burnham and Anderson 2002). Only the most biologically relevant covariates were included in the analysis to avoid over-parameterization. All statistical analyses were conducted using Program R (v3.2.2, www.r-proje ct.org).

Results
Spring hair samples were collected from 80 (n = 25 male, n = 55 female) independent adult bears during 2014-2016. This included 58 unique individuals, and 22 recapture events. Due to uncertainty of the reproductive status of females during the fall prior to their first spring capture, females were not further categorized based on the presence or absence of dependent young. Carbon (δ 13 C) values ranged from − 23.72 to − 17.45 and nitrogen (δ 15 N) Table 1 Isotopic ratios of carbon (δ 13 C) and nitrogen (δ 15 N) and associated discrimination factors of potential food sources for brown bears (Ursus arctos), Gates of the Arctic National Park and Preserve, Alaska, USA a Meta-analysis of Chum salmon (Oncorhynchus keta) isotopic values reported by Johnson and Schindler (2009) b Terrestrial meat values from Arctic and sub-Arctic tundra systems reported by Mowat and Heard (2006) c Generalized plant baseline value reported by Mowat and Heard (2006) (Fig. 2). Although we acknowledge some uncertainty due to our isospace geometry (i.e., our sources do not form a prefect equilateral triangle; Phillips and Gregg 2003), our food categories did demonstrate reasonable separation with consumers falling within the mixing area (Fig. 2). Diets varied between male and female bears, with males assimilating significantly higher amounts of meat (salmon and terrestrial meat combined) than females (F 1, 63 = 7.143, p = 0.0117; Table 2). The contribution of total meat to the diet of brown bears ranged from 3 − 94% ( − x = 62%, SD = 30) for males and from 2 to 96% ( − x = 40%, SD = 29) for females. Conversely, we found that vegetation accounted for 6 − 97% ( − x = 41%, SD = 32 for males and from 4 to 98% ( − x = 59%, SD = 28) for females. We found that salmon alone (Fig. 3) comprised 53% (SD = 31, range = 2-93%) of the diet of males, on average, which was greater (F 1, 63 = 8.017, p = 0.0047) than that of females ( − x = 31%, SD = 27, range = 1-94%). Terrestrial meat assimilation was not different between males and females (F 1, 63 = 0.001, p = 0.4450) with males ranging from 1 to 21% ( − x = 9%, SD = 7) and females from 1 to 29% ( − x = 9%, SD = 7).
When evaluating the influence of the proportion of assimilated meat in the diet on body mass and body fat, there was correlation between continuous variables (|r|≤ 0.621); however, it was insufficient to preclude any variables from analysis as they were all less than the specified threshold. We collected 65 (49 female, 16 male) spring bear weights. Based on AIC c , the highest ranked model for evaluating the effect of assimilated dietary meat intake on body mass included the proportion of total meat in the diet and sex (Table 3). The proportion of meat in the diet was positively correlated with body mass (F 2, 62 = 26.69, p = 0.0012, Table 4), with larger bears consuming higher proportions of meat than smaller individuals (Fig. 4). Additionally, males were larger than females (p < 0.0001). On average, males weighed 138.0 kg (SD = 51.1, range = 64.0-247.5 kg), whereas the average female weighed 88.5 kg (SD = 15.8, range 53.5-136.5 kg).
Body fat measurements were collected from 56 individuals (46 female, 10 male) with measurements ranging from 3.0% to 33.7% ( − x = 12.1%, SD = 6.8). The highest ranked model for evaluating the influence of diet on body fat only   (Table 3). However, we found no relationship between body fat and sex (F 1, 54 = 2.79, p = 0.1000, Table 4). Our second highest ranked model was our null hypothesis which was less than 2 ∆AIC c from our top ranked model. Given that there were no significant variables in our highest ranked model, the null suggests similar findings to those of our highest ranked model. Therefore, the null was not considered as a competing model.

Discussion
Brown bears in Gates of the Arctic National Park and Preserve demonstrated variation in foraging behavior consistent with the NVH (Van Valen 1965) and followed dietary patterns scaled to body size as described by Robbins et al. (2007). Bears exhibit sexual size-dimorphism, with males generally having greater mass than females (Lefranc et al. 1987;Swenson et al. 2007;Hilderbrand et al. 2018b). Due to increased nutritional requirements needed for large-bodied males to maintain their body sizes, relatively higher levels of protein consumption are common . Although mixed diets can be beneficial for brown bears (Robbins et al. 2007), larger individuals cannot obtain a biochemically optimal diet as easily as small individuals due to the energetic costs of foraging on dispersed berries (Welch et al. 1997). Increased meat consumption among largerbodied individuals has been documented in other areas of Alaska, including systems with abundant salmon (Mangipane et al. 2018). Although increased body size imposes higher daily energy requirements, the benefits of larger size can provide reproductive (e.g., increased access to mates or greater recruitment; Noyce and Garshelis 1994;Zedrosser et al. 2007;Hilderbrand et al. 2019) and survivorship (e.g., decreased intra-specific predation) benefits. For example, increased body size can provide a positive relationship on the age of first reproduction for male and female bears, female litter size, and weight of offspring (Rode et al. 2014). Therefore, maintaining a large body size may be a desirable trait despite the associated energetic costs. Variation in diet was not associated with body fat (%), supporting our hypothesis of behavioral plasticity among individuals. By extension, we suggest that ecological flexibility within populations may provide an adaptive advantage by allowing individuals to reduce competition with conspecifics by foraging on alternate food resources (i.e., the NVH). For example, although brown bears on average assimilated > 40% salmon in 2013, there was substantial variation in diet among individual (Fig. 3), suggesting that individuals can subsist on variable diets ranging from highly mixed to highly specialized. Although variation in diet can occur due to home-range specific recourse availability (i.e., home ranges that do not have access to anadromous streams), salmon was available to all bears in our study. Therefore, proportional use of salmon may have been more closely tied to an individual's ability to compete for and defend salmon resources. Previous works have found support for the NVH in resource-rich systems (Lafferty et al. 2015;Mangipane et al. 2018); however, this had not been tested in less productive systems. Although brown bears in Gates of the Arctic may not have the resources required to become as large as bears in resource-rich systems (Hilderbrand et al. 2018b), they should be more efficient at gaining fat since they are smaller and therefore have lower daily energy requirements, allowing them to forage more efficiently on dispersed berries (Welch et al. 1997) and closer to a biochemically optimal diet (Robbins et al. 2007). Brown bears in our study did not assimilate the proposed optimal diet (total meat consumed was 62% for males and 40% for female versus the proposed optimal of 24% meat); however,  they foraged closer to optimal than larger-bodied bears in the interior of Lake Clark National Park and Preserve ( − x = 70% total meat; Mangipane et al. 2018). Additionally, because the optimal diet (Robbins et al. 2007) was calculated based on dry matter intake, whereas our estimates are based on assimilated diet, these values may not be directly comparable, and therefore, may be closer to optimal than we report. Despite having a more biochemically optimal diet, bears in Gates of the Arctic on average had much lower fat levels than bears in systems with abundant marine resources ( − x = 11.6% in Gates of the Arctic versus − x = 18.6% in Lake Clark, − x = 15.6% in Kodiak, and − x = 12.2% in Katmai; Hilderbrand et al. 2018a). This suggests that although bears in our study may be small enough to maximize fat gain with a nutritionally mixed diet (Robbins et al. 2007), they may ultimately be unable to do so due to limited resource availability and a shorter active season. In addition to having lower body fat than brown bears in resource-rich systems, bears in Gates of the Arctic are also smaller bodied (Hilderbrand et al. 2018b). Nutrient availability is the most critical factor influencing brown bear body size and population density Mowat and Heard 2006). Therefore, given that systems at higher latitudes are known to be less nutrient rich (Cramer et al. 1999), bears in our study may be unable to attain the levels of body mass and reproduction seen in more productive systems. Unlike female bears in highly productive systems that have been found to grow throughout their life, female bears in Gates of the Arctic reached an asymptote in lean body mass, indicating that the cost of reproduction may be equal to the available nutrient intake possible for brown bears in the system (Hilderbrand et al. 2018a). Therefore, morphological (i.e., body size) and physiological (i.e., percent body fat) differences between bears in our study and those in more productive systems suggest that although bears can be nutritionally flexible in nutrient poor systems, ceilings on overall nutrient acquisition may ultimately limit body condition and size, and thereby reproduction and potentially population density.
Although limited resources may restrict dietary choices in Gates of the Arctic, we found that the assimilated diet of bears included a substantial salmon component. Past studies of brown bears in Arctic regions have documented extensive use of plant based foods (Hechtel 1985;MacHutchon and Wellwood 2003;Edwards et al. 2011), caribou (Gau et al. 2002), aquatic browsers (i.e., moose and beaver [Castor Canadensis]; Edwards et al. 2011)], and small mammals (Hechtel 1985;Edwards et al. 2011). Salmon is an important food resource for many brown bear populations Mangipane et al. 2018), and our results lend additional support for the recent documentation of salmon consumption by bears in this interior, montane system in the low Arctic (Sorum et al. 2019). Gates of the Arctic has relatively low salmon abundance, with limited spatial and temporal distribution (Alaska Department of Fish and Game Anadromous Waters Catalog https ://www.adfg.alask a.gov/ sf/SARR/AWC/Catal og). The relatively high proportion of salmon consumed by brown bears (53% for males and 31% for females, on average), given their limited availability in the system, demonstrates the high nutritional importance of salmon to these bears, and suggests that salmon is a preferred food when available in the fall. Therefore, some brown bears may be opting to use an energy-maximizing strategy (i.e., consuming high amounts of easily attainable, nutrient dense salmon; Hilderbrand et al. 1999) over a biochemically optimal diet that relies on widely dispersed berries (Robbins et al. 2007). Given these results, emphasis should be given to the importance of science-based salmon management in the system, and the potential impacts reduced salmon runs could have on brown bears in this nutritionally limited system should be evaluated (Hilderbrand et al. 2004).
Our results suggest that the diets of bears in Gates of the Arctic include higher proportions of salmon than previously known and much lower proportions of terrestrial meat than expected. Additionally, plants accounted for > 80% of the diet of bears that consumed < 20% salmon. Given that our study focused on July through late September diets of bears to evaluate diet during the time that is most biologically relevant to fat gain, our results are likely seasonally biased as to reflect food resources that are available during the sampling period. Therefore, if salmon were to be used by brown bears, we would expect a higher proportional contribution to be detected when fish are spawning. Similarly, the highest proportion of use for terrestrial meat resources would likely be in spring when vulnerable neonate ungulates and carrion from winter-killed ungulates become available on the landscape (Reynolds and Garner 1987). Although our results reflect July through late September diets, the high amount of salmon and low (~ 9%) amount of terrestrial meat observed in the diets of male and female brown bears raises pertinent questions that should be further evaluated regarding the annual diets of bears in Gates of the Arctic in relation to the management of prey species, habitat use, development, and harvest. Such issues for Gates of the Arctic include evaluating potential impacts of development on foraging habitats, accounting for areas of heightened bear activity in management planning (such as along anadromous waterways), and increased harvest potential with increased access. Additionally, understanding individual differences in foraging behavior within populations is essential for species management and long-term conservation (Fuller and Sievert 2001). As previously suggested (Edwards et al. 2011), identifying variable foraging behaviors within a population can provide a better understanding of complex population dynamics, and ultimately lead to more informed management decisions. Populations of generalist omnivores often consist of specialist individuals (Vander Zanden et al. 2010).
Therefore, applying a one-size-fits-all approach to generalist populations without acknowledging among-individual variation may ultimately hinder the desired outcome of management actions.