ABSTRACT: Understanding diet composition is fundamental to making conservation and management decisions about depleted species, particularly when nutritional stress is a potential threat hindering recovery. Diet in free-ranging marine mammals is challenging to study, but stable isotope mixing models are a powerful means of estimating the contribution of prey species to diet and can improve precision by leveraging information from multiple data sources. We evaluated diet composition of a fish-eating killer whale population (Southern Resident killer whales, Orcinus orca) using 2 approaches. First, we fit generalized additive models to evaluate seasonal and interannual patterns in isotopic values across age, sex, and pod, which revealed seasonal carbon enrichment for certain pods and a recent increased nitrogen enrichment that could suggest increased Chinook salmon consumption, changing isotopic values of prey, or nutritional stress. Second, we developed a Bayesian stable isotope mixing model that accounts for the different integration times represented by bulk stable isotopes and fecal samples. Results showed that estimated prey contributions are similar between prey data sources, though the precision of estimates from periods with smaller sample sizes was improved by using an informative prior to account for the different consumption windows of the data. This study illustrates the importance of improving our understanding of how killer whale diets vary over time (both seasonally and across years) and uses a novel approach to resolve 2 sources of diet information (stable isotope, fecal samples) with different consumption windows.

中文翻译：

---

使用贝叶斯稳定同位素混合模型和广义加性模型解决食鱼虎鲸Orcinus orca的饮食变化

摘要：了解饮食组成是制定有关枯竭物种的保护和管理决策的基础，特别是在营养压力是阻碍恢复的潜在威胁时。自由放养的海洋哺乳动物的饮食很难研究，但是稳定的同位素混合模型是估算猎物对饮食的贡献的有力手段，并且可以利用来自多个数据源的信息来提高精度。我们评估了以鱼类为食的虎鲸种群（南方虎鲸，Orcinus orca）的饮食组成）使用2种方法。首先，我们拟合通用的加性模型，以评估年龄，性别和豆荚的同位素值的季节性和年际模式，这揭示了某些豆荚的季节性碳富集和近期氮富集的增加，这可能暗示了奇努克鲑鱼的消费量增加，同位素值的变化猎物或营养压力。其次，我们建立了贝叶斯稳定同位素混合模型，该模型解释了以体积稳定同位素和粪便样品表示的不同积分时间。结果表明，尽管通过使用信息性的数据来​​考虑数据的不同消耗时段可以提高样本量较小的时期的估计精度，但估计的猎物贡献在两个猎物数据源之间是相似的。