Institute for Marine and Antarctic Studies, University of Tasmania, Private Bag 129, Hobart, 7001, Tasmania, Australia.
Australian Antarctic Division, 203 Channel Highway, Kingston, 7050, Tasmania, Australia.
Sci Rep. 2019 Jun 20;9(1):8936. doi: 10.1038/s41598-019-44970-1.
Foraging behaviour of marine predators inferred from the analysis of horizontal or vertical movements commonly lack quantitative information about foraging success. Several marine mammal species are known to perform dives where they passively drift in the water column, termed "drift" dives. The drift rate is determined by the animal's buoyancy, which can be used to make inference regarding body condition. Long term dive records retrieved via satellite uplink are often summarized before transmission. This loss of resolution hampers identification of drift dives. Here, we develop a flexible, hierarchically structured approach to identify drift dives and estimate the drift rate from the summarized time-depth profiles that are increasingly available to the global research community. Based on high-resolution dive data from southern elephant seals, we classify dives as drift/non-drift and apply a summarization algorithm. We then (i) automatically generate dive groups based on inflection point ordering using a 'Reverse' Broken-Stick Algorithm, (ii) develop a set of threshold criteria to apply across groups, ensuring non-drift dives are most efficiently rejected, and (iii) finally implement a custom Kalman filter to retain the remaining dives that are within the seals estimated drifting time series. Validation with independent data sets shows our method retains approximately 3% of all dives, of which 88% are true drift dives. The drift rate estimates are unbiased, with the upper 95% quantile of the mean squared error between the daily averaged summarized profiles using our method (SDDR) and the observed daily averaged drift rate (ODDR) being only 0.0015. The trend of the drifting time-series match expectations for capital breeders, showing the lowest body condition commencing foraging trips and a progressive improvement as they remain at sea. Our method offers sufficient resolution to track small changes in body condition at a fine temporal scale. This approach overcomes a long-term challenge for large existing and ongoing data collections, with potential application across other drift diving species. Enabling robust identification of foraging success at sea offers a rare and valuable opportunity for monitoring marine ecosystem productivity in space and time by tracking the success of a top predator.
从水平或垂直运动分析推断海洋捕食者的觅食行为通常缺乏有关觅食成功的定量信息。已知几种海洋哺乳动物物种会进行潜水，在这些潜水过程中，它们会在水柱中被动漂移，称为“漂移”潜水。漂移率由动物的浮力决定，可用于推断身体状况。通过卫星上行链路获取的长期潜水记录在传输前通常会进行汇总。这种分辨率的损失阻碍了漂移潜水的识别。在这里，我们开发了一种灵活的、分层结构的方法，用于从越来越多可供全球研究界使用的汇总时间-深度剖面中识别漂移潜水并估计漂移率。基于南部象海豹的高分辨率潜水数据，我们将潜水分类为漂移/非漂移，并应用汇总算法。然后，我们 (i) 使用“反向”折断棒算法根据拐点排序自动生成潜水组，(ii) 在组之间应用一组阈值标准，确保最有效地拒绝非漂移潜水，以及 (iii) 最后实施自定义卡尔曼滤波器以保留其余潜水在海豹估计的漂流时间序列内。使用独立数据集进行验证表明，我们的方法保留了大约 3%的所有潜水，其中 88%是真正的漂移潜水。漂移率估计是无偏的，使用我们的方法（SDDR）与观察到的每日平均漂移率（ODDR）之间的每日平均汇总剖面的均方误差的上 95%分位数仅为 0.0015。漂流时间序列的趋势符合资本繁殖者的预期，表现出最低的身体状况开始觅食，并随着它们在海上停留时间的增加而逐渐改善。我们的方法提供了足够的分辨率，可以在精细的时间尺度上跟踪身体状况的微小变化。这种方法克服了大型现有和正在进行的数据收集长期存在的挑战，并且具有在其他漂移潜水物种中应用的潜力。通过跟踪顶级捕食者的成功，为在空间和时间上监测海洋生态系统生产力提供了一个罕见而有价值的机会，从而能够可靠地识别海上觅食的成功。
