An air-breathing marine mammal optimises foraging by decreasing average dive residence with increasing energy gain

<b>Abstract</b><b>1.</b><b> </b>Understanding how air-breathing diving animals respond to the quality of their foraging environment, specifically how they balance time at depth feeding and using oxygen and time at the surface breathing, is foundational to studying their foraging ecology.<b>2.</b><b> </b>We quantified the relationships between southern elephant seal dive residence and net energy gain – informed by estimates of at-sea body condition – across sexes, habitats, and temporal scales. We considered net energy gain to be a proxy of the resource quality encountered by seals, integrating information on average foraging patch conditions across a neighbourhood of patches.<b>3.</b><b> </b>After accounting for buoyancy effects and irrespective of sex and foraging habitat, we found that as net energy gain increased, seals consistently increased transit (ascent, descent) rates and decreased relative dive durations for a given depth (<i>i.e.</i> accounting for increasing duration with depth). The homogeneity of these results suggests a common pattern in which seals perform, on average, shorter, sharper dives during periods of high net energy gain.<b>4.</b><b> </b>A key prediction of the marginal value theorem (MVT) is of continued patch use until gains drop below the neighbourhood mean. Our findings align with this prediction, considering net energy gain as a proxy for forage quality integrated over a neighbourhood of multiple prey patches, and dive residence, the average residence within individual patches. Based on this, within a given neighbourhood of patches, seals would spend less time on average <i>per </i>patch when the neighbourhood quality is high (high net energy gain). This is because in areas of high neighbourhood quality, the relative difference in quality between individual patches and the mean for the neighbourhood is small (<i>e.g.</i> due to high patch density and hence low cost of moving); hence patches would deplete to the neighbourhood mean faster than they would in a lower quality neighbourhood.<b>5.</b><b> </b>These findings have important implications for work that draws inferences of foraging habitat quality from dive behaviour alone.<b>Repository description</b>This code and data repository contains the final dataset and code used to run the paper analyses and to plot the figures in the main text.It also contains the workflows in the subdirectory “./preprocessing_workflows_with_data_subset/” (see below) for processing the raw data to get it to the final test dataset stage. Accompanying this workflow, is a subset of the raw datasets. This step uses a custom-built R package called seaTracks (https://github.com/davo-b-green/seaTracks), which internally calls the aniMotum (https://github.com/ianjonsen/aniMotum) and rSRDL (https://github.com/embiuw/rSRDL) packages, which in turn do the heavy lifting. See the seaTracks GitHub repository for more information. The step for estimating drift rates relies on the slimmingDive package (https://github.com/farcego/slimmingDive).See README.docx for more details