Animals are distributed relative to the resources they rely upon, often scaling in abundance relative to available resources. Yet, in heterogeneously distributed environments, describing resource availability at relevant spatial scales remains a challenge in ecology, inhibiting understanding of predator distribution and foraging decisions. We investigated the foraging behaviour of two species of rorqual whales within spatially limited and numerically extraordinary super‐aggregations in two oceans. We additionally described the lognormal distribution of prey data at species‐specific spatial scales that matched the predator's unique lunge‐feeding strategy. Here we show that both humpback whales off South Africa's west coast and blue whales off the US west coast perform more lunges per unit time within these aggregations than when foraging individually, and that the biomass within gulp‐sized parcels was on average higher and more tightly distributed within super‐group‐associated prey patches, facilitating greater energy intake per feeding event as well as increased feeding rates. Prey analysis at predator‐specific spatial scales revealed a stronger association of super‐groups with patches containing relatively high geometric mean biomass and low geometric standard deviations than with arithmetic mean biomass, suggesting that the foraging decisions of rorqual whales may be more influenced by the distribution of high‐biomass portions of a patch than total biomass. The hierarchical distribution of prey in spatially restricted, temporally transient, super‐group‐associated patches demonstrated high biomass and less variable distributions that facilitated what are likely near‐minimum intervals between feeding events. Combining increased biomass with increased foraging rates implied that overall intake rates of whales foraging within super‐groups were approximately double those of whales foraging in other environments. Locating large, high‐quality prey patches via the detection of aggregation hotspots may be an important aspect of rorqual whale foraging, one that may have been suppressed when population sizes were anthropogenically reduced in the 20th century to critical lows. A free Plain Language Summary can be found within the Supporting Information of this article.

Southern Hemisphere humpback whales (Megaptera novaeangliae) generally undertake annual migrations from polar summer feeding grounds to winter calving and nursery grounds in subtropical and tropical coastal waters. Evidence for such migrations arises from seasonality of historic whaling catches by latitude, Discovery and natural mark returns, and results of satellite tagging studies. Feeding is generally believed to be limited to the southern polar region, where Antarctic krill (Euphausia superba) has been identified as the primary prey item. Non-migrations and / or suspended migrations to the polar feeding grounds have previously been reported from a summer presence of whales in the Benguela System, where feeding on euphausiids (E. lucens), hyperiid amphipods (Themisto gaudichaudii), mantis shrimp (Pterygosquilla armata capensis) and clupeid fish has been described. Three recent research cruises (in October/November 2011, October/November 2014 and October/November 2015) identified large tightly-spaced groups (20 to 200 individuals) of feeding humpback whales aggregated over at least a one-month period across a 220 nautical mile region of the southern Benguela System. Feeding behaviour was identified by lunges, strong milling and repetitive and consecutive diving behaviours, associated bird and seal feeding, defecations and the pungent “fishy” smell of whale blows. Although no dedicated prey sampling could be carried out within the tightly spaced feeding aggregations, observations of E. lucens in the region of groups and the full stomach contents of mantis shrimp from both a co-occurring predatory fish species (Thyrsites atun) and one entangled humpback whale mortality suggest these may be the primary prey items of at least some of the feeding aggregations. Reasons for this recent novel behaviour pattern remain speculative, but may relate to increasing summer humpback whale abundance in the region. These novel, predictable, inter-annual, low latitude feeding events provide considerable potential for further investigation of Southern Hemisphere humpback feeding behaviours in these relatively accessible low-latitude waters.

Predators can impact ecosystems through consumptive or risk effects on prey. Physiologically, risk effects can be mediated by energetic mechanisms or stress responses. The predation-stress hypothesis predicts that risk induces stress in prey, which can affect survival and reproduction. However, empirical support for this hypothesis is both mixed and limited, and the conditions that cause predation risk to induce stress responses in some cases, but not others, remain unclear. Unusually clear-cut variation in exposure of Cape fur seals (Arctocephalus pusillus pusillus) to predation risk from white sharks (Carcharodon carcharias) in the waters of Southwestern Africa provides an opportunity to test the predation-stress hypothesis in the wild. Here, we measured fecal glucocorticoid concentrations (fGCM) from Cape fur seals at six discrete islands colonies exposed to spatiotemporal variation in predation risk from white sharks over a period of three years. We found highly elevated fGCM concentrations in seals at colonies exposed to high levels of unpredictable and relatively uncontrollable risk of shark attack, but not at colonies where seals were either not exposed to shark predation or could proactively mitigate their risk through antipredatory behavior. Differences in measured fGCM levels were consistent with patterns of risk at the site and seasonal level, for both seal adults and juveniles. Seal fGCM levels were not correlated with colony population size, density, and geographic location. Investigation at a high risk site (False Bay) also revealed strong correlations between fGCM levels and temporal variation in shark attack rates, but not with shark relative abundance. Our results suggest that predation risk will induce a stress response when risk cannot be predicted and/or proactively mitigated by behavioral responses.

No abstract

Machine learning algorithms are often used to model and predict animal habitat selection—the relationships between animal occurrences and habitat characteristics. For broadly distributed species, habitat selection often varies among populations and regions; thus, it would seem preferable to fit region- or population-specific models of habitat selection for more accurate inference and prediction, rather than fitting large-scale models using pooled data. However, where the aim is to make range-wide predictions, including areas for which there are no existing data or models of habitat selection, how can regional models best be combined? We propose that ensemble approaches commonly used to combine different algorithms for a single region can be reframed, treating regional habitat selection models as the candidate models. By doing so, we can incorporate regional variation when fitting predictive models of animal habitat selection across large ranges. We test this approach using satellite telemetry data from 168 humpback whales across five geographic regions in the Southern Ocean. Using random forests, we fitted a large-scale model relating humpback whale locations, versus background locations, to 10 environmental covariates, and made a circumpolar prediction of humpback whale habitat selection. We also fitted five regional models, the predictions of which we used as input features for four ensemble approaches: an unweighted ensemble, an ensemble weighted by environmental similarity in each cell, stacked generalization, and a hybrid approach wherein the environmental covariates and regional predictions were used as input features in a new model. We tested the predictive performance of these approaches on an independent validation dataset of humpback whale sightings and whaling catches. These multiregional ensemble approaches resulted in models with higher predictive performance than the circumpolar naive model. These approaches can be used to incorporate regional variation in animal habitat selection when fitting range-wide predictive models using machine learning algorithms. This can yield more accurate predictions across regions or populations of animals that may show variation in habitat selection.