* E-mail: sjorgensen@mbayaq.org Affiliations Monterey Bay Aquarium, Monterey, California, United States of America, Hopkins Marine Station, Stanford University, Pacific Grove, California, United States of America ⨯
Affiliation Hopkins Marine Station, Stanford University, Pacific Grove, California, United States of America ⨯
Affiliation Hopkins Marine Station, Stanford University, Pacific Grove, California, United States of America ⨯
Affiliation Hopkins Marine Station, Stanford University, Pacific Grove, California, United States of America ⨯
Affiliation Université Paris VI, Paris, France ⨯ Affiliation Point Reyes National Seashore, Inverness, California, United States of America ⨯Affiliation Hopkins Marine Station, Stanford University, Pacific Grove, California, United States of America ⨯
Elucidating how mobile ocean predators utilize the pelagic environment is vital to understanding the dynamics of oceanic species and ecosystems. Pop-up archival transmitting (PAT) tags have emerged as an important tool to describe animal migrations in oceanic environments where direct observation is not feasible. Available PAT tag data, however, are for the most part limited to geographic position, swimming depth and environmental temperature, making effective behavioral observation challenging. However, novel analysis approaches have the potential to extend the interpretive power of these limited observations. Here we developed an approach based on clustering analysis of PAT daily time-at-depth histogram records to distinguish behavioral modes in white sharks (Carcharodon carcharias). We found four dominant and distinctive behavioral clusters matching previously described behavioral patterns, including two distinctive offshore diving modes. Once validated, we mapped behavior mode occurrence in space and time. Our results demonstrate spatial, temporal and sex-based structure in the diving behavior of white sharks in the northeastern Pacific previously unrecognized including behavioral and migratory patterns resembling those of species with lek mating systems. We discuss our findings, in combination with available life history and environmental data, and propose specific testable hypotheses to distinguish between mating and foraging in northeastern Pacific white sharks that can provide a framework for future work. Our methodology can be applied to similar datasets from other species to further define behaviors during unobservable phases.
Citation: Jorgensen SJ, Arnoldi NS, Estess EE, Chapple TK, Rückert M, Anderson SD, et al. (2012) Eating or Meeting? Cluster Analysis Reveals Intricacies of White Shark (Carcharodon carcharias) Migration and Offshore Behavior. PLoS ONE 7(10): e47819. https://doi.org/10.1371/journal.pone.0047819
Editor: Sharyn Jane Goldstien, University of Canterbury, New Zealand
Received: April 11, 2012; Accepted: September 17, 2012; Published: October 29, 2012
Copyright: © 2012 Jorgensen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This project was funded by the Sloan, Moore and Packard Foundations as a part of the Tagging of Pacific Pelagics (TOPP) program of the Census of Marine Life. The Monterey Bay Aquarium Foundation also provided financial support. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: SJJ is currently employed by the Monterey Bay Aquarium. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
Highly migratory apex marine predators such as tunas and sharks likely impose ecological pressures across a variety of different ecosystems including coastal and open ocean habitats [1]. Such broad movement prohibits continuous direct observation of fine-scale behaviors and ecological interactions. Pop-up satellite archival transmitting (PAT) tags have been developed to describe these unobservable phases. However, technological and cost restrictions have for the most part limited PAT tags to three sensors which archive light, pressure and temperature observations, used to determine geographic position, swimming depth and environmental temperature, respectively. Though these parameters are instrumental in describing core areas and environmental preferences, differentiating the functional significance of habitat use has remained challenging beyond broadly categorizing behaviors such as ‘transiting’ and ‘foraging’ based upon observations of more linear and more tortuous horizontal movement patterns [2]. There remains a strong need for novel analysis methods to provide information on finer scale behaviors within these broad categories, to better understand how mobile ocean predators utilize the ecosystems they inhabit.
The oceanic migrations of white sharks (Carcharodon carcharias) have only recently been described following the advent of PAT tags [3]–[8]. From these tagging studies the seasonal patterns of long-distance migrations have emerged in the northeastern Pacific. Mature white sharks, and some sub-adults (adolescents), spend a portion of time around coastal aggregation sites (approximately August to January), primarily near pinniped rookeries, and another portion offshore in pelagic habitats. The offshore habitat extends from the western coast of North America, between Mexico and Canada, to the Hawaiian Archipelago as far west as Midway Island. White sharks range throughout this pelagic habitat, but much of the activity, particularly for males, is concentrated in a region centered approximately halfway between the Baja peninsula and the big island of Hawaii [8]. This area has been referred to as an offshore focal area [6], a shared offshore foraging area [7], and the ‘white shark Café’ [8].
Animals undertake extensive migrations for a number of reasons including escape, dispersal, foraging and reproduction [9]. The seasonal migratory movements of white sharks from coastal California to offshore waters was first observed by Boustany et al. [3] who hypothesized it was most likely related to some type of foraging or reproductive opportunity. Weng et al. [6] noted that one tagged white shark of unknown sex made rapid vertical movements in the water column for extended period while offshore, and suggested a potential courtship behavior. Domeier and Nasby Lucas [7] noted that white sharks tagged off Guadalupe Island, Mexico, did not visit California but visited the same offshore locations including the Café - the primary area where the two groups overlapped seasonally. They concluded that this must be a common foraging area, given a large geographical expanse, and apparent sexual separation (also see [10]). Jorgensen et al. [8] noted the Café was primarily defined by the presence of males converging during spring within a much smaller core area coincident with an increased rate of vertical movement while females visited the Café center only briefly. They cited this as support for a potential mating area. Recent empirical evidence from stable isotopic analysis of white shark tissues sampled off the coast of California confirmed that white sharks foraged while offshore, but questioned whether foraging was the primary benefit, since the rate of prey consumption offshore was estimated to be approximately half of that occurring at the coast [11]. Unfortunately, difficulty in observing sharks during this offshore phase and limitations in PAT data analyses have made behavioral and physiological conclusions elusive. While foraging or mating remain the primary candidate explanations, specific hypotheses that can be tested with available or foreseeable methods are needed to ultimately elucidate this important aspect of white shark ecology specifically in this offshore location.
Patterns of vertical behavior recorded from electronic tags can be an effective tool to infer behavior in fishes and sharks [12]–[15] such as courtship/spawning behavior [16], [17]. A number of methods have been effectively used to analyze detailed fine-scale archival depth data to differentiate behaviors (reviewed in Bradford et al. [18]). Due to behavioral variability among individuals it is important to have large sample sizes to attain generalizable results, which has been a limiting factor. Large PAT tag datasets exist for white sharks, however, current analyses of vertical behavior have relied on the use of archival data from rarely recovered PAT tags, which are rich in detail but poor in general representation due to low replication of individuals [5]–[8], [19]–[21].
To date, some specific patterns in white shark vertical behavior have been identified, along with hypotheses about their ecological meaning. Goldman and Anderson [22] identified signature swimming depth patterns in white sharks patrolling near seal rookeries at the Farallon Islands from active tracking (primarily between 30 m and just below the surface). Boustany et al. [3] showed this coastal signature changed as individuals left the coast and began migrating, at which point they swam primarily at the surface with infrequent dives to 500 m. Weng et al. [6] showed that while offshore, white sharks engaged in ‘rapid oscillatory diving’ (ROD), noting that one individual made repeated vertical excursions below the surface mixed layer up to 96 times in 24 hours. A subsequent study demonstrated that ROD occurred primarily in the Café as males converge there during spring [8]. A fourth pattern, identified from individuals near Hawaii closely mirrored the diel vertical migration typical of the deep scattering layer (DSL) [23] and was attributed to foraging within that community [8].
An important methodological objective of the present study was to provide a new dive behavior analysis approach using entire PAT tagging datasets, not just recovered tags with detailed archived data, to ascertain more behavioral information than previously possible. To accomplish this we used a previously published C. carcharias PAT tag dataset [8] and applied a clustering analysis to transmitted summary records to objectively differentiate dive behavior modes throughout the population's range in the northeastern Pacific. We then sought to use the subset of records with full archival data to validate the results. Finally, we looked for spatial, temporal and sex-based patterns in the diving behavior to further inform the discussion of the potential for foraging and/or mating in the white shark Café.
This project was conducted with permits from the California Department of Fish and Game, National Oceanographic and Atmospheric Administration, Office of the National Marine Sanctuaries (under permit MULTI-2005-005; MULTI-2009-005), U. S. National Park Service, and under Stanford University animal care protocol 10765 which specifically approved the tagging methodologies used in this study.
White sharks were tagged with PAT tags in central California as previously described in Jorgensen et al. [8]. A table detailing the deployments of all tags analyzed in this study is available in the online data supplement at http://rspb.royalsocietypublishing.org/content/277/1682/679/suppl/DC1.
We then calculated a distance matrix to determine the similarity among the 5571 days based on differences in vertical distribution in the 11 depth bins (‘pdist’ function in Matlab, Mathworks, v 7.9.529). For distance calculations we used the ‘City block’ measure, also known as the ‘Manhattan’ measure. The ‘Euclidian’ distance measure (geometric straight line) is likely the most common distance measure for clustering analysis. However, since distance is computed in multi-dimensional space, the ‘Euclidian’ measure is inappropriate if dimensions are of differing scales. In this case, since the bin data were of varying depth ranges (from 5 to 300 m), the method would tend to be biased by those dimensions with larger scales. Instead we calculated ‘City block’ distance, also a very common measure, which is simply the average distance along each dimension, and is more appropriate for discrete data. To maximize contrast between days we set bin values
From the distance matrix, we created a hierarchical cluster tree using an un-weighted average distance (UPGMA) linkage algorithm (‘linkage’ function in Matlab, Mathworks, v 7.9.529). These resulting clusters were plotted along with a dendrogram using the Matlab ‘dendrogram’ function.
For each day represented by depth histogram data, a median geographic position (latitude and longitude) was estimated by fitting geolocation data to a Bayesian state space model according to previously described methods [24]. These positions were grouped by cluster and plotted to detect geographic, seasonal and sex based behavior patterns.
Archival data in 60 sec resolution were available for nine individuals, and comprised 43% of available histogram days. Following the clustering procedure, the days with archival data became interspersed among the clusters. These data were queried by cluster to examine differences in diving patterns. This step was to investigate diel patterns not discernible from 24-hour histograms, and trends in vertical velocities determined from the change in measured depth between successive (60 sec) readings. To visualize diel patterns we compiled all archival data for each cluster and plotted the density of depth readings over a single 24-hour cycle. Because the range of the white sharks spanned 60° of Longitude, we aligned the data to correspond to local time for the 24 h plots. This was done by interpolating the longitude estimate for each datum along the estimated track using the Matlab function ‘interp1’ with the ‘spline’ method, and then adjusting for local time according to longitude.
To determine the spatial dependence of specific behavior modes in the Café, we calculated the proportion of data from each mode versus their geographic position within the Café. This analysis was done using longitude, since longitude estimation is more precise and latitude estimation can be confounded by non-normal spatially dependent error [24]. The Café center was defined by the local maximum in the distribution of longitude estimates across the entire data set. We used simple linear regression to examine the prevalence of different behavior modes as a function of distance from the Café center.
To determine the fraction of time male and female white sharks engaged in particular behaviors while in the Café (±10 degrees of longitude from the Café center) we calculated the fraction of days each individual engaged in each behavior for each month. We then compared the median of individuals (for males and females) by month.
Four groups emerged as the dominant diving behavior modes from the cluster analysis. These four groups accounted for over 97.9% of the data. Three additional groups were differentiated (clusters 2, 6 and 7), however, they represented only 1.59, 0.48, and 0.02% of the data respectively (Figure 1). The dendrogram plot (Figure 1) illustrates the relative distance between groups, representing differences in diving behavior. These clear differences are evident from the depth-bin histogram plots for each cluster.