Oceanic adults, coastal juveniles: tracking the habitat use of whale sharks off the Pacific coast of Mexico

Study site and electronic tagging

Seventeen whale sharks, ranging from 5 to 12 m total length, were tagged with PSATs (PTT-100 standard rate PSAT; Microwave Telemetry Inc.): 16 between 2008 and 2011, in inshore waters of Bahía de La Paz, the marine protected area (MPA) of Espíritu Santo Island (ESI), and a single individual at Banco Gordo in 2012 (Fig. 1A; Table 1). A spotter airplane was used to locate sharks off ESI. Flights followed pre-defined transects over a duration of three hours. The east and west coasts of ESI (including El Bajo), respectively, were covered over consecutive days. Surveys were conducted weekly in May 2009 and from May to July in 2010. Whale shark positions were communicated to the research vessel (7 m) via radio, allowing the vessel to conduct in-water studies. Shark lengths were measured by a swimmer using a metric band or estimated by repeated on-board observations of the whale shark swimming parallel to the research vessel (Ramírez-Macías, Vázquez-Haikin & Vázquez-Juárez, 2012; Fig. S1A). Sex of the animal was determined by the presence of claspers on males, which are visible from birth. Clasper morphology was used to distinguish juvenile from adult males: claspers are short, soft, and smooth in sexually immature males, but quickly grow and calcify during maturation (Norman & Stevens, 2007). Females were categorized as either juvenile or adult based on their estimated lengths and external indications of pregnancy (i.e., distended abdomen; Ramírez-Macías, Vázquez-Haikin & Vázquez-Juárez, 2012; Acuña Marrero et al., 2014; Robinson et al., 2016). While we acknowledge the difficulty of confirming pregnancy in adult female sharks, the visual assessments made for the present study were based on extensive field observations of adult sharks, both pregnant and non-pregnant, by the first author (Ramírez-Macías, Vázquez-Haikin & Vázquez-Juárez, 2012). Given the uniquely distended nature of the pelvic region in these sharks, pregnancy seems to be the most parsimonious explanation. However, given that pregnancy was not possible to verify by direct methods (e.g., ultrasonography) we refer to putative pregnancy in whale sharks in this study. We have proceeded on that assumption. The flank of each whale shark was also photographed for individual identification (Marshall & Pierce, 2012; Ramírez-Macías, Vázquez-Haikin & Vázquez-Juárez, 2012).

Whale sharks were tagged using a spear gun with a standard rubber band. The full tag setup consisted of the PSAT unit, a 136 kg test monofilament tether and a stainless steel tag anchor. All whale sharks were tagged on the left side at the base of the first dorsal fin. Tags were programmed to pop off after deployment intervals of either 9 (274 days) or 12 (365 days) months (Table 1). During deployment on the study animal, Microwave Telemetry’s standard rate PSATs record temperature, depth and light-level every 2 min (for a detailed description on how PTT-100 standard rate PSATs record, archive and transmit data see Brunnschweiler (2014) and http://www.microwavetelemetry.com/fish). The maximum depth the tag model used in this study can archive is 1,285.7 m. However, an emergency release feature detaches the tag automatically when the shark is deeper than ∼1,250 m for more than 15 min to prevent the tag from being crushed at depth. In all tags attached to adult whale sharks, and juveniles JM1 and JM2 (Table 1), a constant pressure release feature, set at ±10 m for 4 days, was enabled (Brunnschweiler et al., 2009; Brunnschweiler, Queiroz & Sims, 2010). In all other tags attached to juvenile whale sharks (Table 1), the constant pressure release feature was set at ±3 m for 4 days.

Data analysis

Eight of the 17 PSATs attached to whale sharks (47.1%) uplinked to the Argos satellite system and transmitted data (Table 1). Pop-off date and the reasons for detachment were determined by constant 0 m depth readings, start time of transmission to the Argos satellite system, and status information transmitted by the PSAT. After pop-off, PSATs AF2, AF3 and AF4 each floated on the surface for several days (6, 10, and 10 respectively) before uplink to the Argos satellites, and therefore true pop-off positions are unknown.

Whale shark locations were estimated using satellite-relayed data from each tag. Recovered light-level data were used by Microwave Telemetry (MT) to estimate local time of midnight or midday for longitude calculations and day length for latitude calculations using a proprietary algorithm derived from standard celestial algorithms. Light-derived position estimates were subsequently improved by combining: (i) a swim speed filter (to constrain distances), (ii) remotely sensed SST data and tag recorded SST values to constrain probable locations, and (iii) location bathymetry and tag recorded depths to filter positions that would be too shallow. To achieve this, the MT-estimated latitudinal and longitudinal positions and error fields were used to define the area that contained all the possible positions for a given location. A 1 m s⁻¹ swim-speed filter (which is consistent with cruise speeds of pelagic sharks (Gunn et al., 1999; Sims, 2000)) was used for each estimated position to generate a circular area representing the possible positions that could be reached by the shark in the time between the current and the next location estimate. The intersection of the swim speed area and the error field produced a sub-set of possible points which were then filtered to ensure the maximum dive depth did not exceed that which was possible given the bathymetry at that location. All the available filtered points were then checked against the corresponding SST map (using OSTIA high-resolution, spatially complete, global coverage remote-sensing images) for that day and any points within a 2 °C threshold between the tag-recorded temperature and the SST were recorded as possible ‘waypoints’. After all locations were processed, the waypoints at each location were scored according to the difference in SST and distance from the original estimated location with high scores representing poorer SST matches and longer distances. The best (lowest) scoring waypoints were then connected to form a most probable track. Resighting data were also used when available by fitting them as known locations during the filtering process and these were incorporated into the final tracks.

Gaps between consecutive dates were linearly interpolated to one position per day to obtain unbiased estimates of shark space use. Furthermore, to account for the spatial error around individual geolocations, these were randomly resampled 100 times (point density data) along previously reported tag-specific longitudinal and latitudinal Gaussian error fields, 0.16°in longitude and 1.19°in latitude (Hueter, Tyminski & De la Parra, 2013; Fig. S2). Resampled geolocations were then combined with satellite-derived environmental data; the environmental data used were daily (i) sea surface temperature, SST and from NOAA Optimum Interpolation Quarter Degree Daily SST Analysis (OISST) data. Based on the OISST data we also calculated (ii) daily SST maximum gradient maps by calculating, for each pixel, a geodetic—distance-corrected maximum thermal gradient (°C/100 km), and (iii) monthly merged chlorophyll a levels (0.25°spatial resolution), acquired from GlobColour (European Space Agency—ESA). Before further analysis, point density and environmental data were averaged into 0.5°grid cells (Fig. S3).

Point density data were used to calculate shark space use by performing a kernel density interpolation with barriers in ESRI ArcGIS [v. 10.3]. To analyse the spatial relationship between environmental variables and shark space-use, a Generalized Linear Mixed Model (GLMM) analysis with penalized Quasi-Likelihood parameter estimation (PQL; to account for non-normal error distributions) was employed using R (Venables & Ripley, 2002; Austin et al., 2006). In the model, shark space-use was set as a random factor, while environmental parameters were set as fixed effect factors. An autocorrelation structure of order 1 (corAR1) was used to account for the temporal correlation in the dataset (Zuur et al., 2009). GLMM models were fitted with a normal (Gaussian family) distribution. Finally, to evaluate model performance, the concordance index (C-index; Harrell et al., 1984) was calculated using R (Hmisc package). The C-index estimates the probability of concordance between predicted and observed responses, varying between 0.5 and 1.0 with the following classification: excellent if above 0.9; good 0.9–0.8; reasonable 0.8–0.7; poor 0.7–0.6 and unsuccessful 0.6–0.5 (Swets, 1988). Model results are given in the following format: β ± SD, P, C-index, where β is a measure of the slope of the relationship. Presence of sharks in ‘coastal’ waters is defined here as continental shelf (<200 m), while ‘offshore’ refers to the open ocean (>200 m depth).

Vertical movements and diel patterns in behaviour were analyzed for individual whale sharks using archived time–depth and time–temperature data (depth and temperature resolution = ∼5.4 m and ∼0.17 °C, respectively). Due to the non-Gaussian distribution of the data, mean and median were used to summarize the results, and the non-parametric Mann–Whitney U test for comparisons (significance level = 0.05). Except for the analysis of deep dives (see below), delta limited depth and temperature values (Brunnschweiler, 2014) were removed from the raw datasets, with mean and medians used to summarize the results. Vertical and thermal niches were determined using daily minimum/maximum depth and temperature values recorded at 2 min intervals (Brunnschweiler, 2014).

Archived and daily minimum/maximum depth readings were assigned to one of three categories: epipelagic (0 m, <200 m), mesopelagic (>200 m, <1,000 m) and bathypelagic (>1,000 m). Pressure readings of 0 and 5.4 m were defined as the whale shark being at the water surface (Brunnschweiler & Sims, 2012) and deep diving was defined as diving to meso- or bathypelagic depths. To investigate at what time(s) of the day deep diving occurred, all archived depth readings including delta limited values in the meso- or bathypelagic zones were assigned to one of 24 hourly bins. An individual deep dive was defined as the time between the whale shark descending from <50 m to meso- or bathypelagic depths until reaching the surface again. The time-depth series (archived depth values at 15 min intervals) of individual deep dives with maximum non-delta limited archived depth recordings in the meso- and bathypelagic zones were plotted, visually classified based on their time-depth profiles, and their dive geometry characterized following definitions provided by Gleiss, Norman & Wilson (2011). To test the hypothesis that deeper dives (maximum daily depth) occurred in less productive waters, a Spearman Rank correlation was performed (data not normally distributed; Shapiro–Wilk; W = 0.588, p < 0.001) between maximum daily depth (using pooled data for all tracked whale sharks) and mean chlorophyll-a concentration.

Resightings and tag performance

All eight PSATs prematurely detached from sharks after being attached for 14–134 days (Table 1). Of these, four (JM1, AF1, AF2, JM2) and two (JF1, AM1) PSATs released due to activation of the constant pressure or the crush depth emergency release mechanisms, respectively. For two PSATs (AF3, AF4), the reason for premature release was unknown (Table 1). Except for PSAT AF2, for which three days were missing, latitude and longitude positions were transmitted for all days of the respective tracks. PSATs transmitted between 30 and 100% of archived pressure and temperature data (Table 1). Data from six PSATs (AF1, AF2, JF1, AF3, AM1, AF4) contained between 0.1 (AF2) and 2% (AM1), and 2.5 (JF1) and 23.2% (AM1) delta limited depth and temperature values, respectively.

Eight whale sharks were resighted post-tagging using photo-identification (Table 2). Six of them were resighted in the coastal waters of Bahía de La Paz (Fig. 1A). Two whale sharks were resighted away from the tagging site. The juvenile male tagged in Bahía de La Paz on 16 February 2011 was resighted, without the PSAT, in the protected area of Bahía de Los Angeles on 13 August 2011. The pregnant whale shark tagged at Banco Gordo on 8 May 2012 was resighted, without the tag, in the protected area of Roca Partida, the smallest of the four Revillagigedo Islands (Fig. 1B), on 22 November 2012, and was reported to be pregnant.

Horizontal movements

There was a marked difference in the horizontal movements exhibited by juvenile and adult whale sharks. All juveniles remained in the GoC, while all adult animals moved out into the eastern Pacific Ocean. The two juvenile males JM1 and JM2, tagged for 25 and 14 days respectively, remained in the shallow waters of Bahía de La Paz where they were tagged. All six other whale sharks left the tagging area and moved large distances (Fig. 1B). Tagged animals mainly occupied two areas, one in the coastal waters around the southern part of the Baja California peninsula, and the other offshore in the eastern Pacific Ocean (Fig. 2).

Juvenile JF1 spent the first weeks after tagging in and north of Bahía de La Paz before moving north to 28.7°N–113.0°W where the tag popped-off south of Bahía de Los Angeles, on 9 May, after 58 days (Fig. 1B). Pregnant whale sharks AF1 and AF2 both left the tagging area, and after a week the GoC, moved to the same offshore area (Fig. 1B). It appears that both animals moved clockwise in a large circle in June/July 2009 before heading north. Whereas shark AF1 lost its tag after 44 days on 6 July, shark AF2 was tracked for another three months. In that time period, it continued to move north (Fig. 1B). All three females, adults AF1 and AF2 as well as juvenile JF1, remained in productive waters for longer time periods before moving north.

Putative pregnant whale sharks AF3 and AF4, tagged in 2010, left the GoC shortly after tagging, similar to sharks AF1 and AF2 from 2009. The 2010 sharks remained associated with productive coastal waters in July/August, before moving to an oligotrophic offshore area in September where tags popped off (Fig. 1B). Similar to the pregnant females, the only adult male (AM1) tagged in this study left the GoC within a week after tagging and quickly moved south where it stayed in a relatively confined area in August until the PSAT popped off (Fig. 1B).

GLMM analysis showed that shark space-use was significantly influenced by SST and SST gradients (fronts). However, temperature had a negative (albeit weak) effect on space-use, while stronger thermal gradients had a positive effect (Table 3) indicating whale sharks tended to spend more time in frontal regions linked with upwelling (which resulted in colder surface water). No significant relationship was found between shark space-use and chlorophyll a concentration and model performance was reasonable (C-index = 0.7).

Vertical movements

Mean depth and temperature experienced by individual whale sharks ranged between 1.1 (JM2) and 29.4 m (AM1), and 21.64 (AF1) and 25.48 °C (JM1), respectively (Table 4). Significant day and night differences in mean depth were detected in all but one individual (JM1). Mean depth was greater during the night for all sharks except for JF1, the one juvenile that moved significantly away from the tagging area, which stayed deeper during the day (Table 4).

Excluding juvenile whale sharks JM1 and JM2, which did not leave the tagging area over the duration of PSAT attachment and dived to relatively shallow maximum depths of 32.3 and 21.5 m, respectively, time spent at the surface (<6 m) constituted 65 ± 20.7% of total time for the other whale sharks. These six whale sharks, which all moved long distances from the area where they were tagged (Fig. 1B), spent most of their time in 20–26 °C water. All exhibited regular diving throughout their tracks and had broad vertical and thermal niches. Three individuals were recorded in each of the mesopelagic (AF1, AF3, AF4) and bathypelagic (AF2, JF1, AM1) zones several times during PSAT attachment (Fig. 3). Two sharks, JF1 and AM1, recorded dives to or exceeding the maximum depth that the tags could record (1,285.8 m; Table 4).

Individual differences with regards to overall diving behaviour were evident. The four putative pregnant whale sharks (AF1, AF2, AF3, AF4) that left the GoC showed similar diving patterns, with maximum daily depths in the epipelagic zones on most days (Figs. 3A, 3B, 3D and 3F). The juvenile whale shark (JF1) that stayed within the GoC stayed shallow with a maximum diving depth of 32.3 m for the first 49 days of its track, excluding 3 April 2010 (Fig. 3C). Then, for the remaining nine days before the PSAT popped up due to activation of the deep-depth release mechanism, this whale shark showed increased deep diving activity into the meso- and bathypelagic zones during its northwards movement to the upper GoC (Figs. 1B and 3C). This shark (JF1), with a mean depth of 6 m over the entire duration of the track, marked the shallowest end of the spectrum. The adult male whale shark AM1 was the individual with the deepest mean depth (Table 4). This whale shark quickly left the GoC and moved south (Fig. 1B), with daily maximum depths in the meso- and bathypelagic zones on 52.5% of days tracked (Fig. 3E).

Meso- and bathypelagic diving and dive geometry

Overall, sharks spent a relatively small amount of time in the meso- and bathypelagic zones (Fig. 3). Dives to maximum depths in these zones occurred largely during the day, with 72.7% of all >200 m recordings archived between 06:00 and 18:00 (Fig. 4). Most deep diving activity was performed in the morning between 07:00 and 09:00 (28.6%) and in the late afternoon between 15:00 and 18:00 (20.8%) (Fig. 4).

Fifteen dive profiles with maximum depth readings in the mesopelagic zone were available for visual classification. Nine contained delta limited depth values (Brunnschweiler, 2014). Three were characterized as isolated V-dives (Fig. 5A), two as V-dives in series (Fig. 5B), four as U-dives (Fig. 5C), and six dive profiles could not be assigned to any of the geometries described by Gleiss, Norman & Wilson (2011) because of uncertainties with delta limited data points (Fig. 5D). All 15 dive profiles indicate that whale sharks quickly dived from the epipelagic zone to the maximum mesopelagic depth and back up into surface waters after the animal spent little (e.g., V-dives) or longer time periods (U-dives) at depth (Figs. 5A–5D). The pattern of rapid descents and ascents can also be inferred from the only two dive profiles that contain actual depth readings in the bathypelagic zone, but could not be reconstructed with confidence because of missing archived data points and most depth values being delta limited (Figs. 5E and 5F; see Brunnschweiler (2014) for details). Mean dive depth was negatively correlated with chlorophyll a concentration (Spearman Rank correlation; rs =  − 0.16, p <0.05).