The development of an intermediate‐duration tag to characterize the diving behavior of large whales

Abstract The development of high‐resolution archival tag technologies has revolutionized our understanding of diving behavior in marine taxa such as sharks, turtles, and seals during their wide‐ranging movements. However, similar applications for large whales have lagged behind due to the difficulty of keeping tags on the animals for extended periods of time. Here, we present a novel configuration of a transdermally attached biologging device called the Advanced Dive Behavior (ADB) tag. The ADB tag contains sensors that record hydrostatic pressure, three‐axis accelerometers, magnetometers, water temperature, and light level, all sampled at 1 Hz. The ADB tag also collects Fastloc GPS locations and can send dive summary data through Service Argos, while staying attached to a whale for typical periods of 3–7 weeks before releasing for recovery and subsequent data download. ADB tags were deployed on sperm whales (Physeter macrocephalus; N = 46), blue whales (Balaenoptera musculus; N = 8), and fin whales (B. physalus; N = 5) from 2007 to 2015, resulting in attachment durations from 0 to 49.6 days, and recording 31 to 2,539 GPS locations and 27 to 2,918 dives per deployment. Archived dive profiles matched well with published dive shapes of each species from short‐term records. For blue and fin whales, feeding lunges were detected using peaks in accelerometer data and matched corresponding vertical excursions in the depth record. In sperm whales, rapid orientation changes in the accelerometer data, often during the bottom phase of dives, were likely related to prey pursuit, representing a relative measure of foraging effort. Sperm whales were documented repeatedly diving to, and likely foraging along, the seafloor. Data from the temperature sensor described the vertical structure of the water column in all three species, extending from the surface to depths >1,600 m. In addition to providing information needed to construct multiweek time budgets, the ADB tag is well suited to studying the effects of anthropogenic sound on whales by allowing for pre‐ and post‐exposure monitoring of the whale's dive behavior. This tag begins to bridge the gap between existing long‐duration but low‐data throughput tags, and short‐duration, high‐resolution data loggers.

locations and can send dive summary data through Service Argos, while staying attached to a whale for typical periods of 3-7 weeks before releasing for recovery and subsequent data download. ADB tags were deployed on sperm whales (Physeter macrocephalus; N = 46), blue whales (Balaenoptera musculus; N = 8), and fin whales (B. physalus; N = 5) from 2007 to 2015, resulting in attachment durations from 0 to 49.6 days, and recording 31 to 2,539 GPS locations and 27 to 2,918 dives per deployment. Archived dive profiles matched well with published dive shapes of each species from short-term records. For blue and fin whales, feeding lunges were detected using peaks in accelerometer data and matched corresponding vertical excursions in the depth record. In sperm whales, rapid orientation changes in the accelerometer data, often during the bottom phase of dives, were likely related to prey pursuit, representing a relative measure of foraging effort. Sperm whales were documented repeatedly diving to, and likely foraging along, the seafloor. Data from the temperature sensor described the vertical structure of the water column in all three species, extending from the surface to depths >1,600 m. In addition to providing information needed to construct multiweek time budgets, the ADB tag is well suited to studying the effects of anthropogenic sound on whales by allowing for pre-and post-exposure monitoring of the whale's dive behavior.
This tag begins to bridge the gap between existing long-duration but low-data throughput tags, and short-duration, high-resolution data loggers.

K E Y W O R D S
animal-borne data loggers, archival whale tags, biologging, diving and foraging behavior,

| INTRODUCTION
Understanding how animals use their environment at multiple scales is a key goal in behavioral ecology. Data loggers and tracking devices have been used in various forms for over 30 years to monitor animal activity during times when they cannot be observed (Cooke et al., 2004;Mate, Mesecar, & Lagerquist, 2007;Ropert-Coudert & Wilson, 2005). This type of remote monitoring is especially valuable in the field of marine mammal research because the study subjects spend the majority of their time below the surface of the water, where direct observation ranges from very difficult to impossible. Researchers working with pinniped species have had great success attaching data loggers to their subjects' pelage in order to study movement, diving physiology, and body condition over periods of months (Biuw, McConnell, Bradshaw, Burton, & Fedak, 2003;Costa & Gales, 2003;Guinet et al., 2014). Large-whale researchers have faced a much greater challenge due to the impossibility of capturing or otherwise controlling the subject during tag attachment. Two main types of tag attachment are currently used to study whales, each with advantages and disadvantages. Transdermal attachments have been used with increasing regularity for satellite-monitored tags since the mid-1990s to document long-term (months) movements (Andrews, Pitman, & Ballance, 2008; Heide-Jorgensen, Witting, & Jensen, 2003;Mate et al., 2007). Some of these tags can now function for over one year (Mate et al., 2007); however, their data can only be recovered through the Argos satellite system, which drastically limits the amount of information that can be transmitted. On the other hand, suction-cup-attached data loggers are capable of recording dive depth, body orientation, and acoustic data at rates >16 Hz (Burgess, 2000;Johnson & Tyack, 2003), but the large quantities of data generated cannot be sent via satellite, so they are stored on board for download after the tag is recovered, which typically occurs within 24 hr of deployment on large cetaceans (Fais et al., 2015;Simon, Johnson, & Madsen, 2012), with occasional longer deployments reported (34 h: 62 h: Amano & Yoshioka, 2003).
While high-resolution data loggers can record relatively large amounts of information on dive behavior, they cannot be used to characterize how that behavior changes over time due to the short-attachment duration. This knowledge gap represents the next frontier in technology development for whale research, particularly in the face of the growing need to document how these sensitive species might respond to various sources of anthropogenic disturbance such as noise from commercial vessel traffic, mid-frequency naval sonar, or seismic exploration vessels (Nowacek, Thorne, Johnston, & Tyack, 2007;Soto et al., 2006;Southall et al., 2007) beyond short-term responses . In order to better understand whale behavior over a longer temporal scale, and to identify behavioral changes that may result from exposure to anthropogenic noise, a high-resolution data logger is needed that can stay attached to a whale for periods of several weeks or more (Johnson, Tyack, Gillespie, & McConnell, 2013;Nowacek, Christiansen, Bejder, Goldbogen, & Friedlaender, 2016).
Earlier attempts to study whale diving behavior with longerduration tags have been made with some success (Baumgartner, Hammar, & Robbins, 2015;Davis et al., 2007;Schorr, Falcone, Moretti, & Andrews, 2014), although the resolution and types of data collected have remained inferior to those obtained from short-duration, highresolution data loggers. Recent development of methods for longer attachment of high-resolution data-logging packages has provided records for 7.6-16 days (Owen, Jenner, Jenner, & Andrews, 2016;Szesciorka, Calambokidis, & Harvey, 2016), representing progress toward this goal and showing the utility of high-resolution data collected over multiple days.
Here, we describe the development through four generations of the Advanced Dive Behavior (ADB) tag, a spatially explicit, highresolution (1-Hz) data logger for large whales capable of staying attached for intermediate time periods (weeks to >1 month). The design focused on a semi-implantable style that allowed the tag to record data onboard, then release from an attachment housing and float to the surface for subsequent recovery and data download. The greater attachment duration and ability to set a release date for recovery are a significant step toward the goal of measuring large whale behavior.
The data records obtained from this tag will dramatically advance our understanding of cetacean ecology.

| Tag configuration and deployment
The ADB tag is a novel configuration of the Wildlife Computers  (Bryant, 2007) was incorporated into the top of the float along with an Argos antenna, LED lights for recovery, a hydrostatic pressure sensor, and a saltwater conductivity sensor to detect surfacing events. A corrodible link wire in the form of a loop was mounted to the underside of the float for attachment to the deployment housing until the desired release time was reached.
Generation-3 and Generation-4 ADB tags were equipped with three release wires so they could be redeployed after recovery. Tags were also equipped with light level and temperature sensors as part of the original Mk10-PATF configuration. Three-axis accelerometers were incorporated into the tags in Generation 2, while magnetometers and a Fastloc-3 GPS receiver were added in Generation 3 (Table 1).
For deployment, the tag was inserted into an attachment housing constructed from 16-gauge surgical-quality 316 stainless steel consisting of an 18.5-or 14.5-cm-long, 2.6-cm-diameter shaft (Generation 1 and generations 2-4, respectively) affixed to a plate with a raised lip to protect the foam float ( Figure 1). Four cutting blades attached to a Delrin nose cone were affixed to the distal end of the shaft, along with two rows of backward facing petals in a similar configuration to implantable tags described in Mate et al. (2007). The tag was secured to the housing by threading a small screw through the release wire and then through a perpendicular tab below the housing plate.
For deployment, tags were attached to a carrier (a "pushrod") using a sculpted Delrin attachment that held the outside of the float with pressure provided by an O-ring while it applied force to the rim of the attachment housing plate. A tag and pushrod were deployed at close range (2-4 m) from a 6.8-m rigid-hulled inflatable boat using the Air Rocket Transmitter System, a modified line-thrower using compressed air (Heide-Jorgensen et al., 2003;Mate et al., 2007), charged to 125 psi (for blue whales, Balaenoptera musculus and fin whales B. physalus) or 140 psi (for sperm whales, Physeter macrocephalus). Tags were deployed 0.25-4 m forward of the dorsal fin/hump of the whale, depending on the species, and no more than 20 cm down from the midline. Care was taken to place the tag perpendicular to the surface of the whale, so the plate and float would sit flat to minimize drag. The impact of deployment separated the pushrod from the tag for recovery.

| Data collection and transmission
All collected data were stored in an onboard archive, and the complete data record could only be accessed by recovering the tag for download. ADB tags were programmed to collect sensor data (depth, light level, temperature, accelerometers, and magnetometers) at 1 Hz for the duration of all deployments. Collection of Fastloc GPS locations could occur at regular, user-specified intervals (i.e., 1 location per hour), or immediately after the whale surfaced from a "qualifying dive" defined by the user.
Argos messages were transmitted every 45 s while the whale was at the surface. A saltwater conductivity switch prevented transmissions while the tag was underwater. An Argos transmission could contain (1) a location message containing one set of Fastloc GPS pseudo-ranges; (2) one of four types of histogram summary messages for qualifying dives (time at depth, time at temperature, maximum dive depth, and dive duration); (3) a behavior message with summaries of four consecutive qualifying dives listing the dive date/time, maximum dive depth, dive duration, dive shape, and subsequent surfacing duration; or (4) a utility message summarizing battery voltage, number of Argos transmissions, and number of Fastloc attempts. The Argos messages could be assigned differing priorities and allowed same-day monitoring of the whale's diving behavior and location while the tag was attached.

| Programmed release and recovery
Release from the tag housing could be triggered by three possible criteria: Reaching a user-specified release date and time, if the estimated F I G U R E 1 A schematic drawing of the external design of the ADB tag (bottom) with the deployment housing (top)

Sensor (Resolution) Generation 1 Generation 2 Generation 3 Generation 4
Stem Dimensions (cm) 2 × 15 2 × 11.5 2 × 11.5 2 × 11.5 Fastloc version v. When a release was identified, recent locations were downloaded from Argos to define an initial search area and direction of drift for the tag. An uplink receiver and accompanying software on a computer carried onboard the recovery vessel were capable of receiving, decoding, and solving location messages sent by the tags at a range of ≤3 nautical miles. Solved locations from this system were used to focus the search area within 50 m of the floating tag, so that it could be located visually from the vessel. The three LED lights on the float made tags easier to locate at night.

| Predeployment accuracy testing
In 2007, four Generation-1 ADB tags were affixed to a life ring at a variety of angles and allowed to drift on the water for 90 min with the tags set to collect Fastloc GPS locations every 5 min. The results of those locations were compared to locations collected by a Garmin-72 GPS unit that was also affixed to the life ring to assess the accuracy of ADB-generated GPS locations.  (Table 2; Table   S1). Generation-2 ADB tags were deployed on 11 sperm whales in the northern Gulf of Mexico, USA, during summer 2011, and nine Generation-3 ADB tags were deployed in the same area during 2013.
Generation-3 and Generation-4 ADB tags were deployed on both blue whales and fin whales off southern California, USA, during summer 2014-2015. Median attachment duration for the tagged blue whales was 19.8 days (Generation 3, n = 3) and 25.9 days (Generation 4, n = 5; range across both generations: 18.3-29.8 days;  Table S1).
Tag recovery was complicated by the extended attachment duration of the tags, which allowed some tagged whales to travel >500 km from the tagging area before tag release. In such cases, it was most economical to charter a local vessel from the closest port to attempt recovery. Poor weather and the tag's distance from shore (>160 km in some cases) were further limitations to recovery, such as tags 2013_5701 and 2015_5744 that continued transmitting until their batteries were exhausted. Tag 2013_5701 was found >1 year later by beachgoers and returned, as were three others in different years, demonstrating the continuing possibility of tag recovery following field work.

| Assessment of Fastloc GPS location accuracy
In 2007, predeployment testing using four tags showed that the median straight-line distance between a handheld GPS location and  Table S1 for individual tag data. Any submergence >10 m depth and 1 min duration was counted as a dive.
a Fastloc GPS location collected by the tags was 43 m, 83% of distances were less than 100 m and all distances were less than 455 m.
Distance decreased with increasing number of satellites recorded, as has been observed in other studies (Dujon, Lindstrom, & Hays, 2014;Hazel, 2009). In general, distances were normally distributed in both the easting and northing directions, although there was a slight bias in the northwest-southeast direction (Figure 2). The root-mean-square error (RMSE) of these distances for all tags was 92.2 m, but RMSE from one tag (# 4405841) was over twice that of the other three (RMSE = 178.2 m vs. 69.9, 50.2, and 73.6 m). That tag produced all of the locations with only 4 or 5 satellites during the test, which is indicative of poor-quality locations (Dujon et al., 2014;Hazel, 2009), and only produced half the number of locations as each of the other three tags.
In addition to the number of satellites, Fastloc GPS locations provided a "residual value," which indicates the relative spatial accuracy of a location. In other studies, locations with residual values greater than 30 have been excluded (Shimada, Jones, Limpus, & Hamann, 2012;Witt et al., 2010), but in our test, only four locations exceeded this threshold (range = 33.7-39.8) and all were <47 m from the true location. All four locations were also associated with a large number of satellites (≥8), suggesting that the identification of poor-quality Fastloc GPS locations is more complex than indicated by the residual value and/or the number of satellites.

| Depth data
Archived dive profiles from recovered tags were similar to those published for sperm whales (Amano & Yoshioka, 2003;Miller, Johnson, & Tyack, 2004), blue whales, and fin whales (Croll, Acevedo-Gutierrez, Tershy, & Urbán-Ramírez, 2001). Blue and fin whale dive profiles often recorded stereotypical upward excursions during the bottom phase of the dive, which are known to indicate feeding lunges (Calambokidis et al., 2007;Croll et al., 2001). Diel variability in dive depths was recorded with deeper dives occurring during the day (Figure 3), and consecutive dives often ascended or descended near sunset or sunrise, respectively, indicating the whales were following vertically migrating prey layers (Calambokidis et al., 2007).
While many archived sperm whale dives had similar characteristics to documented pelagic foraging dives (Miller et al., 2004;Watwood, Miller, Johnson, Madsen, & Tyack, 2006), some dives from both F I G U R E 2 The easting and northing components of the distance between ADB tag Fastloc GPS locations and a handheld GPS unit for four Generation-1 tags for which accuracy testing was conducted, showing a slight bias in the northwest-southeast direction for all tags and larger errors for tag # 4405841, as discussed in the text. Contours represent the straight-line distance (in m) between Fastloc and handheld GPS locations, interpolated over the easting and northing differences

| Accelerometer data
Accelerometer-derived metrics, such as the "jerk" (the difference in the norm of all three acceleration vectors after removing gravity), can detect rapid changes in orientation and acceleration of a tagged whale associated with foraging events (Miller et al., 2004;Simon et al., 2012).
Peaks in ADB-derived jerks from blue and fin whales showed a close correspondence with stereotypical upward excursions in the depth profile during the bottom portion of a dive, which were previously known to be indicative of a baleen whale lunge-feeding event (Acevedo-Gutierrez, Croll, & Tershey, 2002;Goldbogen et al., 2006). By documenting the location and frequency of jerk events across a multiweek track, the ADB tag was able to identify localized areas of high foraging effort ( Figure 5) and the depths at which it occurred (Figure 3).
While many current data loggers use a substantially higher sampling rate, the 1-Hz sampling rate of the ADB tag was sufficient to detect lunges and has been used in past studies that examined baleen whale foraging behavior (Acevedo-Gutierrez, Croll & Tershy 2002;Goldbogen et al., 2006). More detailed analyses of baleen whale behavior, like the fluking frequency of fin whales, has been examined using a 1-Hz sampling rate (Goldbogen et al., 2006) and should therefore also be possible with the ADB data from blue and fin whales.
However, the fluking rate of smaller species like humpback whales (Megaptera novaeangliae) has been shown to be higher than that of fin whales (up to 0.5 Hz; Simon et al., 2012) and confounding factors like aliasing therefore become more problematic as the size of the study animal decreases and the maximum rate of the signal approaches the sampling rate. Accelerometer data from ADB tags sampled at 1 Hz are therefore best suited to examining low-frequency signals and care should be taken, or a higher sampling rate should be used, when studying smaller species or attempting to examine higher-frequency signals.
Rapid orientation changes from jerk events also were detected in sperm whale ADB records using accelerometers, predominantly during the bottom phase of dives. Increased rates of orientation change during the bottom phase of a dive have been linked to foraging in sperm whales (Aoki et al., 2012;Miller et al., 2004), although their application is less direct compared to baleen whales. In sperm whales, a prey capture attempt is more reliably distinguished acoustically by a rapid clicking (the "buzz," detectable by a hydrophone on the tag), which occurs at close range to the prey (Miller et al., 2004). Multiple rapid orientation changes at varying intensities might occur during a pursuit prior to prey capture so, without an onboard hydrophone, the number of ADB-detected jerk events is not a direct measure of the number of prey capture attempts by the whale during a dive. However, because animals are predicted to forage more intensely in areas of higher prey density (Krebs, 1978;Schoener, 1971), the number of jerk events recorded should be dependent on the number of actual foraging attempts made during a dive. In such a case, the number of jerk events would be a relative measure of foraging effort made by the whale per dive, allowing for the spatial variability of foraging to be examined.

F I G U R E 4
A 20-h dive profile from ADB tag 2013_840 attached to a sperm whale in the Gulf of Mexico in 2013. The solid gray polygon shows the seafloor depth (from GMRT) nearest to the tag's Fastloc GPS location at the beginning of each dive. Note that several dives reach past the reported seafloor depth, as discussed in the text

| Temperature data
Generation-2 through Generation-4 ADB tags were equipped with external temperature probes capable of sampling the water temperature during a dive. While whales do not make completely vertical dives, and there may be small-scale differences in the thermal structure of the water column between where they start and end a dive, the measurements are adequate to identify important aspects of the ocean's thermal regime such as the thermocline, or the daily heating of surface waters, with reasonable accuracy (Figure 6)

| Argos dive summary transmissions
The amount of summary data received was highly variable and dependent on the priority settings of each Argos message type (Table   S2). Behavioral differences between tagged whales likely also affected the chance of a tagged whale being at the surface when a satellite was overhead. Transmission priorities were the same for ADB tags deployed on sperm whales in 2011 and 2013, allowing comparison.
Behavior messages from these tags summarized an average of 62% of all qualifying dives made, while histogram messages summarized an average of 50% of the tracking period across both years. Tags which drifted for extended periods of time before recovery generated a higher rate of data return due to the uninterrupted transmissions while the tag was floating at the surface.

| Complications with attachment and release
Variability in attachment duration of the ADB tags was strongly influenced by the depth of tag penetration upon deployment (Table S1).
The tag float and housing plate produce a substantial amount of hydrodynamic drag as the whale moves through the water, which acts to pull the tag out. Achieving full penetration, so that the housing plate is flush against the skin of the whale, not only allows more time for the housing to be fully extracted from the whale, but it also reduces the area of the tag exposed to hydrodynamic drag, thereby lengthening the attachment duration.
Recovery of the archived data required that the tag triggers a corrodible release wire in order to separate from its housing for subsequent recovery. This process created a number of challenges across

| Applications for ADB tag data
The detailed data collected by ADB tags over periods of multiple weeks have the ability to expand on current research directions and create new opportunities. For example, recent cetacean research is progressing beyond characterization of behavior and into investigations of how behavior relates to foraging ecology, energetics and diving physiology . Collection of such longer-duration data records will dramatically improve temporal and spatial scales of observed trends and quantify individual variability.
Species distribution models for blue and fin whales have been developed with substantial success based on remotely sensed data like sea surface temperature and phytoplankton chlorophyll-a concentration (Becker et al., 2016). In contrast, similar approaches with sperm whale data have been more variable, with some identifying direct relationships to environmental correlates, while others finding weak or negligible associations (Pirotta et al., 2011;Skov et al., 2008;Waring, Hamazaki, Sheehan, Wood, & Baker, 2001). The environmental cues driving sperm whale distribution continue to be elusive, likely due to their foraging at such deep depths. By monitoring the temperature profile of the water column and the spatial variation of foraging effort, ADB data may offer new insights regarding the water masses where the whales are foraging, and the scale on which they are searching for, and foraging on, sparsely distributed patchy prey (Palacios, Baumgartner, Laidre, & Gregr, 2013).
Characterization of cetacean responses to anthropogenic noise (military sonar, seismic surveys, vessel traffic, etc.) is a growing need (Nowacek et al., 2007;Soto et al., 2006;Southall et al., 2007) and currently the subject of substantial research Harris et al., 2016). However, the experimental period of a majority of studies is limited by short-attachment-duration suctioncup tags, preventing the collection of baseline (pre-exposure) data on the subject animal, the duration of experimental exposures, and post-experiment monitoring to estimate the duration of lasting effects (Nowacek et al., 2016). A longer-duration data logger like the ADB tag would allow a better understanding of normal variations in whale behavior and the time scales over which they occur. Such information could be applied to experiments to better identify behavioral responses when they occur and better understand the implications of those responses (Nowacek et al., 2007). Meanwhile, the dive behavior summary messages transmitted via Argos could be used to monitor a whale's behavior in near-real time for responses that exceed a behavioral threshold while an experiment is occurring.
While other transdermal tags (e.g., Argos satellite tags; Mate et al., 2007) and short-duration data loggers will continue to be useful for a wide range of applications with large whales, intermediate-duration archival tags like the ADB tag can bridge the gap between the two types of data. Many of the behavioral analyses developed for shortduration, high-resolution data loggers could be extended to ADB data while also accounting for spatial and temporal variability of those behaviors that previously could not be addressed. Conversely, behaviors and their corresponding movements described by the ADB data could be extended to better inform the more limited, but longer-duration, data produced by transdermal tags. The result will be a dramatic improvement in our ability to study the behavior of large whales and the ecological mechanisms that drive it.

DATA ACCESSIBILITY
The data used in this manuscript are reported in Tables S1 and S2.