We examined the specifications (e.g., size, weight, power consumption, and electronic interface) of four commercially-available fluorometers, including the Seapoint Chlorophyll Fluorometer (Seapoint Sensors, Inc., Exeter, NH, USA), Cyclops Integrator (Turner Designs, Sunnyvale, CA, USA), Pisces Fluorometer (Pisces Design, La Jolla, CA, USA), and a miniature version of the Environmental Characterization Optics series (ECO Puck™; WET Labs, Inc., Philomath, OR, USA), and concluded the ECO Puck™ (http://www.wetlabs.com/products/eflcombo/pucks.htm) was the smallest instrument that would be compatible with a commercially available satellite transmitter manufactured by Wildlife Computers (Redmond, WA, USA). We decided the SPLASH10 (http://wildlifecomputers.com/our-tags/splash/) would serve as the basic architecture for electronic integration because it has a channel designed to log analog voltages from an external sensor and the ability to archive high resolution data, including depth, temperature, and light levels. This instrument has 1 GB of memory and the controller contains up to eight 12-bit analog-to-digital converters, 512 KB of low-power static RAM, and 128 KB of program FLASH memory. This model is also equipped with a real-time clock and a wet/dry sensor, which conserves battery power by limiting transmissions to when the animal surfaces.
For initial data exploration during laboratory and field trials an archival TDR10 equipped with a temperature probe was first interfaced with the ECO Puck™ and configured to measure chl-a (0 to 75 ± 0.02 µg Chl/L) at 4 Hz. To determine if this interfaced prototype functioned properly it was suspended with a calibrated Combo Fluorometer-Turbidity Unit (FLNTU; WET Labs) of similar configuration in a black bucket of distilled water, which was diluted in 20 mL increments with a chlorophyll mixture extracted from frozen spinach. Data from the FLNTU were decoded using ECOView software (WET Labs), whereas all voltage readings from the prototype were downloaded, decoded using a data analysis program (DAP; Wildlife Computers), and converted to chl-a using the algorithm from the characterization sheet supplied by WET Labs after a transfer function (gains/offsets provided by Wildlife Computers) was applied to the raw data. Chlorophyll-a values were averaged per dilution period for each instrument, plotted, examined for proper saturation, and tested for linearity using linear regression (α = 0.05). An analysis of covariance (ANCOVA; α = 0.05) was also used to determine if linear models differed between the two instruments.
To further verify functionality of the prototype, manual casts of both instruments were conducted at two locations in Hood Canal, Washington. Both instruments were mounted to a metal block (separating them by ~15 cm) attached to an 8 m cable (SOOW 600 VAC Service Cord; McMaster-Carr, Los Angeles, CA, USA), which was suspended at different depths (range 1–6 m) for 5 min intervals. Data were processed using methods above, chl-a values were averaged for each depth per instrument, and an analysis of variance (ANOVA) was used to determine if trends differed between the two instruments by examining the interaction term depth × instrument (α = 0.05).
To determine if sensor orientation or animal behavior affected chl-a measurements, the archival prototype was deployed on three trained, open-water Steller sea lions (Eumetopias jubatus) at the Open Water Research Facility, University of British Columbia (UBC). This also allowed us to observe if the sea lions were affected by the fluorometer’s LED. Experiments lasted ~50 min, during which an individual sea lion made two to three trips to depth (11.5–12 m) to feed and was filmed with an underwater camera system. Otherwise, the sea lion remained at the surface within a floating Plexiglas dome while respirations were monitored. The instrument was attached to the harness of the first animal with the optics facing forward (towards the head of the sea lion), whereas the optics faced backwards for the second animal. For the third animal, the instrument was also backwards and the animal was released from a boat into the waters off Port Moody and instructed to swim beside the boat for ~20 min at 2–3.5 knots. After completing the open-water trials, the data were inspected using Instrument Helper (3.0; Wildlife Computers), a data visualization and analysis program.
After the open-water trials indicated the archival prototype was robust to behaviors of concern (e.g., pitch/roll, fast speeds, rapid surfacings, etc.), it was completely cast in epoxy (370 g; 10.8 × 6.4 × 6.2 cm). To collect in situ data on a marine mammal while in its natural habitat, the cast prototype was deployed on a free-ranging, adult female northern fur seal (Callorhinus ursinus) that was captured using a hoop net (Fuhrman Diversified, Seabrook, TX, USA) on 9 September 2013 at Reef Rookery, St. Paul Island, Alaska. The fur seal was manually restrained, weighed to the nearest 0.1 kg (Dyna-Link, Measurement Systems International, Seattle, WA, USA), and the prototype attached to the dorsal pelage between the scapulae using 5-min epoxy (Devcon Products, Riviera Beach, FL, USA). Additionally, a satellite transmitter (SPOT5, Wildlife Computers) and VHF transmitter (Advanced Telemetry Systems, Isanti, MN, USA) were glued to the lower back to monitor the animal while at sea and on shore, respectively.
After the fur seal returned to the rookery on 18 September 2013, she was located via her VHF transmitter, captured, weighed, and the archival prototype retrieved. Data were downloaded, decoded, and smoothed using a median-value filter; chl-a and temperature outliers (i.e., values ±0.1 μg/L or °C from the median) were replaced with the median using a moving window of 20 values, whereas depth outliers (i.e., values ±4 m from the median) were replaced with the median using a moving window of 10 values. Smoothed data were further inspected using Instrument Helper and IgorPro (WaveMetrics, Portland, OR, USA) and anomalous measurements (i.e. data spikes) were omitted. Smoothed data corresponding to the top of the second were merged with the satellite telemetry data after they were obtained through Service Argos, decoded using DAP, and filtered using a maximum transit rate of 2 m/s [20]. The merged data set was then processed with a continuous-time correlated random walk model (CTCRW; [21]) to predict uniformly spaced locations every second so that in situ measurements of chl-a and temperature could be spatially interpolated to locations at sea (R 3.1.2, [22]) and examined in a 3D environment (ArcScene 10.1; ESRI, Redlands, CA, USA).
The archived data set also was used to determine the best approach for collecting, summarizing, and compressing data for transmission through the Argos satellite system, which has bandwidth restrictions of 256 bits per message. For data compression we decided fluorescence and temperature data would be collected at 4 Hz during the ascent of the first dive exceeding a user-defined depth (i.e. ≥9.5 m for this case report and referred to hereafter as a ‘qualifying’ dive) after the top of each hour. After converting fluorescence to chl-a (µg/cL) on board the instrument (via correction coefficients and scale factors applied to the raw data) transmitted data would be reported to the nearest hour and include: (1) the chl-a and temperature (°C) values at 3 m depth (for future comparisons with satellite remote sensing data), (2) the maximum chl-a value (with corresponding temperature value) and the depth at which they occurred, (3) the chl-a and temperature values at the maximum dive depth (including the depth value), and (4) the sum of all chl-a values from the surface to the maximum depth of the dive. This sampling scheme was chosen to conserve battery power and achieve a two-month life expectancy for the tag. This technology is customizable to the user, however, and different forms of data compression can be considered. Additionally, data can be relayed in different formats (i.e., raw data vs. chl-a values).
To determine if the data reduction routine established for the Argos system was a sufficient summary of the data, and to verify functionality of the Argos message generation by a transmitting unit, data from the archival prototype were processed with a ‘simulator’ (i.e., Windows-based software package comprised of the same code installed on the SPLASH10 controller) to produce messages that would be transmitted through Service Argos. Those messages were manually cross-referenced with the archived data set to confirm consistency. A housing was then constructed to support the ECO Puck™ fluorometer with the SPLASH10 and two independent power sources; the fluorometer was powered by 3 AA batteries, whereas the SPLASH10 controller was powered by 2½ AA batteries connected in series. After the final tag was assembled (AM-A320A-AU Fluorometer; 458 g, 11.9 × 5.8 × 6.2 cm; Fig. 1) it was placed in a pressure chamber with a voltage standard (WET Labs) and dives to known depths were simulated to further verify functionality of the Argos message generation.
To observe operation of the new AM-A320A-AU Fluorometer in the field it was deployed on a free-ranging, adult female Steller sea lion at Adak, Alaska on 9 October 2014. That animal was chemically immobilized [23], intubated, and supplemented with isoflurane (range 0.5–1.5 %) in 100 % oxygen using a portable vaporizer to maintain anesthesia while the fluorometer was attached to the dorsum using 5-min epoxy (Devcon). After the transmitter was secure, the sea lion was administered reversal agents, extubated, and monitored before release [23]. After transmissions ceased, data were obtained from Service Argos and chl-a and temperature data were spatially interpolated to locations at sea using Argos positions that were decoded, filtered, and modeled using the same methods detailed for the northern fur seal.