Co-occurrence of beaked whale strandings and naval sonar in the Mariana Islands, Western Pacific

Mid-frequency active sonar (MFAS), used for antisubmarine warfare (ASW), has been associated with multiple beaked whale (BW) mass stranding events. Multinational naval ASW exercises have utilized MFAS offshore of the Mariana Archipelago semi-annually since 2006. We report BW and MFAS acoustic activity near the islands of Saipan and Tinian from March 2010 through November 2014. Signals from Cuvier’s (Ziphius cavirostris) and Blainville’s beaked whales (Mesoplodon densirostris), and a third unidentified BW species were detected throughout the recording period. Both recorders documented MFAS on 21 August 2011 before two Cuvier’s beaked whales stranded on 22-23 August 2011. We compared the history of known naval operations and BW strandings from the Mariana Archipelago to consider potential threats to BW populations. Eight BW stranding events between June 2006 and January 2019 each included 1-3 animals. Half of these strandings occurred during, or within 6 days after naval activities, and this co-occurrence is highly significant. We highlight strandings of individual BWs can be associated with ASW, and emphasize the value of ongoing passive acoustic monitoring, especially for beaked whales that are difficult to visually detect at sea. We strongly recommend more visual monitoring efforts, at sea and along coastlines, for stranded cetaceans before, during, and after naval exercises.

Methods
Acoustic recordings were collected at a sampling rate of 200 kHz at two locations near the islands of Saipan and Tinian from 2010-2014 from High Frequency Acoustic Recording Packages (HARPs). All instruments were bottom-mounted and deployed to seafloor depths of 600-700 m for the “West HARP” location (15° 19.026’ N, 145° 27.463’ E), and 1000 m at the “East HARP” location (15° 2.344’ N, 145° 45.130’ E). From 2011-2013, the temporal coverage of recordings overlapped at both locations; however, recordings were not collected in all months for all years (see duty cycles in Figures 1 and 2). The hydrophone used was an omni-directional sensor (ITC-1042, International Transducer Corporation, Santa Barbara, CA), which had an approximately flat (±2 dB) hydrophone sensitivity from 10 Hz to 100 kHz of -200 dB re V/μPa. Each system contained a custom-built preamplifier board and bandpass filter. The calibrated system response was accounted for during analysis. The ability to assess the presence of beaked whales varies as a function of the recording schedule, the location, and the relative abundance and vocal activity of the beaked whale species of interest. Accordingly, the duty-cycled recording schedules used in this study may result in an underestimation of actual beaked whale presence, and the different recording schedules used throughout the study should be considered when evaluating relative abundance of acoustic activity.

The acoustic activity of beaked whale signals was detected using a multistep detection process following methods described in Baumann-Pickering et al., 2013. All echolocation clicks were detected using a computer algorithm (Soldevilla et al., 2008). Click detections were then classified as Cuvier’s, Blainville’s, or “BWC” signal types based on the spectral and temporal characteristics of the species-specific descriptions provided by Baumann-Pickering et al., 2013.

To identify mid-frequency active sonar events, the acoustic recordings were downsampled to a sampling rate of 10 kHz, and two analysts (AES and RH), trained to recognize MFAS signals, scanned long term spectral averages (LTSAs) over a frequency range of 10-5000 Hz to identify time periods with MFAS present. A “packet” was defined as a tightly spaced cluster of pulses or pings, which occurred within a 1 kHz band between 2.5-4.5 kHz, with a pause between signals of no more than 0.1 seconds. To inspect the packets more closely, the analyst scanned spectrograms (Hann window, DFT=1,000, 50% overlap) in a 20-second window to log the start time, and the lowest and highest frequencies of packet components. When the start time of one packet occurred within five seconds of the start of a previous packet, they were combined into a single packet.

The acoustic energy of the MFAS was characterized based on MFAS packets. To minimize low-frequency ambient noise and focus on the energy band of MFAS, the data was filtered with a 10-pole Butterworth bandpass filter (2-4.95 kHz). The duration of the sonar packet was defined as the interval over which 90% of the sound energy arrived at the receiver, with the start and end points of an event at the 5% and 95% levels of cumulative energy within a time window [5]. A ten second or three second window was used for sonar packets with multiple or single MFAS signals, respectively. Details of the signal level calculations are included in the electronic supplementary materials.