Description

Methods (En)

Field context and data collection

Upon sighting the dolphin group, the vessel’s engine was stopped and the recorder was gently lowered to 5–7 m depth, where it remained until the encounter concluded. The HydroMoth (Open Acoustic Devices) was configured for broadband capture at 250 kHz, Medium gain, a 0.5–125 kHz band-pass, and duty-cycled recording (300 s record / 5 s sleep). As the raw audio is not calibrated to absolute sound pressure, amplitudes are reported as relative dB FS (peak-to-peak) rather than dB re 1 µPa. Field information (vessel, timings, species confirmation, number of individuals, recorder identity and depth, coordinates, sea state, short behavioural notes, and images) is mapped to Event and Occurrence in the associated DwC-A (available in GBIF: https://doi.org/10.15470/77bhey).

Screening and analysis workflow

Recordings were screened in Raven Lite 2.0.5 to verify signal presence and in Audacity 3.7.5 to assess potential saturation. At Fs = 250 kHz two label types were created to annotate clipping: HardClip (0 dBFS) and NearClip (−0.5 dBFS). This clipping-labelling step is new in the present release; downstream analyses therefore exclude ±5 ms around labelled intervals and the resulting metrics differ from earlier versions.

Subsequent processing used PAMGuard v2.02.16 (https://www.pamguard.org/). The spectral configuration for measurement and figure generation was FFT length 2048, Hann window, 50% overlap, yielding Δf ≈ 122 Hz and hop ≈ 4.096 ms. Whistle contours were extracted and reviewed using ROCCA (Northeast Atlantic model) via the graphical interface (see Oswald & Oswald, 2013).

ICI rule and operational click classes

To reduce false positives and summarise temporal structure, PAMGuard Click Detector outputs were ordered by UTC within each minute. For each detection, the representative inter-click interval was computed as ICI_rep = min(ICI_prev, ICI_next). We defined “real clicks” as those with ICI_rep in 0.5–100 ms and outside ±5 ms around Audacity NearClip/HardClip labels. Based on ICI_rep, real clicks were classified into BurstPulse (ICI ≤ 3 ms), ClickTrain (3 < ICI ≤ 10 ms) and EcholocationClick (10 < ICI ≤ 100 ms). Because the HydroMoth is not acoustically calibrated, amplitude statistics are reported as relative dB FS (p-p). Within the eight-minute segment, we validated 92 whistles and identified 20,217 real clicks under these rules.

In supplementary material, a twelve-panel mosaic of spectrograms (display range typically 4–32 kHz, with some panels extending higher) illustrates several whistle types (convex, ascending, and quasi-constant segments with harmonics) together with the co-occurrence of echolocation clicks, click trains, and burst pulses (buzzes), consistent with social and echolocation behaviours of T. truncatus.

Contents and data flow

This Zenodo record preserves:

• the raw audio (WAV) and HydroMoth configuration file (TXT);
• the PAMGuard project (.PSFX) and storage exports (e.g., ClickDetectorClicks,  DetectorClicks_Classified, and ROCCA ContourStats; CSV);
• a mosaic of spectrograms and a photograph of the encounter (PNG); and
• a README clarifying directory layout and variable definitions.

The corresponding Darwin Core Archive (DwC-A), including Event, Occurrence, ExtendedMeasurementOrFact (eMoF), SimpleMultimedia, plus meta.xml and EML, is published through GBIF IPT. SimpleMultimedia rows in the DwC-A point back to Zenodo-hosted media to support persistent, FAIR-compliant links between repositories.

Technical info (En)

Software used

• Audacity (Version 3.7.5). Audacity Team. Free, open‑source, cross‑platform audio software.
• AudioMoth Firmware Basic v1.11.0 (Open Acoustic Devices).
• JASP (Version 0.95.4). JASP Team. Free and open statistics package.
• PAMGuard (Version 2.02.16). Open-source platform for passive acoustic monitoring (Gillespie et al., 2008).
• Raven Lite (Version 2.0.5). K. Lisa Yang Center for Conservation Bioacoustics at the Cornell Lab of Ornithology. Interactive sound analysis software for bioacoustics.

Notes (En)