Data analysis:
C-POD data were processed using the Chelonia CPOD.exe software (V. 2044)
and inbuilt KERNO classifier to detect harbour porpoise click trains.
F-POD data were processed in a similar manner using the custom F-POD.exe
software (V 1.1) and KERNO-F classifier (Chelonia Ltd., 2022). Click
trains were classified as “NBHF” (narrowband high frequency) and all
train quality classes were exported for further examination. Train
quality filters are defined as “Hi” (high),” Mod” (moderate), and
“Lo” (low). All detections were visually verified following guidelines
from the manufacturer (Chelonia Ltd., 2022). Data were exported as
different detection metrics; number of clicks ‘NClx’, detection positive
days ‘DPD’, detection positive hours ‘DPH’ and detection positive
minutes ‘DPM’.
Detection metrics were summarised for each deployment across three
groupings of train quality filters, specifically HiModLo, HiMod, and Hi,
reflecting commonly used groupings in the literature (Sarnochinska et
al., 2016; Clausen et al., 2018). Kendall’s rank (non-parametric)
correlation tests were carried out between the detections on the C-POD
and F-POD at the scale of each temporal detection metric and for each
train quality classification.
Both monthly and seasonal DPH were summarised for both the C-POD and the
F-POD and compared using a detection ratio, expressed as: CF =
Det_C/Det_F. This ratio was used to explore the comparability between
the PODs across time and by what margin the F-POD detects more
echolocation clicks than the C-POD.
Data on echolocation clicks were also exported and used to identify
buzzes, assumed to be foraging behaviour (Verfuß et al., 2009), based on
the duration of the inter-click interval (ICI). Gaussian mixture models
were used to categorise echolocation clicks based on their ICI (Pirottaet al. , 2014). Buzzes were defined as echolocation clicks with an
ICI of less than 10 ms (Carlström, 2005). Detections were then
summarised as foraging buzzes per hour (BPH) for further analysis.