Figure 7. Spectrograms of leopard seal vocalizations recorded in (A, B,
C) underwater and in (D, E, F) the air. (a) triple ascending trill, (b)
ascending trill, (c) hoot, (d) triple ascending trill, (e) low single
trill, and (f) hoot (underwater data: 48,000 fast Fourier transform
points and 4,800-point Hanning window, airborne data: 48,000 fast
Fourier transform points and 2,400-point Hanning window).
According to Thomas and DeMaster (1982), the diurnal pattern of
underwater vocalizations is negatively correlated with the haul-out
pattern. In other words, the number of vocalizations was greater between
19:00 and 06:00 and lower between 09:00 and 18:00, as the species spends
more times in the water at night. According to our results, the
vocalization rate and sound pressure levels were relatively low from
05:00 to 17:00 on December 10, and were higher before and after, which
was similar to the trend observed in the previous study but with a time
shift. The persistence of elevated call rates and sound pressure levels
from 05:00 to 17:00 on December 11 suggests that further validation of
the diurnal pattern through long-term observation is necessary. However,
our measurements were conducted at higher latitudes and later in the
year when day and night conditions are indistinct than those of Thomas
and DeMaster (1982), which may explain the differences compared to the
trend. The results from the Perennial Acoustic Observatory in the
Antarctic Ocean (PALAOA) conducted from January 2006 to January 2007
showed that the leopard seal was most vociferous on 16 December (Van
Opzeeland et al., 2010). Although acoustic data collected over 54 hours
in this study are insufficient for characterizing diurnal patterns and
their correlations with environmental data, they were measured when
vocalization rates might be near the peak so that leopard seal
vocalizations could be the dominant sound source. The proportion of call
types presented in Section 3-2 was subject to limited comparative
analysis due to differences in call type classification from the
previous study. The absolute call counts per minute were 3.7 (± 0.03
SE). By call type, they were 1.2 (± 0.02 SE) for HDT, 0.4 (± 0.01 SE)
for MST, 1.3 (± 0.02 SE) for LDT, 0.5 (± 0.01 SE) for HST, 0.3 (± 0.01
SE) for DT, and 0.1 (± 0.01 SE) for triple ascending trill. These values
were significantly lower than those reported by Shabangu and Rogers
(2021), because we extracted only calls that the structure could be
clearly identified, which likely excluded calls with low signal-to-noise
ratios. Furthermore, we were unable to collect environmental data such
as sound speed profiles and water depths in the field, and long-term PAM
data are required to facilitate a more comprehensive analysis
correlating these environmental variables with acoustic data.