Characterizing offshore polar ocean soundscapes using ecoacoustic intensity and diversity metrics

Abstract

Polar offshore environments are considered the last pristine soundscapes, but accelerating climate change and increasing human activity threaten their integrity. In order to assess the acoustic state of polar oceans, there is the need to investigate their soundscape characteristics more holistically. We apply a set of 14 ecoacoustic metrics (EAMs) to identify which metrics are best suited to reflect the characteristics of disturbed and naturally intact polar offshore soundscapes. We used two soundscape datasets: (i) the Arctic eastern Fram Strait (FS), which is already impacted by anthropogenic noise, and (ii) the quasi-pristine Antarctic Weddell Sea (WS). Our results show that EAMs when applied in concert can be used to quantitatively assess soundscape variability, enabling the appraisal of marine soundscapes over broad spatiotemporal scales. The tested set of EAMs was able to show that the eastern FS, which is virtually free from sea ice, lacks seasonal soundscape dynamics and exhibits low acoustic complexity owing to year-round wind-mediated sounds and anthropogenic noise. By contrast, the WS exhibits pronounced seasonal soundscape dynamics with greater soundscape heterogeneity driven in large part by the vocal activity of marine mammal communities, whose composition in turn varies with the prevailing seasonal sea ice conditions.

Keywords: MANTA software; marine mammals; one-third-octave-level bands (TOL); passive acoustic monitoring (PAM); sea ice; underwater ambient noise levels.

Figure 1.

Geographic position of the recording stations. (a) FS, Arctic, recorder position 79 °N and 5.7 °E and (b) the WS, Antarctica, recorder position 65.7 °S and 36.7 °W. Maps were created with the R-package ggOceanMaps [45].

Figure 2.

(a–d) Seasonal mean (pink line) and median (black-dashed line) power spectral density levels (PSD, dB re 1 µPa Hz⁻¹) and spectral probability density (SPD, colour scale) of the FS recordings. The main contributors to seasonal soundscape patterns are indicated by black arrows, boxes and circles. The indicated wind noise in c is exemplary and also applies to the other panels.

Figure 3.

(a–d) Seasonal mean (pink line) and median (black-dashed line) power spectral density levels (PSD, dB re 1 µPa Hz⁻¹) and spectral probability density (SPD, colour scale) of the WS recordings. The main contributors to seasonal soundscape patterns are indicated by black arrows, boxes and circles.

Figure 4.

Ecoacoustic intensity metrics as SPLs for the broadband (10–3428 Hz) and TOL bands centred at 20 Hz (for the FS station), 25 Hz (for the WS station), and 63, 125, 250 and 500 Hz. The yellow triangles indicate the seasonal mean SPLs and the blue- and red-dashed lines indicate the annual mean and median SPLs, respectively. The box colours indicate the seasons: pink for summer, purple for autumn, blue for winter and green for spring. The lower and upper box borders show the interquartile range (25th to 75th percentiles) and the whiskers indicate the 5th and 95th percentiles. Seasonal medians are given by the inner black lines (see also electronic supplementary material, tables S3 and S4, for mean, median, maximum and minimum values for the FS station and the WS station, respectively).

Figure 5.

Ecoacoustic diversity metrics (ACI, AEI, ADI, HI and BI) for the FS and WS. The yellow triangles indicate the seasonal metric means and the blue- and red-dashed lines indicate the annual metric means and medians, respectively. The box colours indicate the seasons: pink for summer, purple for autumn, blue for winter and green for spring. The lower and upper box borders show the interquartile range (25th to 75th percentiles) and the whiskers indicate the 5th and 95th percentiles. Seasonal medians are given by the inner black lines.

Figure 6.

The PCA biplot for feature selection shows the clustered soundscape characteristics (table 2) explained by the combination of the hourly means (5756 soundscape observations) of the ten EAMs (broadband SPL, 63, 125, 250 and 500 Hz TOL band, ACI, AEI, ADI, HI and BI) along the first two principal components (PCs). Ellipses represent the 95% percentiles of cluster variables. The length of the arrows indicates the importance of the seven EAMs, which together explain approximately 70% of the variance in the first two PCs in differentiating among soundscape characteristics.