Fig 1.
A sample classic call produced by an adult male (ring no. DA48567).
Analysing bandwidth = 93.75 Hz. Spectrogram plotted using the seewave package [12].
Fig 2.
A sample single call produced by an adult male (ring no. DA48567).
Analysing bandwidth = 93.75 Hz. Spectrogram plotted using the seewave package [12].
Fig 3.
Annual (pink) and 40-year average (blue) meteorological data from Hornsund over the little auk’s breeding period (May-August).
Fig 4.
Sound attenuation at different frequencies, calculated from mean May conditions in Hornsund over the monitored period 1983–2021, based on the ISO 9613–1 standard [20].
There is no apparent shift in attenuation profiles over the years.
Table 1.
Raw acoustic parameters extracted from audio files.
Variable explanations as per soundgen package [28].
Table 2.
Maximum and mean SL values of the call types.
All SL values are given in dB re 20 μPa at 1 m.
Fig 5.
A sample classic call attenuated to mimic signal structure at 10 exponential distances in a range of 1–1000 m.
Notice that the signal remains very stable, and harmonics are only lost at extreme distances. Note that to improve readability, the spectrograms are not plotted on the same scale, and the colours should not be interpreted as comparable between the panels. Spectrograms plotted using the seewave package [12].
Fig 6.
A sample single call attenuated to mimic signal structure at at 10 exponential distances in a range of 1–1000 m.
Notice that the signal remains very stable, and harmonics are only lost at extreme distances. Note that to improve readability, the spectrograms are not plotted on the same scale, and the colours should not be interpreted as comparable between the panels. Spectrograms plotted using the seewave package [12].
Fig 7.
S-UMAP classification of the classic call to individual remains efficient in spite of signal filtering and attenuation.
Fig 8.
S-UMAP classification of the single call to individual remains efficient despite signal filtering and attenuation.
Table 3.
Results of the permuted discriminant function analysis for classic calls attenuated to mimic signal structure at distances from 1 to 1000 m (552 calls of 11 individuals per distance), as well as for all distances pooled together (5520 calls of 11 individuals), using the principal components of eigenvalues > 1.
Calls could be reliably classified to individuals above chance levels independently of the attenuation.
Table 4.
Results of the permuted discriminant function analysis for single calls attenuated to mimic signal structure at distances from 1 to 1000 m (264 calls of six individuals per distance), as well as for all distances pooled together (2640 calls of six individuals), using the principal components of eigenvalues > 1.
Calls could be reliably classified to individuals above chance level independently of the attenuation.
Table 5.
Accuracy of classification to individuals using SVM based on S-UMAP reduced data.
Table 6.
Beecher’s statistic’s values in the propagated signals.
Column meaning specifies how many individuals can be theoretically distinguished based on the signal alone.