Temporal evolution of the Mediterranean fin whale song

Abstract

We present an analysis of fin whale (Balaenoptera physalus) songs on passive acoustic recordings from the Pelagos Sanctuary (Western Mediterranean Basin). The recordings were gathered between 2008 and 2018 using 2 different hydrophone stations. We show how 20 Hz fin whale pulses can be automatically detected using a low complexity convolutional neural network (CNN) despite data variability (different recording devices exposed to diverse noises). The pulses were further classified into the two categories described in past studies and inter pulse intervals (IPI) were measured. The results confirm previous observations on the local relationship between pulse type and IPI with substantially more data. Furthermore we show inter-annual shifts in IPI and an intra-annual trend in pulse center frequency. This study provides new elements of comparison for the understanding of long term fin whale song trends worldwide.

Figure 1

Spectrogram of a fin whale pulse sequence recorded by the Bombyx buoy in October 2018. Spectrogram parameters are described in section "Spectro-temporal pulse analysis". Dashes show the center frequencies of the detected pulses, with grey dashed lines showing the IPIs. The blue dashed line denotes the discrimination threshold between A and B pulse types, at 20 Hz.

Figure 2

Map showing the two recording stations used in the analysis. This map was made using the Ocean Data View software.

Figure 3

Calendar of the recorded days (grey cells). Shades of red denote the number of detected pulses normalized by the number of recorded hours (ranging from 0 to 30).

Figure 4

ROC curves for each test set (the two remaining sources serving as training set) and for the template matching method. The ROC curve of the model over the dataset published by Madhusudhana et al. is also displayed.

Figure 5

Histogram of the center frequencies of the detected pulses (post filtering). Black lines denote the fitted GMM, and the dotted line denotes the discrimination threshold between the two pulse types.

Figure 6

Histogram of the IPI for each type sequence (or bi-gram).

Figure 7

Scatter plot of the most frequent IPI per trimester for each type sequence. Fitted linear models are shown as grey dashed lines. Points extracted from Clark et al. and Castellote et al. appear as crosses.

Figure 8

Bi-histogram of the center frequencies against months of the year. The horizontal line shows the separation between type A and type B pulses. The fitted linear model is shown as a black dashed line.

Figure 9

Scatter plot of pulses center frequency against the time until the next pulse (IPI). Colors denote the GMM assignation, whose means are marked with crosses.