Figure 1.
Spectrograms of a female (a) and male (b) distance call, recorded at 2 m.
Both signals were high-pass filtered above 500 Hz to avoid displaying low-frequency background noise. The color scale is in relative dB as shown on the color bar with 100 dB corresponding to maximum amplitude observed.
Figure 2.
Overview of the acoustical analysis.
This figure summarizes the procedure for extracting both sets of parameters (a), and the subsequent discriminant function analysis (b). The discriminant functions were calculated using the fitting dataset, and we tested the validating dataset against these to obtain the percentages of correct classification.
Figure 3.
Spectrograms of the same female (a) and male (b) distance call recorded at various distances.
All signals were high-pass filtered over 500 Hz, and the same color scale was applied to all spectrograms.
Figure 4.
Frequency power spectra of calls at every propagation distance (from 2 m to 256 m).
These were calculated using all female (a) and male (b) distance calls recorded in France. The frequency spectrum of the background noise at the recording sites is shown in black. This curve was obtained by averaging the noise spectrum across all recording sites and can therefore be slightly higher than the noise floor at particular distances, as it can be observed for the lower frequency range.
Figure 5.
Comparison of transfer functions in France and in Australia.
The transfer functions were calculated between the recordings obtained at 2 = 30°C, 15% relative humidity and 1 atm. The Australia site 2 is next to zebra finch nesting sites and includes sparse vegetation and the recordings were performed on a windless evening. The attenuation curve was obtained using T = 25°C, 15% relative humidity and 1 atm. The recordings in France are in an open field and were obtained during the day in low wind conditions. The attenuation curve was obtained using T = 11°C, 50% relative humidity and 1 atm.
Figure 6.
Modulation power spectra (MPS) of female (a) and male (b) calls for each propagation distance.
Mean MPSs are shown for each propagation distance used in this study (from 2 m to 256 m). Projections on the spectral modulation axis (in cycles per kHz) and the temporal modulation axis (in Hz) are shown beside and below each MPS respectively. The major observable features are shown and labeled on one example shown in panel (c).
Figure 7.
Mean percentages of correct classification obtained for each propagation distance and both sexes.
(a) Taking the data for each propagation distance separately, (b) taking the 2 m data as a reference for other distances, and (c) taking all propagation distances into account. SPC parameters are represented as solid lines and envelope parameters as dash-dot lines. Standard deviations are indicated. The chance level, corresponding to 6.25% of correct classification, is shown as a horizontal dotted line.
Figure 8.
Comparison between the importance of spectral and temporal features of the envelope parameters.
Calculations are shown separately for females (a) and males (b). Histograms represent the relative decrease as a percentage in the probability of correct classification (PCC) observed when removing the spectral (left) or temporal (right) parameters from the DFA. Error bars show two-standard errors estimated by bootstrapping. For each distance, except for 128 m for females, the changes observed by removing spectral information vs. the temporal information are significantly different (all p values <10−5 using a two-sided t-test).
Figure 9.
Representation of the discriminant functions (DFs) projected into the spectrographic space for the SPC parameters.
The first 3 DFs, obtained from the SPC parameters, are shown for females (left) and males (right). Each row indicates the type of dataset used to perform the DFA: each distance taken separately for 2 m, 64 m and 256 m (first 3 rows) and all distances taken into account (bottom row). As examples, on DF1 at 2 m for females and DF2 at 64 m for males, full arrows show positive frequency bands (red) and dashed arrows show negative bands (blue). This representation enables a description of the most important features in the spectrogram that can be used to discriminate between individuals at various distances.
Figure 10.
Relative importance of the discriminant functions (DFs) as a function of frequency.
Calculations are shown for females (a) and males (b), for every propagation distance. These importance functions were obtained by calculating the normalized mean of the absolute values of the first 4 DFs for each frequency window. This analysis shows the frequencies that are the most useful to discriminate between individuals at each distance.