The function plots the estimate of the power spectral density in dB/Hz (the estimate is divided by the sample frequency) according to the frequency (linear axis) till the Nyquist/Shannon frequency. For instance, the periodogram of the spectrum Figure 2 is presented in Figure 5. The interesting part is situated under 1 kHz, others frequencies being weakly influenced by the WT noise. Note that the signal is A-weighted before the calculation of the periodogram and only the difference between levels are relevant due to a non calibration of the absolute level of the recording. Both maxima previously found are visible: a large around 400 Hz and a second thinner as a tonality. A zoom of the part circled in red is shown in Figure 6. Many peaks are present but not always at the same regularity and as high as the peak at 120 Hz.1 So a first criterion based on this peak and an additional about the large maximum could be tested
Unfortunately the peak at 120 Hz is not fixed and can move towards a higher frequency or towards 100 Hz...And some other kinds of noise (not too high in dB(A), without a great emergence and positioned in low frequencies) also prese important acoustic energy at 100 Hz as a local maximum. Therefore a fixed and precise criterion around 120 Hz is used and some periods are missed, or a more flexible criterion is applied and non-relevant periods are selected