Estimating whale density from their whistling activity: Example with St. Lawrence beluga

• Published in: Applied Acoustics
• Main Authors: Simard, Y., Roy, N., Giard, S., Gervaise, Cedric, Conversano, M., Menard, N.
• Other Authors: Extraction et Exploitation de l'Information en Environnements Incertains (E3I2), École Nationale Supérieure de Techniques Avancées Bretagne (ENSTA Bretagne)
• Format: Article in Journal/Newspaper
• Language: English
• Published: HAL CCSD 2010
• Subjects: Beluga; Sound; Passive acoustic monitoring (PAM); Estimation; Saguenay; St. Lawrence; [SPI.ACOU]Engineering Sciences [physics]/Acoustics [physics.class-ph]; [PHYS.MECA.ACOU]Physics [physics]/Mechanics [physics]/Acoustics [physics.class-ph]; Canada; Beluga*; Delphinapterus leucas; White whale
• Online Access: https://hal-ensta-bretagne.archives-ouvertes.fr/hal-00522344; https://doi.org/10.1016/j.apacoust.2010.05.013

WOS International audience A passive acoustic method is developed to estimate whale density from their calling activity in a monitored area. The algorithm is applied to a loquacious species, the white whale (Delphinapterus leucas), in Saguenay fjord mouth near Tadoussac, Canada, which is severely affected by shipping noise. Beluga calls were recorded from cabled coastal hydrophones deployed in the basin while the animal density was estimated visually from systematic observations from a fixed-point on the shore. Beluga calling activity was estimated from an algorithm extracting the call events in time–frequency space, while simultaneously tracking the masking intensity resulting from local shipping noise. The activity index was summarized in 15- and 30-min bins using four different metrics. For bins containing more than 40% of valid data, the metrics were compared to the corresponding visual observations. The estimated mean acoustic detection range generally exceeded the fjord width, and extended to the whole not, vert, similar3-km long monitored area under low-noise conditions. The significant linear relations of the visual estimates with the calling activity metrics allowed assessing expected number of visually detected belugas in the basin from a weighted regression model, with a mean standard error of 7.1%.