Mesoscale habitat use by St. Lawrence estuary beluga over the annual cycle from an acoustic recording network.

The spatial-temporal distribution pattern of St. Lawrence Estuary (SLE) beluga is examined with a passive acoustic monitoring network of 13 stations from June 2018 to October 2021. A beluga calling index, correlated with beluga density, is used as a proxy for habitat use by the population at weekly, monthly, and yearly scales. The seasonal pattern along SLE upstream-downstream axis was repeated annually. In summer, beluga habitat was confined to a 150 km segment of the SLE, with higher occurrences in its ∼20 km central portion, including the head of the Laurentian Channel and Saguenay Fjord mouth. During fall, the distribution gradually shifted to the downstream portion of the SLE and into the Northwestern Gulf, leaving low to no occurrences upstream in winter, until the spring return, characterized by the highest upstream occurrences. Occurrences off Ste. Marguerite Bay, 25 km upstream in Saguenay Fjord, were essentially from June to October. This multi-year continuous habitat use pattern provides a baseline for year-round SLE beluga distribution dynamics for assessing and mitigating anthropogenic threats to this endangered population, such as shipping noise. It also provides insights for optimizing the assessments of population size from aerial line transect surveys.

Performance of a deep neural network at detecting North Atlantic right whale upcalls.

Passive acoustics provides a powerful tool for monitoring the endangered North Atlantic right whale (Eubalaena glacialis), but robust detection algorithms are needed to handle diverse and variable acoustic conditions and differences in recording techniques and equipment. This paper investigates the potential of deep neural networks (DNNs) for addressing this need. ResNet, an architecture commonly used for image recognition, was trained to recognize the time-frequency representation of the characteristic North Atlantic right whale upcall. The network was trained on several thousand examples recorded at various locations in the Gulf of St. Lawrence in 2018 and 2019, using different equipment and deployment techniques. Used as a detection algorithm on fifty 30-min recordings from the years 2015-2017 containing over one thousand upcalls, the network achieved recalls up to 80% while maintaining a precision of 90%. Importantly, the performance of the network improved as more variance was introduced into the training dataset, whereas the opposite trend was observed using a conventional linear discriminant analysis approach. This study demonstrates that DNNs can be trained to identify North Atlantic right whale upcalls under diverse and variable conditions with a performance that compares favorably to that of existing algorithms.

Decadal passive acoustics time series of St. Lawrence estuary beluga.

Passive acoustics is used to monitor the threatened St. Lawrence estuary beluga between 2007 and 2017 from a site downstream of the beluga summer habitat. Acoustic metrics of presence and occurrence based on beluga acoustic band activity (BABA) are extracted by a dedicated algorithm adapted for the shipping noise from the St. Lawrence Seaway. A formal optimization process is used to set the algorithm parameters. Results evidence a year-round occurrence of belugas in the region, seasonal and diel patterns, and significant inter-annual variations. This study shows how passive acoustics methodology can be applied to monitor a loquacious species over multi-year periods in a shipping-noise-dominated environment, in order to understand its use of the habitat over the continuum of ecologically significant time scales.

Analysis and modeling of 255 source levels of merchant ships from an acoustic observatory along St. Lawrence Seaway.

An ensemble of 255 spectral source levels (SSLs) of merchant ships were measured with an opportunistic seaway acoustic observatory adhering to the American National Standards Institute/Acoustical Society of America S12.64-2009 standard as much as possible, and deployed in the 350-m deep lower St. Lawrence Seaway in eastern Canada. The estimated SSLs were sensitive to the transmission loss model. The best transmission loss model at the three measuring depths was an empirical in situ function for ranges larger than 300 m, fused with estimates from a wavenumber integration propagation model fed with inverted local geoacoustic properties for [300 to 1 m] ranges. Resulting SSLs still showed a high variability. Uni- and multi-variate analyses showed weak intermingled relations with ship type, length, breadth, draught, speed, age, and other variables. Cluster analyses distinguished six different SSL patterns, which did not correspond to distinctive physical characteristics of the ships. The broadband [20-500 Hz] source levels varied by 30 dB or more within all four 50-m length categories. Common SSL models based on frequency, length and speed failed to unbiasly replicate the observations. This article presents unbiased SSL models that explain 75%-88% of the variance using frequency, ship speed, and three other automatic identification system ship characteristics.

A Seaway Acoustic Observatory in Action: The St. Lawrence Seaway.

A setup for measuring spectral source levels (SSLs) of ships transiting along a seaway, the traffic density and shipping noise, is presented. The results feed shipping-noise modeling that reproduces the actual in situ observations to map shipping-noise variability over space and time for investigating its effects on aquatic organisms. The ship's SSL databank allows sorting the different contributors to total shipping noise for assisting in exploring mitigation approaches (e.g., fleet composition, rerouting). Such an acoustic observatory was deployed since November 2012 for a complete annual cycle of measurements in the deep downstream part of the St. Lawrence Seaway.