Airguns used for offshore seismic exploration by the oil and gas industry contribute to globally increasing anthropogenic noise levels in the marine environment. There is concern that the omnidirectional, high intensity sound pulses created by airguns may alter fish physiology and behaviour. A controlled short-term field experiment was performed to investigate the effects of sound exposure from a seismic airgun on the physiology and behaviour of two socioeconomically and ecologically important marine fishes: the Atlantic cod (Gadus morhua) and saithe (Pollachius virens). Biologgers recording heart rate and body temperature and acoustic transmitters recording locomotory activity (i.e. acceleration) and depth were used to monitor free-swimming individuals during experimental sound exposures (18–60 dB above ambient). Fish were held in a large sea cage (50 m diameter; 25 m depth) and exposed to sound exposure trials over a 3-day period. Concurrently, the behaviour of untagged cod and saithe was monitored using video recording. The cod exhibited reduced heart rate (bradycardia) in response to the particle motion component of the sound from the airgun, indicative of an initial flight response. No behavioural startle response to the airgun was observed; both cod and saithe changed both swimming depth and horizontal position more frequently during sound production. The saithe became more dispersed in response to the elevated sound levels. The fish seemed to habituate both physiologically and behaviourally with repeated exposure. In conclusion, the sound exposures induced over the time frames used in this study appear unlikely to be associated with long-term alterations in physiology or behaviour. However, additional research is needed to fully understand the ecological consequences of airgun use in marine ecosystems.

Underwater source localization is an important task, especially for real-time operation. Recently, machine learning methods have been combined with supervised learning schemes. This opens new possibilities for underwater source localization. However, in many real scenarios, the number of labeled datasets is insufficient for purely supervised learning, and the training time of a deep neural network can be huge. To mitigate the problem related to the low number of labeled datasets available, we propose a two-step framework for underwater source localization based on the semi-supervised learning scheme. The first step utilizes a convolutional autoencoder to extract the latent features from the whole available dataset. The second step performs source localization via an encoder multi-layer perceptron trained on a limited labeled portion of the dataset. To reduce the training time, an interpretable feature selection (FS) method based on principal component regression is proposed, which can extract important features for underwater source localization by only introducing the source location without other prior information. The proposed approach is validated on the public dataset SWellEx-96 Event S5. The results show that the framework has appealing accuracy and robustness on the unseen data, especially when the number of data used to train gradually decreases. After FS, not only the training stage has a 95% acceleration but the performance of the framework becomes more robust on the receiver-depth selection and more accurate when the number of labeled data used to train is extremely limited.

Seismic imaging is essential for detailed characterization and understanding of the subsurface. Traditionally, marine seismic data have been acquired either using towed streamers or ocean-bottom receivers. Distributed acoustic sensing (DAS) is a novel and fast-emerging technology that makes it possible to use fiber-optic cables for acoustic measurements. The technology turns the fiber-optic cable into a densely sampled receiver array. A controlled source experiment comparing DAS and conventional towed streamer data for seismic imaging purposes is described. An existing 130 km telecommunication fiber-optic cable was interrogated to acquire DAS data to analyze the signal at long distances and for subsurface imaging. A seismic source vessel was used to create seismic data along the full cable length, while at the same time acquiring conventional towed streamer data for comparison. In addition, the telecommunication fiber-optic cable was used to acquire passive DAS data. The results show that seismic signals in the telecommunication fiber-optic cable are detectable at distances above 100 km from the source point. Furthermore, comparisons of the DAS data and towed streamer data show that the DAS technology allows for the acquisition of large offset seismic data that are comparable with conventional seismic data. Analysis of the passive DAS data shows that the infragravity waves correlate with the observed weather conditions during the acquisition.

We propose a time-frequency fused underwater acoustic source localization method based on self-supervised learning with contrastive predictive coding. Firstly, two feature extractors are trained to solve the pretext task (predicting the future) based on the unlabeled acoustic signals in the time and frequency domains, respectively. Next, encoders with frozen parameters are taken from the trained feature extractors for extracting the high-level features in the time and frequency domains. During the training stage of the source localizer, features extracted by two encoders are concatenated together as a time-frequency fused feature vector and fed into a 3-layer multi-layer perceptron for solving the downstream task (source localization) based on a tiny labeled dataset. This method is assessed on the SWellEx-96 Experiment and compared with several alternative methods. The performance analysis confirms the promising performance of our proposed method.

Accurate approximations of Green’s functions retrieved from the correlations of ambient noise require a homogeneous distribution of random and uncorrelated noise sources. In the real world, the existence of highly coherent, strong directional noise generated by ships, earthquakes, and other human activities can result in biases in the ambient-noise crosscorrelations (NCCs). We have developed an adapted eigenvalue-based filter to attenuate the interference of strong directional sources. The filter is based on the statistical model of the sample covariance matrix and can separate different components of the data covariance matrix in the eigenvalue spectrum. To improve the effectiveness and make it adaptable for different data sets, a weight is introduced to the filter. Then, the NCCs can be calculated directly from the filtered data covariance matrix. This approach is applied to a 1.02 h data set of ambient noise recorded by a permanent reservoir monitoring receiver array installed on the seabed. The power spectral density indicates that the noise recordings were contaminated by strong directional noise over nearly half of the whole observation period. Beamforming and crosscorrelation results indicate that the interference still exists even after applying traditional temporal and spectral normalization techniques, whereas the adapted eigenvalue-based filter can significantly attenuate it and help to obtain improved crosscorrelations. The approach makes it possible to retrieve reliable approximations of Green’s functions over a much shorter recording time.

Acoustics is important for all underwater systems for object detection, classification, surveillance systems, and communication. However, underwater acoustics is often difficult to understand, and even the most carefully conducted measurements may often give unexpected results. The use of theory and acoustic modelling in support of measurements is very important since theory tends to be better behaved and more consistent than experiments, and useful to acquire better knowledge about the physics principle. This paper, having a tutorial flair, concerns the use of ray modelling and in particular eigenray analysis to obtain increased knowledge and understanding of underwater acoustic propagation.