Characterizing underwater noise during rain at the northeast Pacific continental margin

Approximately 11 400 h of acoustic recordings from two sites off the Oregon coast have been evaluated to characterize and model the frequency and wind dependence of wind noise in the northeast Pacific continental margin. Acoustic data are provided by two bottom-mounted broadband hydrophones (64 kHz sampling frequency) deployed at depths of 81 and 581 m at the continental shelf and slope, respectively. To describe the spectral level versus frequency relation, separate linear models for the 0.2–3 kHz and 3–25 kHz frequency range are fitted to the data. While spectral slopes for the 0.2–3 kHz range generally decrease with increasing wind speed, slopes remain constant (shallow location) or increase with increasing wind speed (deep location) above 3 kHz. The latter is in strong contrast to results from previous studies. The relation between spectral level and wind speed is described by a piecewise linear model where spectral levels are approximately constant below a critical wind speed vc and increase linearly with logarithmic wind speed above vc. It is shown that the critical wind speed and the slopes of the piecewise linear model strongly depend on the acoustic frequency.

Statistical analysis and modeling of underwater wind noise at the northeast pacific continental margin

Schwock

Abadi

2021

An overview of ambient sound using Ocean Observatories Initiative hydrophones

Ragland

Schwock

Munson

et al. 2022

The Ocean Observatories Initiative (OOI) sensor network provides a unique opportunity to study ambient sound in the north-east Pacific Ocean. The OOI sensor network has five low frequency (Fs = 200 Hz) and six broadband (Fs = 64 kHz) hydrophones that have been recording ambient sound since 2015. In this paper, we analyze acoustic data from 2015 to 2020 to identify prominent features that are present in the OOI acoustic dataset. Notable features in the acoustic dataset that are highlighted in this paper include volcanic and seismic activity, rain and wind noise, marine mammal vocalizations, and anthropogenic sound, such as shipping noise. For all low frequency hydrophones and four of the six broadband hydrophones, we will present long-term spectrograms, median time-series trends for different spectral bands, and different statistical metrics about the acoustic environment. We find that 6-yr acoustic trends vary, depending on the location of the hydrophone and the spectral band that is observed. Over the course of six years, increases in spectral levels are seen in some locations and spectral bands, while decreases are seen in other locations and spectral bands. Last, we discuss future areas of research to which the OOI dataset lends itself.

Listening for rain: Principal component analysis and linear discriminant analysis for broadband acoustic rainfall detection

Mallary,

Berg,

Buck

et al. 2023

Rain falling on the ocean creates acoustic signals. Ma and Nystuen [(2005). J. Atmos. Oceanic Technol. 22, 1225–1248] described an algorithm that compares three narrowband “discriminant” frequencies to detect rain. In 2022, Trucco, Bozzano, Fava, Pensieri, Verri, and Barla [(2022). IEEE J. Oceanic Eng. 47(1), 213–225] investigated rain detection algorithms that use broadband spectral data averaged over 1 h. This paper implements a rainfall detector that uses broadband acoustic data at 3-min time resolution. Principal Component Analysis (PCA) reduces the dimensionality of the broadband data. Rainfall is then detected via a Linear Discriminant Analysis (LDA) on the data's principal component projections. This PCA/LDA algorithm was trained and tested on 5 months of data recorded by hydrophones in a shallow noisy cove, where it was not feasible to average spectral data over 1 h. The PCA/LDA algorithm successfully detected 78 ± 5% of all rain events over 1 mm/h, and 73 ± 5% of all rain events over 0.1 mm/h, for a false alarm rate of ≈ 1% in both cases. By contrast, the Ma and Nystuen algorithm detected 32 ± 5% of the rain events over 1.0 mm/h when run on the same data, for a comparable false alarm rate.