Page 69 - Summer2022

Figure 7. a: Seaglider (SG-190) profiles during a field experiment at 10°N, 125°W. b: Acoustic spectra of wind and rain were identified using rule-based algorithms. Color dots in a correspond to color spectra in b and indicate the detection of wind and rain according to various detection rules (wind 1-3, rain 1-2, and drizzle). SPL, sound pressure level; raw, no rain or wind. The challenges for the acoustic method are that it is passive, it relies on correct amplitude calibration, and other sources of ambient sound can affect acoustic data quality. The Future To a large extent, the development of the Ocean Observing System has resolved the perennial problem of available platforms to hang a rain gauge on at sea. In addition, a cabled observing system across the seafloor provides a constant stream of real-time data from passive acoustic sensors located both on the seafloor and in the water column. These systems, which are cabled to shore, allow real-time data access from the comfort of one’s home. Rain- fall signals can be extracted from the acoustic data stream from a cabled hydrophone system (Schwock and Abadi, 2021). These observations, employed for scientific studies or monitoring beyond the rain measurements, align with the concept of multipurpose acoustic systems suggested by Howe et al. (2019). The adoption of general-purpose acoustic receivers can serve the scientific community inter- ested in passive acoustic monitoring. Ocean sound as an essential ocean variable (Miksis-Olds et al., 2018) is gaining more visibility and traction in the global ocean-observing community. For instance, the MERMAID program (Nolet et al., 2019; Simons et al., 2021), a project of the United Nations Decade of Ocean Science for Sustainable Development 2021–2030, devel- oped floats for hydroacoustic monitoring of earthquakes. Float capabilities are now being extended to general pur- pose use, including rain and wind measurements and marine mammal detections. The original PAL was developed to provide a consistent, reliable, and self-contained long-term recorder. It stored spectra and very short sound bites (seconds) only to ease the computational burden of data processing. The lack of complete time series, however, limits the ability to distinguish transient noises either from platform self- generated noises or unidentified sources. A 1-month acoustic time series, sampled at 120 kHz, requires several terabytes of data storage, which is now read- ily available. Such large datasets are helpful to dissect the transient sounds that may harbor new discoveries. Some of the advanced autonomous platforms of Ocean Observing Systems can transmit processed data in near- real time, but they are limited by onboard power storage and data bandwidth. It is an art to balance all the factors to decide when, what, and how to process and trans- mit useful ocean environmental data. The challenge of vast datasets may be addressed with machine-learning applications that may reduce dimensionality and cluster and classify acoustic source data (Bianco et al., 2019), perhaps allowing Nystuen’s vision (hearing) to reach its full potential. We hope to develop next-generation instruments for new passive acoustic rain measurements specifically and for high-frequency sound monitoring generally, exploit- ing all the recent advances. Such instruments would be designed to be deployed on any of the several platforms available from operational Ocean Observing Systems. Acknowledgments We are grateful to the editor, Arthur N. Popper, for con- structive suggestions and comments to make the article in its present form. Barry Ma is supported by Office the Naval Research Grant N00014-19-1-2626. Summer 2022 • Acoustics Today 69