THE FIRST GENERATION

Born from ideas hatched by Guust Nolet in the 1990s, and further developed by Frederik J. Simons at Princeton, MERMAID-001 was built by Jeff Babcock, John Orcutt, Russ Davis, and their groups at the Scripps Institution of Oceanography in 2003, with funding from the US National Science Foundation. A prototype in the traditional sense of the word, it fulfilled the core requirements of seismological functionality. A modified SOLO (Sounding Oceanographic Lagranian Observer) float, the acoustic payload was a High Tech HTI-90-U hydrophone. Data were sampled at 20 Hz, recorded with 16-bit precision, and stored on a flash memory card.

All in all, MERMAID-001 gathered a mere 120 hours of acoustic pressure data from a depth of around 700 m offshore La Jolla, without telemetry. The results were nevertheless encouraging: over the course of three recovered field tests in November 2003, September 2004, and August 2007, several positive earthquake identifications stood out from the noise. One of these was a tremor offshore Colombia, of sufficient magnitude (moment magnitude 6.0) and epicentral distance (46.5°, an equivalent 5170 km measured along the surface of the earth, which qualifies it as teleseismic) to prove the potential of MERMAID ultimately improving seismic coverage in the oceans. Several smaller-magnitude events were, strictly speaking, bycatch: not useful for deep-earth imaging, but suitable for assessments of local and regional seismicity, and to study crustal structure.

Funding from the UK Natural Environment Research Council (NERC) and the Nuffield Foundation, and from Princeton University, proved essential to ensuring continued innovation.

Central to the early development was the crafting of algorithms that enable MERMAID to, completely independently and on an ongoing basis, detect segments of incoming energy, and identify any signals of seismological interest. Detection was carried out by way of traditional STA/LTA analysis, whereby ratios of short-term over long-term average amplitude are monitored for threshold exceedance, as is commonly done for telemetered seismometers worldwide. The discrete wavelet transform (DWT) proved particularly adept at the task of discrimination and identification, with a custom algorithm developed by Frederik J. Simons at Princeton University, later at University College London.

Implemented in integer arithmetic with in-place calculation via the fast lifting algorithm, the streaming time-domain samples are reduced to an evolving bar code whose character quite clearly reveals which records are due to distant earthquakes, compared to arrivals from regional and local events whose high-frequency energy has survived selective removal by intrinsic attenuation and geometrical spreading in the solid Earth. Another distinct acoustic fingerprint is from T phases from very shallow sources, whose energy carries over long distances, bouncing around in the Sound Fixing and Ranging (SOFAR) low-velocity channel of the ocean itself. Most lately, these have been used for seismic thermometry over long spatial and temporal baselines.

Simple pattern recognition (no machine-learning required!) in the two-dimensional space of time versus wavelet scale removes potential triggering due to non-seismological sources (e.g. shipping noise, volcanic eruptions, glacial calving, rain, biological signals—not to mention mysterious plane crashes). The breakdown of time-domain acoustic pressure variations into scale-dependent wavelet coefficients provides useful ways of data denoising and compression, and has revealed its utility in applications for seismometry on land, e.g., for earthquake early-warning studies. Adaptations targeting rather than avoiding such competing sources of energy are straightforward: after all, one maid’s noise is another one’s signal. Indeed, independently developed profiling floats with acoustic packages tailor-made for cetacean (whales!) census research, and wind and rainfall intensity, i.e., meteorology, are to be considered part of the extended family of "Lagrangian" "quasi-Eulerian" hydroacoustic marine devices.

MERMAID-001 retired to the University of Rhode Island, where she went on display on the Narragansett campus in the lab of Harold T. "Bud" Vincent.