The early success of MERMAID belied its drawing power for further funding by US agencies (are these the wages of interdisciplinarity?), but the move of Guust Nolet to Géoazur at Sophia Antipolis, France, proved to be the impetus for the crucial second phase. His collaboration with Yann Hello at Géoazur gave rise to the second-generation model and led to a series of science deployments. A mariner’s dozen APEX (Autonomous Profiling Explorer) floats, built by Teledyne Webb Research, were equipped with a Rafos II Benthos hydrophone, sensitive down to 0.1 Hz, and deployed in a number of different locations from the Mediterranean to the Indian Ocean, and in the Pacific Ocean for a first tomographic science experiment dedicated to imaging the deep roots of the Galápagos plume.

As with the first-generation instrument, algorithms development and their electronic implementation on dedicated low-power boards, carried out by French engineering firm OSEAN, were just as critically important as hardware improvements. Alexey Sukhovich, now at the University of Brest, took the wavelet-based detection and discrimination routines of the first-generation MERMAID to a new level, endowing them with a probabilistic evaluation and scoring system that intelligently guides MERMAID towards sound decision making. In continued use on all instruments built to the present day, the algorithm confidently picks out segments of seismological interest and reports them to the receiving data centers, making MERMAID fully autonomous and capable of reporting data in near real-time.

In this context, one often wonders whether MERMAID technology could be useful for tsunami detection and early-warning systems. The chief trade-off will be with cruising depth: current ascent speeds are on the order of 300 m/h (descent around −100 m/h). MERMAID would have to hold off triggering its ascent (which stops the recording) to report the seismic arrival from the causative earthquake. Issues with hydrophone or pressure-sensor calibration at very low frequencies (below 0.01 Hz) require to be resolved. But, since tsunami waves caused by an earthquake in, e.g., Chile, reach French Polynesia, Hawaii, and Japan about 10, 15, and 22 hours after the main event, respectively, there is, in the recommendation of Masayuki Obayashi from JAMSTEC, a space for a worldwide array of instruments to fill this important niche.

On their various runs roaming at depths between 1500 and 2000 m across the Mediterranean Sea and the Indian Ocean, second-generation MERMAID established itself as a reliable purveyor of signals from earthquakes large and small. While signal-to-noise environments vary vastly between oceans, with the seasons, and according to the parking depth, multiple deployments revealed that between 35 and 63% of worldwide earthquakes with moment magnitudes greater than 6.5 can be recorded—and magnitudes smaller than 6.0 under favorable noise conditions. Over the course of four weeks in 2013, one of the floats unexpectedly reported no fewer than 235 seismograms from an earthquake swarm following a magnitude 5.1 main shock that occurred on 24 November 2013 on the triple junction where the African, Indian and Antarctic plates meet. These aftershocks had magnitudes between 2.7 and 3.4, thereby establishing a lower magnitude threshold in these noisiest of environments. Nearby land stations recorded only 25 of the largest of these aftershocks (all greater than 4.4 and all but two within four days of the main shock). While global seismology continues to focus on larger and more distant earthquakes, scientists interested in the seismic budget of closely collocated crustal earthquakes in the oceans will find MERMAID perfectly capable of being optimized for their recovery.

The first mid-scale coordinated "scientific" experiment was dedicated to imaging the mantle roots of the Galápagos volcanic hot spot. Nine second-generation MERMAIDs launched by Ecuadorean INOCAR vessel LAE Sirius remained in operation for about two years each, altogether returning 580 crucial arrival times sampling different ray paths, illuminating mantle corridors unresolved by any other seismic information. In combination with data from land stations, the tomographic inversion revealed that the Galápagos archipelago is underlain by a deeply-rooted (about 1900 km), 200–300 km wide plume: rocks buoyed up by excess temperature, carrying a heat flux likely much larger than that predicted from the swell bathymetry.

Read on about the Third Generation.