Hydroacoustic Travel Time Variations as a Proxy for Passive Deep-Ocean Thermometry

This study demonstrates the retrieval of travel time variations by time-lapse analysis of a three-component broadband seismometer and a triplet hydrophone station given an underwater source. The repeated hydroacoustic activity of the submarine Monowai Volcanic Centre, observed at the hydrophone station H03S and the seismic station RAR, is used to estimate the relative travel time variation. Variations in travel time are a proxy for spatially averaged changes in the deep-ocean temperature. A proposed pre-processing workflow overcomes the differences in instrument sensitivity and sample rates. The vertical and radial seismic components are cross-correlated with all individual hydrophones. Cross-correlation functions (CCFs) eligible for time-lapse analysis are selected based on source activity and source directionality. Automatic coincidence-trigger analysis of the CCF's signal-to-noise ratio marks the active periods. The source directionality corresponds to the horizontal slowness of a plane wave traversing the hydrophone triplet, applied to the CCFs with the vertical seismic component. The time-lapse analysis is applied to the eligible CCFs by two-dimensional cross-correlation of the power spectral density in two consecutive octave bands, for 3-6 Hz and 6-12 Hz. The resulting peak two-dimensional cross-correlation coefficient and the corresponding time lag and frequency variations yield the travel time variation of use for the deep-ocean thermometry. The medium induced travel time change is assumed to be equivalent to the trimmed weighted mean variation in time lag and assuming no variation of the source signature. The estimated travel time variation per octave band reveals both a complex periodic variation as well as a distinct linear trend. The linear decrease is assumed to be associated with net temperature increase and is larger around the SOFAR channel (higher-frequency octave) compared to the deep ocean (lower-frequency octave). The peak periodicity for the lower and higher frequency octaves is estimated at 1.9 y and 1.1 y, respectively. The methodology applies to other triplet hydrophone arrays or three-component seismometers given a known source. This work intends to be a cookbook study, fostering FAIR data principles. Source code and examples are publicly available on GitLab.