Near real-time calculation of submarine fault properties using an inverse model of acoustic signals

A submarine earthquake that generates a tsunami sends a family of acoustic signals that carry information on the fault dynamics and geometry. These signals travel at the speed of sound in water, much greater than the phase speed of the gravity waves, and thus can act as early warning of tsunamis. To utilise this property in real-time, a semi-analytical inverse approach is employed assuming the fault is slender and the water depth is constant, allowing the calculation of some fault parameters analytically. However, the remaining parameters require a numerical evaluation, which increases the calculation time substantially. In order to overcome this difficulty, a probabilistic inverse model is proposed. The model analyses data at the envelope of a pressure signal, which reduces numerical complexities. More specifically, it selects multiple measurement points, in order to produce several sets of solutions within given ranges of the properties. The model is applied to real hydrophone recordings, where the fault geometry and dynamics are estimated near real-time on a standard PC. Some aspects of the model are general and can be used to estimate simplified geometry and dynamics of various signal sources from violent events in the ocean, such as impacting meteorites, submarine explosions, landslides and rogue waves.