Multi-fidelity data fusion for ocean sound speed

Modelling acoustic propagation in the ocean environment requires accurate sound speed profiles. However, ocean sound speed data are often limited to measurements from in situ probes and historical or forecast climatological data, which inherently differ in accuracy, resolution, and spatiotemporal coverage. Herein, multi-fidelity data fusion is used with disparate data to produce updated estimates of sound speed profiles and their uncertainties. This method is compared to a reference data field, and it is applied to depth-only data fusion and three-dimensional data fusion with multiple expendable probes. Acoustic propagation modelling is performed to demonstrate the importance of accurate and up-to-date sound speed profiles, highlighting how differences, and uncertainty, in sound speed data impact acoustic predictions. Monte Carlo simulation is used to quantify uncertainty in acoustic propagation arising from uncertainty in sound speed. Through a series of applications, it is shown that fused sound speed profiles obtained through this method can significantly affect acoustic propagation modelling. The approach developed here supports predictions of acoustic propagation with quantified uncertainty for critical applications such as underwater communication, sonar systems, and environmental monitoring.