Effect of Frontal Instabilities on Underwater Acoustic Propagation: Study of the Narval 2021 Data

Thermohaline frontal areas can strongly influence acoustic propagation. Due to their dynamics, they are areas difficult to model. In this preliminary work, we propose to use Gaussian process regression to fit a simple structure model to in-situ data. This then allows to model possible samples of the sound-speed field around the Iceland-Faroe front. The Gaussian process framework helps to find the front position, and capture the large scale variability of the area. We then compare ray-tracing simulations on different sound-speed fields, including Gaussian process samples, to in-situ acoustic data. We show that for some source-receiver configurations, range-dependent simulations using the Gaussian process model show the best fit to real data. The structure model lacks mesoscale and sub-mesoscale contributions, which have a strong acoustic impact, but is a good basis for more advanced environmental modelling.