Quantifying the accuracy of acoustic arrival-time predictions using high-resolution ocean circulation models

Accurately modeling acoustic propagation in dynamic ocean environments exhibiting sub-mesoscale variability is challenging. Such environments require high-resolution ocean circulation model to accurately predict local sound speed variability at sub-mesoscales. This study focuses on high-resolution rendition of the De Soto canyon area, located in the Gulf of Mexico. Predicted sound speed profiles were found at times to exhibit very strong vertical gradients (e.g. with sharp “elbows”) reflecting the small-scale vertical variations of ocean layer properties. If used as raw inputs for ray-tracingsimulations, such sound speed profiles were found to generate inconsistent estimation of eigenray paths and spurious paths appearing intermittently caused by the inherent limitations of the ray-tracing methods. Nevertheless, we show that such artefacts can be mitigated by appropriately smoothing the input sound speed profile spatial variability especially if only low to mid-frequency (<1 kHz) propagation is predicted. To do so, the accuracy of arrival-times predictions using ray-tracing predictions with spatially smoothed sound speed profiles as a function of center frequency (<1 kHz) will be quantified against parabolic equations simulations used here as reference solution. We hypothesize that this smoothing approach for the input sound speed variations will enable to retain the numerical advantage of ray-tracing predictions while maintaining sufficient accuracy.