Submarine Sensor Environmental Inference

The Base and Option propose applying machine learning (ML) deep learning (DL) methods to estimate the Waveguide Invariant (WGI). The ML and DL application also provide prior parameterization of the waveguide. Data fusion methods generate Energy Spectral Density (ESD) and Empirical Orthogonal Functions (EOF) to characterize the uncertainty of the sound speed profile in the water column from which to develop a Likelihood function. These quantities then become input into a multi-step Bayesian maximum entropy (BME) method which determines the frequency-dependent source levels (SL) which in turn allows for the vertical Time axis in surface ship spectrograms to be replaced by Range, and for Received Levels (RL) to be replaced by transmission loss (TL). The resulting maximum a priori (MAP) for seabed parameter values then provide point estimates for the waveguide parameter values. With the MAP point estimates, the TL versus range to the SSN array for a shallow source depth can be established. The in situ determined TL can then be used as a measure to test and evaluate TL determined with the Submarine STDA.