default Coupled ocean-acoustic structures and covariances for improved acoustic prediction: exploiting the synergies between physical and acoustic observing networks

Coupled ocean-acoustic structures and covariances for improved acoustic prediction: exploiting the synergies between physical and acoustic observing networks. Storto, Andrea ; Falchetti, Silvia ;  Oddo, Paolo. CMRE-MR-2019-025. March 2020.

This report investigates the feasibility of assimilating underwater acoustic data into ocean predictive models through extended data assimilation schemes. In particular, the impact of transmission loss (TL) data associated with a simple geometry of source and receivers was assessed. The scenario analyzed in this study considers a low frequency acoustic propagation signal (75 Hz), conceived here as a sound source of opportunity for improving oceanographic forecasts. Assimilation of TL data was achieved through implementing an observation operator that maps forward and backward increments of temperature onto increments of TL. Such an operator is based on the canonical correlation analysis (CCA) of physical and acoustic datasets coming from an ensemble of oceanic and acoustic simulations. The use of CCAs approximates to large extent the correlations between temperature and TL data extracted from integrated NEMO (nucleus for European modelling of the ocean) ocean model and RAM (range-dependent acoustic model) acoustic model simulations, respectively. The experiment setup follows the one typical of observing system simulation experiments (OSSE), where a perturbed nature (truth) run provides TL synthetic observations. Control and assimilation experiments were then performed to evaluate the potential benefits of such observations. It turns out that the ingestion of TL data helps to improve the upper ocean structure and the mixed layer depth mean state and variability in the areas in proximity of the underwater propagation path. The largest benefits of the assimilation are found in the vertical region between 100 and 200 m of depth where the sound speed is the smallest and the sensitivity to temperature fields is the largest. In such region, the decrease of temperature root mean square error exceeds 30% compared to the assimilation-blind experiment. The assimilation also leads to improved acoustic predictions if the corrected temperature fields are used within the RAM simulations. The use of multiple outer loops within the analysis-forecast steps to refine the linearization embedded in the variational data assimilation scheme is found positive in the top 200 m of depth. Finally, the report discusses ways to further enhance the impact of such observations and paves the way for future research activities.