Estimating three-dimensional current fields in the Yeosu Bay using coastal acoustic tomography system

Abstract

Observation of current speeds in coastal seas is crucial because it can provide useful information for ship operations, fishing activities, and rapid responses to marine disasters. Coastal acoustic tomography (CAT) is a technology that can continuously monitor environmental changes such as current velocity and water temperature using reciprocal acoustic signals between CAT stations in coastal seas. This technology is different from traditional pointwise or intermittent sectional observations in that it can produce time-varying two- or three-dimensional current fields. The results of previous studies using CAT systems have been limited to reproducing horizontal maps of depth-averaged two-dimensional current fields. Utilizing results from a high-resolution coastal ocean model, this study developed a novel technique for estimating three-dimensional (3-D) current fields by combining the inverse method with an artificial intelligence (AI) model. Following three steps are the procedure for the test of estimating the 3-D current fields. First, utilizing the ray tracing model 'Bellhop,' reciprocal travel times among five CAT stations using the coastal ocean model outputs are computed. These five stations correspond to the locations where in-situ CAT systems were established for continuous monitoring of current changes in Yeosu Bay, Korea. Subsequently, the range-averaged currents at the five layers were estimated by incorporating this travel time difference data into an AI model trained using the same coastal ocean model outputs. Finally, the inverse method is applied to each layer to estimate the 3-D current fields. The validation results revealed that the newly developed method performed well in both summer and winter. Time-varying two-layer-like current fields were reasonably produced, occasionally revealing an out-of-phase relationship between the upper and lower layers depending on the tidal phases. This method yielded average root-mean-squared errors of less than 4 cm/s on six simulation paths for acoustic signal propagation. Furthermore, when the same method was applied to in-situ CAT observations, the average correlation coefficient (R) of the along-channel current of each layer was found to be approximately 0.9 or higher. These results suggest that this novel method can be effectively applied to the continuous monitoring of 3-D current fields in coastal seas using a CAT system.