Inverting sound speed, temperature, and current velocity in waters via coastal acoustic tomography informed Fourier neural operator

Reconstructing hydrological parameters with high temporal-spatial resolution for water resources management and underwater research is significant. As a classical non-contact measure method, coastal acoustic tomography (CAT) technology has been widely used to reconstruct hydrological parameters under different waters. Satisfying the time-efficient requirements under high temporal resolutions is crucial for reconstruction of hydrological parameters. However, traditional method requires extensive time and resource on processing of measurement data. Here, we present a CAT informed Fourier neural operator (FNO) to efficiently reconstruct hydrological parameters. Meanwhile, an experiment utilizing three CAT systems was conducted in a reservoir to test the performance of our model. We reconstruct the hydrological parameters (i.e., sound speed, temperature, and current velocity). The test results illustrate the averaged coefficient of determination of sound speed, temperature, and current velocity is 0.9991, 0.9986, and 0.9648, respectively. Additionally, in inversion stage, the averaged inference time of sound speed, temperature, and current velocity decrease to 0.71 s, 0.72 s, and 0.71 s, respectively. The results verify the CAT informed FNO method exhibits high efficiency. This work provides a new insight into the realization of time-efficient reconstruction for hydrological parameters subsequently.