Abstract
We propose a new method to estimate a 2D detail temperature distribution in the ocean from seismic data. This method is based on a combination of Simulated Annealing (SA) and fast waveform calculation methods. In contrast to the local optimization methods, the SA enables us to optimize the model parameters with less dependency on the initial model. We applied this inversion method to multi-channel seismic reflection data acquired around the axis of Kuroshio Current. Since the survey line was oriented NW-SE direction, we can obtain the velocity / temperature structure across the Kuroshio Current using the seismic data. The 2D acoustic velocity structure derived from the inversion revealed that the multi-layered structures are dominant in the ocean, and these layers are dipping toward the current axis. We estimated the temperature distribution from the obtained acoustic velocity using an empirical formula. Furthermore, we investigate the effect of random noise on the inversion results. We contaminated the data derived from the finite difference method by Gaussian distributed noise. We observed that the estimated velocity contrasts tend to get large compared to the true contrasts as S/N ratio decreases. This suggests that one of the reasons for the large velocity contrasts in the obtained acoustic velocity profiles could be originated from the observation noises.
