A data assimilation technique was applied to analyze the first coastal acoustic tomography data, obtained in the Neko-Seto Channel of the Seto Inland Sea, Japan on March 2--3, 1999. The ensemble Kalman filter technique accompanied by the forecast of error covariances by means of the Monte Carlo method is found to be best fitted to the present problem because currents in the tomography region are dominated by the strongly nonlinear tidal vortices. Furthermore, the analyzed results are in better agreement with the shipboard ADCP results than those obtained by the inverse analysis of the travel time difference data for the pairs of acoustic stations. ¿ 2000 American Geophysical Union