Acoustic reverberation signals generated by an experimental explosive source are analyzed by nonlinear dynamical methods. Three characteristic parameters, i.e., the correlation dimension, the largest Lyapunov exponent, and the Koimogorov en-tropy, are estimated in the reconstructed phase space. The results indicate that the reverberation signals are nonlinear. The Volterra adaptive prediction method is introduced to model the oceanic reverberation signals. The reverberation time series can be predicted in short term with small prediction errors. A preliminary conclusion can be reached that the nonlinear low-dimensional dynamic sys-tem model is more suitable for modeling oceanic reverberation than the classical random AR model.