Abstract: A class of adaptive nonlinear prediction algorithm was proposed to predict ocean reverberation based on the phase space reconstruction of nonlinear dynamic system and track the evolving trajectory in the phase space of the reverberation by developing an adaptive filter.The prediction algorithm was tested through the experimental reverberation time series measured in different sea areas,and the one-step forward prediction results were well fit by the experimental data.If the errors between the predicted and experimental data are considered as the statistical test variable to detect the target in the reverberation series,the reverberation is suppressed and the signal-to-reverberation ratio is improved.