The method for sea clutter modeling with second order Volterra series filter is discussed.According to the phase space reconstruction theory,the embedded dimension is adopted as the length of the filter.The parallel multiplication-coupled structure for second order Volterra series filter is built up to reduce the complexity.The adaptive NLMS algorithm is used for adjusting the filter core coefficient in real time.Contrast experiment shows that the algorithm has faster convergence rate and smaller MSE.The Volterra filter was used for prediction of real sea clutter data.The results indicate that:1) this filter can model and predict sea clutter accurately;and 2) compared with RBF network,it has better performance for large step prediction,while it is slightly inferior in one-step prediction performance.