Abstract: For the OFDM underwater acoustic communication systems, the least square (LS) channel estimation algorithm has the disadvantages of low estimation accuracy and high pilot overhead, while the static compressive sensing channel estimation algorithm has the disadvantages of high computational complexity and poor real-time performance. Aiming at these problems, a new algorithm called dynamic orthogonal matching pursuit (D-OMP) is proposed by establishing a dynamic sparse observation model based on the temporal correlation of the underwater acoustic channel impulse response. The algorithm only performs a complete OMP channel estimation at the initial time to obtain the channel support set, and then tracks the channel by continuously tracking changes in the previous channel support set. The simulation results show that the proposed algorithm has better channel tracking performance and lower algorithm complexity compared with the traditional LS algorithm and the classical OMP algorithm under the same pilot overhead.