Abstract: To address the challenging computational complexity associated with the two-dimensional direction of arrival estimation underwater in curved-surface arrays, this study introduces a fast calculation method jointing two-stage beamforming and deconvolution. Firstly, the method decomposes the high-dimensional steering vector of the array into the Kronecker product of two low-dimensional steering vectors. This favors fast computation of the two-dimensional direction of arrival estimation by horizontal and vertical beam weighting. Then it is combined with one-dimensional deconvolution beamforming through two-dimensional spatial spectrum decomposition and reconstruction. This realizes the fast and high-resolution direction of arrival estimation of the two-dimensional regular area arrays. The simulation results show that the computational complexity of the proposed method is approximately two to three orders of magnitude lower than that of the two-dimensional minimum variance distortionless response (2D-MVDR) beamforming method. Moreover, the method demonstrates superior target resolution and more accurate angle estimation performance. The lake trial data further verifies the performance of this method in the direction of arrival estimation and confirms the feasibility of the technical implementation. Furthermore, the lake trial data also validates that this method offers better target resolution and noise suppression capabilities compared to the 2D-MVDR method.