Abstract: In order to reduce the influence of sound speed error on method based on sound velocity calibration is proposed. Firstly, the the localization accuracy, a distributed localization acoustic velocity is derived as functions of unknown source location, to approximate the acoustic velocity field distribution in the wind field. Then, they are introduced into the TDOA (Time Differences of Arrival) algorithm, to construct nonlinear equations. Finally, the particle swarm optimization algorithm is used to get the source location. The simulated results show that the proposed algorithm can significantly improve the localization accuracy for different wind velocity, source location and sizes of test area. The experimental results show that the localization error of proposed method can be reduced to 41% of original in the four nodes' localization system.