Retrieval of small-scale three-dimensional (3D) wind field based on a single radar is an important issue in various fields, including aviation meteorology and wind resource assessment. To obtain the 3D structure of small-scale wind field with high efficiency, this study proposes a variational retrieval method based on lidar detection. The proposed method comprises local and global variational methods. First, the local variational method is constructed locally in an analysis volume to preliminarily retrieve the 3D wind field, as the initial value. Second, the initial value is tuned using the global variational method in the entire lidar scanning volume. Simulation results show that the proposed method outperforms the traditional two-step variational method. The root mean square error of the algorithm reduces by 1.49 m/s and 1.19 m/s in the case of homogeneous and inhomogeneous wind fields, respectively. Simulation and field detection results both prove that the proposed algorithm can obtain accurate 3D wind field retrieval results with high efficiency when the lidar scanning volume covers a large range of the elevation angle. The proposed method can be applied in aviation safety, wind resource assessment and other fields.