Abstract. Existing water vapor tomographic methods use Global Navigation Satellite System (GNSS) signals penetrating the entire research area while they do not consider signals passing through its sides. This leads to the decreasing use of observed satellite signals and allows for no signals crossing from the bottom or edge areas especially for those voxels in research areas of interest. Consequently, the accuracy of the tomographic results for the bottom of a research area, and the overall reconstructed accuracy do not reach their full potential. To solve this issue, an approach which uses GPS data with both signals that pass the side and top of a research area is proposed. The advantages of proposed approach include improving the utilization of existing GNSS observations and increasing the number of voxels crossed by satellite signals. One point should be noted that the proposed approach needs the support of radiosonde data inside the tomographic region. A tomographic experiment was implemented using observed GPS data from the Continuously Operating Reference System (CORS) Network of Zhejiang Province, China. The comparison of tomographic results with data from a radiosonde shows that the root mean square error (RMS), bias, mean absolute error (MAE), and standard deviation (SD) of the proposed approach are superior to those of the traditional method.