Accurate three-dimensional (3D) atmospheric water vapor distribution, which plays a crucial role in understanding the meteorological phenomena and hazards, has been successfully retrieved using Global Navigation Satellite System (GNSS) tomography technique. Presently, the problem of the ill-posed tomography system, that results from poor GNSS acquisition geometry, remains a vital issue to be solved. In this paper, we develop a new hybrid observation GNSS tomography (HOGT) method, which constructs and introduces the virtual signals, inverted to the real rays, to address the acquisition geometry defect. Within the HOGT, the slant wet delay (SWD) of the virtual inverted signal (VIS) is estimated by means of the tropospheric parameters derived from the hourly ERA5 data using the ray-tracing algorithm. Two designed experiments, based on the dense and sparse GNSS networks (corresponding to Hong Kong and Xuzhou), respectively, are implemented to assess the performance of HOGT for the different networks. The results reveal that HOGT provides a more robust observation geometry and more accurate water vapor distribution than the traditional tomography model, with the mean number of the crossed voxels enhanced by 26.45% and 27.11% for the two networks, and the average root-mean-square error (RMSE) of the tomography solutions improved by 18.18% and 38.28% in the two areas, respectively. Furthermore, HOGT shows a significant improvement close to the Erath’s surface from 0 to 2 km, implying its superior capability to optimize the accuracy of tomography results. An additional experiment to investigate the performance of the proposed method under different time resolutions demonstrates that HOGT can be promised to retrieve more accurate water vapor distribution over short time intervals, especially for rainy days with an interval of 10 min or even shorter, which highlights the interest in tomography solutions for improving the understanding of severe weather.