The Global Navigation Satellite System (GNSS) tomographic technique can be used for remote sensing of the three-dimensional water vapor (WV) distribution in the troposphere, which has attracted considerable interest. However, a significant problem in this technique is the excessive reliance on constraints (particularly in large GNSS networks). In this paper, we propose an improved tomographic method based on optimized voxel, which only considers the voxels passed by GNSS rays. The proposed method can completely prevent the tomographic algorithm interference of constraints that originated from empirical functions. Experiments in Nanjing in the periods of day-of-year (DOY) 182–184, 2019, and 244–246, 2019, show that the mean absolute error (MAE) and root mean square error (RMSE) of the WV density profile obtained using the proposed method are 0.9 and 1.3 g/m³, while those obtained using the conventional method are 1.3 and 1.8 g/m³, respectively, with respect to the radiosonde (RS) method. The numerical results show that the proposed method is reliable and has a superior accuracy to that of the conventional method.

The objective of the study was to put forth an interpolation method (the LZ method) for refining the GNSS-derived precipitable water vapor (PWV) map. We established a regional weighted mean temperature (T_m) model for this experiment, which introduced a minor difference into the resultant GNSS-derived PWV compared to the previous T_m models. The kernel of the LZ method consists of increasing the sample density via the virtual sample points. These virtual sample points originated from the digital elevation model (DEM) were constructed on the basis of the statistically significant correlation between PWV and geographical location (i.e., geographical coordinates and elevation). The LZ method was validated and compared to the conventional interpolation approach only accounting for the original sample points. The results reflect that the PWV maps generated by the LZ method showed more details than through conventional one. In addition, the prediction performance of the LZ method was better than that of the conventional method by using cross-validation.

Global navigation satellite system (GNSS) can provide dual-frequency observation data, which can be used to effectively calculate total electron content (TEC). Numerical studies have utilized GNSS-derived TEC to evaluate the accuracy of ionospheric empirical models, such as the International Reference Ionosphere model (IRI) and the NeQuick model. However, most studies have evaluated vertical TEC rather than slant TEC (STEC), which resulted in the introduction of projection error. Furthermore, since there are few GNSS observation stations available in the Antarctic region and most are concentrated in the Antarctic continent edge, it is difficult to evaluate modeling accuracy within the entire Antarctic range. Considering these problems, in this study, GNSS STEC was calculated using dual-frequency observation data from stations that almost covered the Antarctic continent. By comparison with GNSS STEC, the accuracy of IRI-2016 and NeQuick2 at different latitudes and different solar radiation was evaluated during 2016–2017. The numerical results showed the following. (1) Both IRI-2016 and NeQuick2 underestimated the STEC. Since IRI-2016 utilizes new models to represent the F2-peak height (hmF2) directly, the IRI-2016 STEC is closer to GNSS STEC than NeQuick2. This conclusion was also confirmed by the Constellation Observing System for Meteorology Ionosphere and Climate (COSMIC) occultation data. (2) The differences in STEC of the two models are both normally distributed, and the NeQuick2 STEC is systematically biased as solar radiation increases. (3) The root mean square error (RMSE) of the IRI-2016 STEC is smaller than that of the NeQuick2 model, and the RMSE of the two modeling STEC increases with solar radiation intensity. Since IRI-2016 relies on new hmF2 models, it is more stable than NeQuick2.

Global navigation satellite system (GNSS) tomography can effectively sense the three-dimensional structure of tropospheric water vapor (WV) using the GNSS observations. Numerous studies have utilized a tomographic window to include more epochs of observations, which significantly increases the number of valid signals. However, considering the tomography grid limits, a massive number of valid signals inevitably exhibits linear dependence. This dependence makes it impossible to improve the rank score of the tomography coefficient matrix by blindly introducing a large number of valid rays. Furthermore, excessive valid signals may lead to a high condition number in the coefficient matrix (ill-condition problem), which causes unstable results using the GNSS-WV tomography. Considering these problems, we proposed an improved tomographic approach, which applies a refined tomographic window. It differs from the general tomographic window in that the window is refined to traverse the valid signals available 15 min before and after the target epoch while retaining only the linearly independent parts (characteristic signal). Compared to the conventional method, the proposed method can filter the characteristic signal, which increases the rank score of the coefficient matrix and improves the stability of the tomography model. In this paper, we used GNSS observations from the Hong Kong Satellite Positioning Reference Station Network (SatRef) to validate the performance of the proposed method over the day-of-year (DOY) periods of 130–132, 2019 and 146–148, 2019. The numerical results showed that, by using a refined tomographic window, the proposed method obtained superior WV products in comparison with that of the conventional method.