The current GNSS meteorology literature focuses on ground-based and space-based GNSS observations separately, without exploring potential synergies. In this study, we propose combining the two data sources using GNSS tomography to overcome current limitations in (1) horizontal resolution of GNSS space-based, (2) low vertical resolution of GNSS ground-based tropospheric retrievals when the number of GNSS ground-based observations is limited and (3) instability of the tomography system due to a lack of observations traversing the atmosphere horizontally. Our study on the combination of GNSS ground-based and space-based presents an innovative way for data integration based on uncertainty estimation. The developed integrated tomography operator, based on 3D ray tracing principles, is tested on 30 days of simulated data with 101 ground stations and over 240 radio occultation events, using three different station layouts. The a priori data introduced into the tomography processing is from a deterministic model, while ray tracing uses the ERA5 reanalysis wet refractivity field to obtain input data for individual test cases. The results are verified by comparing tomography output to ERA5 reanalysis. We observed a decrease in tomography RMSE between 2% and 16% in the case of an integrated solution, depending on GNSS station layout and the number and geometry of radio occultation ray paths. We show that a single RO event during one processing epoch can shift the wet refractivity estimates by 2 to 5 ppm closer to the correct solution compared to ground-based-only GNSS tomography.

目前的 GNSS 气象学文献分别关注地基和天基 GNSS 观测，没有探索潜在的协同作用。在本研究中，我们建议使用 GNSS 层析成像将两个数据源结合起来，以克服当前在 （1） GNSS 天基水平分辨率，（2） 当 GNSS 地面观测数量有限时 GNSS 地面对流层检索的低垂直分辨率，以及 （3） 由于缺乏水平穿越大气层的观测结果而导致的层析成像系统的不稳定性。我们对 GNSS 地基和天基结合的研究提出了一种基于不确定性估计的数据集成创新方法。开发的集成层析成像运算符基于 3D 光线追踪原理，使用三种不同的站点布局，在 101 个地面站和 240 多个射电掩星事件的 30 天模拟数据上进行了测试。引入层析成像处理的先验数据来自确定性模型，而光线追踪使用 ERA5 再分析湿折射场来获取单个测试用例的输入数据。通过将断层扫描输出与 ERA5 再分析进行比较来验证结果。我们观察到，在集成解决方案的情况下，层析成像 RMSE 降低了 2% 到 16%，具体取决于 GNSS 站布局以及射电掩星光线路径的数量和几何形状。我们表明，与仅基于地面的 GNSS 断层扫描相比，一个处理时期的单个 RO 事件可以将湿折射率估计值移动 2 到 5 ppm，更接近正确的解决方案。