Impact of Vertical Grid Design on GNSS Tomographic Reconstruction of Tropospheric Wet Refractivity

Atmospheric water vapour drives weather processes and climate variability, yet its strong spatiotemporal heterogeneity makes accurate three-dimensional (3D) monitoring challenging. GNSS atmospheric tomography enables reconstruction of 3D wet refractivity fields from slant tropospheric delays, however reconstruction accuracy is highly sensitive to the design of the tomographic voxel grid, particularly in the vertical dimension, which has received comparatively limited attention.  

We develop a GNSS tomography framework to investigate the impact of vertical grid design on wet refractivity reconstruction accuracy. Horizontal discretization (latitude–longitude) is kept fixed, while multiple vertical grid configurations are tested, including a reference vertical grid adopted from previous work [1], homogeneous layer thicknesses (100, 500, and 1000 m). Furthermore, two adaptive, station-specific vertical grid layouts are derived from radiosonde profiles. The adaptive approach tailors the vertical resolution of the voxel grid to the local moisture gradients obtained from the latest radiosonde observations. This model adapts the vertical resolution of the grids to the closest radiosonde observation both spatially as well as temporarily.  

The methodology is applied over the Carpathian Basin using dense GNSS observations, precise satellite orbits (SP3), VMF1 tropospheric mapping functions, and radiosonde soundings over a period of 10 days with twice-daily epochs. Three-dimensional wet refractivity fields are reconstructed using the Multiplicative Algebraic Reconstruction Technique (MART), with radiosonde profiles used as a priori information and independent profiles for validation. 

The results demonstrate a clear dependence of performance on altitude based on RMS zenith wet delay (ZWD) errors. In the lower troposphere (0–4 km), adaptive vertical grids yield markedly improved reconstruction accuracy, with RMS values of 0.009–0.034 m, whereas the reference and coarse homogeneous grids exhibit substantially larger RMS errors. In the mid-troposphere (4–8 km), errors decrease to the order of 10⁻³ m, with comparable performance between adaptive grids and fine homogeneous discretizations. In the upper troposphere (>8 km), all grid configurations perform similarly, with RMS values generally below 2×10⁻³ m, indicating that adaptive discretization is not necessary in moisture-poor layers. These findings highlight the critical role of adaptive vertical grid design for accurate GNSS wet refractivity tomography in the lower troposphere. 

 

[1] Rózsa, S., Turák, B., and Khaldi, A.: Near Realtime tomographic reconstruction of atmospheric water vapour using multi-GNSS observations in Central Europe, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-4465, https://doi.org/10.5194/egusphere-egu23-4465, 2023.