Development and Validation of an Enhanced GPS Tomography Algorithm for Reunion Island

It is well known that GPS signals are affected by the amount of water vapor contained in the troposphere. This phenomenon creates delays, which can be converted into a corresponding integrated water vapor content along the receiver–satellite path (Slant Integrated Water Vapor, SIWV). Moreover, when a dense network of GPS stations is available, we obtain an ensemble of such SIWV paths that crisscross over the network area. Hence, by defining a three-dimensional regular grid composed of different boxes, called voxels, over our area of interest, and using a tomographic inversion method, we can retrieve the water vapor density in each voxel of the grid. Thus, this allows us to obtain a 3-D field of water vapor density above our area of interest.

Here, we implement this approach on Reunion Island (a South West Indian Ocean Volcanic tropical island about 2500km²), which counts approximately 40 GPS stations. We had take into account for some local specificities: 1°/ the orography of this volcanic island is extremely sharp with high altitude gradients between neighboring stations, and 2°/ the spatial distribution of the GPS stations is very heterogeneous with a high density (about half of the stations) distributed around the active volcano of Piton de la Fournaise. Therefore, two developments were carried out. First, regarding the tomographic geometry, we use Voronoï diagram to implement a grid adapted to the spatial distribution of the GPS stations. Second, the tomographic inversion method itself was improved using the more robust truncated singular value decomposition (TSVD) approach using the L-curve technique to define the analysis threshold (Moeller, 2017).

To validate these developments, the results obtained from the tomographic inversion was compared to 30 water vapor profiles obtained during a radio sounding campaign conducted in Saint-Philippe (SE of the island, close to the Piton de la Fournaise) between May 2025 and July 2025.