Abstract
Atmospheric water vapor is a primary driver of weather and climate, yet its high spatiotemporal variability makes it difficult to model accurately. While GNSS tomography offers 4D insights, it typically requires dense receiver networks. This paper introduces a novel mathematical framework to model water vapor using data from a single GPS receiver. By expanding Slant Wet Delays (SWDs) into a series of trigonometric functions for time and spherical harmonics for space, we track the 4D evolution of water vapor. Applied to the THTI station on Tahiti Island, the model demonstrates high fidelity, capturing seasonal fluctuations and correlations with local wind regimes.
Introduction
Traditional GPS meteorology often simplifies atmospheric water vapor into a 1D vertical column known as Zenith Wet Delay (ZWD). While useful, this approach ignores the "slant" paths of signals that pass through different atmospheric regions. To achieve a 3D or 4D perspective, researchers often employ GNSS tomography, which necessitates expensive, dense sensor grids. For isolated locations like Tahiti, such networks are often impractical. This research bridges the gap by extracting maximum information from a single point of observation, modeling the sky as a dynamic hemisphere of data.
Methodology
The proposed method involves four primary steps. First, Slant Wet Delays (SWDs) are reconstructed from standard GPS products, including Zenith Total Delay and horizontal gradients. Second, a stationarity transformation is applied: SWDs are projected to the zenith ($SE = SWD \cdot \sin(\epsilon)$) to remove geometric bias from satellite elevation. Third, this $SE$ value is modeled using a series expansion where time is represented by trigonometric functions and space by spherical harmonics. Finally, a least-squares fitting process determines the coefficients, effectively mapping the 4D water vapor field above the receiver.
Study Area and Data
The study focuses on the THTI GPS station on Tahiti Island. Tahiti's high humidity and distinct wet and dry seasons make it an ideal laboratory for studying tropical water vapor turbulence.
Data was collected over two four-month periods: the wet season (Nov 2017 – Mar 2018) and the dry season (May 2018 – Sep 2018). Validation was performed using International GNSS Service (IGS) products and independent meteorological data from radiosonde soundings and ground weather stations.
Key Results and Validation
The model achieved high-fidelity fits for the $SE$ signals. Analysis of the post-fit residuals revealed a Gaussian distribution, indicating that the model successfully captured the primary physical signals.
Key findings include a "4-Day/20-Degree Rule," where water vapor signals correlate over approximately four days and 20° of angular distance. Furthermore, a measurable correlation was found between water vapor delays and wind velocity at the 1000m "trade wind layer," particularly during the wet season (0.2216 correlation) compared to the dry season (0.1013).
Discussion and Future Work
The single-station approach proves that the structure of the atmosphere can be inferred without dense networks. However, current spatial resolution is limited by the number of visible satellites to approximately 30°. The presence of a diurnal component in the residuals suggests some high-frequency aliasing. Future integration of multi-constellation GNSS (Galileo, BeiDou) is expected to triple the available slant paths, potentially allowing for spatial resolutions of 10° or less, rivaling multi-station tomography.
Conclusion
This series expansion framework provides a powerful tool for monitoring atmospheric water vapor in regions where dense sensor networks are unavailable. By treating the atmosphere as a fluctuating field rather than isolated rays, it offers a scalable solution for high-resolution meteorological studies.
Summary updated