Global Navigation Satellite Systems (GNSS) rely on ultra-precise timing to determine positions on Earth. However, as satellite signals travel through the atmosphere, they encounter two highly dynamic layers that distort their path and timing: the ionosphere (packed with free electrons) and the troposphere (filled with water vapor).
Traditionally, global positioning systems have relied on broad empirical models to estimate and subtract these atmospheric delays. For example, standard global ionospheric models simplify the entire planet's ionosphere into a handful of mathematical constants updated infrequently. While computationally cheap, these global averages smooth out local variations, leading to significant positioning errors during turbulent space weather or in regions with complex local climates.
The core insight behind recent advancements in regional GNSS correction is that atmospheric disturbances are highly correlated over localized geographic areas. By shifting from rigid global formulas to data-driven regional models, researchers can capture fine-grained spatial and temporal dynamics, dramatically improving both positioning accuracy and local meteorological forecasting.
The Intuition of Regional Spatial Correlation
Why do regional models consistently outperform sophisticated global models? The answer lies in spatial correlation. If two ground receivers are separated by a few hundred kilometers, the satellite signals traveling to them pierce the atmosphere through nearly identical pockets of electron density and water vapor.
Instead of treating each receiver as an isolated point trying to solve a global equation, regional modeling leverages dense networks of local ground stations (such as GNSS reference stations and radiosondes) to sample the local atmosphere simultaneously. By treating these stations as a distributed sensor array, researchers can interpolate or fit mathematical surfaces across the data points. This transforms the problem from predicting atmospheric behavior via generalized empirical equations to mapping it via localized data refinement.
Ionospheric Refinement: Mapping Total Electron Content
The delay a satellite signal experiences in the ionosphere is directly proportional to the Total Electron Content (TEC) along its path. Regional modeling captures the highly volatile fluctuations of TEC through several distinct strategies:
High-Resolution Spatial Grids
Instead of applying a single formula across an entire hemisphere, regional systems divide a geographic area into a high-resolution spatial grid. Yi Jiang et al. conducted a long-term evaluation of this approach for the BeiDou Navigation Satellite System in Long-term Performance Evaluation for Ionospheric Delay Correction of BeiDou Navigation Satellite System. They demonstrated that frequent grid updates capture localized variations that global models average out, boosting the ionospheric error correction rate from 73% to 85% and maintaining stability even during solar disturbances.
Refining Global Baselines
Constructing a regional model does not always require building from scratch. The authors of A Refinement Method of Real-Time Ionospheric Model for China (Yang Wang 0077 et al.) showed that real-time vertical TEC coefficients from global products can serve as a baseline. By incorporating real-time data from even a single local GNSS station, the global model can be dynamically adjusted and corrected, significantly improving regional positioning accuracy.
Advanced Surface Fitting and Sparse Networks
To capture complex regional variations, Changhao Yu et al. utilized Empirical Orthogonal Function (EOF) analysis on dense GNSS observations in Constructing a Regional Ionospheric TEC Model in China with Empirical Orthogonal Function and Dense GNSS Observation, achieving a remarkably low root-mean-square error of 2.45 TECU.
Even when ground networks are sparse, regional modeling remains highly effective. In A Consistent Regional Vertical Ionospheric Model and Application in PPP-RTK Under Sparse Networks, Sijie Lyu et al. developed a Regional Vertical Ionospheric Model (RVIM) that isolates and separates receiver biases. This consistency allows sparse networks to provide rapid convergence and high-precision positioning for end users.
Global Model: [ Broad Empirical Equation ] ---> Smooths out local peaks
|
Regional Model: [ Local Station ] -> [ Grid/Surface Interpolation ] -> Captures local peaks
Bypassing Mapping Functions
Traditional models calculate the ionospheric delay along a vertical path and use a mathematical "mapping function" to project it to the actual slant path of the satellite signal. This projection introduces geometric errors. Wen Jung Li et al. bypassed this limitation in A satellite-based method for modeling ionospheric slant TEC from GNSS observations: algorithm and validation by modeling delays directly along the slant path. Because ionospheric delays for individual satellites are highly correlated over regional distances, treating each satellite as an independent modeling target allows the system to correct over 90% of ionospheric errors.
Tropospheric Calibration: Localizing Water Vapor Retrieval
While the ionosphere is driven by solar radiation, tropospheric delay is governed by local weather conditions—specifically water vapor. By measuring this delay, GNSS can be used in reverse as a meteorological tool to calculate Precipitable Water Vapor (PWV), a critical metric for short-term rainfall and flood forecasting.
To convert the Zenith Wet Delay (ZWD) of a satellite signal into PWV, scientists rely on the atmospheric weighted mean temperature ($T_m$). Global empirical models estimate $T_m$ poorly because they cannot account for local topography, coastal effects, or complex regional microclimates.
Site-Specific Temperature Modeling
To overcome the limitations of global averages, researchers use historical regional data from radiosondes (weather balloons) to establish precise linear relationships between surface temperature ($T_s$) and $T_m$.
In tropical environments, where high humidity complicates signal propagation, Jean-Pierre Barriot et al. demonstrated in Metrology Assessment of the Accuracy of Precipitable Water Vapor Estimates from GPS Data Acquisition in Tropical Areas: The Tahiti Case that site-specific $T_m$ models accounting for seasonal variations significantly outperform global alternatives.
Similarly, Liangke Huang et al. deployed a localized model across an 11-station network in Spatiotemporal characteristics of GNSS-derived precipitable water vapor during heavy rainfall events in Guilin, China. They proved that local tuning is essential for high-precision meteorological monitoring in areas with complex terrain and high humidity, allowing them to use rapid spikes in PWV as reliable precursors to heavy rainfall.
Multi-Parameter and Dense Mapping
To further increase precision, regional models can incorporate additional meteorological dimensions. In Short-Term Rainfall Forecasting by Combining BP-NN Algorithm and GNSS Technique for Landslide-Prone Areas, Zufeng Li et al. developed a regional $T_m$ model that factors in pressure, temperature, relative humidity, and periodic seasonal terms, outperforming simpler empirical approaches in landslide-prone areas.
When constructing continuous maps from discrete station data, the density of the data points matters. Nanshan Zheng et al. addressed this in A New Method for Refining the GNSS-Derived Precipitable Water Vapor Map by introducing a method that leverages Digital Elevation Models (DEM) to inject virtual samples into the interpolation process, significantly improving the spatial resolution of regional PWV maps.
Real-Time Ground Observations
The ultimate expression of localized modeling is the integration of real-time, ground-based meteorological sensors directly into the positioning loop. Qinglan Zhang et al. investigated this in Investigation of the Application of Measured Meteorological Observations in Real-Time Precise Point Positioning, demonstrating that direct, real-time meteorological measurements from dense ground networks provide far more accurate Zenith Tropospheric Delay (ZTD) estimates than any predictive atmospheric model. This real-time localized data drastically accelerates the convergence speed and accuracy of Precise Point Positioning (PPP).
Summary
The transition from global empirical modeling to localized regional modeling represents a paradigm shift in how we handle atmospheric interference in satellite navigation. By exploiting the high spatial correlation of atmospheric properties over regional zones, and feeding dense, real-time local data into grids, surfaces, and site-specific regressions, researchers have unlocked a dual benefit: centimeter-level positioning accuracy for navigation and high-fidelity atmospheric mapping for meteorology.
Key Foundational Work
Metrology Assessment of the Accuracy of Precipitable Water Vapor Estimates from GPS Data Acquisition in Tropical Areas: The Tahiti Case (2018) — contributed 2 concept elements to this direction.
A New Method for Refining the GNSS-Derived Precipitable Water Vapor Map (2019) — contributed 2 concept elements to this direction.
Long-term Performance Evaluation for Ionospheric Delay Correction of BeiDou Navigation Satellite System (2019) — contributed 2 concept elements to this direction.