Spatiotemporal-varying tropospheric stochastic model for GNSS meteorology enhancement

Abstract Accurately estimating the zenith wet delay (ZWD) is critical in Global Navigation Satellite System (GNSS) meteorology. ZWD is typically modeled as a random walk process with spatiotemporal-invariant process noise in the Kalman filter using precise point positioning. However, this approach is not rigorous due to the notable geographical and seasonal variations in water vapor content. In response, we propose a spatiotemporal-varying ZWD stochastic model to improve the ZWD estimation accuracy, thereby enhancing the GNSS meteorology. First, 15 years of ZWD data are employed to derive the process noise. Next, the model is established using a trigonometric-based fitting function. Finally, hierarchical clustering is implemented for computational efficiency improvement. Experiments at 26 test stations indicate that the maximum ZWD accuracy is improved by 34.95% under dynamic mode and 22.67% under static mode. Moreover, the maximum data availability of GNSS meteorology is improved by 10.56% and 4.55% under dynamic and static modes, respectively.