Dynamic stochastic model for estimating GNSS tropospheric delays from air-borne platforms.

Air-borne GNSS can provide vertical tropospheric information on different heights compared to the traditional ground permanent stations. However, the decorrelation of heights and tropospheric parameters for GNSS high-kinematic observation remains a challenge. The general stochastic process noise is set to constant for a certain time duration, inducing obvious errors for air-borne GNSS data processing. We proposed a new GNSS zenith total delay (ZTD) estimation method suitable for high-kinematic air-borne observation, in which process noise considers both the temporal and spatial atmosphere state variation assisted by numerical weather models. An unmanned aerial vehicle (UAV) experiment was designed using both traditional models and the proposed dynamic model. For assessments of ZTD, ERA5-ZTD was used as the reference to evaluate external accuracy, and a closed-loop test was proposed to evaluate inner accuracy. The results show that the proposed method outperforms the traditional estimation method, significantly improving the convergence and the accuracy of ZTD results. It achieved an inner accuracy of 4 mm, with an improvement of at least 55.5% over the traditional method. The proposed method can satisfy the demand for real-time data processing and potentially benefit high-precision meteorological applications. [ABSTRACT FROM AUTHOR]