Empirical Modeling of Tropospheric Delays and Uncertainty

Abstract. Accurate modeling of troposphere delay is important for high-precision data analysis of space geodetic techniques, such as the Global Navigation Satellite System (GNSS). The empirical troposphere delay models provide zenith delays with an accuracy of 3 to 4 cm globally and do not rely on external meteorological input. They are thus important for providing a priori delays and serving as constraint information to improve the convergence of real-time GNSS positioning, and in the latter case, the proper weighting is critical. Currently, the empirical troposphere delay models only provide the delay value, but not the uncertainty of the delay. For the first time, we present a global empirical troposphere delay model, which provides both the zenith delay and the corresponding uncertainty, based on 10 years of tropospheric delays from the Numerical Weather Model (NWM). The model is based on a global grid, and at each grid point a set of parameters that describes the delay and uncertainty by the constant, annual, and semi-annual terms. The empirically modeled zenith delay has an agreement of 36 and 38 mm compared to three years delay values from NWM and four years estimates from GNSS stations, which is comparable to the previous models such as GPT3. The modeled ZTD uncertainty shows a correlation of 96 % with the accuracy of the empirical ZTD model over 380 GNSS stations over the four years. For GNSS stations where the uncertainty annual amplitude is larger than 20 mm, the temporal correlation between the uncertainty and smoothed accuracy reaches 85 %. Using GPS pseudo-kinematic PPP solutions of ~200 globally distributed stations over four months in 2020, we demonstrate that using the proper constraints can improve the convergence speed. The uncertainty modeling is based on a similar dataset as the GPT series, and thus it is also applicable for these empirical models.