Toward a realistic spatio-temporal description of GNSS station position time series

Today, Global Navigation Satellite Systems (GNSS) are widely used to study the kinematics of the Earth's surface and provide a fundamental contribution to the establishment of terrestrial reference frames. GNSS station position time series are derived by different analysis centres, which have achieved tremendous progress over the years. Nevertheless, GNSS station position time series remain polluted by stochastic variations – often called "noise" – which hinder their geophysical interpretation and contribution to terrestrial reference frames.

Many studies have evidenced that these stochastic variations exhibit both temporal and spatial correlations. On the one hand, temporal correlations have been analyzed in detail and are known to be well approximated by a combination of white noise and flicker noise. The presence of flicker noise, whose origin remains unknown, is a major source of uncertainty in the estimation of long-term station positions and velocities. Consequently, most geodetic studies now routinely take these temporal correlations into account to avoid inferring unrealistically optimistic uncertainties. On the other hand, the stochastic variations in GNSS station position time series are also spatially correlated up to distances reaching a few thousand kilometres. The origins of these large-scale correlations, whether non or poorly-modelled ground deformation, positioning errors, or a combination of both, remain to be understood. Although this spatially correlated noise is known to obscure geophysical signals and affect the estimation of parameters of interest, it is hardly ever modelled or accounted for.

This presentation introduces a realistic spatio-temporal correlation model for the stochastic variations in GNSS station position time series. Based on the analysis of position time series from the IGS repro3 campaign (over 1,300 stations) and the Nevada Geodetic Laboratory analyzes (over 11,000 stations), we first provide a diagnosis of the spatial correlations of the white and flicker noise processes separately. We then introduce three-dimensional spatial correlation models for each process. Finally, we discuss the implications of these spatio-temporal correlations on the uncertainty of long-term station positions and velocities.