Determination of Kalman filter TRF solutions based on white noise station coordinate behavior

Kalman filtering has become an established technique for the creation of terrestrial reference frame (TRF) solutions, and has been successfully used in the determination of JTRF2014, the most recent ITRS realization by Jet Propulsion Laboratory. So far, the station coordinates have typically been modeled as random walk processes. Advantages of random walks are, for example, that they are computationally efficient and easy to implement in a Kalman filter environment. Furthermore, they guarantee a certain short-term stability of the station coordinate time series, since the coordinates of one epoch directly depend on those from the previous epoch. However, time series of station coordinates from different space-geodetic techniques as well as non-tidal loading displacements, which are often associated with the non-linear part of coordinate variations, indicate a different stochastic behavior. Most often, the appropriate stochastic model would be white noise. For this reason, we have implemented the option to use white noise as the process noise type for station coordinates, realized by including an additional parameter per coordinate component in the state vector of the Kalman filter (in addition to offset, velocity, and seasonal signals). Based on data from very long baseline interferometry (VLBI), we will compare the performance of white noise Kalman filter solutions with conventional solutions based on random walk processes. In particular, we will investigate their agreement with existing VLBI coordinate time series, as well as their capability of accurately predicting future station coordinates.