Detecting and Removing Repetitive Errors from PPP Time Series by Means of Adaptive Filter

Abstract

PPP (Precise Point Positioning) is a relatively new GNSS (Global Navigation Satellite System) method that enables position determinations using a single receiver. Being able to survey with a single receiver makes this approach more attractive and affordable than other GNSS positioning techniques. Since a single receiver is used, a drawback of this technique is propagation of errors such as orbit and clock uncertainties, atmospheric disturbances etc. Due to these errors, precision of point positioning may reach up to decimeter levels. Fortunately, some of these error sources such as orbital errors, multipath etc. are repetitive errors and therefore can be detected by using digital filters. Adaptive noise canceling filters are used to detect the correlated part of the time series. These filters optimize themselves using the Least Mean Square (LMS) algorithm which is a powerful tool for Geosciences and are also used for many engineering applications. In this study, it is aimed to detect repetitive errors on coordinate time series by using adaptive noise canceling filter. For the application, four stations have been selected from NGS CORS (National Geodetic Survey Continuously Operating Reference Station) network. Two of these stations are highly affected by multipath error and the other two are not. The coordinate time series of the GNSS sites have been obtained using PPP technique on kinematic mode (epoch by epoch) for several consecutive days. The cross-correlation analysis has been performed and up to 92% correlation has been detected between the daily time series. The correlated part of the data set has been captured by means of adaptive noise canceling filter and removed from the data set. After the filtration process, up to 50% precision improvement has been achieved on coordinate time series, especially for the stations affected by multipath.