Filtering GNSS time series using singular spectrum analysis

Modern space geodetic systems, such as Global Navigation Satellite Systems (GNSS), are based on radio navigation technology that provides positioning and time transmission anywhere in the world. GNSS provides a unique opportunity to monitor and model the Earth's surface and environmental dynamics with high resolution and precision. Continuously Operating Reference Stations (CORS) are the receivers that perform continuous GNSS measurements at a specific point. These receivers, installed in global and regional GNSS networks, allow continuous and regular acquisition of GNSS coordinate time series and detection and monitoring of changes related to location. Advanced mathematical and statistical techniques are used during time series analysis to obtain the desired signals and information from these time series. The installation of the stations in the CORS-TR network in Türkiye began in 2006 and was completed in 2009. Most of the stations have approximately 15 years of time series. When the time series are visually examined, many complex structures such as earthquake effects, non-linear trends, periodic movements with variable amplitudes over time, and transient signals are observed. In this study, the time series of stations in CORS-TR network were analyzed. Robust Estimation (RE) and Singular Spectrum Analysis (SSA) methods were used to filter time series. These advanced methods are used to increase the accuracy and reliability of the information obtained from the analysis of GNSS time series. The aim of the study is to understand and detect seasonal effects and ground movements with the help of the information to be obtained from the time series analysis of the stations. Within the scope of the thesis, the time series of CORS-TR stations located in the Marmara region were analyzed with two different methods and the results were compared. As a result of these comparisons, the trend is obtained as a suitable linear line for the time series in RE method, while the trend is obtained as a non-linear curve in SSA method. The use of linear trend may be preferred in earth sciences and especially tectonic studies. It is seen that the SSA method is more successful than the RK method, especially in the extraction of seasonal periodic oscillations. The main reason for this is that the RE method uses a fixed amplitude and frequency value for each seasonal period, while the SSA method does not have such a restriction. Finally, in cases where there are transient signals in the time series, the SSA method gives more successful results. In this thesis study, RE and SSA methods were applied to CORS-TR station time series. The results of these methods were compared and findings were obtained and discussions were performed based on their performances in different situations. It can be stated that the results of the study can assist the studies in the fields of geodetic, geological and geophysical.