Discontinuity Detection in GNSS Station Coordinate Time Series Using Machine Learning

Global navigation satellite systems (GNSS) provide globally distributed station coordinate time series that can be used for a variety of applications such as the definition terrestrial reference frame. A reliable estimation trends gives valuable information about movements during measured period. Detecting discontinuities various origins in is crucial accurate and robust velocity estimation. At present, there no fully automated standard method detecting discontinuities. Instead, discontinuity-catalogues are frequently used, which when device was changed or an earthquake occurred. However, it known these catalogues suffer from incompleteness. This study investigates suitability machine learning classification algorithms data-driven to detect caused by earthquakes without need external information. For this study, Japan selected testing area. Ten different have been tested. It found Random Forest achieves best performance with F1 score 0.77, recall 0.78, precision 0.76. Overall, 525 565 recorded test data were correctly classified. further highlighted splitting into chunks 21 days leads performance. Furthermore, beneficial combine three (normalized) components GNSS solution one sample, adding value range additional feature improves result. Thus, work demonstrates how possible use series.