Global Navigation Satellite Systems (GNSS) is a space geodetic technique capable of precise global positioning. By analysing time series of GNSS coordinates, information regarding the movement and deformation of the Earth's surface can be derived, which is of critical importance in many applications. To obtain positioning results with highest accuracy, it is often required to correct known displacements using dedicated models, typically physics-based. However, in particular, corrections for non-tidal loading (NTL) effects are often not (uniformly) considered. If these loading effects are not compensated, the obtained results will be distorted up to the centimeter level. This study investigates if machine learning (ML) in combination with environmental variables can replace or augment the existing physics-based models by providing data-driven corrections for GNSS displacements. Therefore, vertical displacements of 3553 GNSS stations in Europe are utilized to train and validate XGBoost models. Three different strategies were tested, differing in the preprocessing of the GNSS data. A significant improvement was achieved for all strategies ranging from 1.3% to 19.6%. The improvement is derived based on the root mean squared error (RMSE) reduction of the GNSS residual coordinates w.r.t. a trajectory model, accounting for a linear trend, seasonal signals, and discontinuities in the GNSS coordinate time series. In addition to evaluating the ML models, a thorough feature importance analysis based on SHapley Additive exPlanations (SHAP) is carried out to analyse the driving factors of the model output and to gain insights into what signals could still be found to enhance existing physics-based models.
Crocetti et al. (Fri,) studied this question.