Modeling of regional GNSS network using adaptive boosting algorithm: a case study in the Xinjiang Uyghur Autonomous Region

Analyzing the position time series from the regional global navigation satellite system (GNSS) network can contribute to rationally explaining geophysical phenomena at different spatiotemporal scales. This paper examines the use of machine learning model to reflect the interactions between reference stations from a regional GNSS network. The input and output data to the machine learning algorithm consist of GNSS reference station data. In the forecasting experiments, the Adaptive Boosting (AdaBoost) algorithm is introduced and evaluated with the time series datasets collected from 20 GNSS reference stations over the period of 10 years. The results indicate that the forecasting accuracy of the AdaBoost has improved by 59% compared to the Prophet model using the time variable and show the feature contribution of each GNSS reference station to model the regional GNSS network. Through the modeling of regional GNSS network, interpolation of the GNSS time series can be completed. The interpolation results indicate that the AdaBoost model can capture the trend information and shows better capability than the cubic spline interpolation method. Overall, the AdaBoost model demonstrates superior performance in modeling of GNSS vertical time series, thus impacting the analysis and maintenance of the regional GNSS network.