GNSS coordinate time series noise model selection considering suboptimal noise model

In the field of developing noise models of Global Navigation Satellite System (GNSS) coordinate time series, selecting a suitable noise model has important value for extracting velocity signals from GNSS coordinate time series. For a long time, researchers have focused on how to establish the optimal noise model to describe the noise components in the GNSS coordinate time series, but they often directly ignore the suboptimal noise model that is close to the performance of the optimal noise model. This paper selects eight common noise models and uses Maximum Likelihood Estimation (MLE) for estimating noise model parameters. The Akaike Information Criterion (AIC) is used as an evaluation criterion to select noise models. We select GNSS time series from 36 coordinate components of 12 International GNSS Service (IGS) reference sites in China and 150 coordinate components of 50 IGS reference sites worldwide. The results show that more than half of the site coordinate components have suboptimal noise model whose model performance is close to the optimal noise model. On this basis, for IGS sites in China, 14% of suboptimal noise models have larger velocity uncertainty values than that of optimal noise models in the east direction; in the north and up directions, these values are 33% and 63%, respectively. For global IGS sites, these values are 31%, 39%, and 48%, respectively. This indicates that when selecting GNSS coordinate time series noise models, it is advantageous to obtain more conservative estimation results by considering the suboptimal noise models whose model performance model are similar to that of the optimal noise models.

Automated summary

Selecting a suitable noise model is important for extracting velocity signals from GNSS coordinate time series. This study focuses on both optimal and suboptimal noise models using Maximum Likelihood Estimation (MLE) and the Akaike Information Criterion (AIC) for evaluation. Results show that more than half of the coordinate components have suboptimal noise models that perform similarly to the optimal noise model, emphasizing the importance of considering suboptimal noise models in estimating velocity uncertainty.