Application of an annual phase-augmented clustering approach to annual detection of vertical GPS station deformation

We present an improved clustering approach for grouping regional GPS stations with similar annual signals into four similarity levels. A newly introduced criterion, the annual phase difference among stations, reinforces the clustering process and resolves the discrepancy among the annual signals of stations classified into several clusters. Comparisons show that the improved method enhances the accuracy and representativeness of the land surface deformation features derived from the mean annual signals of clustered stations. With respect to CMONOC stations throughout mainland China, the observed GPS coordinate time series are probably more reliable for acquiring real land surface deformations when employing the 25 regional clusters derived from the improved approach in comparison with some modeled surface loading models (SLMs), which potentially contain mistakes. Based on the results of the improved technique, descriptions of the annual land surface deformation throughout China are given to illustrate the characteristics of the annual signals of GPS station motions in different regions. For the annual signals of the 25 clusters discussed, those of 10 clusters show obvious sinusoidal shapes, including 7 clusters in Southwest China. The peaks and troughs of these annual motion signals happen during different times of a year for northern and southern station signals. In applications to land surface deformation, the improved method can play a preliminary but significant role in searching for the most stable stations before comparing the results of GPS with other geodetic techniques, such as the Gravity Recovery and Climate Experiment and SLMs.