Application of an improved clustering approach on GPS height time series at CMONOC stations in Southwestern China

Abstract

Abstract The hydrological, geological and meteorological conditions in southwestern China are relatively complex, so that the land surface deformation presents various features. Using 58 Crustal Movement Observation Network of China (CMONOC) stations across four provinces in Southwestern China, we adopt an improved clustering algorithm to classify 49 stations into 12 clusters with different similarity levels. Our results show that the average annual signals of GPS stations within each cluster have strong consistency, while obvious differences exist among the 12 clusters, indicating that clustering algorithm helps to describe surface deformation features more accurately in regions with complex conditions. We then combine other earth observation techniques, such as the Gravity Recovery and Climate Experiment (GRACE) satellite datasets and surface loading models (SLM), and observe that GPS, GRACE and SLM have strong correlation in their monthly displacement series at GPS stations. After excluding non-clustered stations according to our previous clustering results, the correlation coefficients of GPS/GRACE and GPS/SLM are enhanced. Also, the RMS reduction rates of GPS coordinate time series have been improved after deducting displacements obtained from GRACE and SLM, thus the clustering algorithm proves to be effective in improving the consistency of three techniques in joint detection of surface deformation. Moreover, the vertical displacements of four riverside GPS stations in the Three Gorges Reservoir (TGR) area show significant negative correlation with water level of TGR, hence we conclude that the Three Gorges Dam (TGD) may directly affect the consistency of GPS annual signals of its upstream and downstream GPS stations. Graphical Abstract

Keywords

• Clustering algorithm
• CMONOC stations
• GPS height time series
• Land surface deformation
• GRACE
• SLM