Application of independent component analysis to GPS position time series in Yunnan Province, southwest of China

Abstract

Pervasive seasonal signals have been found in GPS site position time series. A critical aspect in the seasonal geodetic analysis is the identification and extraction of different sources of deformations in space and time domains. In this work, we applied the independent component analysis (ICA) to discriminate four different seasonal periods in GPS time series: spatial uniformly distributed annual signals, local annual variations with spatial heterogeneity, spurious signals of draconitic terms, and semiannual terms. The annual displacements could be well described by surface mass loadings. As we successfully identified the draconitic terms, we removed these terms from the GPS time series and re-estimated the annual signals from the GPS; the annual amplitude reduced by ∼1 mm on average. Our results show that the GPS draconitic terms bias the annual signal estimation in conventional sinusoidal model fitting and that the ICA can be successfully used to extract different seasonal periods. We recommend removing the draconitic terms before estimating the annual signals, which are intended for high-accuracy applications.