Offset detection in simulated time-series using multivariate analysis

Different application of GPS position time series in geodetic and geophysical studies such as plate tectonics, glacial isostatic rebound, crustal deformation and earthquake dynamics requires proper assessment of the time series. A functional model for GPS time series consists of a linear trend, periodic signals with annual and semiannual periods and probabilistic offset. The undeterministic effects can best be described as a noise. The correct detection of offset requires a proper estimation of noise and the covariance matrix of the data. Towards this end, the least-squares variance component estimation is used. It is shown that how a correct analysis of the noise components can affect the offset detection method. Ignoring colored noise of GPS time series degrade the offset detection power. We first use the univariat time series analysis. To increase the offset detection power, the multivariate analysis is then recommended. In univariat analysis, only one of the coordinate components is used, while in the multivariate analysis all three components, which is assumed to have the same structure of noise and offsets, are simultaneously used. In this paper, some simulated daily time series of positions during 8 years are used. Finally, offsets having different size and positions are located through them and detected successfully, using the multivariate analysis.