BIFROST: Noise properties of GPS time series

Abstract

The BIFROST project uses GPS to observe the intra-continental deformation of Fennoscandia caused predominantly by Glacial Isostatic Adjustment (GIA). The noise in GPS position time series has been proven correlated, so we investigate a fractal model in order to obtain a parameter that can gauge our network stations true velocity uncertainties and utilize an empirical orthogonal function (EOF) to remove the inherent common mode. We employ a Kaiser window to reduce the power spectrum variance and retain independent power estimates based on the window's main lobe width ratio compared to that of a boxcar. As power spectra lack Gaussian distribution properties, we devise a transform that normalizes the power spectrum and subsequently iterate a fractional power law noise model. We find that the spectral indices for the different velocity components in our network are 0.6 for North, 0.5 for East, and 0.7 for the Vertical. As there is no white noise floor in our power spectra to indicate inevitable system noise it is possible that GPS time series should be sampled more frequently then once per day in order to separate between different uncertainty sources. © Springer-Verlag Berlin Heidelberg 2007.