Effects of Surface Mass Loading on the Stochastic Properties of GPS Time Series in the Great Lakes Region
Loading...
Authors
Krcmaric, Jordan A.
Issue Date
2024
Type
Thesis
Language
Keywords
Geodesy , GPS Time Series Analysis , Hydrologic Loading , Surface Mass Loading
Alternative Title
Abstract
Daily Global Positioning System (GPS) time series provide a critical dataset for studying the Earth. In the past decade, an explosion in the number of continuously operating GPS stations combined with improved processing strategies have enabled the detection of ground motions with rates of less than 1 mm/yr. However, detecting such subtle rates in GPS time series is complicated by the presence of errors and other effects that introduce stochastic variability and increase rate uncertainty. Some of this stochastic variability is due to surface mass loading, which causes displacements in GPS time series from atmospheric pressure fluctuations, oceanic mass redistribution, and changes in hydrology. This thesis investigates the effect of surface mass loading on 2,481 GPS station time series around the Great Lakes region of the U.S. and Canada. This region is ideal for studying the effects of surface mass loading because it is covered by a dense network of GPS stations that capture ground displacements due to fluctuations in water levels within the Great Lakes. Uncorrected GPS time series are compared to time series that have been corrected with models of surface mass loading using 4 metrics: variance reduction in corrected residual time series; white, flicker, and random walk noise amplitudes; station velocity uncertainties; and average power spectra of residual time series. Common mode component (CMC) filtering is a method used to remove spatially correlated signals from GPS time series. I also compare CMC filtered time series to see if filtering methods are effective at removing surface mass loading. Finally, to understand the role that GPS station monument type plays in the stochastic variability of the time series, I compare noise amplitudes and velocity uncertainties of 5 categories of monuments: deep-drilled braced monuments, roof mounted, concrete pillars, steel towers, and any monument directly anchored in bedrock. Results show that non-tidal atmospheric and ocean loading (NTAOL) is responsible for the most significant proportion of variance in GPS time series, and correcting for NTAOL reduces the median flicker noise amplitudes by ~50%. However, median white and random walk noise amplitudes increase in NTAOL corrected time series due to a masking effect by NTAOL in the uncorrected time series. Correcting for hydrological loading reduces both variance and random walk noise amplitudes most significantly in GPS stations closest to the Great Lakes. Correcting for both NTAOL and hydrological loading decreased the median velocity uncertainty from 0.35 mm/yr to 0.18 mm/yr. There are no significant differences in variance reduction or stochastic properties between CMC filtered time series whether they have been corrected for surface mass loading or not, indicating that most of the variability due to surface mass loading is removed in the filtering process. When compared to time series that are not filtered, CMC filtering reduces both median white and flicker noise amplitudes, but median velocity uncertainty changes very little due to an increase in random walk noise amplitudes in some filtered time series. A comparison of monument types using the CMC filtered time series shows that out of the two most common monument types in the study region, roof mounted and concrete pillars, roof mounted monuments have a lower median velocity uncertainty. However, concrete pillar monuments in Michigan outperform other concrete pillar monuments and most roof mounted monuments, indicating that the design of concrete pillar monuments plays a crucial role in the stability of the station. A seasonal signal correlated to temperature was found to be present in some stations, particularly in the states of Wisconsin and Minnesota. The implications for this temperature seasonal signal and a potential method for correcting it, in order to extract only the hydrological related seasonal signal from GPS time series, is discussed.