Analysis and Prediction of Long Term GNSS Height Time Series and Environmental Loading Effects
Open access
Author
Date
2021-01-25Type
- Master Thesis
ETH Bibliography
yes
Altmetrics
Abstract
GNSS height residuals often exhibit seasonal amplitudes, that can partly be explained by environmental surface loadings, such as hydrological, non-tidal atmospheric and non-tidal oceanic loading. In this thesis the state of the art procedure to reduce the height component of GNSS residuals in Europe by those environmental loadings is evaluated, with a focus on the residual RMS and amplitude reduction. On the one hand, the environmental loadings are subtracted directly from the residuals and on the other hand their annual component is reconstructed using Singular Spectrum Analysis (SSA), which is then used for the reduction. In the second part of this thesis, more complex relationships between GNSS height residuals and environmental in uences are explored. Temporal Convolutional Networks (TCN) and Random Forests (RF) are used to model and predict the GNSS height residuals, using environmental loadings, raw meteorological data and tropospheric zenith delays as input features. The RMS and amplitude of the GNSS height residuals could be reduced to a median of up to 19% and 23% when using the original loading series, respectively. An RMS reduction of an average of 2% was obtained, by removing the loadings series after rst reconstructing them with SSA. The majority of tested GNSS stations bene t from the inclusion of environmental parameters in their residual prediction. These yield improvements of up to 6% in prediction error, when compared to a prediction using GNSS residual data only. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000519390Publication status
publishedPublisher
ETH ZurichOrganisational unit
09707 - Soja, Benedikt / Soja, Benedikt
More
Show all metadata
ETH Bibliography
yes
Altmetrics