Browsing by Author "Chen, Qiang"

Now showing 1 - 2 of 2

Open Access
Analyzing and modeling environmental loading induced displacements with GPS and GRACE
(2015) Chen, Qiang; Sneeuw, Nico (Prof. Dr.-Ing.)
The redistribution of atmospheric, oceanic and hydrological masses on the Earth's surface varies in time and this in turn loads and deforms the surface of the solid Earth. Analyzing such environmental loading signal and modeling its induced elastic displacements are of great importance for explaining geophysical phenomena. Based on the well-established loading theory, this thesis makes use of two different space-borne measurements, i.e. GPS and GRACE, along with other environmental loading data to investigate three different aspects of environmental loading and its induced elastic deformations: Firstly, an increasing concern is observed recently over time variable seasonal signals in geodesy. Several model based approaches were applied to extract amplitude and phase modulated annual and semiannual signals. In view of this phenomenon, this thesis introduces an alternative approach, namely, singular spectrum analysis (SSA). With respect to these model-dependent approaches, the advantage of SSA lies in data-driven and model-independence. Several aspects regarding the application of SSA, e.g. optimal choice of window size, are investigated before showing its abilities. Through applying SSA to the lake level time series of Lake Urmia (Iran) and the basin averaged equivalent water height time series of the Congo basin, the capabilities of SSA in separating time varying seasonal signals are demonstrated. In addition, we find that SSA is also able to extract the non-linear trend as well as long-term oscillations from geodetic time series. Secondly, we look into the comparison between GPS and GRACE with an emphasis on GRACE data filtering. Three types of deterministic filters and two types of stochastic filters are studied and compared over GPS sites from two regions, i.e. the Europe area and the Amazon area. The comparisons indicate that no single filtering scheme could provide consistently better performance over other considered filters. However, we find that the stochastic filters generally show better performance than the deterministic filters. The DDK 1 filter outperforms other filters in the Europe area and the regularization filter of parameter lambda=4, which follows the concept of the DDK filters, shows optimal performance in the Amazon area. The combination of the isotropic Gaussian filter of a low smoothing radius, e.g. around 300 km with the destriping filter is proved to be optimal filter choice if only the deterministic filters are considered. Thirdly, based on an overview of displacements modeling at various spatial scales, we evaluate three methods, i.e. two types of half-space approaches and the classic Green function approach, by using a high spatial resolution local load data along the lower Mississippi river when a severe flood happened in 2011. The equivalence between the two half-space approaches, i.e. point load approach and surface load approach, are demonstrated with the local load data. However, the point load approach is recommended for practical use in terms of computational efficiency. In addition, within such a limited spatial extent, we investigate the differences between the half-space approach and the Green function approach. It is shown that the half-space approach predicts larger displacements than the Green function approach and agrees better with the observed deformations at 11 considered GPS sites. Meanwhile, strong global environmental loading effects are found via two global hydrological models, i.e. GLDAS and MERRA. Thus, a reduction of these far-field loading effects beforehand is suggested before probing the local crustal structure using the half-space approach. Last but not least, based on the local load data, the effects of site-dependent Green functions are studied with two types of site-dependent Green functions, which were generated by modifying the local crustal structure of the REF Earth model using the CRUST 1.0 and CRUST 2.0 models. A relative RMS of differences of more than 5% in vertical component and 25% in horizontal components are found with respect to the PREM Earth model based Green functions. It indicates that the Green functions could contribute more uncertainties in loading induced displacements modeling than reported in the literature.
Open Access
GPS time-variable seasonal signals modeling
(2015) Chen, Qiang
Seasonal signals (annual plus semi-annual) in GPS time series are of great importance for understanding the evolution of regional mass, i.e. ice and hydrology. Conventionally these signals (annual and semi-annual) are derived by least-squares fitting of harmonic terms with a constant amplitude and phase. In reality, however, such seasonal signals are modulated, i.e. they will have a time-variable amplitude and phase. Recently, Davis et al. (2012) proposed a Kalman filter based approach to capture the stochastic seasonal behavior of geodetic time series. In this study, a non-parametric approach, singular spectrum analysis (SSA) is introduced. It uses time domain data to extract information from short and noisy time series without prior knowledge of the dynamics affecting the time series. A prominent benefit is that obtained trends are not necessarily linear and extracted oscillations can be amplitude and phase modulated. In this work, the capability of SSA for analyzing time-variable seasonal signals from GPS time series is investigated. We also compare SSA-based results to two model-based results, i.e. least-squares analysis and Kalman filtering. Our results show that singular spectrum analysis could be a viable and complementary tool for exploring modulated oscillations from GPS time series. Based on the SSA-derived seasonal signals, we look into the effects of the input noise variances in the framework of Kalman filtering. Two Kalman filtering based approaches with different process noise models are compared over 79 GPS sites. We find that the basic Kalman filtering technique with the input noise model suggested by Davis et al. (2012) turns out to be optimal.