Forootan, Ehsan: Statistical Signal Decomposition Techniques for Analyzing Time-Variable Satellite Gravimetry Data. - Bonn, 2014. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37662
@phdthesis{handle:20.500.11811/5858,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37662,
author = {{Ehsan Forootan}},
title = {Statistical Signal Decomposition Techniques for Analyzing Time-Variable Satellite Gravimetry Data},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2014,
month = sep,

note = {The time-variable gravity fields from the Gravity Recovery and Climate Experiment (GRACE) satellite mission provide valuable information about total water storage variations on a global scale. This quantity is difficult to observe with in-situ measurements but important for understanding regional energy balance, as well as for agricultural, and water resource management. In order to utilize GRACE time-variable level 2 products for studying global mass transport, there are two major problems that users face: 1) the presence of correlated noise in the level 2 potential spherical harmonic coefficients that increases with harmonic degree and causes ‘striping’ in the spatial domain, and 2) the fact that different physical signals are overlaid and difficult to separate from each other. These problems are termed the ‘signal-noise’ separation problem and the ‘signal-signal’ separation problem.
In this thesis, statistical decomposition methods are investigated to perform signal-noise and signal-signal separation using the time series of total water storage changes derived from satellite gravimetry products. In particular, the focus lies on the mathematical foundation of the second order statistical decomposition approach such as the principal component analysis (PCA), and its ordinary extensions, as well as the higher order statistical decomposition method of independent component analysis (ICA). The mathematical relationships between second and higher order statistical signal decomposition techniques are discussed. Uncertainties introduced in the extracted patterns, e.g., due to the limited time span of observations in computing auto-covariance matrices and higher order moment tensors, are addressed. The ICA approach is extended to the Complex ICA technique, which allows extraction of patterns that vary in space and time. Simulations of GRACE-like total water storage time series are used to assess the performance of the introduced statistical approaches. The ICA approach is applied to reduce the spatial leakage over the Australian continent, and to partition total water storage changes into terrestrial and groundwater storage changes over the Middle East. A new statistical approach is also introduced to forecast total water storage changes over West Africa, where it exhibits strong atmosphere-land-ocean interactions.},

url = {http://hdl.handle.net/20.500.11811/5858}
}

The following license files are associated with this item:

InCopyright