The Application Of Least-Squares Collocation In Physical Geodesy

The elementary theory of Least-Squares Collocation (LSC) in Physical Geodesy is described in detail in this thesis, as well as the method defining the covariance function, which is the key problem when utilizing LSC in physical geodesy. Based on these premises, an experiment is implemented by combing the observed gravity data, DTM and EGM96 Global Geopotential Model to construct the local gridded geoid model. Then the approach of combing gravity anomalies and GPS/Levelling data is discussed. The computations are carried out both in China and in United States for ascertaining the conclusions. When compared with the method more commonly used now-polynomial fitting, the feasibility and validity of LSC method are presented. In order to overcome the most primary numerical problem that LSC is a CPU time-consuming method, fast collocation is discussed subsequently due to the special structure of covariance matrix that if the input observations are gridded, Sigal-Noise irrelevant, then the covariance matrix of the observations is a block/block Toeplitz matrix, the spent time and memory can be reduced sharply without the explicit loss of accuracy. The general theory of processing altimetry data is then discussed. After processing the GEOSAT/GM altimetry data, the Sea Surface Heights in the open sea are derived and then used to predict gridded gravity anomalies with the fast collocation; the effects of degree variance model on LSC results are discussed particularly.