A Study Of Theory And Methods On Establishment Of Regional Crustal Motion Model

The study of regional crustal motionl is an important domain in which the motion characteristics and the deforming reason on the earth’s surface are quantitative analyzed. And the building and the implement of regional crustal motion model are very significant in the research of earth science. With the speedy development of modern geodetic techniques, the more and more space geodetic data and new analytical techniques provide us a new way to detect and study the crustal deformation. How to ultilize the data collected by space geodetic techniques effective, and analyse them reasonally in the correct background fields of the crustal deformation are the foundational issues for geodynamics. This dissertation mainly focuses on the theories and algorithms for model construction of regional crustal deformation. The main works and contributions are summarized as follows:1. Conventional function models for study of regional crustal motion are compared and analyzed. Since the local deformation or abnormal observations that are unavoidable may influence the reliability of modeling, robust estimation with high breakdown point is applied to select and identify relatively stable stations. With the proposed method, the weighs of the abnormal observations are reduced even to zero to resist their impact to model parameters and the model that constructed just by relatively stable stations is obtained directly.2. The characteristics both of functional and statistical approximation methods are compared and analyzed. A new family of synthetically approximation by functional model and statistical method is proposed, by which functional models are used to approaching the trends, systematic variations and the statistical methods are used to compensate the fitting remanets (in which the local deformations and stochastic deformations are involved) of function model. As to concrete practice, adaptive collocation method is proposed which adjust the disharmony between covariance matrices of observations and random signals by Helmert variance component estimation to improve the accuracy of the calculation of the variation of the unobserved points. Another model synthesized by functional models and Kriging statistics is also proposed to approximate simultaneously the trended deformation and the local, stochastic deformation.3. In large-scaled or complicated velocity field, faults are often hidden, which can bring about discontinuous deformations on both sides of the faults. Thus, conventional collocation method would produce inaccurate results where only sketchy trend can be drawn and just smooth, continuous deformations are obtained instead of the actual abrupt differences of velocities on the two sides of the fault. So collocation based on regional tectonic deformation background is proposed, which divides the area roughly according to geology information and then establishes collocation models respectively in each division. The points near by the boundary are differentiated by confidence interval of deformation parameter. After the points round the boundary are all judged and divided into correct sub-divisions, collocation models are established again in each renewed sub-division. For those new-built models contain the information of all the reasonable points in each sub-division, it adapts better to the characters of local deformation, and is more reasonable and thus can discover determinate position of the hidden fault.4. Science the datum differences, geophysical model errors, observational model errors, and so on in the result of geophysical method and geometrical (or observational) method, it is reasonable to combine the two kinds of information to collect the crust deformation information. So a hybrid estimation strategy is proposed for evaluating the deformation parameters employing the adaptively robust Bayes method. Adaptive factors are introduced to balance the contribution of the geophysical moded information and the geometrical measurements to the model parameters. And the effects of measurement outliers on the estimated parameters are controlled by robust equivalent weights. In this paper, the displacements predicted from the geophysical model are employed as priori value to implement the crustal deformation analysis by using the GPS measurements and Helmert variance estimation are adopted to balance the effects from the two different types of data sets.5. Bayesian estimation relys on reliable prior imformation for iterative calculation in solve of non-linear model, so synthesis method of stochastic optimization and Bayes estimation and its application in parameter calculating for non-linear model in crustal deformation analysis are discussed. Gaving that the prior information of model parameter may be distorted, Bayes estimation based on robust cyclic searching is proposed, by which deformation predicted by geophysical model as priori information is employed in geodetic (GPS) measurements processing, and cyclic searching is adopted to modify improper priori information. Particle swarm optimization is also discussed for the satiation that the prior information is ambiguous even unknown, by which approximately parameter value can be obtained and served as initial value of model parameter for subsequent iterative calculation in Bayes estimation.