Font Size: a A A

The Study Of3D Inversion Of Full Tensor Gravity Gradiometry Data Based On Geostatistics Method

Posted on:2016-07-28Degree:DoctorType:Dissertation
Country:ChinaCandidate:M X GengFull Text:PDF
GTID:1220330467993954Subject:Solid Earth Physics
Abstract/Summary:PDF Full Text Request
Compared with conventional gravitational exploration which only surveysfirst-order vertical derivative of gravitational field, gravity gradient can survey secondderivative of gravitational field. Therefore gravity gradient survey has betterresolution, and is more affected by near-surface density variations. Moreover, gravitygradiometry has higher resolution of localized density contrasts within regional fieldand a higher signal-to-noise ratio due to its ability to reject common-mode noise.Therefore, gravity gradiometry is widely used in aviation, marine and satellite gravityfield.With the full tensor gravity gradient measurement technology becoming mature,Interpretation of full tensor gravity gradient data becomes the research hotspot.3Dinversion for physical properties, which can outline source scope according to chagesof physical properties, becomes an important inversion method, with which morecomplex sources can be recovered In this thesis, we proposed two inversionapproaches which are cokriging and stochastic simulation based on geostatisticmethod to solve the problem of serious multiple solutions and enormous computationfor3D inversion. The cokriging method minimizes the theoretical estimation errorvariance by using auto-and cross-correlations of several variables. This method caneasily include complex a priori information and the process of inversion does notrequire iterations, so it is effective. Moreover, the smoothing effect in the recovereddensity model can be reduced to a certain extent by using the anisotropy constrain inthe covariance model Based on the inverted result of cokriging, stochastic inversion isimplemented. Both the two method presented a new idea for inversion of gravity andmagnetic data.Kriging is one of the primary methods of geostatistic. Based on the study ofkriging, we established the equation with the density as the primary variable andgravity gradient data as the secondary variables. In order to conteract the near surface effect, the sensitivity matrix was introduced into the cokriging equation. To decreasethe amount of memory required and to speed the computations, we used the sparsematrix techniques by setting a threshold value. We also explored the extent to whichthe sparsity of the covariance matrix in the simple cokriging system affects theinverted results. The model tests show that this proposed inversion method based oncokriging can fully integrate geological information, for example, dip information,physical property constraints, and geophysical data, for example the gravity-gradientand borehole gravity data in this paper by means of the covariance formulation. Thismethod does not rely on iterative minimization of an objective function. Moreover,cokriging inversion also provides estimation error variance for each estimated model,which helps to assess the quality of the inverse models.Based on the results of cokriging, we obtained a variety of probabilistic modelsusing stochastic inversion. All the recovered models fit the observed data and thegeological information contained in the covariance matrix. And the differences amongthe recovered models reflect the heterogeneity and uncertainty of the spatialdistribution of the geological properties.With the advent of full tensor gravity gradiometry (FTG) instrument, richerinformation on the geology body can be obtained. The information contained in eachgradient component is different, joint different components for inversion can help toimprove the accuracy of geological interpretation. In this paper, we tested theinversion results obtained by single component and the combination of multiplecomponents. The results show that inversion using multiple components can improvethe resolution of recovered model, however, which does not mean that the morecomponents are used, the better results are recovered. An ideal combination can notonly reduce the amount of calculation but also improve the speed of inversionprocess.We also studied the ability of joint inversion of borehole gravity data and gravitygradient data using cokriging and stochastic inversion. The model test results showthat cokriging inversion with borehole gravity data allows to recover more accurate,more well-defined density models than is possible just with gravity-gradient data.When the target of interest is intersected by the borehole, ideal3D subsurface densitymodels can be recovered. In particular, the densities in the recovered models can be much closer to their true values. However, the resolution of borehole data falls offrapidly as the distance between the borehole and a feature of interest increases.Therefore, if the borehole is far away from the target of interest, the result recoveredby gravity-gradient data would be better than that recovered by borehole data. Themodel test results also show the combination of all the five independent components,borehole gravity, as well as known densities tends to produce the best result and thestandard deviations calculated from the residuals between the predicated data and theobserved data can be used to predict the noise in each gravity-gradient component.Finally, the method was applied to inverse the airbore FTG data obtained overthe Vinton area. From the inverted result, the space distribution of cap rock of the areawas deduced.
Keywords/Search Tags:Full Tensor Gravity Gradiometry (FTG), 3D inversion, Geostatistic, Cokriging, Stochastic inversion, Joint Inversion, Borehole gravity
PDF Full Text Request
Related items