Research On Three-dimensional Joint Inversion Of Gravity,magnetic And Magnetotelluric Data

Joint inversion is the process of inverting two or more geophysical field data together to ultimately pursue a unified geological model such that all observation datasets satisfy the fitting conditions.In recent years,the multi-perspective and multi-information development trend of geological interpretation has made the joint inversion issue become a hot research topic in the field of geophysical exploration.Gravity,magnetic and magnetotelluric methods are all important techniques commonly used in the geophysical exploration,and the joint inversions between the corresponding data can effectively improve the resolution and reduce the multi-solution problems,which is conducive to a more comprehensive prediction of the subsurface structures and is of great significance for the mineral exploration,geological mapping and tectonic studies.The following key issues in the joint inversions deserve in-depth study.First,the way to construct connections for the mutually independent geophysical datasets has long been a central issue in the joint inversions.Combining the known geological and geophysical information maximally to conduct an integrated inversion interpretation is of great importance for the geophysical exploration.Second,an appropriate regularization is an effective means to further reduce the multi-solvability of the joint inversions and improve the result accuracy.Imposing additional constraints on the physical models while pursing data fitting is conducive to obtaining results that are more consistent with the physical distribution laws,and the research on a more stable and accurate regularization method is important for further improving the inversion quality.In addition,the exploration requirements for the mountainous and hilly areas make the terrain problem in the joint inversions not negligible,and the emergence of the unstructured grid techniques has certainly provided a new idea to solve this problem,which enables the spatial structures of the undulating terrain and its geophysical response to be directly simulated.However,some traditional regularization and joint coupling methods commonly used in the regular grid are not directly applicable to the unstructured grid due to the factors such as irregular arrangement and element volume differences,and it is important to explore their new forms that are suitable for the unstructured grids.Around the above issues,we study the joint inversions between the gravity,magnetic and magnetotelluric data based on the previous work and successively proposed three sets of high-precision joint inversion algorithms,including a cross-gradient joint inversion algorithm for the gravity,gravity gradient and magnetotelluric data based on the smooth-focusing regularization,a cross-gradient joint inversion algorithm for the gravity and magnetic data based on the combined hexahedral grid,and a joint inversion algorithm for the gravity and magnetic data under the tetrahedral grid based on the four-directional-gradients structural constraints.Among the above algorithms,some new regularization and joint coupling methods are proposed to improve the inversion results.First,to address the problems that the boundary of the smoothing constraint result is blurred and the focusing constraint is prone to generate false anomalies,we propose a smooth-focusing regularization method for the magnetotelluric inversion by combining the two constraints.The method achieves more stable focusing by giving a weaker smoothing constraint to the high-anomaly elements and a stronger smoothing constraint to the low-anomaly elements.Model tests show that the smooth-focusing regularization method is useful for obtaining a smooth large-scale physical distribution to reduce the local false anomalies,while effectively capturing the small-scale details to make the anomaly boundary clear and sharp.The above improvements on the resistivity will further improve the overall accuracy of the joint inversion results through the structural coupling.Second,to address the problem that the traditional sensitivity-based weighting method is not directly applicable to the gravity and magnetic inversions under the unstructured grids due to the element volume differences,we propose a new form for this method that re-corrects the weights according to the element volumes.Model tests show that the new weighting form is appropriate,which can effectively avoid the false anomaly problems caused by the traditional method,that is,the tendency of the anomalies to gather in the small elements,and not to induce the anomalies that actually exist in the small elements to their nearby large elements due to the excessive correction.The new weighting form is beneficial to obtain a more accurate inversion result.On the other hand,to improve the error and efficiency limitations of the three structural coupling methods based on physical gradients,including the cross-gradient method,the gradient dot product method and the cosine dot-gradient method,when applied to the unstructured grids,some new forms of the above methods are proposed to avoid their dependence on the Cartesian axial gradients.For the gradient dot product and cosine dot-gradient method,we propose an idea of constructing the constraints directly using the gradients in the four adjacent element directions under the unstructured tetrahedral grid and design the corresponding four-directional forms for the two dot product constraint functions.For the cross-gradient method,which is difficult to construct into a four-directional form,we propose a combined hexahedral grid.The grid uses curved hexahedral divisions in the shallow part of the model to simulate the terrain accurately,and uses regular hexahedral divisions in the deep part of the model to ensure the computational efficiency.With the more regular element arrangement of the grid,we define a geometrical cross-gradient that directly constrains the physical relationships between elements to achieve the structural coupling.By avoiding the indirect process of obtaining the Cartesian axial gradients and its accompanying error and efficiency problems,the new methods obtain results with better accuracy and structural consistency in the model tests,and show higher efficiency in the calculation of the joint coupling correlation terms.Finally,the above-mentioned cross-gradient joint inversion algorithm based on the combined hexahedral grid and the tetrahedral grid joint inversion algorithm based on the four-directional-gradients structural constraints are applied to the processing of measured gravity and magnetic data from a skarn-type mining area in Huzhong,Heilongjiang Province,China,respectively.After establishing a model with terrain based on the elevation data of the study area,the model was divided into a tetrahedral grid and a combined hexahedral grid respectively,and the observation data were inverted separately and jointly under the two grids.It is shown that both the joint inversion algorithms obtain model results with higher structural consistency than the separate inversions,and the deposit location inferred from the joint inversion results is in general agreement with the previously proven one.This successful application case further verifies the effectiveness and practicality of the above algorithms.