UBC Theses and Dissertations
Practical Inversion of 3D time domain electromagnetic data Napier, Scott
The purpose of this thesis is to provide a simple procedure by which time domain electromagnetic (TEM) data, which has been collected in the field, can be inverted to recover a three dimensional distribution of electrical conductivity. An algorithm, EH3DTDinv, which has been developed in order to invert TEM data in 3D is briefly presented. The algorithm has been previously shown to work well on synthetic data. However, its ability to invert real data collected in field surveys was previously unclear. To illustrate the effectiveness of the inversion algorithm when used in conjunction with an inversion procedure, the example of the UTEM survey at the San Nicolas deposit in Zacatecas Mexico is used. The inversion procedure provides a detailed description of the preparations required for 3D TEM inversion. In order to clarify the individual stages in the procedure, the inversion preparations are outlined as they apply to the San Nicolas survey. Problems of non-uniqueness in the inverted models at San Nicolas are encountered and the solution is determined to be a combination of a sensitivity weighting scheme and the use of the data from multiple transmitters simultaneously in a single inversion. The iterative nature of the inversion procedure is demonstrated through the San Nicolas example. The San Nicolas inversion example culminates in the simultaneous inversion of data from 3 transmitters where a priori information from the regional geology has been incorporated into the result. The 3D inversion result is the first large scale 3D TEM inversion for a mineral deposit and agrees well with the geology and conductivities interpreted from drilling. Practical aspects of inversion are also considered and some recommendations on data formats, software utilities, and survey design are made in order to enhance future results. A brief study into inversion efficiency determined that a significant reduction in computation times can be achieved without loss of accuracy or resolution by fine tuning the cooling schedule of the algorithm. Further reductions in computation times are made through parallelization of the forward modelling process. Finally, yet more reductions in computation times are possible through the coarsening of the spatial and temporal discretizations. However, this final method of computational reduction is shown to come at the cost of accuracy and resolution in the inverted model. The thesis concludes with a discussion of the potential for 3D TEM inversion and brief outline of future advancements in the field.
Item Citations and Data