3D modelling and inversion of audio-magnetotelluric (AMT) data from McArthur River, Athabasca Basin, using unstructured tetrahedral grids

The main aim of this study is to do trial-and-error 3D forward modelling and inversion for the AMT data collected in the McArthur River area of the Athabasca Basin. For the forward modelling and inversion, the ideas are not new, but the tools used in this study have not been used previously for the corresponding AMT data set. In the McArthur River area, the P2 fault is a thrust fault in the basement formed within pelitic gneiss, and the mineralization of the unconformity-type uranium found in the Athabasca Basin is often related to the P2 fault - which is originally a graphitic unit. Different geophysical and geological surveys have provided information about parts of the substructure; however, none of the studies so far presented have provided a better understanding of the P2 fault of the McArthur River mine area at depth. All previous studies, including electrical and electromagnetic (EM) surveys, had pointed out the increasing need for imaging the deep parts of the P2 fault. Consequently, a natural source method, the Audio-Frequency Magnetotelluric (AMT) method, was performed in 2002 within the scope of EXTECH IV (EXploration science and TECHnology) project by other researchers to image greater depths at low costs. In this study, a synthetic model is first created by trial and error based on previous studies. The accuracy of the model is checked by comparing the calculated apparent resistivity and phase values with the measurements. According to the results obtained from the forward modelling calculations, the accuracy of the model is satisfactory. iii From a data inversion point of view, this study consists of four inversions. The first is the inversion of the synthetic data for the model constructed by the trial-and-error forward modelling. This allows for the capabilities of the inversion process to be assessed. The other three are the inversions of the real data. For the synthetic data inversion, ten frequencies were used, and the data were fit successfully. The real data inversions were performed using three different forms of the data uncertainties. In the first scenario, the variances estimated from the data processing were considered as uncertainties. In the other two real-data inversions, uncertainties of 3% and 5% were used. Results of the real-data inversion were compared to those of previous inversion studies. The results from all three real-data inversions show good consistency with those of earlier studies