Seismic constrained magnetotelluric inversion for iron orebody at Gällivare, Sweden

The Norrbotten region (Northern Sweden) boasts a rich history of mineral exploration, particularly in the Kiruna and Gällivare areas. Both passive and active geophysical exploration methods, specifically magnetotelluric (MT) and seismic measurements, have played pivotal roles in guiding mineral exploration in this region. In this study, we propose an MT inversion method constrained by interpreted 3D seismic reflections in the Gällivare area of northern Sweden, with a focus on a known iron orebody, primarily composed of magnetite. Constraints were introduced using a covariance matrix created from interpreted 3D seismic reflection surfaces. 

Between 2016 and 2021-2023 (as part of the D-REX project), Luleå University of Technology and LKAB collaborated to collect broadband magnetotelluric data from 116 stations around Gällivare. The survey covered a total area of 15 x 15 km, with a site spacing of approximately 1 km where allowed by terrain conditions. Similarly, in the same area, 3D seismic data was acquired by a service company, made available as 3D stacked data. For this study, a subset of the acquired MT data was used to limit the modelling domain to the area covered by 3D reflection seismic, resulting in a square area of approximately 5 x 5 km, comprising 17 MT stations. Only higher frequencies (10^-3 to 1Hz) were considered to focus on the shallow area where seismic reflections are prominent (approximately 2000 m depth). 

The MT data were inverted using the ModEM code on a 120 x 120 x 50 meters grid with 10 growing boundary cells in horizontal directions and 30 in the vertical direction, facilitating valid electromagnetic boundary conditions during the inversion process. The dense discretization was chosen to better delineate the interpreted seismic surfaces. Subsequently, observed seismic reflection features, expected to be associated with an orebody of interest, were manually picked and interpreted. A Python code was developed to convert the interpreted surfaces into a binary 3D grid, matching the one used for the MT inversion in ModEM. This grid was exported and converted into a covariance matrix, utilized as a smoothing constrain to limit the interaction between cells inside and outside of the observed seismic surface in the inversion model. For comparison purposes, the same MT subset was also inverted using the same parameters but without a covariance matrix constraint.  

Upon comparing the results of both inversions, it was observed that the constrained model converges faster. This is advantageous not only for reduced computing time but also because the mathematical model stabilizes more quickly while achieving similar residual RMS values in both cases.  The resulting electrical resistivity model exhibits a more geological behaviour after including the smoothing constraint, aligning with the surface behaviour observed in the seismic data. 

The successful application of MT inversion methods constrained by seismic data is anticipated to reduce uncertainty in the electrical resistivity models of orebodies in the Gällivare mining area when compared to MT surface measurements alone, thereby enhancing confidence in the final inversion results and exploration efforts.