Probabilistic imaging of sedimentary basins using spatial clustering of magnetotelluric model ensembles change-points

Abstract

Mapping the internal structure of sedimentary basins, including the depth to basement, is valuable for a variety of geological applications, particularly the identification of natural resources such as groundwater and minerals, or of the geological structures associated with them. The magnetotelluric (MT) method has proven to be a reliable technique for imaging complete sedimentary sequences, especially in areas where thick sedimentary cover makes imaging challenging. However, due to the high regularisation required by MT inversion procedures, it is limited in its ability to precisely locate geological interfaces, which is crucial for exploration. Building on previous work where we imaged the basement using interface probabilities derived from 1D probabilistic inversion of MT data, we present a new method which allows for the simultaneous classification of multiple interfaces across an entire survey, incorporating constraints on the spatial relationships between models. As a result, we obtain spatially consistent model ensembles and classified interfaces corresponding to transitions between layers of consistent electrical resistivity. This classification effectively reduces the size of the model ensemble, decreasing uncertainty in the estimated depth of the interfaces of interest. We apply this method to the Eucla sedimentary basin in Western Australia, analysing 550 MT sites distributed across 12 profiles. The ensemble clustering clearly identifies three interfaces which are consistent with the sedimentary succession expected from this basin. The results show great consistency across all the survey, providing valuable insights into the geological setting of the area. This research demonstrates the capability to reliably image the structure of a sedimentary basin using MT, within a workflow that integrates probabilistic inversion and model ensemble classification.