Search

In this document I show a few synthetic examples using the method of: Jérémie Giraud,Mark Lindsay, andMark Jessell, (2021), "Generalization of level-set inversion to an arbitrary number of geologic units in a regularized least-squares framework,"GEOPHYSICS86: R623-R637, https://library.seg.org/doi/full/10.1190/geo2020-0263.1 to furhter test the robustness of level set inversion to noise and errors in the starting model.&nbsp

Taking advantage of all available information including geological and geophysical datasets is important to understand the subsurface. Integration of geological and geophysical datasets is commonly limited to the interpretation of areaspecific case studies and has seldom been addressed by automated techniques. Considering, the geometry of rock units as the unknown for both problems, one advantage of the automation of integration approaches is the parametrization of the subsurface which can be identical for both geological modelling and geophysical inversion. In this study, we introduce an approach and application that allows us to automate geophysical inversion workflow with an implicit geological modelling engine. We consider the contacts between units as the quantity linking geology and geophysics and generate a model from the geometry of the rock unit boundaries. We use a generalized level-set inversion algorithm integrating the implicit geological modelling engine called LoopStructural, which is used to define a regularisation correction term for geophysical inversion. The geological plausibility of the inverted model is continually evaluated and ensured during inversion based on correction factors from geology. Constraints from other geophysical datasets or existing knowledge are also encapsulated in weighting matrices through constraining terms. The introduced methodology has been tested in the Boulia region (Southern Mount Isa, Queensland), using several 2D seismic transects, bore-hole density measurements, and high-resolution gravity datasets. This study focuses on recovering the geometry of major structures of the basement units on a regional scale. The utilisation of seismic profiles and gravity datasets with geological modelling allows us to generate a 3D model of the area containing distinct rock units and boundary geometries that realistic gravity and seismic responses. The legacy datasets and existing results from the previous modelling have been utilised and adjusted ensuring that the final model follows all known information about the region., Open-Access Online Publication: May 29, 2023

The uploaded file contains four subdirectories: 1- Starting model The starting model belongs to the Boulia Region in Queensland, Australia, generated from existing interpretations. The size of the model is (nx=100, ny=90, and nz=40) 2- Inverted model This file contains the inverted model and grid from the cooperative inversion using the Level-set algorithm. 3- Seismic traces The file contains CDP coordinates of reflection seismic profiles (the first column shows Easting coordinates and the second column shows northing coordinates) that could be used for constraining purposes. 4- Gravity datasets The file contains the coordinates (first three columns) and measured spherical cap Bouguer anomaly gravity datasets (fourth column). The model and datasets in this repository have the same format as input models and datasets in Tomofast-x software (https://github.com/TOMOFAST/Tomofast-x)

Magnetotelluric (MT) and the passive seismic Horizontal to Vertical Spectral Ratio (HVSR) methods are commonly used to characterize cover thickness. While MT is sensitive to resistivity contrasts in the subsurface, the MT data inversion process is affected by non-uniqueness, noise, and sometimes sparsely sampled data, all which tend to increase uncertainty in the inverted models. Likewise, HVSR models are also affected by uncertainty. In this study, we test a new approach to exploit the complementarity between HVSR and MT modelling, using structural information from HSVR to reduce the uncertainty on the cover thickness recovered using MT data inversion. To this end, we adjust the roughness constraints applied to MT inversion using the depth range predicted by the HVSR method. New approach can recover sharp resistivity contrasts at the cover-basement interface and the cover depth estimation uncertainty could be reduced to 5%., Open-Access Online Publication: March 03, 2023

We introduce a sequential inversion workflow where we integrate magnetotelluric and magnetic data. We first perform probabilistic MT inversions, from which we derive 2D or 3D probabilities of observing an interface between rock units of contrasting electrical resistivity. Secondly, we use these probabilities to partition the model into domains where different rock units, or combinations thereof, can be observed. Using these domains, we define constraints for magnetic data inversion where the corresponding intervals (joint or disjoint) of magnetic susceptibility are used as spatially varying bound constraints for inversion. After introducing the methodology, we investigate the proof-of-concept using a geologically realistic synthetic model and conclude that the proposed methodology is applicable to field data. We then present ongoing investigations in the Cloncurry area (Queensland) on a 2D line to image the thickness of the sedimentary cover and to reduce interpretation uncertainty., Open-Access Online Publication: March 03, 2023

Integrating complementary geophysical datasets has proven to be a powerful tool for constraining the subsurface properties and also generating a model compatible with available datasets. As different geophysical surveys are designed with specific targets, their respective coverage usually do not overlap in all regions of interest. This prompts the development of new techniques that enable the integration of geophysical datasets with different spatial coverage in a single workflow. In this study, we aim at introducing a workflow that allows quantitative integration of geophysical datasets with different surface coverage, resolution, and levels of sparsity. We focus on constraining gravity inversion with seismic data sparsely distributed within the model space using a generalized level-set approach. An illustration of the applicability of the technique on 2D and 3D models with lower-dimensional constraints from seismic data is presented using two examples. We show that the uncertainty in target positioning can be quantitatively appended to the regularization terms allowing level-set to correct the boundary positioning. Furthermore, the flexibility of the approach in terms of including spatially distributed constraints from seismic interpretation in the level-set inversion is demonstrated. Finally, the primary results of the constrained level-set inversion on the Yamarna region are presented. The inverted density contrast model of the subsurface follows the detectible features of main greenstones in the seismic section. The resulting model encourages introducing new geological units and new structural constraints to be applied during modelling process of the region in future studies., Open-Access Online Publication: March 01, 2023