Joint inversion, the inversion of multiple geophysical data sets containing complementary information about the subsurface, has the potential to significantly improve inversion results by reducing the nonuniqueness of the inverse problem. One of the challenges of joint inversion is deciding how to couple the multiple physical property models. If a coupling approach is used that is inconsistent with the physical truth, then inversion artifacts can occur and may lead to incorrect interpretations. In this paper, we investigated the fuzzy c-means (FCM) clustering approach to provide a lithological coupling of the seismic velocity and density models in joint 2D inversions of first-arrival traveltimes and gravity data. Even though this coupling approach has been used in previous works, recommendations for its effective use have not yet been developed. We conducted a suite of joint inversion tests on synthetic data generated from a geologically realistic model based on magmatic massive sulfide deposits. There is a known relationship between seismic velocity and density for the silicate rocks and sulfide minerals involved; this lithological relationship was used to design a clustered coupling strategy in the joint inversions. The tests we conducted clearly exhibited the benefits of joint inversion using FCM coupling. Our work revealed the effects of including inaccurate a priori physical property information. We also evaluated approaches to assess whether such inaccurate information may have been used.

Three-dimensional geological Earth models typically comprise wireframe surfaces of connected triangles that represent geological contacts. In contrast, Earth models used by most current 3D geophysical numerical modeling and inversion methods are built on rectilinear meshes. This is because the mathematics for computing data responses are simpler on rectilinear meshes. In such a model, the relevant physical properties are uniform within each brick-like cell but possibly different from one cell to the next, producing a pixellated representation of the Earth. In principle, arbitrary spatial variations can be represented if a sufficiently fine discretization is used. However, no matter how fine the discretization of the rectilinear mesh, such a mesh is always incompatible with geological models comprising wireframe surfaces. Also, because the computational resources required by 3D numerical modeling and inversion methods increase dramatically as the discretization of a model is refined, it is never really possible to achieve as fine a discretization as one would like. This exacerbates the mismatch between models that comprise wireframe surfaces and those built on rectilinear meshes. To address this incompatibility, we are using unstructured tetrahedral meshes to specify 3D geophysical Earth models. We hope that working with unstructured meshes will facilitate the construction of common Earth models consistent with both the geological and geophysical data available.

No abstract

Self-organizing maps (SOMs) are a type of unsupervised artificial neural networks clustering tool. SOMs are used to cluster large multi-variate datasets. They can identify patterns and trends in the geophysical maps of an area and generate proxy geology maps, known as remote predictive mapping. We applied SOMs to magnetic, radiometric and gravity datasets compiled from multiple modern and legacy data sources over the Baie Verte Peninsula, Newfoundland, Canada. The regional and local geological maps available for this area and the knowledge from numerous geological studies allowed for assessing the accuracy of the SOM-based predictive mapping. Proxy geology maps generated by primary clustering directly from the SOMs and secondary clustering using a k-means approach reproduced many geological units identified by previous traditional geological mapping. Of the combinations of datasets tested, the combination of magnetic data, primary radiometric data and their ratios, and Bouguer gravity data gave the best results. We found that using reduced-to-the-pole residual intensity or analytic signal as the magnetic data were equally useful. The SOM process was unaffected by gaps in the coverage of some of the datasets. The SOM results could be used as input into k-means clustering as k-means clustering requires no gaps in the data. The subsequent k-means clustering resulted in more meaningful proxy geology maps than were created by the SOM alone. In regions where the geology is poorly known, these proxy maps can be useful in targeting where traditional, on-the-ground geological mapping would be most beneficial which can be especially useful in parts of the world where access is difficult and expensive.

No abstract