Advances on the implicit structural modeling benchmark.

Guillaume Caumon and Julien Renaudeau and Modeste Irakarama and Lachlan Grose and Miguel Varga and Michael Hillier and Eric A. Kemp and Pauline Collon and Florian Wellmann and Laurent Gautier and Laurent Ailleres. ( 2020 )
in: 2020 RING Meeting, ASGA

Abstract

Implicit methods have gained significant popularity in recent years as an alternative to surface-based structural modeling methods (also known as contouring, gridding or explicit structural modeling). These approaches represent geological interfaces as equipotentials of a scalar field. Whereas existing formulations share the same principle, they have not yet been quantitatively compared beyond theoretical discussions. We propose a new review of existing methods and highlight their main features in terms of data handling, basis function, mathematical principles and relationships to meshing. We propose three benchmark data sets which present specific challenges for implicit modeling: a folded turbidite (Hecho), a carbonate buildup displaying large thickness variations (Claudius) and a convoluted synthetic surface representing a hydrothermal body (Moureze). We applied several implementations of the various classes of implicit methods on these data sets, using one single scalar field and minimal parameter tuning. Results are available on an open repository. We propose metrics to analyze the obtained results in terms of data accuracy, ability to predict some features not present in the data, and topological complexity. We highlight, however, the risk of abusing these metrics to provide a definitive global ranking of methods or codes. The first results highlight, for all methods, inconsistent stratigraphic features on the Claudius data. Even though this can be addressed using several scalar fields, this illustrates the difficulty to come up with a universal geological modeling method, and highlights the need to build additional geological knowledge to cope with uncertainty in future structural modeling methods.

Download / Links

    BibTeX Reference

    @INPROCEEDINGS{Caumon_RM2020,
        author = { Caumon, Guillaume and Renaudeau, Julien and Irakarama, Modeste and Grose, Lachlan and Varga de la, Miguel and Hillier, Michael and Kemp de, Eric A. and Collon, Pauline and Wellmann, Florian and Gautier, Laurent and Ailleres, Laurent },
         title = { Advances on the implicit structural modeling benchmark. },
     booktitle = { 2020 RING Meeting },
          year = { 2020 },
     publisher = { ASGA },
      abstract = { Implicit methods have gained significant popularity in recent years as an alternative to surface-based structural modeling methods (also known as contouring, gridding or explicit structural modeling). These approaches represent geological interfaces as equipotentials of a scalar field. Whereas existing formulations share the same principle, they have not yet been quantitatively compared beyond theoretical discussions. We propose a new review of existing methods and highlight their main features in terms of data handling, basis function, mathematical principles and relationships to meshing. We propose three benchmark data sets which present specific challenges for implicit modeling: a folded turbidite (Hecho), a carbonate buildup displaying large thickness variations (Claudius) and a convoluted synthetic surface representing a hydrothermal body (Moureze). We applied several implementations of the various classes of implicit methods on these data sets, using one single scalar field and minimal parameter tuning. Results are available on an open repository. We propose metrics to analyze the obtained results in terms of data accuracy, ability to predict some features not present in the data, and topological complexity. We highlight, however, the risk of abusing these metrics to provide a definitive global ranking of methods or codes. The first results highlight, for all methods, inconsistent stratigraphic features on the Claudius data. Even though this can be addressed using several scalar fields, this illustrates the difficulty to come up with a universal geological modeling method, and highlights the need to build additional geological knowledge to cope with uncertainty in future structural modeling methods. }
    }