A Hitchhiking Foray into the Structural Uncertainty Space

Laurent Gautier and Lachlan Grose. ( 2020 )
in: 2020 RING Meeting, ASGA

Abstract

“Space is big. You just won't believe how vastly", hugely, mind-bogglingly big it is. I mean, you may think it's a long way down the road to the chemist's, but that's just peanuts to space.” \\ ― Douglas Adams, The Hitchhiker's Guide to the Galaxy “All models are wrong, but some are useful.” \\ ― Georges E. P. Box, Robustness in the Strategy of Scientific Model Building The geomodelling community has long reached a consensus about the necessity to go beyond single deterministic models to account for structural uncertainties. Nonetheless, advocating for a probabilistic approach is the easy part of the trip. The difficulties arise when trying to propose a proper approach to building alternative hypotheses, which comes down to generating ensembles of plausible models. Several strategies have been proposed to that aim, mainly: (1) tweaking existing deterministic models, (2) perturbing datasets before generating new models, or (3) sampling geometrical and/or kinematical parameters of geological structures. However, if all models are essentially wrong, the struggle is to generate all and only the useful ones. That being said, evaluating the efficiency of existing strategies for exploring uncertainties appears highly difficult, even in relatively simple real geological cases. In fact, only trying to imagine the complexity of the space of uncertainty, i.e., the architecture of the model space, is a real challenge. A rapid combinatorial evaluation shows that this universe of possible models is mind-bogglingly vast. This contribution proposes to address this issue by transposing the space of possible geological models into a simpler, better controlled geological universe. This surrogate universe is generated without any geological a priori but the number of different geological units considered and the resolution. A geodiversity approach, which combines Principal Component Analysis and clustering," is then applied for studying the organisation of this universe and identify families of models. We use this exploration to evaluate the capability of different uncertainty modelling approaches to explore the space of possible models when given a particular dataset.

Download / Links

BibTeX Reference

@INPROCEEDINGS{LAURENT_RM2020,
    author = { Gautier, Laurent and Grose, Lachlan },
     title = { A Hitchhiking Foray into the Structural Uncertainty Space },
 booktitle = { 2020 RING Meeting },
      year = { 2020 },
 publisher = { ASGA },
  abstract = { “Space is big. You just won't believe how vastly", hugely, mind-bogglingly big it is. I mean, you may think it's a long way down the road to the chemist's, but that's just peanuts to space.” \\ ― Douglas Adams, The Hitchhiker's Guide to the Galaxy “All models are wrong, but some are useful.” \\ ― Georges E. P. Box, Robustness in the Strategy of Scientific Model Building The geomodelling community has long reached a consensus about the necessity to go beyond single deterministic models to account for structural uncertainties. Nonetheless, advocating for a probabilistic approach is the easy part of the trip. The difficulties arise when trying to propose a proper approach to building alternative hypotheses, which comes down to generating ensembles of plausible models. Several strategies have been proposed to that aim, mainly: (1) tweaking existing deterministic models, (2) perturbing datasets before generating new models, or (3) sampling geometrical and/or kinematical parameters of geological structures. However, if all models are essentially wrong, the struggle is to generate all and only the useful ones. That being said, evaluating the efficiency of existing strategies for exploring uncertainties appears highly difficult, even in relatively simple real geological cases. In fact, only trying to imagine the complexity of the space of uncertainty, i.e., the architecture of the model space, is a real challenge. A rapid combinatorial evaluation shows that this universe of possible models is mind-bogglingly vast. This contribution proposes to address this issue by transposing the space of possible geological models into a simpler, better controlled geological universe. This surrogate universe is generated without any geological a priori but the number of different geological units considered and the resolution. A geodiversity approach, which combines Principal Component Analysis and clustering," is then applied for studying the organisation of this universe and identify families of models. We use this exploration to evaluate the capability of different uncertainty modelling approaches to explore the space of possible models when given a particular dataset. }
}