Constraining stochastic 3-D structural geological models with topology information using approximate Bayesian computation in GemPy 2.1

Abstract

Abstract. Structural geomodeling is a key technology for the visualization and\nquantification of subsurface systems. Given the limited data and the resulting\nnecessity for geological interpretation to construct these geomodels,\nuncertainty is pervasive and traditionally unquantified. Probabilistic\ngeomodeling allows for the simulation of uncertainties by automatically\nconstructing geomodel ensembles from perturbed input data sampled from\nprobability distributions. But random sampling of input parameters can lead to\nconstruction of geomodels that are unrealistic, either due to modeling artifacts\nor by not matching known information about the regional geology of the modeled\nsystem. We present a method to incorporate geological information in the\nform of known geomodel topology into stochastic simulations to constrain\nresulting probabilistic geomodel ensembles using the open-source geomodeling\nsoftware GemPy. Simulated geomodel realizations are checked against topology\ninformation using an approximate Bayesian computation approach to avoid the\nspecification of a likelihood function. We demonstrate how we can infer the\nposterior distributions of the model parameters using topology information in two\nexperiments: (1)\xa0a synthetic geomodel using a rejection sampling scheme\n(ABC-REJ) to demonstrate the approach and (2)\xa0a geomodel of a subset of the\nGullfaks field in the North Sea comparing both rejection sampling and a\nsequential Monte Carlo sampler (ABC-SMC). Possible improvements to processing\nspeed of up to 10.1\xa0times are discussed, focusing on the use of more advanced\nsampling techniques to avoid the simulation of unfeasible geomodels in the first\nplace. Results demonstrate the feasibility of using topology graphs as a summary\nstatistic to restrict the generation of geomodel ensembles with known\ngeological information and to obtain improved ensembles of probable geomodels\nwhich respect the known topology information and exhibit reduced uncertainty\nusing stochastic simulation methods.\n