On the influence of the correlation path in incremental multi-well stratigraphic correlation

in: International Geostatistics Congress 2024, A Springer book series Quantitative Geology and Geostatistics, pages 12 pp, Springer

Abstract

Stratigraphic correlation between several one-dimensional borehole sections is generally achieved by expert-based interpretation and sedimentological reasoning to produce a best-case scenario. To gain productivity and make the process reproducible, automation methods defined since the 1980's have mainly translated the correlation (or alignment) problem as an optimization task. Therefore, stratigraphic layering, which determines rock unit volumes, connectivity and ways to measure distances in geomodels, is most often deterministic. However, the stratigraphic correlation problem is highly non-unique, because of sparse sampling, lateral variability, complex interactions between tectonic and sedimentary processes and because the types of data and the stratigraphic principles are many. In this work, we discuss some of the computational strategies to sample this uncertainty. N-best solutions are a possible option for pairwise correlations but become computationally intractable when the number of stratigraphic sections increases. Moreover, they explore the space of solutions only locally around the optimum. In multi-well correlation, finding the global optimum itself is highly challenging when the number of sections and samples is large. We use a hierarchical and incremental correlation which guarantees the consistency of the solution and naturally prevents crossings. However, the correlation path has an impact on the solution, so we propose strategies to randomize this path based on well distances. This is achieved either using geographic distances between boreholes or distances coming from local pairwise correlations. We test these strategies on a synthetic benchmark data set generated from a forward stratigraphic model, and we evaluate the effect of the randomization strategies on the produced correlation outcomes. Finally, we discuss the implications of this work both for finding the optimal correlation and for the generation of a diverse but acceptable set of scenarios.

Download / Links

BibTeX Reference

@inproceedings{caumon:hal-05066225,
 abstract = {Stratigraphic correlation between several one-dimensional borehole sections is generally achieved by expert-based interpretation and sedimentological reasoning to produce a best-case scenario. To gain productivity and make the process reproducible, automation methods defined since the 1980's have mainly translated the correlation (or alignment) problem as an optimization task. Therefore, stratigraphic layering, which determines rock unit volumes, connectivity and ways to measure distances in geomodels, is most often deterministic. However, the stratigraphic correlation problem is highly non-unique, because of sparse sampling, lateral variability, complex interactions between tectonic and sedimentary processes and because the types of data and the stratigraphic principles are many. In this work, we discuss some of the computational strategies to sample this uncertainty. N-best solutions are a possible option for pairwise correlations but become computationally intractable when the number of stratigraphic sections increases. Moreover, they explore the space of solutions only locally around the optimum. In multi-well correlation, finding the global optimum itself is highly challenging when the number of sections and samples is large. We use a hierarchical and incremental correlation which guarantees the consistency of the solution and naturally prevents crossings. However, the correlation path has an impact on the solution, so we propose strategies to randomize this path based on well distances. This is achieved either using geographic distances between boreholes or distances coming from local pairwise correlations. We test these strategies on a synthetic benchmark data set generated from a forward stratigraphic model, and we evaluate the effect of the randomization strategies on the produced correlation outcomes. Finally, we discuss the implications of this work both for finding the optimal correlation and for the generation of a diverse but acceptable set of scenarios.},
 address = {Ponta Delgada A{\c c}ores, Portugal},
 author = {Caumon, Guillaume and Antoine, Christophe},
 booktitle = {{International Geostatistics Congress 2024, A Springer book series Quantitative Geology and Geostatistics}},
 doi = {10.1007/978-3-031-92870-3},
 editor = {Leonardo Azevedo et al.},
 hal_id = {hal-05066225},
 hal_version = {v1},
 keywords = {Multiple sequence alignment ; Uncertainty ; Dynamic Time Warping},
 month = {September},
 pages = {12 pp},
 pdf = {https://hal.science/hal-05066225v1/file/Caumon_ExtAbstract_CorrelationPath_GEOSTATS2024_submitted.pdf},
 publisher = {{Springer}},
 series = {Quantitative Geology and Geostatistics},
 title = {{On the influence of the correlation path in incremental multi-well stratigraphic correlation}},
 url = {https://hal.science/hal-05066225},
 volume = {20},
 year = {2024}
}