Uncertainty method and visualization in stratigraphic correlations - method and example in a sequence stratigraphic setting.

A. Pons and Jonathan Edwards and Guillaume Caumon. ( 2015 )
in: 35th Gocad Meeting - 2015 RING Meeting, ASGA

Abstract

Basin analysis is based on the integration of a lot of various data and concepts in order to build geological consistent models. Data have different scales and resolutions, and their interpretation generates uncertainties that are needful to apprehend. In data integration, stratigraphic correlation is one of the first steps. It consists in the matching of stratigraphic markers identified along wells to build the geometry of the different boundaries of the structures in an area of interest. Some automatic methods have been developed to allow to have quick reproducible correlations and mitigate the interpreters bias. In this article, the method chosen is the Dynamic Time Warping algorithm (DTW). It con- sists in finding for each marker of a well the best corresponding marker on a second well by finding the least expensive path in a cost table defined by geological rules. In this article the method is included within the sequence stratigraphy. For a better consideration of uncertainties a stochastic method of correlation is needed. The n-best correlation, in other words the n least expensive models, were already the output of algorithms in previous studies. In addition, a stochastic algorithm has been implemented. The main idea is to choose a path stochastically at each step of the DTW path building instead of using the least cost. The stochastic draw follows a uniform distribution. It is also useful have a good vizualisation tool to better interpret the results and better apprehend the uncertainties. The vizualisation of the percentage of realization of the correlation of each pair of markers also permits to compare the n-best models and the stochastic methods implemented. Another way to see the uncertainties is to study the DTW map to see all the possibilities of each markers correlation.

Download / Links

BibTeX Reference

@INPROCEEDINGS{PonsGM2015,
    author = { Pons, A. and Edwards, Jonathan and Caumon, Guillaume },
     title = { Uncertainty method and visualization in stratigraphic correlations - method and example in a sequence stratigraphic setting. },
 booktitle = { 35th Gocad Meeting - 2015 RING Meeting },
      year = { 2015 },
 publisher = { ASGA },
  abstract = { Basin analysis is based on the integration of a lot of various data and concepts in order to build geological consistent models. Data have different scales and resolutions, and their interpretation generates uncertainties that are needful to apprehend. In data integration, stratigraphic correlation is one of the first steps. It consists in the matching of stratigraphic markers identified along wells to build the geometry of the different boundaries of the structures in an area of interest. Some automatic methods have been developed to allow to have quick reproducible correlations and mitigate the interpreters bias. In this article, the method chosen is the Dynamic Time Warping algorithm (DTW). It con- sists in finding for each marker of a well the best corresponding marker on a second well by finding the least expensive path in a cost table defined by geological rules. In this article the method is included within the sequence stratigraphy. For a better consideration of uncertainties a stochastic method of correlation is needed. The n-best correlation, in other words the n least expensive models, were already the output of algorithms in previous studies. In addition, a stochastic algorithm has been implemented. The main idea is to choose a path stochastically at each step of the DTW path building instead of using the least cost. The stochastic draw follows a uniform distribution. It is also useful have a good vizualisation tool to better interpret the results and better apprehend the uncertainties. The vizualisation of the percentage of realization of the correlation of each pair of markers also permits to compare the n-best models and the stochastic methods implemented. Another way to see the uncertainties is to study the DTW map to see all the possibilities of each markers correlation. }
}