--> A new approach to refine and quality control correlations in shale and siltstone formations based on Principal Component Analysis of XRF data

2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

A new approach to refine and quality control correlations in shale and siltstone formations based on Principal Component Analysis of XRF data

Abstract

Correlating stratigraphic surfaces and parasequence sets at regional scale in fine-grained marine deposits is challenged by limited lithological contrasts and commonly sparse biostratigraphy data. Recent publications on the Montney Formation of Western Canada demonstrate that despite a large body of work and a high density of well control, uncertainties remain even for major stratigraphic surfaces such as the Triassic Dienerian-Smithian third-order sequence boundary. A new workflow has been devised to help the geologist refine and quality control his correlation. The new approach has been first tested on a very detailed chemostratigraphy scheme established between two Montney cored wells (1,310 and 785 feet of cores) distant of some 5 miles in the basin dip direction. The units have been defined by sudden and substantial changes of the Ca/Mn ratio. The results, for the interval common to both wells, have been quality controlled by a new Principal Component Analysis approach comparing the Eigen Values of the first principal component (ev1) of all 26 elements measured with the ITRAX core scanner. Perfect match (i.e. Rsquared >90%) was achieved when comparing the same units in both wells but not when comparing successive units in any of the two wells. The various elements are grouped into 3 categories (carbonate, clastics and TOC linked elements) based on their affinity as seen by a normal PCA analysis (EV1 vs EV2). The regression lines of the ev1 per category can then be compared between units and used to assess the similarity between them (successive units in a single well or same units in different wells). Comparison between successive units exhibits changes in slopes of the category-based regression lines whereas no or very minor changes in slopes are seen in the same unit of the two distant wells. Practical applications of this new PCA approach using EV1 comparison was then performed on vertical and horizontal wells with cuttings analyzed every 5 m to 20m in the Montney, Duvernay, Lorraine and Utica Formations. Our analysis demonstrates that multiple ITRAX measurements per cutting vial provides a better sampling of lithological heterogeneity and a useful tool to refine or QC well to well correlations.