Chemostratigraphic Automations: Chemostratigraphic Data Analytics and Interpretations

Summary Chemostratigraphy and the interpretation of geological boundaries is a very time consuming and multivariate and multidimensional problem. Standard manual workflows in chemostratigraphy can be fully automated, reducing uncertainty in the interpretation and use of all the data. This paper presents a new workflow that is based on a five-step approach that automates this process. It begins with data QCing and Principle Component analysis that discards the bad data and creates new set of a few new variables that include all the available variability. It then performs Statistical picking and a Quartile Analysis, that picks stratigraphic tops and identifies chemical components that are important for the correlation. Through this process the zones and robust variables are used for training through mashine learning and Discriminant Function Analysis (DFA). The classifiers are used for training and validation of the models, prediction and correlation of stratigraphic boundaries in uninterpreted wells. The approach was used on an eleven well geological study and produced similar results to those of the standard workflow in minutes rather than weeks. This method, used in combination with the standard workflow, can deliver more robust and accurate geological correlations and assistance for the correlation when there is uncertainty.