Cluster Analysis Based Subdivision of the Clinton Sanstone in East Canton Ohio

Abstract

The Silurian-aged Clinton Sandstone reservoir has been a prolific oil and gas producer for fields in eastern Ohio. Conventional subdivision of the Clinton sandstone into subunits of Clinton 1–5 is often challenging in building reservoir models, hence statistical means was employed for the subdivision process.

Wireline data (Gamma ray, bulk density, and neutron porosity) for 25 wells within an 87 square mile study area in East Canton Ohio were compiled. The Clinton Sandstone was initially subdivided into 5 subunits (Clinton 1 – 5) and four flooding surfaces based on examinations of wireline data and whole core. A cluster analysis was conducted using two common algorithms in Rgui, Mclust and Kmeans to statistically determine the subdivisions of the Clinton Sandstone for static earth and reservoir modeling. Mclust applies several model based clustering algorithms and chooses the method which best fits the data. KMeans minimizes the sum of squares for each data point and clusters them around a center mean for the defined number of clusters.

Gamma ray and bulk density data best represented the variations between the Clinton Sandstone layers and were used in the cluster analysis. The cluster analysis determined that the Clinton Sandstone fell into three statistically different clusters. Cluster 1 contained Clinton 5, 2, and 1 with elevated gamma ray and density. Clinton 3 and 4 fell into separate clusters representing the main sandstone deposits in the depositional sequence. Cluster analysis was successful in determining appropriate subdivisions for the use in static earth and reservoir modeling.

AAPG Datapages/Search and Discovery Article #90218 © 2015 Eastern Section Meeting, Indianapolis, Indiana, September 20-22, 2015