Abstract

Elemental datasets are increasingly being used to model or act as proxies for miner-alogical datasets and depositional environments. Just as elements such as Ni, Mo and U are commonly used as a proxy for TOC, there is an increasing application whereby elemental data are used to model mineralogy. Here we look at the pitfalls and also the advantages of this approach.

Acquiring elemental data from cuttings samples in unconventional plays has become almost routine analysis. The advantages of collecting elemental data are numerous: data collection is quick and cost effective; instrumentation allows good quality data to be gathered at well-site; data acquired are absolute values and require little interpretative processing; the data provides information on lithofacies, paleoredox, TOC and aids with sweet spot identification.

It is also possible to comment on clay, quartz and carbonate content based on Al2O3, SiO2 and CaO concentrations, further enhancing the usefulness of elemental data. However, elemental data are being used to calculate “bulk” mineralogy using a stoichiometric approach. This approach can and does provide meaningful mineralogical data, but without some knowledge of the system you are working with and careful calibration this approach can also provide highly erroneous results.

In this presentation we will demonstrate how using two different stoichiometric models will produce different results from the same elemental data and how those results would impact on looking at rock brittleness. Additionally we will also discuss some techniques to decide which model is nearest to the “true” mineral composition and also demonstrate possible alternative methods using unsupervised machine learning statistics to generate mineralogy from elemental datasets.