How data analysis and 1D models help understand messy vitrinite data

In my last column, I hinted at a case study, where I was able to inter­pret likely fluid migra­tion based on temperature and vitrinite data in a single well. This study was a high­light of my career, in which our team could work with pretty much all of the data available in the basin.

At first, the vitrinite data were all over the place. A colleague from one op­erator company told me he just could not make any sense of those data in their license block and had to ig­nore them. We had the ben­efit of a much wider dataset and a wider context.

It turned out, we had reports for some wells that included original reflec­tance measurements with statistics and photographs for a number of wells. We had kerogen organoparti­cle analyses. We even had overlapping VRo data from two and even three differ­ent labs. Even though the data volume seemed over­whelming and messy I had a feeling we had a chance to make something out of it. So I decided to ignore the reported interpretations and dived into the files. I start­ed organizing and typing all data in spreadsheets and began cross checking all the different pieces against each other. An incredibly tedious and time consuming effort. But it paid off.

Gradually, I started see­ing patterns and relation­ships. Some measurements reported as VRo started to show up as clearly measured on different particle types. I could see which data were more reliable. I found that two of the labs consistently reported reliable or at least reasonable data while the third lab consistently report­ed incorrect data. So in the wells where I did not have underlying measurements, I could flag the reported data based on the lab of origin.

I also did similar de­tailed analysis on temper­ature and RockEval data. Gradually trimming the noise and high-grading the dataset.

Then I proceeded to build 1D thermal models. I defined fairly detailed li­thology columns from the well logs. After I calibrated the first model, I contin­ued building others with similar heatflow settings. The results just made my eyes pop! One model after another came out with an excellent fit to temperatures and VRo that I never saw again.

The models did not only serve as a temperature history estimate for matu­rity calculations, they were important tests of the data integrity and overall geolog­ical model validity. It also confirmed my previous ex­perience that the more mod­ern EasyRo%-DL model is the best one to calculate VRo in a basin model.

The story shows just how important it is to have original and overlapping data, to share data among companies and not to rush through the model build­ing. Take your time, invest a little bit and you’ll be re­warded!