Despite their public stance, Exxon’s internal science team had a clear view on climate change.

Ars Technica looks at the way Exxon executives decided to bury their own company’s very accurate findings on petroleum and climate:

Exxon’s scientific climate work was shut down a long time ago. But the new paper relies on the work of two science historians (Geoffrey Supran and Naomi Oreskes), who dug through old documents and a number of peer-reviewed papers to identify cases where Exxon’s scientists made projections about how carbon dioxide emissions might alter the future climate. In several cases, we know that the graphs that they found were part of documents that were sent to Exxon executives.

The historians collaborated with a noted climate scientist (Stefan Rahmstorf), who then compared their projections to the behavior of the actual climate. This allowed the team to determine whether the projections made internally at Exxon were in line with those produced by academic climate scientists and whether they were “skillful,” meaning that they were in line with what the actual climate ended up doing.

Overall the historians came up with 16 different climate projections produced by Exxon scientists between 1977 and 2003. For most of this era, computer power wasn’t sufficient to run the sort of complete-atmosphere climate models that are used today. Instead, the models they used were generally physics-based energy distribution models that tracked things like where and how much radiation gets absorbed, and how air circulation moves that energy around.

Nevertheless, the projections of future warming done by Exxon scientists were remarkably good. Most of them (nearly two-thirds) had error bars that overlapped with the errors in the temperature record. And a couple of the exceptions were simply due to graphs that lacked error bars, narrowing down the potential to overlap with anything considerably. Two of the remaining ones also forecast more warming than actually occurred.

A climate projection’s skillfulness is a measure of how closely it agreed with the historic record. And again, Exxon scientists performed well. The aggregate skillfulness of their internal climate models is over 70 percent. By that measure, they outperformed contemporary models from the scientific community.

You can read more of the analytical team’s findings here, in Science.