JoulesBurn, I love your posts. Remote sensing is my specialty and I'm thrilled to see it applied to peak oil in such an intelligent way.

I'm having trouble understanding their automatic feature detection (template matching perhaps), the approach is necessarily semi-supervised with potentially a lot of false alarms/false positive that need to be filter out manually. You cannot blindly applied an automatic algorithm without double checking the results. In particular, the inclusions of high voltage towers is quite telling and prove that Bernstein work was done by amateurs and should have been done by hobbyists instead :).

They don't go into much detail. This is from their first report:

However, a black rectangle feature was found to be present at nearly all production sites. This feature was been selected as an indication for production and drilling sites using ASTER data as it is visible in these datasets, as well.
...
It was decided to digitize black features on the ASTER images for 2004/7, while in parallel checking their
validity as indication on oil production sites in GoogleEarth.

Part of the problem is that ASTER is B&W and color differences at wells vs. the surroundings are not apparent.

JB

I'm a SAR-specialist myself. The features they've detected are most probably some stable objects/terrain features (so-called "Persistent Scatterers" or PS in differential SAR interferometry) that reflect incident radiation well back towards the radar. A high-voltage tower can easily make an excellent PS that can be used to track terrain motion.

You are exactly correct however it is not clear if those "objects" are used as PS only or also used in order to draw conclusions about the evolution in drilling activity.

It's impossible to say without seeing the report - I'd like to know what the "activity density" is all about.