wilmott.com

So, here is the pitch.

- Domain is oil & gas exploration
- It is all satellite imagery and machine learning-based
- A new satellite image is available every 12 days
- Provider claims it can detect whether exploration at a site has been 'successful'
- Provider adds they have a rough idea of the magnitude of the new oil & gas field discovered

Any ideas on how do they do that?

I think all this technology has been know since the Telstar era, yes?
"not because they are easy, but because they are hard."

Is it information that can be extracted from the images? maybe you see lots of cheering engineers?

There are also companies that have high res live video instead of images every 12 days http://googlesightseeing.com/2007/04/li ... gle-earth/

http://www.dailymail.co.uk/sciencetech/ ... g-off.html

I don't know. Possibly down to the holes drilled and the landscape deformation post the survey explosion? The surveys are certainly available to see on google earth. The 12 days is the satellite revisit time I guess.

If it's pure machine learning, then they did this:
1. collected time series satellite images of wells with known success/failure and production flow outcomes
2. use your favorite supervised learning method to build a model that estimates these outcome variables
3. Profit!

As to how satellite images have enough latent information to judge well success and magnitude of a well, I'd guess it involve changes in the appearance of the well such as:

1. Drilling Phase: the image shows a drilling rig and the vehicles of the roughnecks.
2. End Phase: Failure: the drill rig disappears, the vehicles disappear, vegetation around the wellhead and access road regrows
3. End Phase: Success: the drill rig disappears, a pumping station and tank(s) appear, vegetation around the wellhead and access road remains somewhat suppressed
4. Flow estimation: pumping station size, tank size & number, amount of vegetation suppression induced by repeated tanker-truck or maintenance visits.

It'll be more sensible to do that by human. It's not like there are 1000 drillings rigs popping up ever 12 days. Maybe they use ML to find drilling activity locations and then forward the images to an analyst?

What does the provider charge for this service?? It's an interesting business idea. Build the infrastructure to collect satellite images. Build detector model for various types of objects. Phone tagoma and sell this info for 2k/month?

as others have said, success or failure can be guessed from the activity at the scene.
As to the size of the field, I was taught seismology long ago in college, but there are more recent techniques.
Google gave me this from 2009,
New Oil Deposits Can Be Identified Through Satellite Images
in that article, they look at the earth's gravitational field.
I also vaguely remember seeing a talk by someone who claimed you could do it as an inverse problem using EM fields.

The point would be that they are not necessarily looking at visible spectrum photographs (for want of a better term, and I'm sure that is not the correct term). They may be looking at other data like gravitational field, heat signatures, EM field, or something else altogether

one more thought re: gauging success
when you strike oil or gas, what exactly happens? maybe certain hydrocarbons (or other chemicals) are spewed out into the atmosphere and perhaps these can be detected by a satellite?

It'll be more sensible to do that by human. It's not like there are 1000 drillings rigs popping up ever 12 days. Maybe they use ML to find drilling activity locations and then forward the images to an analyst?

But how accurate and repeatable is the human? Humans are terribly inconsistent and easily biased by recent patterns?

And although there may not be 1000 new rigs every 12 days, there probably are a few thousands drilling rigs and recently drilled wells world wide that need evaluation/reevaluation every 12 days. (The inflight magazine of the airline we most often fly has numerous ads for drilling rigs and oil field services)

The rigs tend to be in inhospitable places.

It'll be more sensible to do that by human. It's not like there are 1000 drillings rigs popping up ever 12 days. Maybe they use ML to find drilling activity locations and then forward the images to an analyst?

But how accurate and repeatable is the human?  Humans are terribly inconsistent and easily biased by recent patterns?

And although there may not be 1000 new rigs every 12 days, there probably are a few thousands drilling rigs and recently drilled wells world wide that need evaluation/reevaluation every 12 days.  (The inflight magazine of the airline we most often fly has numerous ads for drilling rigs and oil field services)

You're flying B52 or something?? My airline has  sunglasses, perfume and peanut ads.

It'll be more sensible to do that by human. It's not like there are 1000 drillings rigs popping up ever 12 days. Maybe they use ML to find drilling activity locations and then forward the images to an analyst?

But how accurate and repeatable is the human?  Humans are terribly inconsistent and easily biased by recent patterns?

And although there may not be 1000 new rigs every 12 days, there probably are a few thousands drilling rigs and recently drilled wells world wide that need evaluation/reevaluation every 12 days.  (The inflight magazine of the airline we most often fly has numerous ads for drilling rigs and oil field services)

You're flying B52 or something?? My airline has  sunglasses, perfume and peanut ads.

It's Southwest Airlines.

US airline inflight magazines tend to have a mix of consumer stuff (jewelry, matchmaking services, hair replacement & plastic surgery, steakhouse restaurant chains) but also a mix of industrial products/services (oil field stuff, high-efficiency light bulbs for factories & warehouses, injection molding services, portable projectors, enterprise software).

But how accurate and repeatable is the human?  Humans are terribly inconsistent and easily biased by recent patterns?

And although there may not be 1000 new rigs every 12 days, there probably are a few thousands drilling rigs and recently drilled wells world wide that need evaluation/reevaluation every 12 days.  (The inflight magazine of the airline we most often fly has numerous ads for drilling rigs and oil field services)

You're flying B52 or something?? My airline has  sunglasses, perfume and peanut ads.

It's Southwest Airlines.

US airline inflight magazines tend to have a mix of consumer stuff (jewelry, matchmaking services, hair replacement & plastic surgery, steakhouse restaurant chains) but also a mix of industrial products/services (oil field stuff, high-efficiency light bulbs for factories & warehouses, injection molding services, portable projectors, enterprise software).

So they are targeting the Tony Starks of this world: flamboyant, vain, billionaire scientist/engineers.