Objectifying thermal convective lift

Us glider pilots have gotten pretty good at guessing where lift should be, but it's very subjective and I think we actually rely on luck more than skill more than we want to admit.  But what if we could take the guess work out of it... The "money-ball" theory of soaring?  How could a ground heat map inform a convective thermal prediction? 

I am planning to experiment with thermal cameras on a 3-axis gyro stabilized gimbal.  I have grand plans, but first I want to run this by a engineer/pilot and set some good basic ground rules for a research project.

I can't find good concrete research on thermal-scale meteorology.  But.. from what I've learned gliding and from soaring books; here's the key hypotheses I'd like to postulate.

The tera-sourced convection trigger theory;

Given; Dry air with lots of sun and disproportionately ground IR absorption

When; regions of heat contrast form

Then; The probability and strength of thermal triggering events are proportional to the heat contrast.

So; like isobars on a prog chart predicting wind; In my mind I see heat contrast contours tells us something about thermal triggering.  But this isn't enough on it's own.  I think there's another aggregation effect that occurs as well.

The convection accumulation theory;

Given; A air volume with non-homogeneous thermal bubbles (mixing of many small thermals)

When; these bubbles bump into one another

Then; they will join to create larger/stronger thermals.

There's one more effect you mentioned to me last year that I had never though of, so I'll give you credit for this one;

The cloud vacuum effect

Given; A building cumulus cloud and an unstable air volume under it

When; the volume gets closer to the cloud base

Then; that volume of air accelerate into the cloud's vacuum cause by a latent heat of condensation effect inside the cloud.

The air volume under a cloud may not have even been rising, but will generate lift once under the cloud.  There's yet another for cloud streeting (lifted wind finds parallel lines) , but I can save it. It won't help this discussion.

So how can I image process a detailed thermal map into locations  that will likely trigger convective thermals?  If I collect lost of data, can I use machine learning to improve the odds finding thermals?

Let's say you're flying over western Kansas and you had this map...

Which field would you look for thermals over first?