We have geothermal, engineering and construction clients who often need accurate estimates of ground temperatures for project sites before they have physical access to the sites (for feasibility studies, basic system design, etc.).

We have built a hybrid Machine Learning and physics-based model, which we've made available to the public at groundtemperatures.com, to provide these estimates for our clients.


I am just wondering if any of you have found other methods for estimating ground temperatures for locations that you do not have physical access to and for which there are no nearby geo-tech/borehole studies?

Our model is generally accurate to 0.5 degrees Celcius globally, from depths between 0.4m and 200 meters, when compared to real-world ground temperature measurements. So obviously we would be looking for something more accurate than that, which can handle similar depth ranges and also accounts for the time of year (seasonal variability), which our model also does.

  • $\begingroup$ I noticed your associated question at Matter Modeling: Soil Modelling: Getting accurate Thermal Conductivity values. I'm assuming a variography study was undertaken of the thermal data for the kriging part of the analysis & that the spacing of the thermal data points was sufficiently close for good variography results. If not you might need a greater data density. If that's not the issue then you might need to look at different kriging techniques. $\endgroup$
    – Fred
    Sep 27, 2021 at 4:38


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.