Ground Temperature Predictor

Accuracy Estimates

The temperature difference (or error) between measured and predicted values are shown in the bar graph as average and maximum errors for the depths of 0.4 - 180 m. Validation results are divided into the following four geographic zones:
arid desert accuracy iconnorthern arctic accuracy iconcities and urban areas iconall other ground on earth
*Excluding depths below lakes and oceans, and Antartica. These areas are not able to be predicted for with current data.
** Areas with less than 200 mm average annual rainfall that are also south of 50° Latitude.

Accuracies by Location
(0.4-180m depth)

error chart
depth below soil surface accuracy

Download the raw data from our Validation Study:

View map of locations we tested in our Validation Study:

More Information

This Ground Temperature Predictor (GTP) application is a hybrid physics and machine learning software.
Using datasets collected from governments and researchers around the globe, with total datapoints in the millions, the algorithm was trained with Artificial Intelligence (AI)/Machine Learning (ML) using extensive hyperparameter searching, statistical data processing, and traditional thermodynamic modeling to ensure robust data projections.
GTP leverages large quantities of data for key geospatial, topographical, geophysical, thermophysical, climatic, and human impact variables to generate predictions.
This software has been thoroughly tested against measured data and is comparable to traditional calculation methods for ground temperature when the thermal properties of the ground in question have already been measured.
When the values of ground thermal diffusivity/conductivity (or historical climate data) have not been directly measured, then the GTP far outperforms traditional methods, and not only in terms of accuracy: acquiring ground temperature estimates with this tool is many times faster and easier than conventional methods.
This software is continuously updated by the experts at Umny Inc. to further improve its accuracy and speed. If you have questions, feedback, or experimental data you would like to test, please feel free to send us an email.

All Ground - Medium Depth

ground temperatures soil accuracy

Urban Areas - Medium Depth

urban cities villages construction ground temperature

Arctic Areas - Medium Depth

arctic permafrost temperatures accuracy

True-Desert Areas - Medium Depth

arid desert sand and rock temperatures accuracy
umny engineering logo
Built with Umny AI/Machine Learning.