How Bird Geophysical Uses Machine Learning for Heat Flow Prediction

Bird Geophysical uses machine learning to integrate magnetic data with thermal properties in order to predict heat flow. The approach is described on the company's site as combining magnetic measurements with thermal information through machine learning algorithms, and it is applied within the company's consulting work in geothermal and carbon capture, alongside petroleum exploration. This is a data-integration method for estimating subsurface heat flow, not a standalone software product or a published, downloadable model.

What the method combines

The core idea is to bring two different data types into one prediction:

• Magnetic data — measurements of the Earth's magnetic field, which reflect variations in the magnetization of rocks below the surface.
• Thermal properties — the heat-related characteristics of subsurface materials.

Bird Geophysical's site notes that the Earth's gravity and magnetic fields can be measured with extreme accuracy — often better than 1 ppm and 20 ppm respectively — and that density and magnetization changes in rocks produce gravity and magnetic anomalies. Careful interpretation of those anomalies allows subsurface geology to be inferred. The heat flow work extends this logic: magnetic information is used as one input, thermal properties as another, and machine learning ties them together to produce a heat flow prediction.

Why machine learning is used here

Heat flow is not measured everywhere. Direct thermal measurements are sparse and expensive, while magnetic data coverage is often broader. A machine learning model can learn the relationship between the two where both are known, then apply that relationship where only magnetic data exists.

This matters for:

• Geothermal exploration — identifying areas with higher heat flow potential.
• Carbon capture — understanding subsurface thermal conditions at candidate storage sites.
• Petroleum exploration — thermal history affects source rock maturity and basin modeling.

What this is and is not

The site does not publish the specific algorithms, training datasets, accuracy figures, or pricing. Anyone needing those details would have to ask directly.

How to evaluate whether this fits your project

If you are considering this type of work, the practical questions are:

1. Do you have magnetic data over the area of interest? The method depends on magnetic coverage as an input.
2. Are there enough thermal measurements nearby to train or calibrate a model? Without paired data, the machine learning step has nothing to learn from.
3. Is heat flow prediction the actual deliverable you need? If your goal is structural mapping or basin architecture, gravity and magnetic interpretation alone may be the more direct route.
4. What scale and setting? Bird Geophysical reports interpretation experience across platform, rift, forearc, and foreland basins on every continent, including Indochina, Southeast Asia, Indonesia, passive margins along both flanks of the Atlantic, east Africa, India, China, Australia, and New Zealand.

Background on the provider

Bird Geophysical was started in 1997 by Dale Bird, who works in Houston, Texas. The company consults with companies in petroleum exploration, geothermal, and carbon capture; interprets gravity and magnetic data; conducts training courses; and applies machine learning to integrate magnetic data with thermal properties for heat flow prediction. Contact is listed as [email protected].

For a specific project, the most reliable next step is to describe your data coverage and target area directly and ask what inputs the heat flow workflow would require.