In-context learning enables continental-scale subsurface temperature prediction from sparse local observations

TL;DR

This paper introduces In-Context Earth, a transformer-based model that predicts subsurface temperature fields at continental scale using sparse borehole data, outperforming traditional models and adapting to new regions without retraining.

Contribution

The work demonstrates that in-context learning with transformers can effectively utilize sparse geological observations for large-scale subsurface temperature prediction, with calibrated uncertainty and interpretability.

Findings

Achieves a mean absolute error of 4.7°C in the US, outperforming existing models.

Maintains high accuracy in new regions with minimal local observations.

Learns internal representations of subsurface properties without direct observation.

Abstract

Continental-scale knowledge of subsurface temperature is limited by the cost and sparsity of borehole measurements, but such information is essential for geothermal resource assessment and for understanding heat transport in the shallow crust. The thermal field reflects the interaction between lithology, crustal structure, radiogenic heat production, and advective fluid flow, sometimes producing sharp anomalies that are smoothed by conventional interpolation or difficult to capture with physical models. Here we introduce In-Context Earth, a transformer-based model that uses sparse local borehole observations as geological context to predict continuous temperature-at-depth fields with calibrated uncertainty. In the contiguous United States, the model achieves a mean absolute error of 4.7 {\deg}C, outperforming the physics-informed Stanford Thermal Model, a model based on AlphaEarth embeddings, the multimodal Transparent Earth model, and universal kriging, while resolving sharper thermal gradients in geothermal provinces. Its uncertainty estimates are well calibrated, with a Kolmogorov-Smirnov statistic of 2.5%. Without finetuning, the model adapts to Alberta, Australia, and the United Kingdom (UK) using only 20 local observations at inference time, maintaining high accuracy in geologically distinct test regions with a mean absolute error of 2.2 {\deg}C in Alberta, 6.2 {\deg}C in Australia, and 5.4 {\deg}C in the UK. Interpretability analyses show that the model learns internal representations of subsurface properties it never observes during training, including seismic velocities, geochemistry, and crustal structure, and uses these representations in physically consistent ways. More broadly, this work shows that in-context learning can use sparse borehole observations for continental-scale subsurface characterization, without requiring dense measurements or region-specific retraining.