by Xiang Huang 8 May 2026

Maps of the (a) study region and (b) locations of the DTS and DAS cables and borehole; and (c) a photo of the embankment road in Utqia?vik, Alaska. These observations were used to build and test a digital twin of near-surface permafrost conditions. Credit: Gou et al. [2026], Figure 4

Editors’ Highlights are summaries of recent papers by AGU’s journal editors.

Source: Journal of Geophysical Research: Earth Surface

Permafrost beneath Arctic roads is warming and becoming less stable, creating growing risks for northern infrastructure. Yet predicting how frozen ground will evolve remains difficult because subsurface conditions vary sharply over short distances, observations are sparse, and conventional process-based models are not easy to update as new field data arrive. In a new study, Gou et al. [2026] address that challenge at an embankment road in Utqia?vik, Alaska, using fiber-optic temperature measurements collected along a 100-meter transect to track how shallow ground conditions change through time. Rather than treating monitoring and modeling as separate tasks, the authors link them in a framework designed to evolve with the physical system itself.

What stands out here is not simply the use of machine learning, but the way the authors build a physics-informed digital twin for permafrost under infrastructure. Their framework embeds a neural network within a heat-transfer solver, so the governing physics remain central while the model can still update uncertain soil properties as new observations arrive. This study moves beyond black-box prediction toward an interpretable, updateable system that can reconstruct subsurface temperature fields, infer thermodynamic properties such as unfrozen water content and thermal conductivity, and then test those inferences against independent DAS data, borehole temperatures, and laboratory measurements. This makes the work more than a site-specific modeling exercise; it offers a credible pathway toward near-real-time permafrost forecasting and infrastructure monitoring in a rapidly warming Arctic.

Framework of the proposed digital twin model. The neural network (NN) takes soil temperature at each lateral position as input and outputs six unknown parameters that vary laterally with distance. These parameters are embedded in the heat?transfer equation through constitutive relationships, and the resulting system is solved using a finite difference method (FDM). The difference between predicted and observed temperatures is computed and defined as “loss,” and the loss gradients are backpropagated to update the NN parameters. Credit: Gou et al. [2026], Figure 2

Citation: Gou, L., Xiao, M., Zhu, T., Martin, E. R., Wang, Z., Rocha dos Santos, G., et al. (2026). Physics-informed digital twin for predicting permafrost thermodynamic characteristics under an embankment road in Utqia?vik, Alaska. Journal of Geophysical Research: Earth Surface, 131, e2025JF008787. https://doi.org/10.1029/2025JF008787

—Xiang Huang, Associate Editor, JGR: Earth Surface