Sjur Barndon

Mountain glacier evolution since the last interglacial remains poorly constrained, with limited spatial and temporal coverage. Conventional modelling approaches typically target major climatic events, operate at coarse spatial resolution over limited spatial domains, or employ simplified representations of ice dynamics. Here, we address these limitations by applying the Instructed Glacier Model (IGM), to reconstruct, for the first time, mountain glacier evolution since ∼130 ka at 500 m resolution across nine mountain ranges in North America, South America, Eurasia, and Africa. We perform 707 parameter-calibration simulations by varying paleoclimate and ice-dynamic parameters, and validate model performance by assessing glacier extent and ice thickness. The model outputs are evaluated using a spatial frequency map approach, which identifies a set of acceptable model results rather than a single best-fit simulation. As a result, we identify areas of robust agreement and those sensitive to parameter choices, providing a systematic way to visualise spatial uncertainty and glacier–climate-topography interactions. Together, our framework is a scalable foundation for next-generation, uncertainty-aware reconstructions turning glacier modelling at orbital-timescales into a reproducible, expandable workflow that can be deployed across mountain ranges worldwide. All data are publicly available at https://archive.sigma2.no/dataset/evolution-of-mountain-glaciers-since-the-last-interglacial (Barndon et al., 2026b) and the animations at https://av.tib.eu/series/2022 (Barndon et al., 2026a).

The dynamic behaviour of ice-sheet motion over rough landscapes is poorly understood, with most ice-sheet models prescribing a bed smoother than reality, which does not fully capture topographic features. Subglacial fjords striking obliquely to the palaeo flow direction are an end member of this misrepresentation, but are ubiquitous beneath the western margin of the palaeo Scandinavian Ice Sheet, and provide a useful proxy for areas of the present-day Greenland Ice Sheet. Here, we consider Veafjorden as a characteristic western Norwegian fjord where striations clearly evidence palaeo perpendicular ice flow, and perform 3D thermodynamically-coupled ice-motion simulations across a range of orientations. For perpendicular flow, Moffatt eddies, or spiralling flows, occur within a thick layer of temperate ice in the fjord hollow with reverse-direction slip at the fjord base. Area-averaged driving stress in simulations with high-resolution topography and fjord-perpendicular flow is ∼41 %–89 % greater than in simulations with smoothed control topography for an equivalent surface velocity. In comparison, simulations with fjord-perpendicular flow show ∼28 %–45 % greater area-averaged driving stress than simulations with fjord-parallel flow. The steep slopes of fjords and other similar features also provide a clear physically-based example for why bounded basal traction relationships may not hold at the macro scale in rough settings. Similar topographic features may explain surface velocity variations at many locations towards the margins of the Greenland Ice Sheet, and imply that the role of anisotropic roughness in resisting ice-sheet motion may be under-represented in models.

For most glaciated areas, detailed mountain glacier evolution since the last interglacial is largely unknown. Due to limitations of traditional numerical modelling, previous studies have typically operated at a coarse spatial resolution, limited study area size, or focused on major climate events. Here we address these limitations by applying the Instructed Glacier Model (IGM), a deep-learning ice-flow emulator enabling efficient GPU-accelerated transient simulations. We model mountain glacier evolution since 130 ka and up to the present-day at 500 m resolution across eight broad mountain ranges in North America, South America, Eurasia, and Africa. Paleoclimate variables are approximated and regionalised using a combination of global and regional climate proxy datasets. We perform 617 parameter-calibration experiments varying paleoclimate and ice-dynamic parameters, with an average runtime of 21 hours per experiment. Model performance is assessed using combined areal and volumetric validation at two known glacial states; the last glacial maximum and the present-day. We also introduce a reproducible probabilistic model-evaluation framework combining confusion matrix validation score and simulation rank to identify sets of acceptable model parameters rather than a single best-fit solution. Our results show that IGM can model realistic ice-flow patterns, glacier geometries, and transient evolution across full glacial-interglacial cycles, demonstrating that machine-learning models of ice dynamics generalise to new domains and conditions, although performance can decline at coarser spatial resolutions. Together, these results demonstrate the feasibility of global-scale, high-resolution, transient glacier modelling over orbital timescales using a deep learning instructed model, while providing a 100-year interval dataset including glacier extent, ice thickness, and ice flow patterns for the last glacial cycle.

The thermodynamic behaviour of ice-sheet motion over rough landscapes is poorly understood, with most ice-sheet models prescribing a bed smoother than reality, which will not fully capture topographic features. Subglacial fjords striking obliquely to the palaeo flow direction are an extreme case, but are ubiquitous beneath the western margin of the palaeo Scandinavian Ice Sheet, and likely provide a useful proxy for areas of the present-day Greenland Ice Sheet. Here, we consider Veafjorden as a characteristic western Norwegian fjord where striations clearly evidence palaeo perpendicular ice flow, and perform 3D thermodynamically-coupled ice-motion simulations across a range of orientations. For perpendicular flow, surface velocity above the fjord is reduced substantially while a thick layer of temperate ice occupies the fjord. Moffatt eddies, or spiralling flows, occur in the fjord hollow with reverse-direction slip at the fjord base. When compared to smoothed topography, perpendicular flow over real topography requires ∼41–89 % greater area-averaged driving stress, dependent on the overlying ice thickness, while a switch from fjord-parallel to fjord-perpendicular flow for real topography requires a ∼28–45 % area-averaged driving-stress increase. These results may explain surface velocity variations at many locations towards the margins of the Greenland Ice Sheet, and imply that the role of anisotropic roughness in resisting ice-sheet motion may be significantly underappreciated. Last, we note that deep fjords provide a clear physically-based example for why bounded basal traction relationships (i.e. regularised-Coulomb) may not hold at the macro scale in rough settings.

The thermodynamic behaviour of ice-sheet motion over rough landscapes is poorly understood, with most ice-sheet models prescribing a bed smoother than reality, which will not fully capture topographic features. Subglacial fjords striking obliquely to the palaeo flow direction are an extreme case, but are ubiquitous beneath the western margin of the palaeo Scandinavian Ice Sheet, and likely provide a useful proxy for areas of the present-day Greenland Ice Sheet. Here, we consider Veafjorden as a characteristic western Norwegian fjord where striations clearly evidence palaeo perpendicular ice flow, and perform 3D thermodynamically-coupled ice-motion simulations across a range of orientations. For perpendicular flow, surface velocity above the fjord is reduced substantially while a thick layer of temperate ice occupies the fjord. Moffatt eddies, or spiralling flows, occur in the fjord hollow with reverse-direction slip at the fjord base. When compared to smoothed topography, perpendicular flow over real topography requires ∼41–89 % greater area-averaged driving stress, dependent on the overlying ice thickness, while a switch from fjord-parallel to fjord-perpendicular flow for real topography requires a ∼28–45 % area-averaged driving-stress increase. These results may explain surface velocity variations at many locations towards the margins of the Greenland Ice Sheet, and imply that the role of anisotropic roughness in resisting ice-sheet motion may be significantly underappreciated. Last, we note that deep fjords provide a clear physically-based example for why bounded basal traction relationships (i.e. regularised-Coulomb) may not hold at the macro scale in rough settings.

The thermodynamic behaviour of ice-sheet motion over rough landscapes is poorly understood, with most ice-sheet models prescribing a bed smoother than reality, which will not fully capture topographic features. Subglacial fjords striking obliquely to the palaeo flow direction are an extreme case, but are ubiquitous beneath the western margin of the palaeo Scandinavian Ice Sheet, and likely provide a useful proxy for areas of the present-day Greenland Ice Sheet. Here, we consider Veafjorden as a characteristic western Norwegian fjord where striations clearly evidence palaeo perpendicular ice flow, and perform 3D thermodynamically-coupled ice-motion simulations across a range of orientations. For perpendicular flow, surface velocity above the fjord is reduced substantially while a thick layer of temperate ice occupies the fjord. Moffatt eddies, or spiralling flows, occur in the fjord hollow with reverse-direction slip at the fjord base. When compared to smoothed topography, perpendicular flow over real topography requires ∼41–89 % greater area-averaged driving stress, dependent on the overlying ice thickness, while a switch from fjord-parallel to fjord-perpendicular flow for real topography requires a ∼28–45 % area-averaged driving-stress increase. These results may explain surface velocity variations at many locations towards the margins of the Greenland Ice Sheet, and imply that the role of anisotropic roughness in resisting ice-sheet motion may be significantly underappreciated. Last, we note that deep fjords provide a clear physically-based example for why bounded basal traction relationships (i.e. regularised-Coulomb) may not hold at the macro scale in rough settings.

The thermodynamic behaviour of ice-sheet motion over rough landscapes is poorly understood, with most ice-sheet models prescribing a bed smoother than reality, which will not fully capture topographic features. Subglacial fjords striking obliquely to the palaeo flow direction are an extreme case, but are ubiquitous beneath the western margin of the palaeo Scandinavian Ice Sheet, and likely provide a useful proxy for areas of the present-day Greenland Ice Sheet. Here, we consider Veafjorden as a characteristic western Norwegian fjord where striations clearly evidence palaeo perpendicular ice flow, and perform 3D thermodynamically-coupled ice-motion simulations across a range of orientations. For perpendicular flow, surface velocity above the fjord is reduced substantially while a thick layer of temperate ice occupies the fjord. Moffatt eddies, or spiralling flows, occur in the fjord hollow with reverse-direction slip at the fjord base. When compared to smoothed topography, perpendicular flow over real topography requires ∼41–89 % greater area-averaged driving stress, dependent on the overlying ice thickness, while a switch from fjord-parallel to fjord-perpendicular flow for real topography requires a ∼28–45 % area-averaged driving-stress increase. These results may explain surface velocity variations at many locations towards the margins of the Greenland Ice Sheet, and imply that the role of anisotropic roughness in resisting ice-sheet motion may be significantly underappreciated. Last, we note that deep fjords provide a clear physically-based example for why bounded basal traction relationships (i.e. regularised-Coulomb) may not hold at the macro scale in rough settings.