The dynamics of ice sheet collapse in deglaciation periods – Ice-Collapse

To understand what happens beneath ice sheets, this project relies on an original combination of two rarely combined approaches: the exploration of morphological and sedimentary archives from the Last Glacial Maximum, and the laboratory reproduction of sediment-transfer processes occurring beneath ice sheets.

1) Analysis of morpho-sedimentary archives

Vanished ice sheets have left behind vast landscapes, sculpted by the interaction of ice, meltwater, and sediments. Using high-resolution digital elevation models and satellite imagery, complemented by field observations, we have been able to map these imprints with unprecedented precision. A key innovation was to go beyond classical morphometric measurements by introducing new indicators capable of capturing subtle transitions between different types of subglacial bedforms. Statistical tools were then applied to detect spatial patterns that reveal the dynamics of ice flow and subglacial floods. One major outcome was the development of a semi-automated method for extracting landforms to process very large datasets. Unlike traditional approaches, this method does not classify landforms a priori but detects and characterises them directly from their shape. This makes it possible to analyse millions of bedforms quickly and to produce continuous maps of subglacial landscape characteristics.

2) Simulating subglacial processes in miniature

To complement these archives, we developed a novel experimental device capable of simulating the interactions between ice, meltwater, and sediments. Cameras and optical tracking tools enabled us to measure, in real time, ice velocities, strain rates, and the evolution of bedforms that developed beneath our experimental ice sheet. The result was the generation of a wide range of realistic structures (bedforms, channels, …) while simultaneously recording the coupled dynamics of ice deformation and meltwater flow. These experiments confirmed hypotheses about the links between basal shear, subglacial drainage, and the genesis of subglacial landforms.

3) Linking archives and experimental modelling

The originality of the project lies in bridging these two worlds: landscapes analyses, observable at continental scale, and laboratory experiments, where parameters can be manipulated and processes directly observed. This complementarity has allowed us to propose new methodological frameworks, paving the way for more systematic and quantitative palaeoglaciological reconstructions. Ultimately, these advances will help integrate morpho-sedimentary archives into numerical models, thereby reducing uncertainties in ice-sheet modelling and improving projections of their future evolution.

The project has provided new insights into what happens beneath ice sheets. By combining large-scale mapping, laboratory experiments, and field analyses, we developed innovative ways to interpret the landscapes left behind by former ice sheets. These morphological imprints reveal how ice, water, and sediments interacted during phases of rapid deglaciation. The main results are:

(1) A continuum of glacial bedforms. In several regions (Ireland, Canada, Scandinavia), we demonstrated that features such as drumlins, MSGLs, hummocks, ribbed bedforms, and murtoos are not separate categories but part of a continuous spectrum of bedform evolution and formation processes. Their diversity reflects progressive transitions linked to variations in ice flow and meltwater dynamics.

(2) Bedforms as archives of deformation. Laboratory experiments reproduced the continuum of subglacial bedforms under increasing ice stress or evolving drainage systems. Their morphology records both the intensity and orientation of forces transmitted by ice, as well as the degree of hydraulic connectivity. These findings confirm that subglacial landscapes are genuine archives, enabling reconstructions of past ice-sheet velocity, flow direction, and hydrological properties.

(3) Traces of subglacial floods. Analyses and experiments showed that recurrent floods, triggered by the drainage of subglacial or supraglacial lakes, transformed certain bedforms into triangular hills (murtoos) or chaotic corridors (hummocks). Aligned along former meltwater routes, these landforms reflect alternations between catastrophic floods and quiescent phases, providing direct evidence that meltwater dynamics played a crucial role in shaping landscapes and destabilising ice sheets.

(4) A semi-automated mapping protocol. The project developed a novel method able to automatically extract landform outlines and crestlines from high-resolution satellite images. When compared to expert manual mapping, it reached ~75% agreement while enabling the rapid analysis of vast regions previously inaccessible. This innovation paves the way for systematic and objective reconstructions of subglacial processes across entire ice sheets.

(5) Towards improved models. By integrating automated mapping, laboratory experiments, and field evidence, the project established the foundations of a unified framework for interpreting fossil landscapes. These results make it possible to directly link the imprints of past ice sheets to processes still active today beneath Greenland and Antarctica, thereby enabling more accurate modelling and more reliable projections of their future response to climate change.

The next step is ambitious: to reconstruct, at continental scale, the ice–water–sediment dynamics of ice sheets with kilometre-scale resolution. Achieving this would provide an unprecedented picture of how ice sheets collapse and would allow scientists to better test and improve the numerical models used to simulate both past and present ice-sheet evolution.

(1) Decoding the archives of the past. The mechanisms that drive ice sheets to collapse remain poorly understood, yet their signatures are preserved in the landscapes shaped beneath the ice. Among these features, subglacial bedforms hold the highest potential as archives. Systematic mapping of these bedforms will make it possible to test hypotheses on the pace and mechanisms of collapse, reaching far beyond what is observable today.

(2) Shifting scale. Moving from local case studies to continental-scale reconstructions requires new tools. With tens of millions of preserved bedforms, manual mapping is impossible. Automation, big-data processing and advanced spatial statistics are the only way forward. Recent progress in satellite imagery, high-resolution digital terrain models and, above all, artificial intelligence now make this shift feasible.

(3) Artificial intelligence and Big Data. The future lies in building vast databases of millions of mapped features measured through parameters such as size, elongation, curvature or roughness. Big-data techniques and geostatistics (multivariate analysis, clustering, spatial autocorrelation) will help identify and characterise the dynamics of ancient ice streams and meltwater flow corridors. Artificial intelligence, particularly neural networks, will speed up mapping, cut processing times and reveal semi-quantitative palaeoglaciological information hidden in the landscapes.

(4) Comparing data and models. These reconstructions will then be compared with ice-sheet model outputs. Morphometric maps showing flow velocity or drainage routes will serve as semi-quantitative benchmarks to assess model performance. Advanced statistical approaches will even make it possible to invert landform archives into estimates of past ice thickness, velocity and meltwater fluxes.

(5) Towards recalibrated models. By progressively integrating palaeo-archives, laboratory experiments and modelling, this research will refine the representation of subglacial processes in numerical models and improve future projections. Better knowledge of the pace and magnitude of past collapses will ultimately provide scientists and decision-makers with more reliable tools to anticipate the response of Greenland and Antarctic ice sheets to ongoing climate change.

During periods of climate changes, ice sheet shrinking is controlled by the collapse of vulnerable ice shelves at the tip of fast-flowing ice streams. Ice stream collapse has not yet been observed and we therefore lack solid historical data and models to predict accurately their consequences, thus limiting our ability to include such events in sea level projections. In this project, we aim to combine for the first time palaeoglaciological data acquired from former ice stream beds of the Laurentide Ice Sheet with a new experimental setup we developed in our lab to investigate processes of collapse. We here address a hot topic to investigate the ice dynamics/landforms linkage to establish a process-based spatial and temporal model of ice sheet collapse. The project integrates three complementary approaches: (WP1) detailed regional-scale mapping of the geomorphological record left by short-lived ice streams, (WP2) analyses of key lithological, stratigraphical and deformational characteristics of the soft-bed of ice streams and (WP3) physical modelling of unstable glacial systems. In this project, we first aim to reconstruct the origin, timing and processes of past ice stream collapse along the Laurentide Ice Sheet. The combination of experimental and palaeoglaciological data will also contribute to establish new semi-empirical laws that will relate soft-bed changes (erosion, sedimentation, deformation), development of subglacial drainage systems and their efficiencies, ice flow velocities, meltwater production rates and porewater pressure. The numerical modeling community will then be able to implement these new parametrization laws to include ice stream collapse in their ice sheet models.