Search

For reconstructing paleoclimate or studying glacial isostatic effects on the Earth’s lithosphere, increasingly more studies focus on modeling the large-scale ice cover in mountainous regions over long time scales. However, balancing model complexity and the spatial extent with computational costs is challenging. Previous studies of large-scale ice cover simulation in mountain areas such as the European Alps, New Zealand, and the Tibetan Plateau, typically used 1-2 km spatial resolution. However, mountains are characterized by high peaks and steep slopes - topographic features that are crucial for glacier mass balance and dynamics, but poorly resolved in coarse resolution topography. The Instructed Glacier Model (IGM) is a novel 3D ice model equipped with a Convolutional Neural Network which is trained from high order ice models to simulate ice flow. This results in a significant acceleration of run times, and thereby opening the possibility of higher spatial resolution runs. We use IGM to model the glaciation of the European Alps with different resolutions (2 km, 1 km and 200 m) over a time period of 160,000 years. We apply a linear cooling rate to today’s climate until 6 °C colder to mimic ice age conditions. Preliminary results indicate systematic, resolution-related differences: At the beginning of cooling the 2 km resolution yields slightly more ice volume. However, this trend reverses after the large valleys are filled with thick ice. When the Alps are fully ice covered, we find up to 15% more ice volume in the higher resolution models., The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)

For more than a century, a major focus of Quaternary research has been the investigation of glacier advances and retreats during ice age cycles. The cradle for these studies are the European Alps. In recent years, glacier modelling permitted to produce long-term transient simulations of the European Alps glacier evolution. However, only sparse empirical data, e.g. geological reconstructions of glacier margins in the foreland of the Alps of the Last Glacial Maximum, are available to validate these simulations.Speleothems from the Alps are a widespread palaeoclimate archive. They provide stable carbon isotope records, which can potentially inform about soil and vegetation conditions above a cave site but also about the lack of soil during times of glacier coverage. In addition, speleothem growth in cold, high-elevation cave sites during glacials are a strong indicator of temperatures in the soil-karst-cave system above the freezing point, which is only likely to occur if the cave is covered by a temperate glacier. We here utilise existing speleothem data (growth histories and stable carbon isotopes) from Alpine caves to infer glacier coverage and thermo-dynamical state during the last glacial cycle and to assess the compatibility with modelled reconstructions. We compare data from the last glacial cycle from multiple cave sites located at different elevations (785 to 2319 m a.s.l.) in the Alps with simulations obtained with the Parallel Ice Sheet Model (PISM). We find a general agreement between speleothem-derived soil presence or absence and modelled glacier coverage. However, the speleothem data provide evidence of temperate glacier coverage which is not shown by all of the PISM simulations. Our work demonstrates the value of speleothem-based reconstructions from the Alps as proxies for assessing performance of palaeo-ice flow models.

Mountain glaciers are a major source of sea-level rise and also represent an important freshwater resource in many mountainous regions. Thus, accurate estimations of their thickness and, therefore, the total ice volume are important both in predicting and mitigating the global and local effects of climate change. However, to date, only 2% of the world’s glaciers outside the ice sheets have any thickness data, due to the logistical difficulties of obtaining such measurements, creating a large and policy-relevant scientific gap.The recent development of a global-scale ice-velocity dataset, however, provides an ideal opportunity to fill this gap and determine ice thickness across the 98% of glaciers for which no thickness data is available. This can be done by inverting an ice-dynamics model to solve for the ice thickness. For accurate thickness results, this needs to be a full-Stokes model, but such a model is far too computationally cumbersome to apply on a global scale, and simpler, quicker methods usually based on the shallow ice approximation (SIA) are too inaccurate, particularly where sliding dominates glacier motion. The only attempt that has been made to leverage the global velocity dataset to retrieve ice thickness has, though, used the SIA, simply because higher-order approaches are not computationally realistic at this scale. Consequently, most of the widely-used global glacier models have made no concerted attempt to invert for global ice thickness, owing to these limitations.As an additional related problem, failing to fully assimilate ice-velocity data into an ice-flow model necessarily introduces a shock when initialising prognostic glacier simulations, resulting in model glaciers and predictions that may diverge substantially from their real-world counterparts.As a solution to these problems, we present results from a deep-learning-driven inversion model that emulates the performance of state-of-the-art full-Stokes models at a thousandth of the computational cost. This model, by solving an optimisation problem, can fully use and assimilate all available input datasets (surface velocity and topography, ice thickness, etc.) as components of its cost function to simultaneously invert for and optimise multiple control parameters (here, we focus on ice thickness). This approach also gives us the possibility of using the same ice-velocity field for inversion and forward modelling, reducing the magnitude of the shock inherent in traditional modelling approaches. With a view to a large-scale application to all the world’s 200,000 glaciers, we present initial thickness-inversion results for the relatively well-documented European Alps to help constrain model parameters and provide a test bed for extension to other glaciated regions, with initial extension to the Caucasus and the Southern Alps.

Outlet glaciers and ice streams transport ice from the ice sheets to the ocean, where the glaciers lose mass by iceberg calving. Sermeq Kujalleq (SKK, Jakobshavn Isbræ) is one of the largest and most dynamic ice streams of the Greenland Ice Sheet with ice flow velocities up to 40 m/day. With extensive fieldwork and detailed repeated UAV surveys we aim at understanding the complex processes occuring at the ice stream margins and at the calving front of SKK. Such processes are often neclected in numerical models inducing uncertainties in projections of the ice sheet evolution.Within the framework of the COEBELI project we conducted drone photogrammetry surveys in July 2022 at SKK along other field measurements including in-situ GPS, GPRI, seismometers, and time-lapse imagery. Despite challenging weather conditions and constraints due to flying restrictions, we acquired 17 repeated flight surveys over the calving front of SKK during two weeks. As a result, we produced a large imagery data set, which was processed to infer high-resolution ortho-images and digital elevation models (DEM). Comparing the different products enables us to estimate changes in surface topography and ice dynamics. During the observation period several large calving events occurred allowing us to investigate the interaction between frontal processes and ice flow dynamics. With the very detailed data we can study crevasse opening, acceleration at the front, weaknesses in the ice and their origin, and the reaction of the glacier to large calving events.

Outlet glaciers and ice streams are the main channels through which ice sheets transport their mass towards the ocean. One of Greenland’s largest outlet glaciers Sermeq Kujalleq in Kangia (Jakobshavn Isbrae) has been broadly researched after experiencing a rapid retreat of the terminus and accompanying speedup to up to 40 m/day in the early 2000’s. However, such short-term ice dynamic variations remain poorly understood making numerical models difficult to constrain and predictions on future sea-level rise uncertain.The short-term ice dynamics of Sermeq Kujalleq consists in transient states and can only be captured by in-situ measurements of high spatial and temporal resolution. Glacier seismology has proven to be a valuable tool to study these dynamics, it provides data with a high temporal resolution and can provide information on processes happening below the ice surface. Within the COEBELI project we combine passive glacier seismology with global navigation satellite system (GNSS) receivers, long-range drones, time-lapse cameras and terrestrial radar interferometry to capture processes such as calving and basal sliding at their respective timescales.Here, we present results from a multi-array seismic deployment at Sermeq Kujalleq in Summer 2022. From May until September two arrays were deployed in the upstream part of the fast-flowing ice stream (>22 km from calving front) and one array on slower moving ice North of the main trunk. For a 3-week period in July, four more arrays were deployed on the fast-flowing ice stream closer to the calving front (50 Hz) icequakes. By studying these different signals, we are able to better constrain the processes and forces that control fluctuating ice-flow velocity and calving events.

Sermeq Kujalleq in Kangia (Jakobshavn Isbræ), Greenland is one of the most studied glaciers in the world mainly due to its recent retreat associated with extremely fast ice stream flow and high solid ice discharge. However, large limitations remain in the understanding of its short-term ice dynamics as the study of sub-daily variations, generally undetectable in spaceborne observations, requires high-rate field measurements that are challenging to acquire. Here, we present glacier surface velocities determined in Post-Processed Kinematic (PPK) mode from eight autonomous Global Navigation Satellite System (GNSS) stations deployed in July 2022 along the ice stream at a distance of 4 to 30 kilometers from the calving front. During this field campaign, we identified an 8-hour-long glacier speedup which was recorded at all GNSS stations and reached up to 11% of the pre-event velocity, followed by a 12-hour-long slowdown of similar magnitude. We further found the peak velocity was first measured at a GNSS station 16 kilometers away from the calving front, then recorded consecutively at each of the three other downstream GNSS stations with a positive time lag corresponding to a ~3 km/h wave propagation speed. At the station closest to the calving front, the timing of peak velocity corresponded to the occurrence of large-scale calving events. We further present line-of-sight glacier surface velocities measured along three shear margin transects with a terrestrial radar interferometer deployed simultaneously with the GNSS array. Across all profiles, we observed a widespread and simultaneous response of fast- and slow-moving ice suggesting a strong coupling between the main trunk and the shear margins of the ice stream.

Data assimilation in high-order ice flow modeling is a challenging and computationally costly task, yet crucial to find ice thickness and ice flow parameter distributions that are consistent with ice flow mechanics and mass balance while best matching observations. Failing to find these distributions that are required as initial conditions leads to a disequilibrium between mass balance and ice flow, resulting in nonphysical transient effects in the prognostic model. Here we tackle this problem by inverting an emulator of the Stokes ice flow model based on deep learning. By substituting the ice flow equations using a convolutional neural network emulator, we simplify, make more robust and dramatically speed up the solving of the underlying optimization problem thanks to automatic differentiation, stochastic gradient methods and implementation of graphics processing unit (GPU). We demonstrate this process by simultaneously inferring the ice thickness distribution, ice flow parametrization and ice surface of ten of the largest glaciers in Switzerland. As a result, we obtain a high degree of assimilation while guaranteeing an equilibrium between mass-balance and ice flow mechanics. The code runs very efficiently (optimizing one large-size glacier at 100 m takes < 1 min on a laptop) while it is open-source and publicly available.

We present the Instructed Glacier Model (IGM) -- a new framework to model the evolution of glaciers at any scale by coupling ice dynamics, surface mass balance, and mass conservation. The key novelty of IGM is that it models the ice flow by a Convolutional Neural Network (CNN), which is trained from physical high-order ice flow mechanical models. Doing so has major advantages in both forward and inverse modelling. In forward modelling, the most computationally demanding model component (the ice flow) is substituted by a very cheap CNN emulator. Once trained with representative data, IGM permits to model individual mountain glaciers several orders of magnitude faster than high-order ones on CPU with fidelity levels above 90 \% in terms of ice flow solutions leading to nearly identical transient thickness evolution. Switching to Graphics Processing Unit (GPU) permits additional significant speed-ups, especially when modelling large-scale glacier networks and/or high spatial resolutions. In inverse modelling, the substitution by a CNN emulator does not only speed up but facilitates dramatically the data assimilation step, i.e. the search for optimal ice thickness and ice flow parameter spatial distributions that match spatial observations at best (such as ice flow, surface topography or ice thickness profiles) while being consistent with the high-order ice flow mechanics. The reason is that inverting a CNN can take great benefit from the tools used for its training such as automatic differentiation, stochastic gradient methods, and GPU. IGM is an open-source Python code (https://github.com/jouvetg/igm), which deals with two-dimensional gridded input and output data. Together with a companion library of trained ice flow emulators, IGM permits user-friendly, computationally highly-efficient, easy-to-customize, and mechanically state-of-the-art glacier forward and inverse modelling at any scale. We illustrate its potential by replicating a simulation of the great Aletsch Glacier, Switzerland, from 1880 to 2100, based on a Stokes model.

Glacier modelling of the Alpine region during past ice ages has received increasing attention in the recent years. Considering the complexity of the Alpine topography, high spatial resolution climate information is required for running glacier models over the area. However, continuous climatic signal at resolution higher than 10 km and covering several hundred thousand of years cannot be directly derived using dynamical climate models. Alternative strategies must be considered.Here, a climate emulator providing monthly temperature and precipitation over the Alpine region, with a horizontal resolution of 2x2 km and covering the last 400’000 years at intervals of 100 years, is presented. The emulatorcombines a dynamical modelling chain of Earth System Models (ESMs) and a Regional Climate Model (RCM) with different statistical modelling methods. The dynamical modelling chain delivers climate information at highest accuracy based upon physical prognostic equations for specific time slices. The subsequent statistical modelling uses these time slices as physically consistent boundaries and estimates the climate conditions in between them, thus generatinga long-term climate evolution.A total of 19 climate model experiments are conducted for different time-slices of the considered study period, including also sensitivity tests with changes in the ice sheet height of the Northern Hemisphere and land cover type. Starting from a simple linear regression, a series of different statistical approaches is tested for building the emulator. An evaluation of the different versions is then conducted against one of the RCM time-slice experiments, left out in turns from the training set of the statistical model. Results show robust skills of the emulator in the representation of temperature, whose changes are mainly driven by smooth variations in the seasonal pattern of insolation at different time-steps, already using a simple linear regression. For precipitation, non-linearities associated to changes in the large-scale atmospheric circulation seem to dominate, making the use of more complex statistical approaches more appropriate. Additional evaluation tests conducted using glacier modelling driven by the outputs of the developed emulator confirm its potential for reconstructing the ice extent over the Alpine region during the last ice ages.

Sermeq Kujalleq in Kangia (Jakobshavn Isbræ), Greenland has been extensively investigated over the past decades due to its recent retreat associated with extremely fast ice stream flow and high solid ice discharge. However, its short-term dynamics still remain poorly understood as they consist in transient states that can only be captured by high spatial and temporal in situ measurements. In the new COEBELI project, we aim at combining high resolution field data sets from seismic arrays, global navigation satellite system (GNSS) receivers, long-range uncrewed aerial vehicles and terrestrial radar interferometers (TRI) to achieve a comprehensive and detailed study of the short-term ice stream dynamics. Here, we present TRI and GNSS retrievals of surface velocity and elevation acquired during the first, exploratory field campaign of the COEBELI project in summer 2021. Seven kilometers away from the calving front, we specifically identified a slowdown of 1.12 m d-1 within a single day in the main trunk of the ice stream. While the absolute slowdown is larger in the main trunk than in the outer area of the shear margin (1.12 m d-1 versus 0.75 m d-1), it corresponds to a larger fraction of the pre-slowdown velocity in the latter zone (-4.48% versus -7.94%). We further discuss the challenges associated with the acquisition, processing and analysis of high-resolution data sets for the study of such complex and dynamic environments.