MONARCHS: a 3-D model of ice shelf surface hydrology
Sammie Buzzard, Jonathan Elsey, Alex Robel
Corresponding author: Sammie Buzzard
Corresponding author e-mail: sammie.buzzard@northumbria.ac.uk
Recently remote sensing and modelling studies have shown several Antarctic Ice Shelves to be vulnerable to damage from surface meltwater. Understanding the surface hydrology of ice shelves in the present and the future is an essential first step to reliably project future sea level rise from ice sheet melt due to the buttressing effect provided by ice shelves on the grounded ice sheet. Here we present results of modelling studies using MONARCHS: a 3-D model of ice shelf surface hydrology. MONARCHS is the first comprehensive model of surface hydrology to be developed for Antarctic ice shelves, enabling us to incorporate key processes such as the lateral transport of surface meltwater. This community-driven, open-access model has been developed with input from observations, and allows us to provide new insights into surface meltwater distribution on Antarctica’s ice shelves. This enables us to answer key questions about their past and future evolution under changing atmospheric conditions and vulnerability to meltwater driven hydrofracture and collapse. Recent developments include experiments investigating the impact on ice shelf surface flexure in response to meltwater loading, and case studies of individual ice shelves thought to be vulnerable to meltwater induced hydrofracture.
Using the ensemble Kalman inversion to calibrate ice sheet models
Alex Bradley
Corresponding author: Alex Bradley
Corresponding author e-mail: aleey@bas.ac.uk
Projections of sea level rise from ice sheet models are highly sensitive to specific choices of modelling parameters. Many studies attempt to overcome this issue by running ensembles of simulations, with many different parameter choices spanning likely ranges. However, the expense of running ice sheet models means that the ensemble size is limited and often few simulations actually lie in the region of parameter space which gives agreement with observations. As a result, only a small number of simulations dominate these projections. The Ensemble Kalman Inversion is an iterative procedure which allows the space of parameters which agree with observations to be discovered, by comparing model output to observational data. Here, we present this method applied to ice sheet models. Although it has been successfully applied in many areas of earth science and physics, it has yet to be applied to an ice sheet model. The resulting calibrated parameters provide a set of simulations which agree well with observational data, allowing the true uncertainties in sea level rise projections to be probed.
Increased rotational coupling between Antarctic sea ice and the atmosphere over the last 30 Years
Wayne de Jager, Marcello Vichi
Corresponding author: Wayne de Jager
Corresponding author e-mail: djgway001@myuct.ac.za
Antarctic sea ice has been characterized by high temporal and spatial variability since the inception of reliable satellite records. The complex oceanic and atmospheric mechanisms driving this variability present ongoing challenges in determining their respective contributions. In this study, we examine the cyclonic and anticyclonic rotation dynamics within the sea ice and overlying atmosphere at daily timescales from 1991–2020 using the new generation of remote-sensing product for sea-ice drift. A two-dimensional pattern similarity comparison between the ice and atmospheric vorticity fields demonstrated a noteworthy increase in pattern similarity over the past three decades, despite the absence of any discernible trends within each component over the same period. This escalating coupling suggests an increasing susceptibility of sea ice to atmospheric forcing, a phenomenon observed across all regions of the Southern Ocean and independent of the sea-ice extent. Notably, the Weddell Sea experienced a sudden regime shift after 2001, marked by a sharp decline in the intensity of sea-ice rotation, persisting in this weakened state from 2002 onwards. The increased coupling of sea-ice drift at the synoptic scale with no discernible trends in the atmospheric forcing points to a plausible role of the ocean. These features are analyzed in relation to the Southern Annular Mode (SAM) trends from atmospheric reanalysis. SAM’s impact is evident in the vertically integrated atmospheric eddy kinetic energy but revealed no distinct differences during positive or negative phases on surface levels, explaining the absence of long-term sea-ice rotation trends. Our findings underscore once again the predominant role of the atmosphere in driving rotation within Antarctic sea ice, while highlighting a knowledge gap on the possibly increasing influence of the upper ocean on ice drift dynamics, a contribution that may require a dedicated ocean–ice coupled model investigation to better understand.
Improving modeled ice dynamics in Northwest Greenland with transient calibration: from multi-decadal trends to seasonal cycles
Jessica Badgeley, Helene Seroussi, Mathieu Morlighem
Corresponding author: Jessica Badgeley
Corresponding author e-mail: jessica.a.badgeley@dartmouth.edu
State-of-the-art ice sheet model simulations used in the Ice Sheet Model Intercomparison Project for CMIP6 (ISMIP6) show a mismatch with recent observations of ice sheet change. In particular, the difference between the modeled and observed cumulative mass balance trend over the last two decades calls into question the accuracy of current projections of ice sheet contribution to sea level rise. Here, we use one of these models, the Ice-sheet and Sea-level System Model (ISSM), to investigate how applying transient calibration and resolving sub-annual ice dynamics may improve model hindcasts of mass balance and impact projections. Transient calibration is a relatively new approach to initialize ice flow models; it uses time series of observations to invert for uncertain model parameters, such as basal friction and ice rheology. With more observational constraints than the common snapshot inversion method, transient calibration has been shown to better capture trends and to have the ability to estimate how parameters evolve through time. In addition, recent dataset developments provide sub-annually resolved observations of ice dynamics – e.g. ice front positions, surface velocities and surface elevation – gridded for straightforward ingestion into ice sheet models. Here we apply transient calibration and leverage these high temporal resolution datasets to model northwest Greenland, a region undergoing rapid changes. We find that using transient calibration to assimilate surface velocity data brings hindcast simulations to within the uncertainty of unassimilated cumulative mass balance observations. Future simulations to 2100 using the ISMIP6 protocols show that the use of transient calibration leads to greater mass loss and has a greater impact on mass balance than the choice of climate forcing scenario over the next three decades. To reproduce sub-annual ice dynamics, we impose monthly-resolved ice front positions and use transient calibration to vary the basal friction, mimicking the impacts of subglacial hydrology. We find that failing to resolve sub-annual velocity variations can bias modeled cumulative mass balance on decadal timescales.
The state and fate of global permafrost: a time series of modelled permafrost extent (1960–2020)
Harley R. McCourt, Will Roberts, Mike Lim, Matt Westoby, Stuart Dunning
Corresponding author: Harley R. McCourt
Corresponding author e-mail: harley.mccourt@northumbria.ac.uk
Permafrost is ground that is perennially frozen (≤0°C) for a period of at least 2 consecutive years. To monitor the state and fate of permafrost, we require knowledge of its spatial distribution and how it responds to environmental change through time. However, permafrost is difficult to measure over large areas because of complications in obtaining data through direct observation, particularly in remote areas of such as the high Arctic or sites at great altitude. Existing permafrost models cover narrow windows of time, and this limits their use for comparison with satellite time-series data, e.g. the analysis-ready Landsat data from 1982 onward. We present a new global time series from an improved high-resolution (>1 km²) global model of permafrost extent and zonation from 1960–2020. Modern reanalysis datasets (JRA-55 and ERA5) are downscaled onto a global digital elevation model (ALOS 3D World) and permafrost extent is calculated as a function of mean annual ground temperature following Gruber (2012). Between 1960 and 2020 modelled permafrost extent globally has decreased by 15±11% which equates to loss of 25.9×10⁵ km², (0.33×10⁵ km², 1.29%, pertains to the Southern Hemisphere). This implies a loss in permafrost extent of ~4.32×10⁵ km² per decade. This decline has not been spatially consistent and is localized within high-mountain and coastal areas of the globe. This agrees with observed linear trends in anthropogenic warming in these regions.
Tracking glacial sediment transport using a Lagrangian approach with luminescence rock surface burial dating of englacial clasts
Audrey Margirier, Julien Brondex, Ann Rowan, Georgina King, Christoph Schmidt, Vivi Pedersen, Benjamin Lehmann, Leif Anderson, Remy Veness, Scott Watson, Darrel Swift
Corresponding author: Ann Rowan
Corresponding author e-mail: ann.rowan@uib.no
Constraining the time scales of glacial sediment transport is important for understanding the processes controlling sediment dynamics through glacierized catchments, and because the accumulation of supraglacial sediment influences the response of glaciers to climate change. However, glacial sediment transport processes can be difficult to observe; sediment grains (clasts) can be transported englacially, subglacially, supraglacially or at the ice margins, and may be stored on headwalls or within moraines before being (re-)entrained and transported by glacier ice. We developed a novel approach combining luminescence rock surface burial dating of clasts collected from glacier ice with an ice-flow model that includes Lagrangian particle tracking to quantify rates of sediment transport through the Miage Glacier catchment in the Italian Alps during the Holocene (~12 ka to present). Luminescence rock surface burial ages for seven samples embedded in the near-surface ice in the ablation area range from 0.2±0.1 ka to 4.7±0.3 ka and are consistent with results obtained from the ice-flow model. Our results show that the transport durations of individual clasts varied by an order of magnitude, with rapid clast transport near the glacier surface and longer transport histories for clasts transported lower in the ice column. In some cases, clasts were stored on the headwalls for several thousand years in cold-based ice aprons or ice-marginal moraines before being entrained in the glacier. These results illustrate the different routes by which glaciers transport sediment and provide the first direct measurements of the duration of sediment transport within an alpine glacier.
Monitoring and modeling glacier and snowmelt water resources in complex high-mountain terrain
Hongyi Li, Xiaohua Hao, Guang Li
Corresponding author: Hongyi Li
Corresponding author e-mail: lihongyi@lzb.ac.cn
The stable supply of mountain snow and glacier water resources is threatened by highly uncertain climate change. This threat is particularly evident in cold and arid regions facing water scarcity issues. However, accurate assessment of mountain snow and ice water resources faces several challenges. These include significant uncertainties in traditional remote sensing data in complex terrain, unknown losses of ground precipitation measurements due to wind in high-altitude mountains, and the influence of frozen soil on the assessment of meltwater contributions.
To overcome these challenges in high mountains, we developed an integrated cryosphere hydrological model to understand the complex hydrological contributions of snowmelt and blowing snow. We also developed machine learning-driven data fusion methods to overcome the uncertainty in monitoring snow and glaciers. This investigation is conducted in the ecologically and economically important geographical corridor known as the Hexi Corridor. Located in northwest China, the Hexi Corridor is a long and narrow corridor spanning approximately 1000 km. It covers about 270 000 km² of vast snow and ice zones and oasis deserts at the intersection of multiple climatic zones. We will introduce the progress made in the investigation from the following highlights of our work:
1. We conducted observations of unmanned aerial vehicle (UAV) lidar and the ICESAT-2 altimetry satellite to investigate snow accumulation and river ice water resources. We have also developed a machine learning-based method to integrate ground precipitation observations and satellite observations.
2. We developed a daily cloud-free snow-covered area fraction product for complex mountainous terrain, focusing on improving the accuracy of remotely sensed snow data in forested areas and under cloud-cover conditions.
3. A snow and ice water resource assessment model is being developed. This model traces meltwater paths and evaluates the historical changes and future trends of snow and glacier water resources. It incorporates a wind-driven snow drift dynamics module to analyze water resource losses caused by wind-blown snow under typical terrain conditions in high-altitude mountains.
Available and possible datasets based on seven season’s airborne ice-penetrating radar survey in East Antarctica through the Chinese ’Snow Eagle 601‘
Xiangbin Cui
Corresponding author: Xiangbin Cui
Corresponding author e-mail: cuixiangbin@pric.org.cn
Insufficient subglacial data in Antarctica seriously limit our understanding of subglacial conditions and processes of the ice sheet, leading to large uncertainty in prediction of ice sheet instability and contribution to sea level. Airborne survey has been the most efficient way to collect sub-ice data until now. In 2015, China deployed its first fixed-wing airplane, named Snow Eagle 601 for both scientific and logistical activities in Antarctica. Ice-penetrating radar (IPR, similar to HiCARS), GT-2A gravimeter and CS-3 magnetometer were configured and integrated in the airplane. Since 2015, the field campaign of International Collaborative Exploration of the Cryosphere through Airborne Profiling in Princess Elizabeth Land (ICECAP/PEL) has been initiated, and so far, more than 200 000 km of flight lines have been surveyed in seven austral seasons in East Antarctica, covering Princess Elizabeth Land, Amery Ice Shelf, West Ice Shelf, Ridge B, and ice sheet margin along the Enderby Land and Dronning Maud Land, etc. With airborne IPR data collected in the first four seasons, ice thickness and bedrock topographic DEM with 500 m spatial resolution in PEL was interpolated. The DEM reveals the continent topography in PEL in detail for the first time. In Amery, we accurately identified the grounding points along the flight lines using IPR data and compared the results with existed grounding line products developed by remote sensing data revealing a mean separation of 1.00±1.16 km. Moreover, IPR data are also used to constrain the inversion of bathymetry of Amery Ice Shelf from airborne gravity data. Ridge B is one of the least studied areas in Antarctica but has been considered to be a potential location for the oldest ice on Earth. We inferred geothermal heat flux (GHF) in Ridge B based on distribution of subglacial dry–wet zones identified from airborne IPR data. The results show GHF in RB varies locally and ranges from 48.50–65.10 mW m⁻², providing a regional GHF product in the region may benefit estimation ice age in future. Now, we are identifying subglacial lakes and water distribution in PEL with basal reflectivity and specularity methods. China is also a key member in SCAR RINGS Action Group and regional RINGS projects. Airborne IPR data collected along the ice sheet margin will contribute our knowledge of ice thickness and bedrock topography in these data-sparse areas.
The future of Upernavik Isstrøm: sensitivity analysis and bayesian calibration
Eliot Jager, Fabien Gillet-Chaulet, Nicolas Champollion, Romain Millan, Heiko Goelzer, Jérémie Mouginot
Corresponding author: Eliot Jager
Corresponding author e-mail: eliot.jager@helsinki.fi
The future contribution to sea level rise from the Antarctic and Greenland ice sheets remains highly uncertain, as demonstrated by recent multi-model intercomparisons (ISMIP6). This uncertainty arises from ice flow models, atmospheric and oceanic forcing and Shared Socioeconomic Pathways (SSP). For the Greenland Ice Sheet, model uncertainty contributes a significant portion to the ensemble spread (40 mm of sea-level rise by 2100), comparable to atmospheric forcing (36 mm) and SSP uncertainty (48 mm), and double that of oceanic forcing (19 mm). Ice flow model uncertainties stem from numerical model assumptions, parameterization, and initialization processes. We investigate the sensitivity of the Elmer/Ice model to various sources of uncertainty for Upernavik Isstrøm, a tidewater glacier in northwest Greenland. Through a 200-member ensemble, we explore parameters related to the ice sheet model (e.g. ice stiffness, calibration) and forcing such as SSP, climate models and downscaling. Each ensemble member undergoes historical simulation from 1985–2015, initialized close to 1985 and forced using observed front positions and surface mass balance. Projections to 2100 are conducted following the ISMIP6 Greenland protocol. We initially analyse the ability of the ensemble to replicate historical mass variations and assess projection uncertainties associated with different sources (ISM, SSP, climate models, oceanic sensitivity). Subsequently, we confirm Elmer/Ice’s capability to accurately replicate historical mass loss, instilling confidence in its future projection abilities. While all uncertainties significantly contribute to projections until 2100, SSP-related uncertainty dominates by the century’s end. Furthermore, we evaluate various Bayesian calibration strategies using cross-validation, considering datasets including ice mass change, discharge, velocity, and surface fields. Our goal is to identify effective calibration methods and integrate them into an a posteriori ensemble, mitigating overlearning. Results indicate that Bayesian calibration based on spatial data produces more robust posterior ensembles, albeit with a marginal reduction in uncertainty of future mass loss.
Design and performance of ELSA v2: an isochronal model for ice-sheet layer tracing
Therese Rieckh, Andreas Born, Alexander Robinson, Robert Law, Gerrit Gülle
Corresponding author: Therese Rieckh
Corresponding author e-mail: trieckh@gmail.com
We provide a detailed description of the ice-sheet layer age tracer ELSA – a model that uses a straightforward method to simulate the englacial stratification of large ice sheets – as an alternative to Eulerian or Lagrangian tracer schemes. ELSA’s vertical axis is time and individual layers of accumulation are modeled explicitly and are isochronal. ELSA is not a stand-alone ice-sheet model, but requires uni-directional coupling to another model providing ice physics and dynamics (the ‘host model’). Via ELSA’s layer tracing, the host model’s output can be evaluated throughout the interior using ice core or radiostratigraphy data. We describe the stability and resolution-dependence of this coupled modeling system using simulations of the last glacial cycle of the Greenland ice sheet using one specific host model. Key questions concern ELSA’s design to maximize usability, which includes making it computationally efficient enough for ensemble runs, as well as exploring the requirements for offline forcing of ELSA with output from a range of existing ice-sheet models. ELSA is an open source and collaborative project, and this work provides the foundation for a well-documented, flexible, and easily adaptable model code to effectively force ELSA with (any) existing full ice-sheet model via a clear interface.
Greenland-scape: Assessing analytic and numerical models for improving representation of subglacial topography in slow-flowing regions
Allison Chartrand, Joseph MacGregor, Mathieu Morlighem, Olga Sergienko
Corresponding author: Allison Chartrand
Corresponding author e-mail: allisonchartrand@gmail.com
As mass loss from the Greenland Ice Sheet (GrIS) accelerates, observations of the ice surface are ever improving in both spatial and temporal resolution. However, observations of the interface between the ice sheet and the solid Earth remain comparatively limited to boreholes and radar sounding surveys, limiting knowledge of ice thickness and hence subglacial topography. These quantities are critical boundary conditions for models projecting mass loss and sea level contribution from the GrIS. We seek to improve upon present methods of interpolating ice thickness between observations in the ice-sheet interior by developing a method that is consistent with modeled ice physics, subaerial surface observations, and radar thickness measurements. We compare thickness estimates in the GrIS interior derived from an analytic Shallow Ice Approximation (SIA) model (following an earlier implementation for glaciers) and the Mono-layer Higher-order (MOLHO) ice flow model, implemented in the Ice-sheet and Sea-level System Model (ISSM), with thicknesses from ordinary kriging and anisotropic diffusion as reported in BedMachine Greenland versions 3 and 5, respectively. For ISSM-MOLHO, we use ISSM’s inversion capabilities to solve for ice thickness, initially assuming that there is no basal sliding present at the ice–bed interface. We select several large test regions in the GrIS interior to test the strengths and weaknesses of each model. We use the GrIMP surface DEM and MEaSUREs surface speed data to constrain the models, and compare the modeled results to these and radar observations to further tune key model parameters, including ice softness. SIA and ISSM-MOLHO inversions result in thicknesses that fall within the observational uncertainty but also better reflect subaerial variability in topography than currently applied methods. These results provide a basis for improving representation of ice thickness across the ice-sheet interior and in turn an updated subglacial topography map that may be used for improving estimates of ice–bed interactions and projections of ice-sheet change.
Validation of effective subglacial hydrology models
Jeremie Schmiedel, Thomas Kleiner, Angelika Humbert, Roiy Sayag
Corresponding author: Jeremie Schmiedel
Corresponding author e-mail: schmieje@post.bgu.ac.il
Subglacial networks at the ice–bed interface are considered a key component of ice sheet dynamics with the potential to facilitate rapid ice flow and the formation of surges and ice streams. Various numerical methods have been developed to simulate the effects of such networks on ice flow. Validation of these models is crucial to ensure that the important subglacial physical processes are accurately captured. One modelling approach is to model subglacial networks as an effective porous medium (EPM), having a spatio-temporal varying transmissivity that represents a range of subglacial and groundwater processes. An important component of such models is a nonlinear diffusion equation for the subglacial water pressure. We present theoretical axisymmetric solutions for a generalized nonlinear diffusion equation of this type that can model a wide range of flows. We find generalized similarity solutions, and solutions with explicit time-dependent transmissivity. We use this method to validate the parallel implementation of the Confined–Unconfined Aquifer System model (CUAS-MPI) for subglacial hydrology. The model solves in two spatial dimensions and is based on an effective porous media (EPM) approach. Our results show that CUAS-MPI is capable of accurately solving highly nonlinear flows that can represent cavity opening and creep closure in subglacial hydrology. Due to their generality, our solutions are readily applicable to other subglacial hydrology models based on the EPM approach. We anticipate that a validated hydrology model with our solutions can achieve more credible results in the simulation of subglacial networks and consequently in the prediction of ice sheet evolution.
New observational uncertainties for sea-ice model evaluation
Andreas Wernecke
Corresponding author: Andreas Wernecke
Corresponding author e-mail: andreas.wernecke@uni-hamburg.de
The net Arctic and Antarctic sea-ice areas (SIA) and sea-ice extents (SIE) are routinely estimated from passive microwave sea-ice concentration (SIC) satellites products. To be truly useful metrics for the sensitivity of the Arctic sea-ice cover to global warming or as evaluation targets in model assessments, we need, however, reliable estimates of the uncertainties. The ability of observational uncertainties in SIA and SIE to obscure links between the sea-ice cover and the climate system is not well understood. We have retrieved such estimates for SIA and SIE by encompassing the spatial and temporal SIC error correlations into a stochastic model. With this Monte Carlo approach we can derive dynamic single-product observational SIA and SIE uncertainties and even the corresponding probability distributions. In a case study, we find that the observational uncertainties in 2015 of both Arctic SIA and SIE are about 300 000 km² for daily and weekly estimates and 160 000 km² for monthly estimates and further quantify the uncertainty in decadeal trend analysis. I will report on activities to implement these uncertainties in the upcoming OSI SAF Sea Ice Index release.
Ice–ocean coupled modelling for Nioghalvfjerdsbræ (79NG), Greenland
Jo Zanker, Jan De Rydt
Corresponding author: Jo Zanker
Corresponding author e-mail: jo.zanker@northumbria.ac.uk
The Northeast Greenland Ice Stream (NEGIS) drains approximately 12% of the Greenland Ice Sheet’s surface area, containing an ice volume of 1.1 m sea-level equivalent. Nioghalvfjerdsbræ (79NG) is one of two main outlet glaciers of NEGIS, extending into a large floating ice tongue, one of few remaining in Greenland. It is currently not well understood how 79NG will respond to the changing atmosphere and warming oceans, with possible implications for the catchment’s surface mass balance (SMB) and ocean-induced ablation. This research aims to assess the importance of feedbacks between ice-sheet geometry, SMB and ocean-driven melt by having a mutually evolving dynamic ice sheet with evolving SMB parameterization and a 3D ocean circulation model utilizing the ice–ocean coupled model Úa-MITgcm. The potential feedbacks between changes in ice-sheet surface geometry, ice-tongue cavity geometry and the atmosphere/ocean mass balance are as yet poorly understood, especially in the context of Greenland. Of particular interest for NEGIS is the potential for geometry-induced changes in melting of the ice tongue, as found for some Antarctic ice shelves. Development of the Úa ice-flow model will begin with a Greenland-wide setup and experiments based on the ISMIP6 protocol, before focussing on a regional setup of the NEGIS catchment and coupling to a regional configuration of the MITgcm ocean model of the adjacent fjord and continental shelf. The coupled approach of this project aims to improve the representation of the feedbacks between different climate components at a regional scale and draw conclusions about the fidelity of projections of ice sheet-wide mass loss and sea-level rise from ISMIP.
Weak relationship between remotely detected crevasses and inferred ice rheological parameters on Antarctic ice shelves
Cristina Gerli, Sebastian Rosier, Hilmar Gudmundsson, Sainan Sun
Corresponding author: Cristina Gerli
Corresponding author e-mail: cristina2.gerli@northumbria.ac.uk
Over the past decade, a wealth of research has been devoted to the detection of crevasses in glaciers and ice sheets via remote sensing and machine learning techniques. It is often argued that remotely sensed damage maps can function as early-warning signals for shifts in ice shelf conditions from intact to damaged states and can serve as an important tool for ice sheet modellers to improve future sea-level rise predictions. Here, we provide evidence for Filchner–Ronne and Pine Island ice shelves that remotely sensed damage maps are only weakly related to the ice rate factor field A derived by an ice-flow model when inverting for surface velocities. This technique is a common procedure in ice flow models, as it guarantees that any inferred changes in A relate to changes in ice flow measured through observations. The weak relationship found is improved when investigating heavily damaged shear margins, as observed on Pine Island Ice Shelf; yet, even in this setting, this association remains modest. Our findings suggest that many features identified as damage through remote sensing methods are not of direct relevance to present-day ice-shelf flow. While damage can clearly play an important role in ice-shelf processes and thus be relevant for ice-sheet behaviour and sea-level rise projections, our results imply that mapping ice damage directly from satellite observations may not directly help improve the representation of these processes in ice-flow models.
Development of a new snow model in the framework of the ERC IVORI ERC project
Basile de Fleurian, Kévin Fourteau, Marie Dumont, Mathias Bavay, Julien Brondex, Neige Calonne, Pascal Hagenmuller, Henning Loewe
Corresponding author: Basile de Fleurian
Corresponding author e-mail: basile.defleurian@uib.no
The ERC IVORI project aims to fill a modelling gap in the current existing snow models. Most of the existing snow models focus on the description of alpine snowpacks. This excludes a large part of the globe’s snow cover, namely the Arctic and other high-latitude snowpacks, which have a distinctive structure due to their environments. The model is implemented in Julia in a modular structure that can incorporates different physical descriptions of given processes as well as ad-hoc parameterizations when needed. We propose a fully coupled numerical solution to represent heat diffusion, water vapor transport, liquid water percolation, phase changes and settlement. Special care was taken to ensure numerically stable solutions enabling the use of large time steps. The model has also been designed with future coupling with soil, canopy and atmospheric models in mind to ease its integration in larger frameworks or the implementation of new components of the earth system on the same model structure.
Calving MIP – Idealized experiments into calving algorithms and laws
Jim Jordan
Corresponding author: Jim Jordan
Corresponding author e-mail: j.r.jordan@swansea.ac.uk
Calving MIP is an ongoing model intercomparison project investigating the current state of ice shelf calving implementation in numerical ice models. We make the distinction between a calving algorithm (the process by which a numerical model represents ice shelf calving) and a calving law (a process with a physical justification for how much should be calved using a calving algorithm in a model). The first phase of Calving MIP focuses on the implementation of calving algorithms within different models using a variety of idealized geometry with predefined calving rates. Future phases will investigate the effect of different calving laws and realistic bedrock geometry on model behaviour. We present here results from nine different modelling groups (at the time of submission) performing the first phase of Calving MIP experiments and comment on the ability of current numerical ice models to accurately represent the calving process, as well as discussing future planned phases of experiments.
Firn densification in two dimensions: modelling the collapse of snow caves and enhanced densification in ice-stream shear margins
Jon Arrizabalaga-Iria, Lide Lejonagoitia-Garmendia, Christine Hvidberg, Aslak Grinsted, Nicholas Rathmann
Corresponding author: Nicholas Rathmann
Corresponding author e-mail: rathmann@nbi.ku.dk
Accurate modelling of firn densification is necessary for ice-core interpretation and assessing the mass balance of glaciers and ice sheets. In this paper, we revisit the nonlinear-viscous firn rheology introduced by Gagliardini and Meyssonnier (1997) that allows the posing of multi-dimensional firn densification problems subject to arbitrary stress and temperature fields. First, we extend the calibration of the coefficient functions that control firn compressibility and viscosity to five additional Greenlandic sites, showing that the original calibration is not universally valid. Next, we demonstrate that the transient collapse of a Greenlandic firn tunnel can be reproduced in a cross-section model, but that anomalous warm summer temperature during 2012–14 reduces confidence in attempts to independently validate the rheology. Finally, we show that the rheology can explain the increased densification rate and varying bubble close-off depth observed across the shear margins of the North-East Greenland Ice Stream. Although we suggest that more work is needed to constrain the model’s near-surface compressibility and viscosity functions, our results strengthen the rheology’s empirical grounding for future use, such as modelling horizontal firn density variations over ice sheets for mass-loss estimates or estimating ice–gas age differences in ice cores subject to complex strain-rate histories.
What can we learn about ice sheet dynamics by investigating geothermal heat flow in East Antarctica?
Felicity McCormack, Jason Roberts, Christine Dow, Tobias Stål, Tyler Pelle, Jacqueline Halpin, Anya Reading, Martin Siegert
Corresponding author: Felicity McCormack
Corresponding author e-mail: felicity.mccormack@monash.edu
Geothermal heat flow is an essential boundary condition for modelling the thermal structure of ice sheets and glaciers, influencing ice motion through the processes of deformation and sliding. However, geothermal heat flow estimates are generally low resolution (>20 km), not resolving fine-scale spatial structure in geothermal heat flow. Here, we investigate the influence of fine-scale geothermal heat flow anomalies on subglacial meltwater production in the Aurora Subglacial Basin, East Antarctica, using the Ice-sheet and Sea-level System Model. We show that incorporating fine-scale geothermal heat flow anomalies increases subglacial melt rates by up to 5% compared with model simulations with the same overall mean but as a constant background geothermal heat flow. Excluding regions where the frictional heating is sufficient to produce melt, we determine the minimum basal heating required to bring the basal temperature to the pressure melting point (QSmin); where QSmin is relatively low (i.e. <20 mW ⁻²), small anomalies in geothermal heat flow are naturally likely to influence where subglacial meltwater is present. We compare regions where QSmin is low and where radar derived quantities predict that subglacial meltwater may be present, finding significant overlap between these regions.
Multi-physics ensemble modelling of Arctic tundra and taiga snowpack properties
Georgina Woolley, Nick Rutter, Leanne Wake, Vincent Vionnet, Chris Derksen, Richard Essery, Philip Marsh, Rosamond Tutton, Branden Walker, Matthieu Lafaysse, David Pritchard
Corresponding author: Georgina Woolley
Corresponding author e-mail: georgina.j.woolley@northumbria.ac.uk
Sophisticated snowpack models fail to simulate density and SSA profiles of Arctic snowpacks due to an underestimation of wind-induced compaction, misrepresentation of basal vegetation impacting compaction and metamorphism, and omission of water vapour transport. Parameterizations of Arctic snow physical processes were implemented into a 120-member ensemble version of the Soil, Vegetation and Snow version 2 (SVS2-Crocus) land surface model, creating Arctic SVS2-Crocus. Arctic SVS2-Crocus was driven and evaluated using measurements of snowpack properties (snow water equivalent, depth, density and SSA) at Trail Valley Creek (TVC), NWT, Canada over 32 years (1991–2023). The inclusion of wind-induced compaction in Arctic SVS2-Crocus increased surface snow density and reduced the RMSE by 41% (176 kg m⁻³ to 103 kg m⁻³). Parameterizations of basal vegetation were less effective in reducing compaction of basal snow layers (default RMSE: 67 kg m⁻³; Arctic RMSE: 65 kg m⁻³). Nearly all ensemble members of Arctic SVS2-Crocus outperformed default SVS2-Crocus simulations of snow density profiles at TVC, as evidenced by lower continuous ranked probabilistic scores. The recent addition of a forest canopy model to Arctic SVS2-Crocus will be evaluated across a 40 km transect of the Northwest Territories, Canada, where tundra to taiga ecosystems transition. Quantifying the impact of forest and shrub canopies on simulated snow properties will be crucial for the evaluation of satellite microwave radiative transfer models for snow water equivalent retrieval (Terrestrial Snow Mass Mission, TSMM) and coupling of Arctic SVS2-Crocus with CryoGrid to evaluate simulated soil temperatures and their impact on permafrost.
Drivers on ongoing changes Thwaites and Pine Island Glaciers, West Antarctica.
Hilmar Gudmundsson, Mathieu Morlighem, Daniel Goldberg, Jowan Barnes, Sainan Sun
Corresponding author: Hilmar Gudmundsson
Corresponding author e-mail: hilmar.gudmundsson@northumbria.ac.uk
We summarize recent findings on the contemporary dynamics of Thwaites and Pine Island Glacier, West Antarctica, resulting from the Prophet ice-sheet modelling project, a part of the International Thwaites Glacier Collaboration project. We suggest that current retreat rates of Thwaites Glacier are probably not driven by current ocean conditions and may reflect a transient response to previous changes in external forcings, and possible ungrounding of past pinning points. A similar judgement can be made for the neighbouring Pine Island Glacier where ungrounding from a subglacial ridge may have led to a recent phase of an irreversible grounding line retreat, further amplified by high melting rates due to the intrusion of CDW under its ice shelf. Using three different ice-sheet models, and various forcing scenarios we find the existence of multiple tipping points for both Pine Island and Thwaites glacier. This appears to be a very robust aspect of the dynamics of the region, and such tipping points are found for a wide range of basal conditions and frontal retreat scenarios.
Revisiting the implications of cliff-height-dependent calving law on West Antarctic glaciers
Sainan Sun, Hilmar Gudmundsson
Corresponding author: Sainan Sun
Corresponding author e-mail: sainan.sun@northumbria.ac.uk
High-end estimates of sea-level change from Antarctica have been derived from simulations using upper-end forcing scenarios and ice-cliff height dependent calving laws. Those have been hypothesized to cause collapse of glaciers in West Antarctica through marine ice cliff instability (MICI). However, some previously published high-end estimate are based on results from a limited number of ice-sheet models, or even only a single ice-flow modelling study. There is, furthermore, low agreement on the implications of some of those calving laws for the West Antarctic Ice Sheet, and limited evidence of MICI having occurred in the past. Here we investigate the dynamic response of West Antarctic glaciers to high-end calving laws using the Úa ice-sheet model. Specifically, we conduct ice-shelf collapse experiments as defined in ABUMIP (Sun et al., 2020) with and without cliff failure mechanism in transient simulations conduced over centennial time scales. We find that the ice-cliff-height-dependent calving laws can cause glaciers to retreat and collapse from both fast and slow flowing regions. Furthermore, we find that the results are sensitive to numerical resolution near the grounding line. We suggest therefore that ice-sheet modellers always conduct convergence studies when implementing high-end calving laws.
Physical controls on the ocean circulation beneath ice shelves revealed by a simple diagnostic model
Adrian Jenkins, Ole Anders Nøst
Corresponding author: Adrian Jenkins
Corresponding author e-mail: adrian2.jenkins@northumbria.ac.uk
While the circulation beneath ice shelves is often modelled as a buoyant plume flowing along the ice shelf base above a quiescent lower layer, primitive equation ocean models give a fundamentally different picture. Either approach can provide an adequate representation of ice shelf basal melting, given bespoke tuning of the turbulent boundary layer parameterization, and the melt rate is frequently the model output of primary interest. The apparent lack of sensitivity to the dynamical framework of the model used to simulate melting has arguably led to far greater emphasis being put on improving representations of ice–ocean turbulent transfer than on advancing our understanding of the essentially unknown and largely unobserved sub-ice-shelf circulation. To address this apparent oversight, we have developed the simplest possible model of the three-dimensional circulation beneath an ice shelf based on the viscous planetary geostrophic approximation. That theory supplements the basic geostrophic balance with vertical shear stresses, has historically provided powerful insight into the structure of the ocean’s wind-driven gyres, and forms a natural extension of the plume concept. Although dynamically simpler than many two-dimensional implementations of plume theory, our model captures a critical element missing from all plume formulations. The flow of water along the sloping ice shelf base generates a vertical velocity that must be matched in the deeper waters, while the vertical velocity there must be compatible with flow over and around seabed topography. That requirement to match the vertical velocities generated by surface and basal topography gives us a fundamentally different picture of the flow of water at the ice shelf base that drives ice shelf melting. An improved understanding of the physical controls on the sub-ice-shelf circulation brings many advantages, including but not limited to: a conceptual framework for interpreting sparse observations and for designing future targeted observational campaigns; a route to more effective parameterization of melting in standalone ice sheet models; a simplified dynamical core for a new generation of reduced complexity sub-ice circulation models; test cases for the verification of primitive equation models.
Glacier calving: observations and modelling
Richard Parsons, Sainan Sun, Hilmar Gudmundsson
Corresponding author: Richard Parsons
Corresponding author e-mail: richard.parsons@northumbria.ac.uk
A major source of uncertainty in projections of mass loss from ice sheets is calving, the process of icebergs separating from a glacier’s terminus. Calving can influence the dynamic response of glaciers by reducing backstresses that resist upstream ice flow, thus impacting rates of ice discharge. Accurately capturing calving processes within ice sheet models is therefore essential in modelling the future evolution of glaciers and reducing uncertainty in sea level rise projections. Ice-flow models describe calving in a number of different ways but there is still no consensus on the best approach. Recent work suggests that calving can be caused by exceeding a critical threshold of cliff height at the glacier terminus, giving rise to an unstable run-away process. Some of the higher-end predictions of near-future global sea level rely on models implementing such an unstable frontal retreat, a process termed ‘marine ice cliff instability’ (MICI). In this project, both observational and numerical methods are used to better understand the calving process and how best to implement calving numerically. Of particular interest is Crane Glacier, located on the Antarctic Peninsula, where, following the breakup of the Larsen-B ice shelf in 2002, tall ice-cliffs were exposed at the terminus. This is the only such event to have occurred within the timeframe of abundant, high-resolution remote sensing datasets, providing a unique opportunity to validate and constrain the MICI hypothesis.
Reconstructing the Holocene thinning of the Greenland Ice Sheet
Mikkel Lauritzen, Nicholas Rathmann, Bo Vinther, Guðfinna Aðalgeirsdóttir, Aslak Grindsted Grinsted, Anne Solgaard, Brice Noël, Mikkel Lauritzen, Christine Hvidberg
Corresponding author: Mikkel Lauritzen
Corresponding author e-mail: mikkel.lauritzen@nbi.ku.dk
During the Holocene, the Greenland Ice Sheet (GrIS) experienced substantial thinning, with some regions losing up to 600 m of ice. Geological evidence, including moraine lines in western Greenland, suggests that the GrIS once extended far beyond its current boundaries and was bridged with the Innuitian Ice sheet in the northwest. To investigate these long-term geometry changes and interpret their driving factors, we combine a Bayesian framework with the three-dimensional Parallel Ice Sheet Model (PISM). Here, we use thinning data derived from ice cores to inform PISM and simulate observed GrIS geometry changes as well as assess the model sensitivity to key parameters. We find that since the onset of the Holocene, GrIS mass loss has contributed 6.27 m to global sea level rise, which is consistent with ice-core-derived thinning curves spanning the time when the Greenland Ice Sheet and the Innuitian Ice Sheet were bridged. Our results suggest that this bridge collapsed approximately 7–10 ka before present.
Seasonal observations of the microstructure of snow in an Arctic environment
Pascal Hagenmuller, Neige Calonne, Julien Brondex, Kévin Fourteau, Rémi Granger, Pierre Lhuissier, Francois Tuzet, Louis Vedrine, Oscar Dick, Mathieu Fructus, Alvaro Robledano, Laurent Arnaud, Vincent Vionnet, Henning Loewe, Julien Méloche, Daniel Kramer, Alexandre Langlois, Florent Domine, Yannick Deliot, Jacques Roulle, Yves Lejeune, Lisa Bouvet, Basile de Fleurian
Corresponding author: Basile de Fleurian
Corresponding author e-mail: basile.defleurian@uib.no
The microstructure of snow controls its effective properties and consequently influences the snowpack physics. Snow cover across different regions of the Earth exhibits a diverse range of microstructures, leading to varied snowpack formations. Notably, there are significant differences between Arctic and Alpine snowpacks. Despite its importance in Earth’s climate system, Arctic snow microstructure has been less frequently studied compared to Alpine snow. Within the context of the ERC IVORI project, we conducted extensive measurement campaigns in an Arctic environment. Our observations included daily assessments of snow microstructure over a full snow season at the Canadian High Arctic Research Station in Nunavut, Canada. This presentation focuses on the preliminary datasets derived from these campaigns and offers insights into the evolution of snow microstructure across contrasting environments.
Transient reconstruction of Younger Dryas to present-day glacier evolution in the Alps constrained by the geological record
Andreas Henz, Tancrède P. M. Leger, Sarah Kamleitner, Samuel U. Nussbaumer, Guillaume Jouvet, Andreas Vieli
Corresponding author: Andreas Henz
Corresponding author e-mail: andreas.henz@geo.uzh.ch
Several studies have reconstructed specific Alpine glaciers using geomorphological, geochronological, and modelling tools. However, there has never been a detailed 3D reconstruction with model data fit for all glaciers in the entire Alps. This study aims to fill that gap. For this, we use the Instructed Glacier Model (IGM), which combines a high-order 3D ice flow with high computational efficiency through its deep learning accelerated emulator. The IGM enables us to model all glaciers of the European Alps in one simulation at 200 solution. To begin with, the IGM was applied to the advances of the Little Ice Age (LIA), as detailed and comprehensive glacier outlines are available. These LIA outlines were used in our static modelling approach, where each glacier was modelled and fitted to its corresponding LIA extent. The model allows for the derivation of ice surface geometries and volumes that are consistent with glacier physics and the principles of mass conservation. This approach infers equilibrium line altitudes for over 3000 individual glaciers and reveals spatially variable ELA patterns across the Alps. In contrast to the LIA, fewer empirical glacier outlines and ice marginal positions are available for the Younger Dryas and the Holocene glacier fluctuations. Nevertheless, there are numerous dated and interpreted palaeo-glacial landscape features across the Alps. These geomorphological and geochronological data are being compiled in an Alps-wide database (AlpIce) as part of the RECONCILE project. This new dataset will form the basis for constraining a transient glacier modelling reconstruction prior to the LIA. An automated workflow and scoring tool is introduced to evaluate and calibrate the modelled glacier history with the empirical record. The proposed workflow and scoring tool are first applied to selected catchments in the Alps with well dated and well mapped geomorphological features. Preliminary modelling results demonstrate the importance of transient simulations for glacier reconstruction and, therefore, the advantage of an automated scoring process. The modelling will eventually provide a fully non-stationary, high-resolution (200–100 m) reconstruction of glacier fluctuations from the Younger Dryas to the present, consistent with the geological record.
Assessing the role of machine learning in glacier mass balance modelling: a case study over large Himalayan glaciers
Ritu Anilkumar, Basile de Fleurian, Dibyajyoti Chutia, Shiv Prasad Aggarwal
Corresponding author: Basile de Fleurian
Corresponding author e-mail: basile.defleurian@uib.no
Understanding glacier mass changes is vital in assessing a glacial system’s response to climate change and the resulting water stability and hazard vulnerability. Incumbent techniques to measure and model mass balance include, but are not limited to, direct glaciological methods, geodetic methods, remote sensing methods, temperature index models and physics-driven models. Machine learning algorithms have recently been widely utilized in several Earth science applications. In light of these recent trends, it is crucial to assess the strengths and weaknesses of machine learning techniques from a mass balance modelling perspective. In this study, we consider the cases of (a) machine learning used directly in estimating mass balance using time series meteorological datasets and (b) aiding/improving existing techniques, specifically the indirect remote sensing estimation using equilibrium line altitude and the temperature index model. Each case study is systematically evaluated across spatial domains (glacier and multiple basin levels) for various model settings. Our primary experimental setup used the gradient boosted regression model as the machine learning model in line with the recommendations of past studies. We used meteorological drivers such as solar radiation, thermal radiation, albedo, latent heat flux, sensible heat flux, surface pressure, total precipitation and temperature derived using ERA5 Land Reanalysis datasets; topographic and geometric drivers such as equilibrium line altitude, aspect and area of the glacier derived using the MERIT DEM and MODIS datasets. The analysis is performed over all Himalayan glaciers with a size greater than 2 km². We use the performance metrics of root mean squared error, median absolute error and coefficient of determination. A statistically rigorous comparison is performed using the Paired Student’s T test for the models with and without machine learning integrations. We demonstrate that (1) in the case of the direct estimation, the machine learning model effectively captures the feature parameterization indicated by sensitivity studies over the region, (2) machine learning models outperform incumbent simple models at all spatial scales, (3) feature selection plays a primary role in the performance of machine learning models in a limited data setting, and (4) the effectiveness of machine learning models reduces when extending to larger spatial scales.
Representation of snow thermal conductivity controls future simulated winter carbon emissions in shrub-tundra
Johnny Rutherford, Leanne Wake, Alex Cannon, Nick Rutter
Corresponding author: Johnny Rutherford
Corresponding author e-mail: jonathan.a.d.rutherford@northumbria.ac.uk
The Arctic plays a key role in the Earth’s climate system, as in recent decades, air temperatures are increasing around three times faster than the global average. It is estimated that ~1700 Gt of carbon is stored in the permafrost ecosystems of the northern latitudes, accounting for half of the global soil organic carbon storage. Recent warming in the Arctic, which has been amplified during the winter, greatly enhances microbial decomposition of soil organic matter and subsequent release of carbon dioxide (CO²). Winter emissions are low but persistent, and cold season losses may exceed growing season uptake under future climate scenarios RCP 4.5 and 8.5. As such, the contribution of Arctic soils to the global annual carbon budget is expected to increase in a warming world. Assessing model simulations of heat and gas fluxes to subnivean soils is critical in understanding how well models, such as the Community Land Model (CLM5.0), can be expected to simulate future carbon cycling in the Arctic. In this study, we apply a bias-corrected NA-CORDEX meteorological ensemble of RCP 4.5 and 8.5 scenarios to CLM5.0, and present future simulations of snow water equivalent (SWE), ground temperature (TSOI and GT10), soil moisture (SOILLIQ) soil respiration (SR) and methane flux (FCH4) to 2100AD. Current CLM5.0 parameterization of snow thermal conductivity (Keff) produces colder winter soil temperatures than observed and little to no respiration during the winter. Changes to Keff increases winter CO₂ emissions by 120% compared to default parameters by 2100AD, under RCP 8.5. Additionally, we address CLM5.0 parameter uncertainties relating to the temperature sensitivity of soil respiration (Q10) and moisture threshold for soil decomposition (Ψmin) which contribute to poor model performance during the winter. CLM5.0 simulations show an increased duration of the autumn zero-curtain (when soil is at or around 0°C) by up to a month, which suggests that recent increases in both zero-curtain and winter CO₂ emissions are set to continue to 2100AD.
Rift growth and calving triggered by ocean tides and resulting in rapid acceleration of the Brunt Ice Shelf
Oliver J. Marsh, Rob Arthern, Jan De Rydt, Adrian Luckman, Dominic Hodgson
Corresponding author: Oliver J, Marsh
Corresponding author e-mail: olrs@bas.ac.uk
Tabular iceberg calving modifies ice-shelf extent, affecting ocean circulation and ice-sheet stability. Here we present observations of the growth of a rift on the Brunt Ice Shelf, East Antarctica, from 2017–23 and its behaviour in the lead up to calving in January 2023. We show that the timing of crack propagation is controlled by slowly varying ice dynamics and rapidly varying external influences including rate of change of ocean tide height, wind speed, and an iceberg collision in August 2021. Sub-hourly observations of crack opening rates at millimetre resolution reveal the effect of rising tides and strong winds. A viscoelastic rheological model is used to estimate the magnitude of stresses acting on the rift and to determine a critical threshold for fracture, which is exceeded during a sequence of propagation events in early 2019. The eventual calving on 22 January 2023 occurred at the peak of a spring tide, supporting the conclusion that tides directly influence the timing of crack growth. We also present the immediate response to the calving, observed as a change to the rate of acceleration of the ice shelf, not to velocity directly, with acceleration from a velocity of 900 m a⁻¹ to 1500 m a⁻¹ during 6 months following calving. Acceleration initially increased by a factor of 10, with a second, smaller calving at the end of June 2023 leading to further tripling of acceleration. The acceleration was caused by reduction of buttressing at the McDonald Ice Rumples, leading to high localized strain rates, which reduce the strength of the remaining ice shelf.
The simulated response of Antarctic ice flow to observed perturbations in ice-sheet geometry.
Jan De Rydt
Corresponding author: Jan De Rydt
Corresponding author e-mail: jan.rydt@northumbria.ac.uk
Numerical models of Antarctic ice flow routinely use measurements of surface velocity and ice thickness to estimate the unknown distribution of ice viscosity and basal slipperiness. The solution depends on the measurement errors and a range of uncertain model parameters. Most prominently, choices about prior information for viscosity and slipperiness, regularization parameters, and the functional form of the basal sliding law and ice rheology influence the solution. Often, a non-evolving best estimate of viscosity and slipperiness is subsequently used in transient simulations, which introduces largely unquantified uncertainties in the future evolution of the ice sheet. Here we take a step towards quantifying the importance of uncertain optimization parameters (prior information and regularization) and physical constants (flow law and sliding law exponents), by constructing a large ensemble of possible solutions for the viscosity and slipperiness distributions around Antarctica for the year 2000. We then perform diagnostic simulations for each ensemble member, whereby we perturb the ice sheet geometry in line with the observed changes in ice front location, ice thickness and grounding line location between 2000 and the present-day. The simulated changes in ice flow are compared to observed changes in ice velocity, which allows us to construct a posterior distribution of model parameters and their associated viscosity and slipperiness solutions. By construction, the maximum a posteriori contains the parameter set that best reproduces the relationships between observed changes in Antarctic Ice Sheet geometry and ice dynamics, and could be interpreted as the optimal observationally calibrated choice of model parameters. In future work, the posterior distribution will be incorporated in an ensemble of transient simulations with a larger set of uncertain physical parameters, such as basal melt parameters. The reduced uncertainty in the basal sliding law, ice rheology and inversion parameters will allow us to more confidently address the importance of other physical parameters and forcing scenarios for the future evolution of the ice sheet.
Transient electromagnetic imaging of basal marine ice in the Larsen C Ice Shelf, Antarctic Peninsula
Bernd Kulessa, Siobhan Killingbeck, Bryn Hubbard, Adrian Luckman, Katie Miles, Sarah Thompson, Suzanne Bevan, Stephen Cornford, Eduardo De Souza Neto, Glenn Jones
Corresponding author: Bernd Kulessa
Corresponding author e-mail: b.kulessa@swansea.ac.uk
Suture zones are present in all large and numerous smaller Antarctic ice shelves, stabilizing them by containing rifts within meteoric ice units derived from tributary glaciers. Basally accreted marine ice within such zones contains seawater and is warmer than surrounding meteoric ice, allowing suture zones to arrest rifts by accommodating strain and preventing brittle fracture. The presence of basal marine ice can be inferred from digital elevation models and hydrostatic considerations, and ocean models can simulate processes of frazil ice accretion and compaction. Many assumptions underly such indirect inferences, however, and field measurements are therefore required to ground-truth their presence, thickness and physical properties. However, detecting and delineating basal marine ice is notoriously difficult as radar energy cannot penetrate it, and seismic imaging or borehole observations are labour-intensive. Transient electromagnetic (TEM) soundings are relatively readily acquired, requiring only a transmitter with cable loop (100×100 m in our case) and an easily-deployed receiver (a small 3-D coil system in our case). Here we present a TEM-derived image of the basal marine ice layer in the Joerg Peninsula suture zone, in the southern Larsen C Ice Shelf (LCIS), which is modelled to be particularly critical to LCIS stability. The image is derived from 22 ground-based TEM soundings, acquired every 500 m along a 10 km long profile across this zone in the 2022/23 austral summer. Up to four such soundings were acquired per day, and basal marine ice thicknesses compare favourably with those inferred recently from ocean modelling, with ice thicknesses exceeding ~100 m in central areas and tapering out towards the suture zone margins. Complete ice-shelf anatomies can therefore be derived from combined TEM and radar imaging, as we demonstrate here for the LCIS. Indeed, akin to deployments of radar imagers on aircraft, airborne TEM imaging has also recently reached maturity for the imaging of ice sheets and glaciers. Future integrated applications of airborne TEM and radar sounding therefore promises to facilitate imaging of Antarctic ice shelf anatomies over large spatial scales; well suited for ground-truthing of ice-shelf and ocean models.
Validation of a new coupled ice–ocean model of the Amundsen Sea sector
Brad Reed, Kaitlin Naughten, Katherine Turner, Jan De Rydt
Corresponding author: Brad Reed
Corresponding author e-mail: brad.reed@northumbria.ac.uk
The Amundsen Sea sector of the West Antarctic ice sheet has undergone dramatic changes in recent decades, with increased ice loss, widespread thinning and retreating grounding lines. This has led to concerns about the current and future stability of the region and of the wider ice sheet, which has the potential to raise global mean sea level by several meters. In this sector of West Antarctica, mass loss is predominantly driven by basal melting at the coast, where vulnerable floating ice shelves are exposed to warm ocean waters below. However, internal ice dynamics can also play a huge role in how the ice sheet responds to ocean-induced melting. Therefore, to understand and predict how future forcing will impact the ice sheet, we must consider changes in both the ice and ocean systems and how they affect each other. Here we show preliminary results from a coupled ice–ocean modelling study of the Amundsen Sea sector. For the ice component, we use the ice flow model Úa to produce an initial ‘present-day’ configuration of the ice sheet. This is done by using a new two-stage optimization procedure which uses observations of ice velocities and thickness changes. This is then coupled offline to the MIT general circulation model, which has been forced with historical atmospheric conditions. This new coupled model will be run forward in time using idealized and projected conditions, and will evolve through exchanges of basal melt rates and ice shelf geometry. Validation of the model will be done using the most recent observations of ice velocities, thickness changes and integrated basal melt rates. This will help us to better understand the complex interplay between ice dynamics and ocean conditions in the Amundsen Sea sector and what impact this will have in future scenarios.
Self organization in tidewater glaciers and ice shelves: implications for calving laws
Iain Wheel, Douglas Benn
Corresponding author: Iain Wheel
Corresponding author e-mail: iw43@st-andrews.ac.uk
The complexity and range of calving processes at marine-terminating glaciers and ice shelves make it hard to develop models that can accurately simulate such behaviour across a wide range of environments. In higher-order models it is often tempting to include every process, making model use and development cumbersome, while in lower-order models the lack of certain processes increases the uncertainty of the modelled outcome. Within the complexity of calving processes two scales of order emerge from both observations and 3D deterministic models. The first is the order of the ice cliff, where oversteepening due to melt undercutting and ice flow lead to localized collapse. This system produces a calving magnitude–frequency distribution with a power-law form typical of self-organizing criticality. Secondly, order is seen through the scale of the ice tongue. Glaciers and ice shelves oscillate around stable positions between period of rapid transition to new meta-stable positions. Stable positions are often facilitated through pinning points while periods of transition can be invoked when environmental factors push the glacier off a pinned position. The more stochastic nature of ice tongue scale calving produces calving magnitude–frequency distributions that have an exponential form. Using the deterministic crevasse-depth calving law in Elmer/Ice, 3D modelling work at both Jakobshavn Isbræ and Thwaites Glacier has reliably predicted stable positions and calving magnitude-frequency distributions. Although 3D models produce two orders of calving self-organization, the order of the ice cliff can be parameterized by melt, meaning the concept is equally applicable to lower-level models. Longer relaxation times within an Antarctic setting mean that purely deterministic calving laws may not be able to replicate the decadal cycles associated with large ice shelves. We present the case for promoting a stochastic calving law that could more readily predict calving on longer timescales, potentially opening the possibility for a universal calving law that can be widely implemented across models.
A spatio-temporal ice loading model for Mýrdalsjökull Icecap, Iceland
Jonas Liebsch, Guðfinna Aðalgeirsdóttir, Joaquín M. C. Belart, Eyjólfur Magn&uacture;sson, Finnur Pálsson
Corresponding author: Jonas Liebsch
Corresponding author e-mail: liebschjonas@gmail.com
Changes in glacial loading impact the dynamic behavior of the subglacial volcano Katla, which is completely covered by Mýrdalsjökull. A thorough analysis of the changes in glacial loading during the last decades will improve the understanding of Katla’s response to both long-term and seasonal changes. Data from ICESat-2 and ArcticDEM offer valuable constraints on the temporal evolution of glacier elevation. Weather models can also contribute by providing essential information on mass fluxes and consequent changes in glacier elevation over time. However, weather reanalysis and forecasts such as CARRA can exhibit significant biases. We investigate the statistical relationship between weather variables and observed glacier elevation changes. The first steps to building a spatio-temporal model of glacier elevation will be to assess which variables are the best predictors. The unprecedented spatial and temporal resolution of glacial loading provides a unique opportunity for studies of crustal response and volcanic activity.
Glacier area and mass change along the South American Andes over the last five decades.
Owen King, Robert McNabb, Jonathan Carrivick, Daniel Falaschi, Iñigo Irarrazaval, Bethan Davies, Jeremy Ely
Corresponding author: Owen King
Corresponding author e-mail: owen.king@newcastle.ac.uk
Long-term observations of glacier area and mass balance changes are important prerequisites in the modelling of glacier extent, volume and ultimately meltwater yield over multi-decadal or centennial timescales. In the South American Andes, observations of glacier mass balance are sparse outside of the contemporary satellite era, with knowledge of glacier mass change over the period prior to 2000 generally restricted to individual glaciers or small samples of the local glacier population. Glacier inventory coverage and quality is also inconsistent, with global glacier inventories (RGI) suffering from methodological inconsistencies and a broad timestamp in the region. Our understanding of the response of glaciers across markedly different climatic zones along the Andes to recent anthropogenic climate change can therefore be improved. Here we present a record of glacier area and mass change between the 1960s and the present day in ten glacierized catchments in various climatic zones along the length of the mountain range. Our geodetic mass balance time series is based on digital elevation models (DEMs) generated from aerial photography (1960s), declassified Hexagon KH-9 spy satellite images (1972–83) and contemporary (1999–2023) ASTER imagery. We produced new glacier inventories from associated orthoimages, documenting the recession of >3000 glaciers in the region. Our results provide a substantial extension to the record of glacier fluctuations in South America and will be used in the calibration of glacier evolution models to project glacier recession and meltwater yield in the region towards 2150.
An investigation into observed summer colour changes of Icelandic proglacial lakes
Natasha Lee, Andrew Shepherd, Emily Hill, Rachel Carr
Corresponding author: Natasha Lee
Corresponding author e-mail: natasha.j.lee@northumbria.ac.uk
Proglacial lakes often form due to the availability of meltwater at a glacier margin. The greatest increase in proglacial lake area and volume is currently occurring in the Arctic. Surrounding M&uacture;lajökull outlet glacier, southeast Hofsjökull Iceland, a large number of proglacial lakes have formed as the glacier retreated. In this region a wide variety of the colour of the proglacial lakes has been observed within a small region. Seasonal and annual changes in proglacial lake colour have been observed using Landsat and Sentinel satellite images. Investigating differences between the sediment within these proglacial lakes is expected to provide a clearer understanding on the causes of the (change in) colour of proglacial lakes and their hydraulic connection with the outlet glaciers that supply water to them.
Comparing fully-coupled subglacial hydrology models and glaciofluvial landforms beneath the Fennoscandian Ice Sheet
Adam J. Hepburn, Christine F. Dow, Shivani Ehrenfeucht, Joni Mäkinen, Antti Ojala
Corresponding author: Adam J. Hepburn
Corresponding author e-mail: ajh24@aber.ac.uk
The subglacial hydrological system is an extremely dynamic environment and one that exerts a first-order control on the frictional resistance to ice-sheet flow. An accurate representation of subglacial processes in fully coupled models is critical if we are to faithfully bound future ice mass loss. However, considerable uncertainty persists in the parameterization of basal hydrology models because the subglacial environment remains largely inaccessible. As a result, important parameters, such as those describing the hydraulic conductivity of the system, are kept spatially and temporally constant. Conductivity, in particular, is likely to be highly variable over short distances as a function of sediment variability. As is already common in regional ice-sheet modelling, palaeo ice-sheet terrains offer a valuable means of testing and validating subglacial hydrology models and the variability of critical parameters. Glaciofluvial landforms are nearly ubiquitous across Northern Hemisphere glaciated terrain, and landforms such as eskers, tunnel valleys, murtoos and ribbed moraines preserve a signature of former drainage pathways that can be used to inform and guide modelling efforts during periods of rapid ice loss. Existing work modelling palaeo subglacial hydrology has typically prescribed a water pressure at or near ice overburden. Though extremely efficient to run, these models describe steady state conditions applicable over interannual–millennial scales, providing limited insight into subglacial processes known to operate over much shorter periods. Here, we run the Glacier Drainage System model (GlaDS), a subglacial hydrology model coupled to ice dynamics in the Ice-sheet and Sea-level System Model (ISSM). We apply this coupled framework to the Fennoscandian Ice Sheet (FIS) to i) explore the GlaDS parameter space by way of comparison to mapped geomorphology and ii) investigate the role of subglacial hydrology in the rapid retreat of the FIS. By comparing the output ice velocity, channel discharge and distributed water discharge to mapped geomorphology we can improve our understanding of the role that spatially variable subglacial conditions play in modulating ice dynamics.
Towards predictive modelling of Antarctica using the Úa-FESOM coupled ice–ocean framework
Jowan Barnes, Hilmar Gudmundsson, Ralph Timmermann
Corresponding author: Jowan Barnes
Corresponding author e-mail: jowan.barnes@northumbria.ac.uk
Melt rates under Antarctic ice shelves are a major driver of ice sheet dynamics and the sea level contributions resulting from mass loss. Interactions and feedbacks between the ocean and ice are critical to an accurate representation of this interface in modelling. Ice–ocean coupled models are the most accurate, yet also most expensive, approach. Many ice sheet models use unstructured horizontal meshes, giving them the ability to focus resolution on target areas while keeping the resolution coarse elsewhere, limiting the size of the computational mesh. However, most ocean models require a regular rectangular grid, so resolving the ocean in great enough detail while keeping a reasonable computational cost is a challenge for existing coupled setups. The Finite Element Sea ice–Ocean Model (FESOM) is a hydrostatic ocean model designed to run on an unstructured mesh, and therefore a very useful tool in addressing this issue. Work has been undertaken already on coupling FESOM with the ice sheet model Úa. The unique pairing of ice and ocean models both using unstructured horizontal meshes has the potential to run more efficiently than other coupled frameworks, and resolve critical regions in greater detail. Here we further demonstrate the capabilities of this framework in coupled pan-Antarctic simulations.
Scientific rationale for the RINGS efforts facilitating airborne geophysical surveys and relevant resarch of the Antarctic Ice Sheet margin
Kenichi Matsuoka, Tom Jordan, Geir Moholdt, Felicity McCormack, Kirsty Tinto, Xiangbin Cui, Fausto Ferraccioli, Rene Forsberg
Corresponding author: Kenichi Matsuoka
Corresponding author e-mail: kenichi.matsuoka@npolar.no
RINGS is an emerging and urgently needed international collaborative effort aimed at conducting airborne geophysics and other field surveys of the Antarctic Ice Sheet margin. This initiative utilizes field data, satellite data and models to better estimate the Antarctic contribution to global sea-level rise, both at present and in the future. According to the latest IPCC report, the Antarctic Ice Sheet accounts for nearly 50% of the ensemble uncertainty in projections of net sea-level rise by 2100, irrespective of emission scenarios. This uncertainty largely stems from our inadequate understanding of the interconnected, complex system – particularly concerning tipping points – of the Antarctic ice margin, where the ice sheet meets the ocean. The first international RINGS workshop in June 2022 identified scientific priorities along three rings: the primary ring at the current grounding line, the seaward ring across ice shelves and ice rises, and the landward ring located within a few tens of kilometres inland from the current grounding line where the future grounding line would be situated. To further develop the scientific rationale for the high-priority science topics, RINGS has analyzed a range of published data, including the Bedmap3 FAIR database, IBCSO version 2, modelled surface mass balance reconciliation, and results from ISMIP6. Here, we present an overview of these new findings to clarify the scientific rationale for conducting further bed topography surveys in the ice sheet margin.
How much does model weighting alter projections of ice sheet evolution?
Sophie Nowicki, Xiao Luo
Corresponding author: Sophie Nowicki
Corresponding author e-mail: smj_nowicki@hotmail.com
How the Greenland and Antarctic ice sheets will respond to current and future climate change in coming centuries remains the most uncertain component of sea level projections. One approach to sample the possible range of ice sheet evolution is to create an ensemble of ice sheet simulations using as many ice models as possible. The idea is that as each ice sheet model includes different processes or uses different ways to parameterize processes, the resulting ensemble and spread in projection will then capture some of the uncertainty in future sea level. The next question arises of how to combine these ensemble simulations: should the ‘one model, one vote’ approach be adopted, or should models be assigned some kinds of weight based on how well they reproduce certain observations, or should models’ similarity be explored and used to reduce the numbers of models used in the projections, or should something entirely different be used? Would any of these choices make a difference in terms of projections? We explore these ideas with the recent rich dataset resulting from the Ice Sheet Model Intercomparison Project for CMIP6 (ISMIP6).
Gaussian process emulation of a subglacial drainage model
Tim Hill, Gwenn Flowers, Derek Bingham, Matthew Hoffman
Corresponding author: Tim Hill
Corresponding author e-mail: tim_hill_2@sfu.ca
Models of subglacial drainage are key tools for understanding basal processes given the challenge of directly observing the subglacial environment. For the same reason, subglacial drainage models are subject to unquantified uncertainty in model outputs arising from uncertainty in inputs and parameterizations. Moreover, ice-sheet models do not routinely dynamically model effective pressure for coupling via a sliding law due to the long computation time associated with physics-based subglacial drainage modelling. Our goal is to develop a statistical emulator for the Glacier Drainage System (GlaDS) model, an example of a channel-resolving model, to address these limitations. The Gaussian process (GP) emulator, chosen for its uncertainty quantification properties, takes hydrology model parameters as inputs and predicts spatially resolved, seasonally varying water pressure based on a training set of GlaDS simulations spanning the input parameter space. We investigate the impact of emulator architectural choices and the size of the training dataset on prediction performance. Considering the broader applications of subglacial drainage models, we define additional scalar quantities of interest, including the fraction of total subglacial discharge routed through the channelized system, and compare the prediction error for spatiotemporally resolved and integrated quantities. The fast predictions obtained with the GP emulator, taking just a few seconds and with 1–8% error in predicting spatially resolved and seasonally varying water pressure, can enable future ice-sheet modelling studies to include a seasonally varying basal boundary condition without significantly increasing computation time. The GP emulator further provides a Bayesian framework for integrating observations with physics-based drainage models to construct calibrated subglacial drainage predictions.
Quantifying meltwater infiltration mechanisms in firn on the Greenland Ice Sheet with observational time series and numerical simulations
Joel Harper, Taylor Moon, Neil Humphrey
Corresponding author: Joel Harper
Corresponding author e-mail: joel@mso.umt.edu
Current model estimates are that 40–50% of meltwater generated on the surface of the Greenland Ice Sheet is retained by infiltration into the firn layer and refreezing. Meltwater infiltration occurs by a) top-down propagation of a wetting front, and b) inhomogeneous breakthrough of deeper piping events. There is a scarcity of available observations to constrain melt fluxes in firn, and no operational model at the ice sheet scale prescribes inhomogeneous deep infiltration processes. Consequently, the relative partitioning of meltwater fluxes between wetting fronts and heterogeneous piping is highly uncertain. Here, we place observational constraints on meltwater infiltration mechanisms, and we use data to test model simulations of meltwater infiltration. We utilize time series measurements of temperature during several summer melt seasons, which were collected in a network of 32 m deep boreholes drilled across a transect of Greenland’s percolation zone. The thermal signatures of meltwater penetration and refreezing yield time/space gradients of propagating wetting fronts, as well as transient piping events that can reach depths of 8 m. Using the dataset, we first use a numerical scheme based on the heat equation to partition refreezing quantities between the wetting front and deep piping events. Second, we test process-level theory for the rate of wetting front propagation. Finally, we test existing numerical simulations of firn evolution which are forced by regional climate models. Our findings have implications for modeling the present fate of meltwater and for ongoing transformation of the firn layer and impacts on future runoff/retention processes.
How can physics-informed deep learning help reveal the flow law of ice?
Ching-Yao Lai
Corresponding author: Ching-Yao Lai
Corresponding author e-mail: charlottelai007@gmail.com
Antarctic Ice shelves buttress the grounded ice sheet, mitigating global sea-level rise. However, their fundamental mechanical properties, such as flow law and viscosity structure, are largely unknown. Here, we develop a deep-learning-based framework to solve the inverse problem for inferring the viscosity field of ice shelves. Leveraging the remote-sensing data and physics-informed deep learning, we report evidence over several ice shelves that the flow law follows composite rheology in the compression zones. In contract, in the extension zone ice exhibits anisotropic properties. We construct the ice-shelf-wide anisotropic viscosity maps, which clearly capture the suture zones that were known to inhibit rift propagation. The inferred stress exponent near the grounding zone dictates the grounding-line ice flux and the grounding line stability. The viscosity maps can influence the prediction of rifts. Both are essential for improving ice-sheet models’ ability to predict the future mass loss of the Antarctic Ice Sheet.
Historically constrained projections of freshwater fluxes from Antarctica
Violaine Coulon, Frank Pattyn
Corresponding author: Violaine Coulon
Corresponding author e-mail: frank.pattyn@ulb.be
As global temperatures rise, Antarctica’s grounded ice sheet and floating ice shelves are experiencing accelerated mass loss, releasing meltwater into the Southern Ocean. This increasing freshwater discharge poses significant implications for global climate change. Despite these consequences, interactive ice sheets and ice shelves have generally not been included in coupled climate model simulations, such as those in CMIP6. Consequently, CMIP6 projections lack a detailed representation of spatiotemporal trends in ice-sheet freshwater fluxes and their impact on the global climate system, introducing major uncertainties in future climate and sea-level projections. To address this, we provide future Antarctic freshwater forcing data and uncertainty estimates for climate models. These are derived from an ensemble of historically calibrated standalone ice sheet model projections, produced with the Kori-ULB ice flow model, under different climate scenarios up to 2300. Here, we analyse spatiotemporal trends in calving rates, ice shelf basal melt and surface mass balance for all Antarctic ice shelves.
On the theoretical limitations of joint inversion for basal slipperiness and viscosity in ice-flow models
Camilla Schelpe, Hilmar Gudmundsson
Corresponding author: Camilla Schelpe
Corresponding author e-mail: camilla.schelpe@northumbria.ac.uk
Using linear perturbation analysis of the shallow ice stream equations, we demonstrate that in general it is possible to uniquely retrieve ice viscosity and basal slipperiness within ice flow models through inversion of surface data, with the quality of the retrieval dependent on the surface data errors and length scale of the features to be resolved. In the special limiting case where there is no horizontal variation in the velocity fields, we show that there is no response in the surface fields to variations in viscosity to first order, and so viscosity cannot as easily be resolved. We follow the theoretical framework devised by Gudmundsson & Raymond (2008) to determine the resolving power of Bayesian inversion methods applied to ice flow models. We find that there is a hierarchy in how well different model inputs can be retrieved through inversion: features in basal topography are very well resolved; basal slipperiness is well resolved at wavelengths above 10 ice thicknesses in the absence of viscosity mixing effects; while features in viscosity can be resolved at intermediate wavelength (10–10³ ice thicknesses) in regions of the ice sheet that are experiencing some horizontal stress. In the presence of data errors, the ability to uniquely determine both slipperiness and viscosity becomes increasingly challenging at longer wavelengths (of greater than 10³ ice thicknesses).
Evaluating the consistency of subglacial overdeepenings derived from different digital elevation models and ice thickness models
Ipseeta Nayak, Duncan Quincey, Jonathan Carrivick
Corresponding author: Duncan Quincey
Corresponding author e-mail: d.j.quincey@leeds.ac.uk
Glacial lakes govern the glacier dynamics of a region and are important water resources within the glacier environment. These lakes evolve from subglacial overdeepenings. Satellite imagery, ice thickness models and digital elevation models (DEMs) can be combined to estimate the position of the overdeepenings. Studies have used this method to estimate the overdeepenings where the DEM and model are presumed perfect, with limited consideration for uncertainty assessment or exploration of a range of plausible values. Our study aims to test the consistency of different DEM and ice thickness products in identifying the locations, sizes and depths of subglacial overdeepenings within mountain glacial environments. The study area consists of 170 glaciers in the Central Himalayan region. Shuttle Radar Topography Mission (SRTM), Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), Advanced Land Observing Satellite (ALOS) Phased Array L-band Synthetic Aperture Radar (PALSAR) and Copernicus DEMs were used with four ice thickness models to generate overdeepenings. These four models were: the Open Glacier Global Model (OGGM), GlabTop2, Millan’s ice thickness model and Farinotti’s ensemble. Each DEM was used with each ice thickness, which resulted in 16 sets of overdeepenings. Parameters such as the total number, total surface area, mean surface area and mean depth of the overdeepenings were calculated and compared for different combinations. We found that the differences in the parameters were governed by the ice thickness models rather than the DEMs. The total number of overdeepenings ranges from 30 t–87 and the total surface area of the overdeepenings ranges from 6.56–19.09 km² for all the combinations. OGGM overestimated the total area and the number of overdeepenings whereas, Farinotti’s ensemble underestimated both the parameters and also the mean depth. However, both these models overestimated the mean surface area. This shows that, although there is inconsistency among different ice thickness models, it is not uniform throughout and would depend on the parameter being examined.
Deplete and retreat in the Andean water towers: data-model comparisons for improving understanding of glacio-hydrological processes
Bethan Davies, Owen King, Jeremy Ely, Jan Baiker, Claudio Bravo, Wouter Buytaert, Jonathan Carrivick, Alejandro Duissaillant, Ines Duissaillant, Fabien Drenkhan, Dina Farfan Flores, Juan Luis-Garcia, Thomas Gribbin, Iñigo Irarrázaval, Tom Matthews, Nilton Mariano Montoya Jara, Robert McNabb
Corresponding author: Bethan Davies
Corresponding author e-mail: bethan.davies@newcastle.ac.uk
The aim of the NERC-funded ‘Deplete and Retreat’ project is to assess the sensitivity of water resources across the Andes. The food and water security of 90 million people depends on the Andean mountain water tower, and is at risk in several regions, because climate change is depleting the stores of water held in snow and glacier ice. These cryospheric changes are spatio-temporally complex; snowfall amounts are declining but are increasingly delivered in extreme events, and glacio-climatic feedbacks modulate glacier mass loss. Understanding the changing role of the cryosphere in catchment hydrology is vital for long-term water resource management. However, there is an absence of data across the Andes, especially at high altitudes and latitudes. The spatial and temporal coverage of observational data limits our ability to understand cryospheric changes across the Andes and to evaluate model accuracy. The temporal record is especially limited, with the majority of studies focussing on the past two decades. In this project, we are upskilling model ability by combining state-of-the-art regional climate, ice flow and hydrology models, evaluated against newly collated datasets. Data for model evaluation collected in Work Package (WP) 1 of ‘Deplete and Retreat’ focuses on expanding spatial and temporal data coverage on the changing climate, cryosphere and hydrology across the Andes. To date, we have conducted four field campaigns across Peru, Bolivia and Chile to gather direct measurements in data-poor catchments, and are generating a data-rich record of historical glacier change from 1850 CE to present. Together, these data will improve our understanding of climate change impacts on the cryosphere and hydrology across the Andes and will be used to evaluate models in WPs 2–4. Achievements to date include the installation of: four automatic weather stations capable of measuring solid precipitation; three smart ablation stakes; numerous hydrological sensors; generation a new reconstruction of glacier extent across the Andes at ~1850 CE, a new and internally consistent glacier inventory in 2022 CE, and reconstruction of glacier thinning from the 1960s to the present in each catchment. Here, we focus on the design, methods of collection and preliminary results of empirical data designed to be used in data-model comparisons.
Impact of model initialization on future projections of the Greenland ice sheet evolution
Ana Carolina Moraes Luzardi, Sophie Nowicki, Denis Felikson, Joshua Cuzzone, Beata Csatho, Jason Briner, Kristin Poinar
Corresponding author: Ana Carolina Moraes Luzardi
Corresponding author e-mail: amoraesl@buffalo.edu
Model initialization comprises one of the main sources of uncertainty in simulations that project the contribution of the Greenland ice sheet (GrIS) to future sea level rise. There are two main approaches when it comes to initializing ice sheet models. On the one hand, data assimilation captures well the ice sheet geometry at the start of the projections period, but can have significant model drift due to inconsistency of datasets and fails to capture the memory of the ice sheet of past climate variability. On the other hand, glacial cycle spin-ups start projections with self-consistent variables, resulting in a reduced model drift, and capture the response of the ice sheet to past climatic changes, but often begin projections with a geometry that is significantly different from that observed. Currently, there is no consensus in the ice sheet modelling community on which is the optimum approach. Initializing simulations in the 19th century with constraints of past ice sheet geometry could be a solution to minimize model drift in the projections period while maintaining a realistic geometry. However, there are no available constraints of Greenland-wide ice sheet configuration that are consistent in space and time, and state-of-art ice sheet projections are initialized to present-day observations even when simulations start in 1850 onwards. Therefore, it remains unresolved whether constraining ice sheet geometry in the past would help to derive more credible projections of ice sheet change. To help elucidate this question, we explore here the interplay between initial geometry, length of simulations and model drift on simulations of the GrIS evolution. We run GrIS simulations with the Ice-sheet and Sea-level System Model (ISSM) that start at 1850, 1980 and 2015 with different initial geometries to evaluate the impact of initial state and start date at future projections. The different geometries used for initialization stem from present-day observations, the product of a glacial-cycle spin up and a combination of past constrains, mass balance estimates from literature and numerical methods, to derive a Greenland-wide past ice elevation product. Ultimately, our results will help to inform the protocol for ISMIP7.
Modeling the impact of stochastic iceberg calving on ice sheet dynamics
Aminat Ambelorun, Alexander Robel
Corresponding author: Aminat Ambelorun
Corresponding author e-mail: ambelorunaminat@gmail.com
Iceberg calving is one of the dominant sources of ice loss from the Antarctic and Greenland Ice sheets. Iceberg calving is still one of the most poorly understood aspects of ice sheet dynamics because of its variability at a wide range of spatial and temporal scales. Despite this variability, most current large-scale ice sheet models assume that calving can be represented as a deterministic flux. Failure to parameterize calving accurately in predictive models could lead to large errors in warming-induced sea-level rise. In this study, we introduce stochastic calving within a one-dimensional depth-integrated tidewater glacier and ice shelf models to determine how changes in the calving style and size distribution of calving events cause changes in glacier state. We apply stochastic variability in the calving rate by drawing the calving rate from two different probability distributions. We also quantify the time scale on which individual calving events need to be resolved within a stochastic calving model to accurately simulate the probabilistic distribution of glacier state. We find that incorporating stochastic calving with a glacier model with buttressing ice shelves changes the simulated mean glacier state, due to nonlinearities in the terminus dynamics. This has important implications for the intrinsic biases in current ice sheet models, none of which include stochastic processes. Additionally, changes in calving frequency, without changes in total calving flux, lead to substantial changes in the distribution of glacier state. This new approach to modeling calving provides a framework for ongoing work to implement stochastic calving capabilities in large-scale ice sheet models, which should improve our capability to make well-constrained predictions of future ice sheet change.
Impact of assumptions for Glen’s flow law exponent: 30% greater Amundsen Sea Embayment ice loss by 2100 if n = 4
Benjamin Getraer, Mathieu Morlighem
Corresponding author: Benjamin Getraer
Corresponding author e-mail: benjamingetraer@gmail.com
Projections of future sea level rise from Antarctica require assumptions of ice rheological properties, which dictate how sensitive ice deformation is to changes in the stress field. Ice is typically modeled using a power law stress–strain rate relation known as Glen’s Flow Law, with a power n being assumed to be 3. Recent research has shown that n = 3 may not be uniformly true everywhere, and that a value of n = 4 may be a more accurate assumption. However, we still do not have a quantitative understanding of how much this number matters in terms of future contribution of the ice sheet to sea level rise, and its relative importance compared to other sources of uncertainty such as climate forcings. Here, we present a series of experiments using the Ice-sheet and Sea-level System Model (ISSM) to conduct forward simulations of the Amundsen Sea Embayment glaciers in West Antarctica, demonstrating the impact of assuming n = 3 as opposed to n = 4, in comparison with climate forcing. The models are designed to have identical initial conditions, within machine precision, in order to capture the effect of the power-law exponent alone. Each simulation is run until 2300, and the total change in volume above flotation is used as a metric for comparison. Each of the top six targeted CMIP experiments selected for ISMIP6 are used as climate forcings in the simulations, leading to a total of 12 experiments: six climate forcings for n = 3, and the same six climate forcings for n = 4. Our results show that changing n from 3 to 4 yields an increase in ice loss of ∼30% by 2100 and ∼67% by 2300. The difference in ice loss outcomes due to choice of n reaches a similar magnitude to the spread in ice loss outcomes from different ISMIP6 climate forcings by about 2100.
Magnetotelluric imaging of deep subglacial conditions beneath Thwaites Glacier and WAIS Divide
Siobhan Killingbeck, Bernd Kulessa, Alex Brisbourne, Rebecca Pearce, Louise Borthwick, Felipe Napoleoni, Sridhar Anandakrishnan, Martyn Unsworth
Corresponding author: Siobhan Killingbeck
Corresponding author e-mail: s.f.killingbeck@swansea.ac.uk
Magnetotelluric (MT) data were acquired on Thwaites Glacier (TG) to determine the composition and temperature of subglacial material and crust. In 2022/23, four MT stations were acquired at WAIS Divide, a control site away from TG, where the ice flow is relatively slow. In 2023/24, five MT stations were acquired on the main trunk of TG, with the furthest downstream station located on Ghost Ridge (70 km inland of the current grounding line) and four stations located up to 80 km upstream, towards WAIS Divide. Here, we present results from the recently acquired MT data at TG and WAIS. The MT data were collected using the Phoenix Geophysics MTU-5C system with titanium sheet electrodes and specialized high-impedance amplifiers. At all stations, the high-frequency data (100–1 Hz) are characterized by apparent resistivity that does not vary with rotation angle, suggesting that the near surface resistivity structure is one-dimensional within and immediately below the ice. At mid to low frequencies (10–0.001 Hz) apparent resistivity varies with rotation angle, suggesting a two-dimensional or three-dimensional resistivity structure within the subglacial material and crust. A 1D inversion was applied to the apparent resistivity and phase at each station and produced a profile of resistivity with depth. The MT data show a significant variation in resistivity between TG and WAIS. This difference in resistivity can be caused by 1) a lithological change, 2) changes in fluid saturation, 3) changes in the salinity of the fluids, and/or 4) changes in crustal temperatures. Future work will aim to separate the impacts of these four factors on our resistivity models, by joint interpretation of relevant additional geophysical observations acquired in the survey locations. Our observations will allow inferences to be made of the subglacial conditions, sediment distribution, groundwater and geothermal heat flow (GHF) beneath TG, a location critical for the accurate prediction of Antarctic ice mass loss. Further, current models of GHF beneath Antarctica differ significantly depending on the geophysical technique used. The heat flow beneath TG, is proposed to be amongst the highest in Antarctica (>120 mW m⁻²), and significantly higher than the average value estimated for East Antarctica (66 mW m⁻²). Therefore, MT data observations at TG are highly valuable for constraining heat flow models and ice dynamic models to reliably predict the future contribution of TG to Antarctic ice mass loss and resulting sea level rise.
Stability of radially spreading extensional flows and ice shelves
Lielle Stern, Roiy Sayag
Corresponding author: Roiy Sayag
Corresponding author e-mail: roiy@bgu.ac.il
Ice shelves that spread into the ocean can develop rifts, which can trigger iceberg calving and enhance ocean-induced melting. Fluid mechanically, this system is analogous to the radial propagation of a non-Newtonian, strain-rate-softening fluid representing ice that displaces a relatively inviscid and denser fluid that represents an ocean. Laboratory experiments showed that rift patterns can emerge in such systems and that the number of rifts declines in time. Such a dynamics was confirmed theoretically, but only for the earlier stage of the flow and for a fluid layer of uniform thickness. We investigate numerically the stability and late-time evolution of radially spreading, axisymmetric fluid layer of non-uniform thickness. We validate the two dimensional finite-element Úa model using similarity solutions of radially spreading layers of Newtonian fluid that were found to be consistent with laboratory experiments. We then explore the stability of the flow by introducing geometric perturbations to the initial front and tracing their evolution. Our simulations show that the front of Newtonian fluids is stable, although memory of the perturbation spectral form persists.
Impact of topography and meteorological forcing on snow simulation in the Canadian Land Surface Scheme Including Biogeochemical Cycles (CLASSIC)
Libo Wang, Lawrence Mudryk, Joe R. Melton, Jason Cole, Paul Bartlett
Corresponding author: Libo Wang
Corresponding author e-mail: Libo.Wang@ec.gc.ca
This study evaluates two snow cover fraction (SCF) parameterization schemes on snow simulation in the Canadian Land Surface Scheme Including Biogeochemical Cycles (CLASSIC). In the first scheme, snow cover is considered complete when the diagnosed snow depth (SD) reaches 0.1 m; when SD is less than this threshold SCF is calculated as SD/0.1 m. This scheme (Control) has been used in the operational version of CLASSIC, the land component of the Canadian Earth System Models (CanESM2&5). Both versions of CanESM have been found to overestimate SCF with a cold temperature bias (Tas) in the mountains. The second scheme was developed by Swenson & Lawrence (2012, SL12), who proposed an analytical SCF parameterization that reproduces the general features of the observed SCF-SD relationship, including the topographic effects of subgrid terrain. Our evaluation focuses on western Canada (WCA) and the high mountain Asia (HMA) region where most ESMs exhibit large SCF overestimation. To account for uncertainties in precipitation (Pr) data, which can be large over mountains, three reanalysis-based meteorological forcing datasets are used to drive CLASSIC: ERA5, CRUJRA, and GSWP3-W5E5. We evaluate the forcing and the model outputs separately over WCA and HMA using observation-based datasets. As expected, the spatial distribution patterns in Pr combined with the Tas biases explain most of the snow mass patterns simulated by CLASSIC in both regions. The results show that in WCA, differences in simulated SCF are minimal using either scheme during the accumulation season. During the ablation season using the SL12 scheme substantially reduces the SCF overestimation compared to the control simulations (RMSE from 0.2–0.14). In the HMA, the SL12 simulations show much reduced SCF bias year-round compared to the control simulations (RMSE from 0.2–0.15). Although the SL12 simulations show smaller SCF biases than the control simulations when driven by any of the three forcing datasets, the monthly SCF in the ERA5-driven simulation results in the closest agreement with the observation-based datasets over WCA, while SCF in the GSWP3-W5E5-driven simulation shows better agreement for HMA. These results underscore the importance of the topographic effects in SCF parameterizations and meteorological forcing in offline simulation of snow cover in land surface models.
Impact of geothermal heat flow choice on Greenland ice sheet spin up
Dominik Fahrner, Tong Zhang, William Colgan, Agnes Wansing, Anja Løkkegaard, Gunter Leguy, William H. Lipscomb, Cunde Xiao, Joseph MacGregor, Shfaqat Abbas Khan
Corresponding author: William Colgan
Corresponding author e-mail: wic@geus.dk
There is currently poor scientific agreement on whether the ice–bed interface is frozen or thawed beneath approximately one third of the Greenland ice sheet. This disagreement in basal thermal state results, at least partly, from differences in the subglacial geothermal heat-flow basal boundary condition used in different ice-flow models. We present seven widely used Greenland geothermal heat-flow maps in both nudged and transient spin-ups of the Community Ice Sheet Model (CISM). Across the seven heat-flow maps, and regardless of nudged or transient spin-up, the spread in basal ice temperatures exceeds 10°C over large areas of the ice–bed interface. Under transient spin-up, thawed-bed area ranges from 33.5 % to 60.0 % across the seven heat-flow maps. We compare simulated ice-bed temperatures to observations at 27 ice-sheet boreholes. Choice of heat flow map and spin-up type impact the fidelity with which these observations are reproduced. The majority of the heat flow maps yield a thawed North Greenland, which is difficult to reconcile with the Last Glacial Period ice outcrops preserved there today. This highlights the direct, and non-trivial, influence of the heat-flow boundary condition on the simulated equilibrium thermal state of the ice sheet. Of the 21 Greenland model submissions to ISMIP6, 12 prescribed geothermal heat flow from a 20-year-old global heat flow product with limited evaluation data in Greenland. Consequently, the basal thermal state depicted by the ISMIP6 ensemble is biased towards this single geothermal heat flow product with a known warm bias in south Greenland. We suggest that ISMIP7 should only employ newer and better validated geothermal heat-flow maps, namely those exhibiting high agreement against comprehensive observation datasets. Similar to climatic forcing, ISMIP7 should explore the influence of geothermal heat flow forcing on simulated thermal state and ice flow.
Topographically constrained tipping point for complete Greenland Ice Sheet melt
Michele Petrini, Meike D.W. Scherrenberg, Laura Muntjewerf, Miren Vizcaino, William H. Lipscomb, Gunter Leguy, Heiko Goelzer
Corresponding author: Michele Petrini
Corresponding author e-mail: mpet@norceresearch.no
A major impact of anthropogenic climate change is the potential triggering of tipping points, such as the complete loss of the Greenland Ice Sheet (GrIS). Currently, the GrIS is losing mass at an accelerated pace, mainly due to a steep decrease in its surface mass balance (SMB, snow accumulation minus surface ablation from melt and associated runoff). Here, we investigate a potential SMB threshold for complete GrIS melt, the processes that control this threshold, and whether it exhibits characteristics commonly associated with tipping points, such as a non-linear response to external forcings. To do this, we adopt a semi-coupled approach, forcing the Community Ice Sheet Model v.2 (CISM2) with different SMB levels previously calculated at multiple elevation classes with the Community Earth System Model v.2 (CESM2). The SMB calculation in CESM2 and the elevation class method allow us to account for the SMB–elevation feedback based on snow/firn processes and energy fluxes at the ice sheet surface. We find a positive SMB threshold for complete GrIS melt of 230±84 Gt a⁻¹, corresponding to a 60% decrease from the GrIS simulated pre-industrial SMB. The ice sheet shows a highly non-linear response to sustained melt, and its final state is determined by the effect of the SMB-height feedback in response to surface melt and Glacial Isostatic Adjustment (GIA). The GrIS is tipping from nearly 50% equilibrium volume towards complete melt when the ice margin in the central west unpins from a coastal region with high bedrock elevation and SMB. We find that this relatively small coastal region is important to determine the ice sheet stability in response to sustained warming. On the basis of the ice sheet geometry in previous modelling studies of the GrIS during the last interglacial, we suggest that a stabilizing effect of the bedrock topography in the central West might have occurred in the past.
Improving, evaluating and sharing projections of global mean sea level change to 2300
Tamsin Edwards, Fiona Turner, Victor Malagon Santos, Aimee Slangen
Corresponding author: Tamsin Edwards
Corresponding author e-mail: tamsin.edwards@kcl.ac.uk
Projections of the ice sheet and glacier contributions to sea level rise to 2100 in the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report were made by representing physical models with statistical ‘emulators’ or machine learning techniques. This allowed estimation of the impacts of several kinds of model uncertainty on sea level projections: multiple models for the ice sheets and glaciers, multiple settings determining ice sheet sensitivity to climate change, and multiple estimates of global warming, as well as uncertainty from the emulators themselves. However, there were some limitations, including: predicting each year of the century independently (i.e. not providing smooth time series or the possibility to assess rates of change), beginning physical model simulations in 2015 (not allowing evaluation with observations), and exploring a small number of possible model settings. Projections beyond 2100 also had to be estimated for the IPCC with other methods. These limitations presented difficulties for users. We improve on these projections here in their usefulness and robustness for coastal impacts communities and decision-makers. Usefulness by: predicting ice sheet and glacier changes to 2300, not 2100; providing smooth time series; and incorporating the emulators into the community FACTS sea level calculation framework for use by others. Robustness by systematically exploring many more model settings than before (including, for the first time, those for glacier models), and beginning in the past to allow calibration of the projections with observations. The result is more meaningful trajectories of sea level contribution from each land ice source, in which we have greater confidence. We combine these in FACTS with estimates of thermal expansion and land water changes and show new projections of global mean sea level change to 2300. This work was carried out by the EU Horizon 2020 project PROTECT (https://protect-slr.eu).
Long-lasting and irreversible Antarctic ice loss caused by warming overshoots
Ann Kristin Klose, Ricarda Winkelmann
Corresponding author: Ann Kristin Klose
Corresponding author e-mail: akklose@pik-potsdam.de
Earth’s climate will probably exceed a warming of 1.5°C in the coming decades. Maintaining such warming levels for a longer period of time may pose a considerable risk of crossing critical thresholds in Antarctica and thus triggering self-sustained, potentially irreversible ice loss, even if the forcing is reduced in a temperature overshoot. Due to the complex interplay of several amplifying and dampening feedbacks at play in Antarctica, the duration and magnitude of such warming overshoots as well as their eventual ‘landing’ climate will determine the long-term evolution of the ice sheet. Using the Parallel Ice Sheet Model, we systematically test for the reversibility of committed large-scale ice-sheet changes triggered by warming projected over the next centuries. We demonstrate crucial features of the Antarctic Ice Sheet’s stability landscape for its long-term trajectory in response to future human actions: given ice-sheet inertia, an early reversal of climate may allow for avoiding self-sustained ice loss that would otherwise be irreversible (for the same reduction in warming) due to multistability of the ice sheet at the basin- and continental scale. While we show that such ‘safe’ overshoots of critical thresholds in Antarctica may be possible, it is also clear that limiting global warming is the only viable option to evade the risk of widespread ice loss in the long term.
Explainable AI aids the development of a dynamically enhanced temperature index glacier mass balance model that outperforms traditional empirical models
Ritu Anilkumar, Rishikesh Bharti, Dibyajyoti Chutia
Corresponding author: Ritu Anilkumar
Corresponding author e-mail: r.anilkumar@iitg.ac.in
Temperature index models have been extensively utilized to estimate glacier mass balance. Most temperature index models use parameterizations associated with temperature and precipitation. Studies have tried to improve these models by including additional dependencies that better represent the physical processes associated with glacier mass balance modelling. This includes the development of enhanced temperature index and related empirical models. Sensitivity analysis of such models shows spatio-temporally varying relevance of different meteorological parameters, indicating that a single model with fixed parameter calibrations might not well represent the mass balance estimates. In this study, we aim to develop a dynamic, physically relevant empirical formulation to model glacier mass balance using explainable artificial intelligence (XAI) techniques. We propose integrating the XAI algorithm, specifically feature importance-based methods such as Shapley Additive Explanations, with machine-learning-driven mass balance modelling and reanalysis datasets to identify feature importance for varying spatio-temporal considerations. The feature importance values are then used in an unsupervised learning setting (DBScan clustering) to identify clusters of glaciers with similar feature importance values associated with the input meteorological and topographic forcing. The next step is integrating these clusters with the enhanced temperature index models. For this, we generate an ensemble of models using multiple sets of input forcing. The XAI-derived clusters are matched to the input forcing within the ensemble of empirical models, and the appropriate model is selected on a case-by-case basis. Finally, we compare the performance of such models with existing temperature index and enhanced temperature index models. Our study is relevant from the context of understanding the spatially and temporally varying role of meteorological variables impacting the glacier mass balance process. It moves one step closer to an improved physical representation in empirical models and can be easily integrated with established modelling frameworks such as the Open Global Glacier Model and Instructed Glacier Model to permit regional to global scale analysis.
Quantifying suspended sediment export from the Kangerlussuaq region of West Greenland (2017–23)
Holly Bartlett, Kate Winter, John Woodward, Neil Ross, James Lea
Corresponding author: Holly Bartlett
Corresponding author e-mail: holly.bartlett@northumbria.ac.uk
The Greenland Ice Sheet (GrIS) is a hotspot for sediment transport, where glaciers and their meltwaters deliver sediments and associated nutrients to vast proglacial outwash plains and into the ocean. This outflux supports highly productive ecosystems where primary production can drawdown atmospheric CO₂ with resultant negative carbon feedback loops. As field measurements of sediment export are limited both spatially and temporally, we interrogate Sentinel-2 satellite images with Google Earth Engine code to generate the first region-wide estimates for annual sediment flux from the GrIS using high-resolution 10 m imagery. Initially, we focus on proglacial meltwater exports from Leverett Glacier, Russell Glacier and the Ørkendalen Glaciers in the Kangerlussuaq region of West Greenland, where ice melt flows into the Ørkendalen and Sandflugtdalen tributaries and Watson River. We mask out proglacial meltwater in cloud-free satellite data (collected over the period 2017–23) using MNDWI, and use the associated surface reflectance values of the water to calculate suspended sediment concentrations (SSC) across the system. We find that total annual sediment flux across our Kangerlussuaq study site is ~15×10⁹&thikg and that sediment export has increased by approximately 1.1×10⁹ kg each year over our 6 year study period. Whilst Ørkendalen tTributary provides on average 18% more meltwater to Watson River (and therefore the Kangerlussuaq Fjord system) than Sandflugtdalen Tributary, the increased water flow is only responsible for transporting on average 8% more sediment flux. The highest sediment export in our study area belongs to the outflows of Leverett Glacier, which contributed >9 times the sediment flux of Russell Glacier to Sandflugtdalen Tributary over our 6 year study period. Our study provides the first high-resolution remote sensing analysis of meltwater and sediment flux through the Kangerlussuaq foreland system. Following field validations of SSC and meltwater flux we plan to upscale our research to the entirety of the GrIS. Results will provide baseline measurements of export and change over time, helping to ground glacial sediment transport models and predictive forecasts of change.
Sensitivity of coupled climate and ice sheet simulations of modern Greenland to atmospheric, snow and ice sheet parameters
Charlotte Lang, Tamsin Edwards, Jonathan Owen, Sam Sherriff-Tadano, Jonathan Gregory, Ruza Ivanovic, Lauren Gregoire, Robin S. Smith
Corresponding author: Charlotte Lang
Corresponding author e-mail: charlotte.lang@reading.ac.uk
As part of a project working to improve coupled climate-ice sheet modelling by studying the response of ice sheets to changes in climate across different periods since the Last Glacial Maximum, we present an analysis of an ensemble of coupled climate and ice sheet simulations of the modern Greenland using the FAMOUS-BISICLES model and statistical emulation. FAMOUS-BISICLES, a variant of FAMOUS-ice (Smith et al., 2021), is a low-resolution global climate model that is two-way coupled to a higher-resolution adaptive mesh ice sheet model, BISICLES. It uses a system of elevation classes to downscale the lower-resolution atmospheric variables onto the ice sheet grid and calculates surface mass balance using a multilayer snow model. FAMOUS-ice is computationally affordable enough to simulate the millennial evolution of the coupled climate–ice-sheet system as well as to run large ensembles of simulations. The ice sheet volume and area are sensitive to a number of parametrizations related to atmospheric and snow surface processes and ice sheet dynamics. Based on that, we designed a perturbed parameters ensemble using a Latin Hypercube sampling technique and ran simulations with climate forcings appropriate for the late 20th century. Gaussian process emulation allows us explore parameter space in a more systematic and faster way than with more complex earth system models and make predictions at input parameter values that are not evaluated in the simulations. We find that the mass balance is significantly correlated to six parameters:
• rho_threshold, setting the minimum value the dense firn albedo can possibly reach
• n, the exponent in Glen’s flow law, and
• beta, the basal drag coefficient, both influencing the amount of ice lost through discharge
• AV_GR and
• daice, involved in the parametrization of snow and bare ice albedo
• CT, the conversion rate of cloud liquid water droplets to precipitation
Using a history matching approach, we built an implausibility metric (based on Greenland integrated and regional (surface) mass balance) to identify the regions of the parameter space that produce plausible runs. We compare them to those producing plausible LGM North American ice sheets. Finally, we compare projections of Greenland’s future sea level contribution constrained by the modern Greenland plausible parameter space only and the intersection of the modern Greenland and the LGM North American ones.
What’s in a number? (implications of n& =&thin4)
Daniel Martin, Samuel Kachuck, Joanna Millstein, Brent Minchew
Corresponding author: Daniel Martin
Corresponding author e-mail: DFMartin@lbl.gov
Ice is a non-Newtonian fluid the rheology of which is typically described using Glen’s flow law, a power-law relationship between stress and strain rate with a stress exponent, n, generally taken to be 3. Recent observation-based work suggests that a more accurate choice for the Glen’s law exponent in high-strain regions like ice shelves may be n& =&thin4, implying that ice viscosity is more sensitive to changes in stress than is generally assumed. The implications of a higher stress exponent for ice sheet models and their projections of ice sheet response to climate forcing are unclear and are likely to be complex. Rheological parameters, such as ice viscosity, are fundamental to ice sheet dynamics and influence the evolution of marine ice sheets. Here, we present work that explores the rheological parameter space within the idealized MSIMIP+ marine ice sheet configuration using the BISICLES model. We explore the impacts of increasing the stress exponent from 3 to 4, highlighting the considerable changes to the ice sheet system caused by increasing the stress exponent. Beyond dynamic changes in the ice behavior, changes become necessary to the other flow law parameters generally computed during initialization. For example, it may be that viscosity modifiers typically interpreted as ‘damage’ may instead be indications of mismatches in rheology. This study underscores the dynamic sensitivity of glacial ice to changes in the rheological parameters and calls attention to the key variables influencing ice sheet evolution.
Greenland mass balance from laser altimetry between 1995 and 2020
Hui Gao, Beata Csatho, Anton Schenk, Nicole-Jeanne Schlegel, Brooke Medley, Max Brils, Brice Noël, Michiel van den Broeke, Surendra Adhikari, Michael Croteau, Bryant Loomis, Ivan Parmuzin, Kristin Poinar, Sophie Nowicki, Denis Felikson
Corresponding author: Hui Gao
Corresponding author e-mail: hgao7@buffalo.edu
The Greenland ice sheet has been one of the major contributors to sea-level rise and is projected to continue losing mass in the 21st century. Improved mass balance reconstructions, partitioned into mass change components caused by surface and dynamic processes, are critical for ice-sheet model calibration and validation. We combine surface elevations acquired by NASA’s airborne and satellite laser altimetry missions, including ATM, LVIS, ICESat, and ICESat-2, to reconstruct the annual mass balance of the ice sheet at 1 km horizontal resolution since 1995. Three different sets of modeled SMB and firn height change from state-of-the-art climate models and firn models, namely RACMO2.3p2 and IMAU-FDM v1.2G, MERRA-2 and GSFC-FDM v1.2.1, corrected ERA5 and GEMB v1.0, are used to partition the total mass change into surface and dynamic components. We found significant differences among the total mass change obtained using different firn datasets at the drainage basin scale, and reconstructions using the input–output method. In central west Greenland, we estimate ice loss ranging from 536–787 Gt between 1995 and 2020, depending on the firn model used. In comparison, Mankoff et al. (2020) estimate 1125 Gt of ice loss using the input–output method. In central east Greenland, we estimate ice loss between 808 and 994 Gt, whereas Mankoff et al. (2020) estimate only 88 Gt of loss. Aggregating the mass balance for the whole ice sheet cancels out the differences in drainage basins, resulting in good agreement of mass balance among laser altimetry reconstructions using different FDMs and reconstructions using the input–output method between 1995 and 2020. In addition, we found that a much larger fraction of mass loss originates from dynamic mass change when applying the GSFC-FDM (80%) and GEMB (69.6%) firn models, compared to only 41.3% using IMAU-FDM. Aggregating dynamic mass change in the fast-flowing regions of each drainage basin, we found a good correlation with ice discharge from the input–output reconstructions. Our results highlight the necessity to reconcile mass balance reconstructions on the drainage basin level, which is required to calibrate ice sheet models at the regional scale. The ambiguity in partitioning the mass change into surface and dynamic changes suggests that further SMB and firn model developments remain essential.
Ice sheets and surface energy balance: observational opportunities and key model uncertainties
Nicole-Jeanne Schlegel
Corresponding author: Nicole-Jeanne Schlegel
Corresponding author e-mail: schlegel@jpl.nasa.gov
Over the last decade, the Greenland Ice Sheet has contributed significantly to sea-level rise, and this contribution is expected to accelerate. Together, surface energy balance and numerical ice sheet flow models offer a physically based method for projecting ice sheet contribution to sea-level change. However, Greenland’s future contribution to sea level is highly uncertain, in particular due to confidence in model estimates of ice sheet surface mass balance. For instance, it is difficult to quantify the capability of Greenland’s deep snowpack to retain and/or refreeze surface melt, as these processes dictate whether the enhanced meltwater will run off into the ocean (contributing to sea level) or refreeze (releasing additional heat into the snowpack and altering its thermal regime, but not contributing to sea level). This is because observations are sparse, and we must rely on models to simulate the evolution of the ice sheet snowpack density and radiation budget. Such models carry a number of biases – including the assumption that the surface emits longwave radiation as a black body – contributing to uncertainty in model estimates of Greenland’s surface radiation budget, snowpack temperature, meltwater retention, and consequently surface mass balance. NASA’s PREFIRE (Polar Radiant Energy in the Far InfraRed Experiment) and ESA’s FORUM are examples of satellite missions slated to monitor far-infrared emission over the Greenland Ice Sheet and provide observations to improve model representation of energy and mass balance. PREFIRE was selected by NASA to fly two miniaturized Thermal InfraRed Spectrometers, aboard CubeSats, capable of distinguishing the spectral signatures of surface and atmospheric properties in the Polar regions at wavelengths never before measured. The CubeSats are scheduled to launch on 22 May and 5 June 2024 from the Rocket Lab Launch Complex in New Zealand, and the mission is commissioned for 10 months. In anticipation of the PREFIRE data products, a spectral representation of emissivity has been incorporated into state of the art energy balance models, as the PREFIRE science plan focuses on assessment of model sensitivity to assumptions about snow and ice emissivity. Here we provide an update on PREFIRE’s first 2 months of flight and the progress on model integration of mission products, as well as insight into ice sheet model uncertainty related to the representation of surface emissivity across Greenland.
Observations and modelling the long-term development of a perennial firn aquifer on the Lomonosovfonna ice cap, Svalbard
Tim van den Akker, Ward van Pelt, Rickard Petterson, Veijo Pohjola
Corresponding author: Tim van den Akker
Corresponding author e-mail: timvandenakker96@gmail.com
An uncertain factor in assessing future sea level rise is the melt water runoff buffering capacity of snow and firn on glaciers and ice caps. Perennial firn aquifers can store surface melt, thereby acting as aforementioned buffer for sea level rise, and influence the thermodynamics of the firn layer and the ice dynamics through drainage. We present here in-situ observations of the PFA found on the Lomonosovfonna ice cap in central Svalbard. We then use these observations to tune a groundwater flow model, MODFLOW 6, to simulate the long term evolution of the PFA over the period 1957 – 2019. We find an excellent fit between modelled and observed PFA geometry, and a very good fit between modelled and observed PFA depth below the snow surface. We find furthermore that the aquifer was likely present in 1957, and that it steadily grew over the modelled period with about 11% in total water content and 15% in water table depth. The inferred hydraulic conductivity is 0.62 mm a⁻¹, close to values found in the literature. Meltwater input and the total water content of the aquifer have a stronger influence on the water table height than firn density changes at this location.
Greenland surface mass balance using physical-informed deep learning
Anna Puggaard, Anne M. Solgaard, Louise S. Sørensen, Ruth Mottram, Sebastian B. Simonsen
Corresponding author: Anna Puggaard
Corresponding author e-mail: annpu@dtu.dk
Understanding the surface mass balance (SMB) of the Greenland Ice Sheet is crucial for assessing the response to climate change and its implications for rising sea levels. This study presents a novel approach by integrating the ERA5 reanalysis atmospheric model with mass change observations from the GRACE and GRACE-FO satellite missions, alongside discharge estimates derived from Sentinel-1, to train a deep learning model for SMB estimation. Leveraging the individual fields of daily temperature, surface pressure, surface short-wave and long-wave radiation, and precipitation from the ERA5, the model provides high temporal resolution insights into the daily fluctuations in SMB. Training a physically informed deep learning model against independent satellite observations from GRACE and GRACE-FO ensures that the estimates of SMB realistically represent the SMB component of the mass balance of the Greenland ice sheet. We introduce a physics-informed architecture that integrates prior glaciological knowledge to constrain the deep learning model, allowing the model to learn from data and adhere to established physical laws and processes. The research underscores the utility of combining advanced machine-learning techniques with traditional physical modelling approaches to improve the accuracy and interpretability of climate models.
Do ice–ocean feedbacks influence a regime shift of the Filchner–Ronne ice shelf cavity?
Ronja Reese, Jan De Rydt, Kaitlin Naughten
Corresponding author: Ronja Reese
Corresponding author e-mail: ronja.reese@northumbria.ac.uk
The Filchner–Ronne Ice Shelf currently has a ‘cold’ cavity with comparably low melt rates or refreezing at the ice–ocean interface. However, it has been shown that a switch to ‘warm’ conditions under a very strong climate warming scenario is possible within this century. In this case, modified Circumpolar Deep Water that resides at intermediate levels offshore enters the cavity and fuels a 21-fold increase in aggregated melt rates, with implications for ice-shelf buttressing and thereby the dynamics of tributary ice streams and glaciers. Interactions of resulting cavity changes with the ocean could furthermore amplify or weaken the increase in ice-shelf melting. Here we investigate the influence of ice–ocean feedbacks on sub-shelf melt rates and the regime shift from a ‘cold’ to a ‘warm’ ice-shelf cavity using standalone and coupled configurations of the ice sheet model Úa and the ocean model MITgcm. Furthermore, we test their influence on reversibility back to ‘cold’ conditions, and the impact of a regime shift on grounded ice dynamics.
Using observations of surface fracture to address ill-posed ice softness estimation over Pine Island Glacier
Trystan Surawy-Stepney, Stephen Cornford, Anna Hogg
Corresponding author: Trystan Surawy-Stepney
Corresponding author e-mail: eetss@leeds.ac.uk
Many numerical models used to simulate ice streams require the specification of control fields representing the slipperiness of the ice–bed interface and local deviations in the assumed rheological properties of the ice. These poorly constrained components of the system are often found by solving an inverse problem given observations of model state variables – typically ice flow speed. However, these inverse problems are generally ill-posed, resulting in degenerate or error-dominated solutions. The clearest way to improve this is to take advantage of additional prior information regarding the control fields. In this study, we investigate two ways of using maps of surface fracture, derived from Sentinel-1 satellite imagery, to provide prior information to the inverse problem. We first consider a prior that assumes values of effective viscosity significantly different from Glen’s flow law are, for the most part, due to observable fractures. Using Pine Island Glacier as a case study, we investigate the solutions and conditioning of this data-informed inverse problem and compare with a typical heuristic regularization technique. We find that the inclusion of fracture data results in softness fields that resemble fracture features on floating ice. On grounded ice, despite the prevalence of surface crevassing, the softness fields look no more plausible when fracture data is included – suggesting that the presence of surface fracture is not the largest contribution to our uncertainty in the ice rheology. We go on to investigate the use of timeseries of fracture maps to constrain the evolution of the softness field on ice shelves through time, making the assumption that changes to ice rheology occurring on annual timescales are dominated by the fracturing of ice. We show that this method can result in softness fields that visually mimic fracture patterns on floating ice without significantly affecting the quality of the misfit. Such softness fields could be used to constrain evolution equations in isotropic damage models.
Modelled ice sheet sensitivity to basal friction parameterizations is determined by the amount of buttressing and the flow factor inversion.
Tim van den Akker, William H. Lipscomb, Gunter R. Leguy, Willem Jan van de Berg, Roderik S.W. van de Wal
Corresponding author: Tim van den Akker
Corresponding author e-mail: timvandenakker96@gmail.com
A major source of uncertainty in projecting Antarctic ice mass loss by ice sheet models arises from the parameterization of the basal friction. Typically, these parameterizations describe the relation between the ice basal velocities and the basal friction, the so-called basal friction law. Previous studies suggest that basal friction laws with a stronger dependence on the basal velocities lead to lower ice fluxes and therefore to less ice mass loss, compared to basal friction laws with a weaker or even no dependence on basal velocities. The former friction laws are referred to as ‘Power Laws’, the latter as ’Coulomb Laws’. We use the Community Ice Sheet Model (CISM) to show that the sensitivity to the choice of basal friction law on the projected ice mass loss depends on the geometric setting, the perturbation applied and the inversion procedure for the basal friction and a flow enhancement factor. We find a geometric controlled negative feedback between buttressing and basal sliding in the Amundsen sea embayment. This feedback hinges on the existence of a buttressing ice shelf. If the perturbation is sufficiently strong to remove the ice shelf, this negative feedback disappears. For smaller perturbations, it leads to similar sea level rise contributions from different basal sliding laws. Furthermore, we present a two-steps inversion procedure where we first tune simultaneously ocean temperature perturbations under floating ice and free parameters in the basal friction law to nudge the modelled ice thickness towards observations, and secondly we tune a flow enhancement factor in the flow law to match modelled ice surface velocities to observations. This leads to an improved initialized state with a lower sensitivities to the choice of the basal friction law.
Data-driven modelling of satellite radar altimetry for Greenland ice sheet mass balance
Sebastian B. Simonsen, Anna Puggaard
Corresponding author: Sebastian B. Simonsen
Corresponding author e-mail: ssim@space.dtu.dk
We are in a unique era for satellite altimetry, with an unprecedented number of satellites able to track the dynamic polar regions. This diverse satellite fleet employs various measurement techniques, including photon-counting lidar and radar across different bands (Ku-band and Ka-band), as well as different operational modes such as low-resolution mode (LRM), synthetic aperture radar (SAR), and SAR interferometry (SARIn). Each of these methods has unique advantages and limitations when assessing changes in ice sheet volume. Moreover, converting satellite-altimeter data into ice sheet mass balance is complex. It requires a deep understanding of the satellite’s specific characteristics and model efforts of non-ice-related volume changes (e.g. firn compaction). Using ICESat lidar-altimetry data demonstrates a more straightforward way to convert measurements into mass balance for the Greenland ice sheet, as this altimeter directly measures the ice sheet’s actual surface, reducing uncertainties associated with intermediate factors. This study examines a 30-year record of Greenland ice sheet volume changes using data from the European Commission’s Copernicus Climate Change Service. This record is derived from European Ku-band radar missions (ERS-1, ERS-2, ENVISAT, CryoSat-2, and Sentinel-3). However, the longer wavelength of Ku-band radar altimetry introduces variable surface penetration, complicating the conversion from volume changes to mass changes. To address this issue, we leverage machine learning to uncover hidden patterns in existing earth observation data. This approach allows us to combine mass balance estimates from ICESat and ICESat-2 with the extended time series of Ku-band elevation changes. Combining these data sources provides a comprehensive perspective on the underlying factors affecting ice sheet mass balance, offering a more robust and reliable analysis. This data-driven modelling of the conversion from volume change to mass change enables operational estimates of ice sheet mass balance without the latency of regional climate models.
Evaluation of coupled Earth system model icesheet simulations of Greenland against observational products
Yiliang Ma, Robin Smith, Steve George, Ines Otosaka, Jennifer Maddalena, Dan Hodson
Corresponding author: Robin Smith
Corresponding author e-mail: robin.smith@ncas.ac.uk
The mass of the Greenland ice sheet represents around 7 m of potential global mean sea level rise. The rate at which this contribution will be added to the global ocean under future climate change is dependent on the climate forcing and feedbacks between the changing ice sheet and the climate, so it is important that Earth system models (ESMs) represent Greenland climate and feedbacks realistically. UK Earth System Model (UKESM) is one of a handful of CMIP ESMs that is capable of including dynamic models of the Greenland ice sheet and includes state of the art climate–ice-sheet coupling based on the explicit exchanges of water and energy. To gain confidence in the results of future climate simulations of the Greenland ice sheet in UKESM we compare key variables from the model’s Historical simulations with observational products. We evaluate two versions of UKESM1, one used for CMIP6 (UKESM1, 16 Historical members) and one coupled to a dynamic model of Greenland ice sheet (UKESM1.1, 4 members). To investigate the regional climate around Greenland, we first compare the seasonal cycle and multi-year trends in Arctic regional sea ice extent and volume in March and September with the Sea Ice Concentration Climate Data Record (CDR) from 1978–2014 and PIOMAS Arctic Sea Ice Volume Reanalysis from 1979–2014, respectively. For the ice sheet itself we compare surface elevation change products derived from satellite altimetry with the evolution of the BISICLES ice sheet component of UKESM in seven individual Greenland drainage basins from 1994–2014. We further decompose these elevation changes into those due to surface mass balance and ice dynamics using output from a calibrated regional model to represent real-world surface mass balance. These results help us assess which areas of Greenland change in UKESM simulations are most robust and prioritize areas of the model for development.
On the use of dh/dt observations in ice sheet model initialization
Tom Mitcham, Hilmar Gudmundsson
Corresponding author: Tom Mitcham
Corresponding author e-mail: tom.mitcham@bristol.ac.uk
Many ice sheet models use data assimilation to constrain unobservable, spatially variable parameters such as those related to ice viscosity and basal friction. Typically using observations of ice sheet surface velocity, the optimization of these parameters can generate representative initial ice sheet states for forward simulations. However, discrepancies between velocity measurements and other data sets used in models can result in unphysical dh/dt (rate of change of ice thickness) signals in subsequent runs. Techniques used (and under development) to tackle this issue include relaxation methods, iterative optimization procedures, and transient calibration. In this work, we examine the impact of incorporating dh/dt observations into the cost function of the optimization problem using the ice sheet model Úa. We test whether assimilating this additional information both reduces the artefacts mentioned above and helps generate an initial state with a trend in ice thickness change closer to observations. We undertake a case study on the Larsen C Ice Shelf and its tributaries and find a strong trade-off between fitting observed dh/dt values and matching observed present-day ice velocities in this setting. We then extend our analysis of this approach with studies on idealized ice stream/ice shelf setups and other regions of the Antarctic Ice Sheet.
Towards isochronal calibration of continental scale ice sheet models
Antoine Hermant, Christian Wirths, Vjeran Visnjenic, Julien Bodart, Johannes Sutter
Corresponding author: Antoine Hermant
Corresponding author e-mail: antoine.hermant@unibe.ch
The Antarctic Ice Sheet’s (AIS) contribution to sea-level rise under future scenarios remains uncertain. Simulations of the AIS covering past climate periods offer valuable insights into its response to various climatological background states and transitions, as well as its past contributions to sea-level change. However, the scarcity of data to constrain modeled ice-flow and paleo-climate forcing often results in significant uncertainties regarding paleo ice-sheet evolution. Fortunately, a growing pool of traced and dated internal layers (isochrones) of the AIS presents a valuable opportunity. These isochrones record past changes in ice flow and surface mass balance that have resulted in the current state of the AIS. Here, we present and test a framework for isochronal calibration of an ice sheet model. To simulate the AIS over glacial–interglacial cycles, we employ a climate index approach using ice-core records of past climate and regional accumulation changes, in conjunction with Global Climate Model snapshots. Isochrones are then computed online via an isochronal model or via Lagrangian tracers in postrpocessing. By minimizing the discrepancies between observed and modeled isochrones, we optimize the parametrization of ice flow, refine the paleoclimate forcing, and better constrain geothermal heat flux. Initially, we apply our methodology on a regional scale, focusing on sectors where the isochronal structure of the AIS is well-mapped, such as around Dome C (East Antarctica). The observed traced and dated layers in this region cover approximately the last 200 ka. Isochrones in more dynamic sectors of the AIS are mostly much younger (approximately Last Deglaciation–Holocene) but provide valuable constraints on paleo ice stream dynamics. Further constraints over this period, such as grounding-line advance and retreat, will enable us to develop an accurately isochronally calibrated ice sheet model on a continental scale. Our work seeks to deepen our understanding of Antarctic Ice Sheet dynamics on glacial–interglacial timescales and provide improved paleo-informed initializations for Antarctic Ice Sheet projections.
Simulation of crevasse field evolution using a phase-field approach
Juan Michael Sargado, Joachim Mathiesen
Corresponding author: Juan Michael Sargado
Corresponding author e-mail: michael.sargado@nbi.ku.dk
Simulating the evolution of a crevasse field presents several computational challenges. One consideration is that the mesh resolution required to resolve multiple fractures and obtain accurate stresses leads to a computational problem with very large numbers of unknowns. Secondly, the fracture network may itself be topologically complex due to criss-crossing fractures. In this preliminary study, we explore the possibility of using a phase-field approach to model the formation and growth of surface crevasses in ice. The phase-field method has several advantages in this regard, as its algorithmic complexity is independent of the number of fractures. Furthermore it can naturally handle crack nucleation, branching and coalescence. However, full three-dimensional simulations are still computationally prohibitive at the resolutions required by such an approach. Instead, we investigate ways to couple phase-field modeling of fracture with depth-integrated ice sheet models in order to allow simulation of fracture in larger domains.
Chemical weathering products in seasonally diverse proglacial waters as tracers for glacial hydrologic and geochemical modelling
Rebecca McCerery, Ankit Pramanik, Joseph Graly, William Gilhooly, Trinity Hamilton, Kathy Licht
Corresponding author: Joseph Graly
Corresponding author e-mail: joseph.graly@northumbria.ac.uk
The weathering of rocks and minerals on Earth’s surface and their weathering products sustain life and regulate climate. Due to their production of finely crushed sediments, glaciers are an important yet understudied component of these global biogeochemical cycles. Where conditions allow for surface melt, the routing of sediment and weathering products to glacial forefields is highly seasonal. The chemical and biological fluxes of glacial systems are most often characterized from summer outflows. Compared to winter, melt season outflow has limited ice–bed contact and short residence times. We have sampled water and accreted ice from the forefield of Isunnguata Sermia, an outlet glacier in western Greenland, over a range of seasons. Flowing water was collected in late summer/post-melt and in late winter/early spring, including the first eruption of subglacial water. We also sampled accreted ice formed from supercooled water during summer and naled ice formed from upwelling water during winter. In some cases, water beneath naled ice was also sampled. We present preliminary results from the geochemical analyses of this range of water sources from the Isunnguata Sermia forefield. The geochemical signatures of each season show large changes in water residence times, saturation states, and mineral reaction pathways. Flow path length, storage in a distributed subglacial hydrological system, input of ground water, and the development of the mineral substrate of the subglacial environment, all control the final geochemical output. We employ the subglacial hydrology model GlaDS (Glacier Drainage System) to explore the development and interplay of effective and ineffective drainage systems. We integrate GlaDS with a one-dimensional reaction-transport model to simulate the distribution of geochemical tracers beneath glaciers. This assesses the connectivity within the subglacial environment’s interior and the potential shifts in geochemical fluxes under changing melt regimes. Glaciers and ice sheets play an important role in global biogeochemical cycles and these results further our understanding of weathering conditions during winter, as well as allowing for extrapolation of conditions beneath interior of the Greenland Ice Sheet where surface melt is absent. Geochemical data from outlet glaciers also has the potential to be used to inform biogeochemical models for past and future predictions of glacial contributions to global mineral and nutrient cycles.
Ice shelf and glacier grounding line delineation with synthetic aperture radar in low coherence regions using tidal motion correlation: a new grounding line dataset for the Antarctic Peninsula
Benjamin J. Wallis, Anna E. Hogg, Yikai Zhu, Andrew Hooper
Corresponding author: Benjamin J. Wallis
Corresponding author e-mail: eebjwa@leeds.ac.uk
The boundary between ice that is grounded on the bedrock and floating ice, the grounding line, is a key attribute of marine ice sheets and ice shelves. Accurate knowledge of grounding zone configuration is essential to quantify ice sheet mass loss, understand the stability of marine ice sheets and initialize ice sheet models. An established technique for measuring grounding line position is differential synthetic aperture radar interferometry (DInSAR), where the vertical displacement caused by tidal motion of floating ice is precisely measured. A significant limitation of this method is that it relies on interferometric coherence between SAR image acquisitions, making measurements difficult in regions of high ice speed, ice deformation, surface accumulation and melting. Intensity feature tracking measures ice motion without the requirement for interferometric coherence and due to the off-nadir viewing geometry of SAR sensors vertical tidal motion of floating ice creates an apparent, but erroneous, horizontal motion in the range direction of the satellite viewing geometry. This is usually considered an error term when measuring ice velocity, but a limited number of studies have exploited this effect by differencing range velocity results from multiple image pairs to measure grounding line location in the differential range offset tracking method. Here we significantly build on this methodology to develop a full time-series approach to map grounding line position by measuring the correlation between modelled tidal motion and velocity tracking anomaly using the full timeseries of Sentinel-1 IW mode imagery. We validate this methodology in the Antarctic Peninsula region by comparison with existing grounding line products and Sentinel-1 DInSAR measurements concurrent with our period of observation. We demonstrate that this method is suitable for measuring the grounding line position of both large ice shelves and glaciers as narrow as 3 km. Our results provide new dataset for the grounding line of the Antarctic Peninsula, showing grounding line retreat of up to 16.3 km in locations where tidally sensitive grounding line measurements have not been possible since the 1990s. We also highlight how this product has recently been used when modelling the flow of glaciers in the Larsen-B Embayment.
New estimates of englacial and basal thermal conditions of the Antarctic ice sheet
Olivia Raspoet, Frank Pattyn
Corresponding author: Olivia Raspoet
Corresponding author e-mail: olivia.raspoet@ulb.be
The thermal state of the Antarctic ice sheet affects the ice flow regime by modulating ice rheology and subglacial processes such as basal melting, subglacial hydrology, and basal sliding. Basal thermal conditions might also evolve in the future, and the thawing of some parts of the bed could lead to feedback loops and increased ice mass loss. However, basal thermal conditions of the Antarctic ice sheet remain uncertain due to the lack of direct observations of subglacial environments and uncertainties inherent to the factors influencing them. In this study, an ensemble of simulations was performed using different combinations of boundary conditions and ice sheet model uncertainty to determine the likely basal thermal state and calculate basal melt rates of the Antarctic ice sheet. A sensitivity analysis was conducted, allowing us to evaluate the influence of the main factors of uncertainties on model estimates. Observations, including borehole-derived temperature profiles and the distribution of subglacial lakes, were used to validate the ensemble results and determine the most accurate simulations. Results reveal that approximately 54% of the grounded part of the Antarctic ice sheet exhibits temperate basal conditions, with basal temperatures reaching the pressure-melting point. The mean basal melt rate was estimated at 5.5 mm a⁻¹, generating about 60 Gt a⁻¹ of subglacial water. Amongst all uncertainties, geothermal heat flux remains the largest one. Furthermore, since high basal melt rates also result from frictional heating in regions hosting fast-flowing glaciers, basal slipperiness and concomitant sliding strongly influence the model sensitivity, leading to uncertainties of the same order as those related to the geothermal heat flux. This work was funded by an FRIA scholarship from the Fonds de la Recherche Scientifique de Belgique (F.R.S.–FNRS).
Evaluating a regional climate model ensemble of surface mass budget to understand diverging future projections
Ruth Mottram, Charles Amory, Fredrik Boberg, Willem Jan van de Berg, Michiel van den Broeke, Christiaan van Dalum, Alison Delhasse, Xavier Fettweis, Christoph Kittel, Quentin Glaude, Heiko Goelzer, Nicolaj Hansen, Brice Noel, Anna Puggaard, Martin Olesen, Kirk Scanlan, Sebastian Simonsen
Corresponding author: Ruth Mottram
Corresponding author e-mail: rum@dmi.dk
We analyse a new ensemble of surface mass balance projections for both Greenland and Antarctic ice sheets produced by three regional climate models (RCMs), HIRHAM5, MAR3.12 and RACMO2.3p2. The full ensemble includes results from simulations dynamically downscaled from several different global climate models (GCMs) covering a range of emissions pathways. Here, we focus on a subset, an intercomparison of simulations forced by the GCM CESM2 for SSP585 over both ice sheets out to 2100. We also evaluate ERA5 reanalysis forced simulations from the three RCMs with in-situ and satellite observations. Although all three RCMs represent present day climate similarly, and show a similar trend for SMB over Greenland, we identify important differences in present-day surface melt production. We also identify a wide divergence in melt-temperature relationship and estimates of future surface mass budget when forced by CESM2. All models show high mass loss in Greenland with a similar downward trend. HIRHAM5 has the highest melt at the present day. In the projections MAR starts off with less melt than HIRHAM over Greenland but has a steeper increase and ends the 21st century with a lower SMB than HIRHAM5. In Antarctica, the HIRHAM5 model also has much more melt at the present day than the other two models and shows a steeper increase but all models show increasing melt in particular over ice shelves. The RACMO model shows the lowest melt at the present day over both ice sheets and a shallower trend of increasing melt under SSP585 than the other two models over both ice sheets. The end of century surface mass budget projections diverge widely between the highest and lowest estimates, suggesting that while model state is similar at present day when forced by reanalysis, SMB has different sensitivities to climate forcing between the RCMs. This means there may be higher than previously identified uncertainties in estimates of future sea level rise from the ice sheets and implies that ice sheet models need to use a wide range of climate forcings to capture these uncertainties. However, better constraints on present-day meltwater volumes from observations are also needed to evaluate RCMs and quantify the likelihood of the related uncertainties in SMB projections. We suggest that there is therefore an important role for regional climate model evaluation and calibration, but datasets and and parameterizations need to be carefully chosen and tested to avoid model overtuning.
Modelling the recently observed evolution of Helheim Glacier
Emily Hill, Leanne Wake, Hilmar Gudmundsson
Corresponding author: Emily Hill
Corresponding author e-mail: emily.hill@northumbria.ac.uk
Mass loss from the Greenland ice sheet has increased in the last two decades, and ice loss from Greenland is the main contributor to current global mean sea level rise. Future projections of ice loss from Greenland are associated with large uncertainties, both due to uncertainties in climate forcing projections, and due to poorly constrained processes in ice sheet and climate models. By understanding the history of the Greenland ice sheet and ice-sheet–climate interactions we can better constrain unknown parameters in our models and ultimately reduce uncertainties in future projections. This new project will produce new estimates of the Greenland ice-sheet surface mass balance and dynamic ice loss from 1600–2024, using physically based climate and ice-sheet models. Here, we present an initial step towards this goal, focusing on Helheim Glacier, southeast Greenland, in which we explore calibration techniques for initializing our ice-sheet model to ensure that the model is able to replicate the recent trends in observed ice loss since the 1990s.
How much global glacier mass loss is committed under policy-relevant global warming scenarios?
Regine Hock, Harry Zekollari, Lillian Schuster, Fabien Maussion, Ben Marzeion, GlacierMIP3 Consortium
Corresponding author: Regine Hock
Corresponding author e-mail: rehock@alaska.edu
Glaciers around the world are rapidly retreating, directly affecting global sea level, streamflow and hazards. Due to delayed response of glaciers to climate change, glaciers can be expected to continue to lose mass even if the global air temperature were to stabilize at present-day levels. As part of the Glacier Model Intercomparison Project – Phase 3 (GlacierMIP3), a global glacier modelling community effort, we simulate how all glaciers on Earth outside the ice sheets will stabilize under a wide range of stabilized global mean air temperature scenarios. Using eight large-scale glacier models, we estimate global committed losses for 19 glacierized regions under current climate conditions (corresponding to +1.2°C above pre-industrial levels) and their long-term stabilization under various policy-relevant global warming scenarios, including a +1.5°C and +2°C scenario (aligned with the Paris agreement), and a +2.7°C scenario based on current policy trajectories. The stabilized temperature scenarios are derived from historical and future simulations from several GCMs and are repeated over several thousand years to allow the glaciers to equilibrate. We find that committed glacier mass losses are substantial, with about 40% of global glacier ice to be lost eventually, even under current climatic conditions, indicating that a substantial portion of glacier volume is already committed to be lost. The steady-state global glacier volume strongly depends on the assumed temperature stabilization scenario. Regional differences and the time scales to achieve glacier equilibration vary strongly between glacier regions. Topographical features such as the elevation range and the surface slope of glaciers play an important role.
Implementing a Greenland marine terminating glacier melt parameterization within an Earth system model framework
Steve George, Robin Smith
Corresponding author: Steve George
Corresponding author e-mail: s.e.george@reading.ac.uk
The Greenland ice sheet has experienced rapid ice loss since the 1990s, with an average contribution to increased global sea level of ~0.5mm a⁻¹. This loss of mass is split evenly between (atmosphere forced) surface melt and (assumed ocean forced) accelerated ice flow. With a sea-level equivalent of 7 m it is important that we understand how the ice sheet may evolve in the future. UK Earth System Model (UKESM) is one of a handful of CMIP6 Earth system models (ESMs) that is capable of including a dynamic model of the Greenland ice sheet (BISICLES). State of the art atmosphere–ice-sheet coupling provide explicit measures of the ice-sheet surface mass balance (SMB), which can be directly measured against the observed SMB (surface melt). The existing version of UKESM has no mechanism for allowing thermal interaction between the ocean and the ice sheet; marine terminating glaciers are located inland, along long narrow fjords which are unresolved in the ocean model. Calving flux is a function of ice velocity at a defined, dry gate. This study presents the results of the introduction of a new ocean–ice parameterization into UKESM. Building on the work of the ISMIP6 offline studies of ocean forcing of the Greenland ice sheet, we introduce similar parameterizations into the fully coupled model. Rather than advecting the farfield (model) ocean temperatures as a simple thermal forcing (TF) of the glacier termini, increased melt is modelled as a function of both TF and freshwater runoff from surface melt. The parameterization represents runoff released as subglacial discharge, the ensuing buoyant plumes entraining warm ocean water onto the ice face (increasing melt). Within UKESM this melt is implemented as a scaling factor to the calving algorithm, with the ice terminus allowed to retreat (or advance) if marine conditions continue. Farfield ocean forcing is split into representative zones, and runoff is filtered by hydrological drainage basins. We present the results of the impact of the new parameterization on the modelled future evolution of the Greenland ice sheet under different forcing scenarios.
Model development and integration in the Integrated Digital East Antarctica program
Lenneke Jong, Aleks Terauds, Jason Roberts, Ben Raymond, Michael Sumner
Corresponding author: Lenneke Jong
Corresponding author e-mail: lenneke.jong@aad.gov.au
The Integrated Digital East Antarctica (IDEA) program is a new effort within the Australian Antarctic Division which aims to deliver an authoritative digital representation of East Antarctica and the Southern Ocean, integrating data from across research disciplines and providing access to data, tools and data products to underpin science and inform decision making. Climate and geophysical modelling efforts form an important part of this program, through improvements in the integration of data to provide the most up-to-date input available for model input, and by developing tools to improve the utility of modelling outputs for further research and end-users. The success of IDEA is reliant on connection and collaborations across the Antarctic science and climate modelling communities. One new project seeks to make data from long-running airborne geophysical survey projects data more accessible and inter-operable, building tools to develop to better integrate the radar and gravity data collected for input for ice sheet models and conversely, how to use modelling and statistical methods to better inform where to target our future data collection efforts.
How stable are the ice divides in the northern Greenland ice sheet?
Christine S. Hvidberg
Corresponding author: Christine S. Hvidberg
Corresponding author e-mail: ch@gfy.ku.dk
The mass loss from the Greenland ice sheet has increased over the last decades, and is now a major contributor to the global mean sea level rise. While the interior of the Greenland ice sheet has remained relatively stable, the mass loss from ice sheet margins have spread to the north and since 2007 propagated into interior north Greenland. The ice thickness changes and ice velocities in the ice sheet interior are orders of magnitude smaller than along the margins, and the effect on ice divides and interior ice dynamics are not well known. We present here an assessment of the interior stability in north Greenland using GPS data, remote sensing data, climate model output, ice core data and ice flow modelling. We compile GPS survey data from interior ice core sites in north Greenland at GRIP (1992–1996), NorthGRIP (1996–2001), NEEM (2007–2015), and EastGRIP (2015–2022), and compare with surface mass balance estimates and remote sensing data to assess changes over the last decades. While the surface elevation has remained relatively stable at the northern ice divide sites, an inferred northward migration of the ice divide in Northwest Greenland observed in 2007–15 coincided with the onset of thinning along the ice margin in the Baffin Bay area. Preliminary results suggest that the surface elevation near the summit of the Greenland ice sheet observed in 1992–96 lowered slightly over the last 40 years, during a period of widespread thinning along the western margin. Interior elevation changes can be due to local changes in surface mass balance, or an interior dynamical response to mass loss at the margin and changes in ice stream configurations. They represent a challenge for realistic projections of ice sheet mass loss and sea level rise, and are here discussed in the context of both the recent climate warming in Greenland and the long-term simulated evolution of the North Greenland ice sheet interior over the last several thousands of years.
The AWI Earth System Model with interactive ice sheets for simulations on millennial timescales
Uta Krebs-Kanzow, Lu Níu, Lars Ackermann, Gerrit Lohmann
Corresponding author: Uta Krebs-Kanzow
Corresponding author e-mail: Uta.Krebs-Kanzow@awi.de
We present simulations with the Earth system model AWI-ESM with interactive ice sheets covering paleo time scales and future prjections and focus on the interface between ice sheets and atmosphere. A notable feature of the coupled model set-up is the application of an advanced surface mass balance scheme dEBM (diurnal energy balance model,Krebs-Kanzow et al., 2021). In many cases previous ice sheet simulations have used the semi-empirical positive degree-day scheme, which only takes surface air temperature into account for surface melt. The positive degree-day scheme is often calibrated based on modern observations from the Greenland Ice Sheet, which appears not suitable for the glacial ice sheets with substantially different background climate and radiative forcing. The dEBM scheme used here accounts for changes in the Earth’s orbit and atmospheric composition, implicitly accounts for the diurnal melt–freeze cycle and uses physics-based, non-empirical parameters. Only the albedo scheme has been calibrated specifically for today’s Greenland surface mass balance: it distinguishes three surface types (new snow, dry snow and wet snow/bare ice). The classification of surface types is conducted based on energy balance, accumulation and snow height. The dEBM scheme is particularly suitable for paleo simulations because it is computationally inexpensive and requires only monthly input (that is, near-surface temperature, precipitation, cloud cover, short- and longwave radiation). We present applications for both the glacial period (North American Ice Sheets and Eurasian Ice Sheet) and for interglacial and future warm climates (Greenland Ice Sheet) and evaluate the effect of differences in radiative forcing on the surface mass balance. We also evaluate the dEBM scheme for the Antarctic Ice Sheet and we find that surface melt rates simulated by dEBM using RACMO climate forcing compare well with RACMO melt rates
Scaling rheological insights from laboratories to ice sheets using remote sensing and ice-flow models
Brent Minchew, Joanna Millstein, Meghana Ranganthan, Caroline Mouchon, Justin Linick, Sarah Wells-Moran
Corresponding author: Brent Minchew
Corresponding author e-mail: minchew@mit.edu
Modeling glaciers and ice sheets requires insights into the rheology of glacier ice, accurate calibrations of numerous physical parameters, and estimates of how these processes and parameters vary in space and time. Decades of laboratory experiments have provided insights into the physical processes that facilitate the flow, deformation and fracture of glacier ice, but these insights are often difficult to scale to natural glaciers due to practical limitations in experimental setups and run times. Remote sensing observations, on the other hand, provide insights into the kinematics of glaciers and ice sheets from synoptic to meter scales but lack the resolution to connect to laboratory scales. Here, we discuss recent efforts that use laboratory experiments and remote sensing observations to constrain physical models, allowing us to glean deeper insight into the mechanics of natural glacier flow and to calibrate key physical parameters. We will focus the discussion on various projects aimed at better constraining the viscous and fracture properties of glacier ice. We will present several independent lines of evidence that suggest n = 4 in Glen’s Flow Law in rapidly deforming ice, along with a new, modified viscous flow law that can be readily plugged into existing ice-flow models to calculate changes in n and viscosity in space and time. To complement these results, we will highlight ongoing efforts to estimate and map fabric in Antarctica. Finally, we will transition from viscous to brittle by discussing work that constrains the tensile and compressive strengths of ice that underscore a new observationally constrained, simple parameterization of ice-shelf rift propagation.
Novel framework for reconstructing the evolution of Earth’s land ice cover with examples from Greenland
Beata Csatho, Toni Schenk, Hui Gao, Ashlynn Narkevic, Mohammad Salmani, Eric Kabe, Ivan Parmuzin, Sophie Nowicki
Corresponding author: Beata Csatho
Corresponding author e-mail: bcsatho@buffalo.edu
Earth-orbiting satellites provide a multidecadal record of surface changes in ice sheets, glaciers and ice caps with ever-increasing resolution and accuracy. The satellite record can be extended to the past using aerial photographs, historical records and glacial geological evidence of past ice extent. Numerous products, such as ice velocity, elevation and ice sheet boundary, are available to characterize the evolving land ice surface. However, most of these products are snapshots, depicting conditions at a specific time, or changes between two dates. We propose combining observations through a data fusion process, taking advantage of the domain knowledge of ice sheet processes and a priori estimates of data uncertainties. For example, ice sheet margins could be extracted simultaneously from images taken at different times using constraints for continuity and 3-D shape. This presentation will introduce the general framework of formulating the fusion problem to reconstruct changes in different elements of the land ice system. We will demonstrate the expected results with Greenland examples. In the first example, a Bayesian fusion method is introduced to combine data from laser and radar altimetry and digital elevation models to densify elevation and elevation change records derived by our Surface Elevation Reconstruction And Change detection (SERAC) and Mosaic Utility and Large Data set Integration for SERAC (MOULINS) methods. Next, ice velocity and elevations are determined directly from high-resolution, repeat digital stereo pairs by matching ice surface features in 3-D space. This approach eliminates the need to generate orthoimages and thus provides robust, accurate and dense velocity fields and DEMs simultaneously. The third example will explore combining multiple observation types based on different physical and geometric properties to track ice sheet boundary changes using multispectral satellite imagery, panchromatic historical aerial photographs, and high-resolution DEMs. The fusion framework will allow for extending and densifying the data record by adding observations from new data acquisitions (e.g. on-going and future satellite missions) or historical data (e.g. historical stereo aerial photographs). Updates will propagate through the entire data set as they could provide additional clues to improve accuracy and robustness. The evolving, ice-sheet-wide reconstructions will improve model initialization and calibration.
Collision with seamount triggers breakup of Antarctic iceberg
Xianwei Wang, Hilmar Gudmundsson, David Holland
Corresponding author: Xianwei Wang
Corresponding author e-mail: wangxianwei0304@163.com
The calving of the Larsen C ice shelf off the Antarctica Peninsula in 2017 led to the formation of the exceptionally large tabular iceberg A68. At the time of its formation, the future of A68 was subject to considerable speculation, and estimates of its future trajectory and lifespan varied considerably. In describing the lifecycle of A68 from its formation to its ultimate disintegration, we show for the first time that the decisive factor in its demise was a collision with the seafloor to the south of South Georgia in late 2020. By treating the iceberg as a deformable body, in an established ice-flow model, we show how the iceberg’s collision with the seafloor created stresses within the iceberg that led to its disintegration. We find that the drifting and rotating of the iceberg while grounded further enhanced its breakup. Drifting over a grounded shoal increased the tensile stresses by a factor of almost 100 more than grounding alone, and rotational motion about the pinning point increased the stresses by another 20%. The A68a climbed further over the seamount when the ocean surface was elevated by a polar cyclone and a high tide and grounded more when the surface elevation subsequently fell, leading to larger tensile stresses in the iceberg, and was a contributing factor to the breakup. Modeling the fracture and breakup of a large tabular iceberg is an important step toward better understanding an iceberg’s life cycle. It further opens the possibility of using large tabular icebergs as a natural laboratory for studying ice fracture and disintegration.
The relationship between cliff-height and calving rates for Hektoria Glacier in the Larsen B Ice Shelf
Rebecca Goodison
Corresponding author: Rebecca Goodison
Corresponding author e-mail: Rebecca.goodison@northumbria.ac.uk
Cliff-type calving law has been suggested to give rise to an unstable self-sustaining frontal retreat, named marine ice cliff instability (MICI). Simulations from ice-flow models implementing the cliff-type calving law gave rise to some of the higher-end predictions of the global sea level rise near future. However, there is limited evidence that MICI has occurred in the past. The collapse of Larsen B offers the closest recent analogue for testing the concept of cliff-type calving. Hektoria Glacier, a tributary glacier feeding the Larsen B ice shelf, was observed with sevenfold acceleration and a 5–10 km retreat after the collapse of Larsen B. This study uses high-resolution remote-sensing observational datasets on surface velocities, frontal positions, and surface elevation changes from Hektoria Glacier to constrain the cliff-type calving law.
The EOLIS dataset: monitoring land ice from CryoSat-2 swath processing
Carolyn Michael, Livia Jakob, Noel Gourmelen, Sophie Dubber, Andrea Incatasciato, Julia Bizon, Tristan Goss, Martin Ewart, Alex Horton, Alessandro Di Bella, Jerome Bouffard
Corresponding author: Carolyn Michael
Corresponding author e-mail: carolyn.michael@earthwave.co.uk
Satellite radar altimetry has been routinely used to monitor land ice heights since the 1990s. However, the launch of CryoSat-2 – the first altimetry mission to carry a synthetic aperture radar interferometer on board – has allowed several technical breakthroughs and led to many new applications that were previously unforeseen. One such breakthrough is swath processing of CryoSat’s SARIn mode, making full exploitation of the information contained in CryoSat’s waveforms and leading to one to two orders of magnitude more measurements than the conventional so-called point-of-closest-approach (POCA) technique. Following on from the early demonstration of the technique and its potential impact, the CryoTEMPO EOLIS (Elevation Over Land Ice from Swath) dataset now routinely provides elevation measurements over land ice at high resolution on a monthly basis. The dataset allows the use of radar altimetry in new environments such as the more complex terrain over glaciers and ice caps, as well as new applications thanks to the superior spatial and temporal resolution, such as the more precise quantification of subglacial lake drainage events. Currently, the EOLIS dataset is provided at monthly intervals over both ice sheets as well as all larger glacier regions, with an expansion of the dataset to the Antarctic ice shelves recently released and high-resolution annual DEMs covering the full extent of the two ice sheets coming later this year. With the aim of making CryoSat-2 altimetry data available to non-altimetry experts and to encourage its use more broadly by the community, the platform CS2EO (cs2eo.org) provides advanced data access to the EOLIS suite datasets. In CS2EO, users can query coincident data with other altimetry sensors, as well as explore and download custom elevation change time series over desired areas on ice sheets and glaciers, without having to download the EOLIS data first.
Modelling ice cliff collapse with the Material Point Method
Sam J.V. Sutcliffe, William M. Coombs, Wangcheng Zhang, Ravindra Duddu
Corresponding author: Sam J.V. Sutcliffe
Corresponding author e-mail: sam.sutcliffe@durham.ac.uk
The marine ice cliff collapse instability (MICI) is a plausible positive feedback loop of calving that could be catastrophic if realized. The MICI hypothesis is reliant on two main assumptions: there exists a height of sheer ice cliff that is unstable, and once this ice cliff collapses if will form another unstable cliff. The first assumption can be tested in a continuum framework with analytic methods, and classic numerical tools such as the Finite Element Method (FEM). These methods allow evaluation of the self-weight induced stresses against the performance of ice to give an estimate for a maximum stable cliff height. The second assumption is harder to evaluate, requiring modelling to extend out into the post-failure regime of a collapse. The behaviour of the collapse material may help stabilize the cliff front and prohibit the formation of a new cliff. Few numerical tools are robust when modelling the large deformations experienced during a cliff collapse, for example in the FEM mesh distortion may necessitate difficult re-meshing steps. Some methods can handle these large deformations such as the Discrete Element Method (DEM) but at the cost of abandoning a continuum approach. This work evaluates the use of the Material Point Method (MPM), a continuum numerical method that can handle large deformations and geometry changes, for combined pre and post-failure cliff collapse simulation. A non-local continuum plastic-damage model has been used to model ductile–brittle fracture, while avoiding the mesh-dependency issues exhibited by local models.
Climatic drivers of ice slabs and firn aquifers in Greenland
Max Brils, Peter Kuipers Munneke, Michiel van den Broeke, Nicolas Jullien, Andrew Tedstone, Horst Machguth, Willem Jan van de Berg
Corresponding author: Max Brils
Corresponding author e-mail: brils.max@gmail.com
In the past decade, observations evidenced the existence of ice slabs and aquifers on the Greenland ice sheet. These features affect the ice sheet’s hydrology by, respectively, facilitating surface runoff and modulating drainage to the bed. Recently, it has been observed that the extent of both aquifers and ice slabs has expanded. Longer time series are required to identify the climatic drivers of these changes. However, most observations cover only the last two decades. Here, we use the firn model IMAU-FDM to simulate the evolution of ice slabs and aquifers in the period 1958–2020, which we evaluate using radar measurements. The results suggest that both firn aquifers and ice slabs were already present in the late 1950s, and that their extent remained relatively constant until the beginning of this century, after which increased melt led to their expansion. The model shows that the amount of available liquid water (melt and rain) together with accumulation is a good predictor for the occurrence or absence of ice slabs and aquifers. Interestingly, the model also suggests that, if melt increases, a region with an aquifer could turn into an ice slab region, leading to a shift in regional ice sheet hydrology.