Antarctic calving, ice shelf collapse, and “MICI” protocol for ISMIP7

ISMIP7 webinar series

28 January 2026

Our group

• Luke Trusel, Penn State (co-lead)
• Catherine Walker, WHOI (co-lead)
• Helene Seroussi, Dartmouth College (steering committee liaison)
• Mahsa Bahrami, Penn State
• Jeremy Bassis, Michigan
• Maya Fields, Michigan
• Emily Glen, Penn State (postdoc starting Feb 2026)
• Danielle Grau, Georgia Tech
• Peter Kuipers Munneke, IMAU
• Caroline Needell, MIT-WHOI
• Ben Reynolds, U Buffalo (prior to graduation)
• Sanne Veldhuijsen, IMAU (prior to graduation)

Summary of proposed protocol

Some aspects are similar to ISMIP6:

• Consideration of surface meltwater’s role in ice shelf stability

​

Some aspects are new or different from ISMIP6:

• Greater consideration of firn conditions and less prescriptive collapse implementation
• We know that ice shelf collapse does not solely depend on surface melt: We recommend implementing Lai et al. (2020) stress condition criteria in addition to surface melt
• MICI may be dramatic but we know “high-end calving” scenarios are possible: We provide an observation-based parameterization applicable to post-ice shelf collapse retreat and ice cliffs exposed by other means (not MICI specifically, but MICI-esque)
• Given the important impacts of ice shelf collapse, particularly in high-emissions scenarios, our working group considers it best to include collapse in all simulations.

The last time: ISMIP6 ice shelf collapse protocol

Projected changes in surface melt using summer air temperature – melt relationship

Specified ice shelf removal after 10 years of melt >725 mm/yr (pre-collapse Larsen A/B)

​

Trusel et al (2015) Nature Geoscience

ISMIP6 collapse protocol

Nowicki et al. (2020) The Cryosphere

Without collapses: less mass loss

With collapses: more mass loss

​

“The inclusion of ice shelf collapse, here assumed to be caused by large amounts of liquid water ponding at the surface of ice shelves, yields an additional simulated mass loss of 28 mm compared to simulations without ice shelf collapse.”

​

Collapse area depends on future melt rates, which depend on driving GCM.

Seroussi et al. (2020) The Cryosphere

Room for improvement

Kuipers Munneke et al. (2014) Journal of Glaciology

A couple key considerations:

1. Melt and lakes are not the same
2. Not all lakes mean vulnerable ice shelves

​

How do we account for these?

​

Point 1 can be addressed by more fully considering the prerequisites for lake formation

• Lakes can only form if the near-surface firn density is sufficiently high and firn air content is low
• Surface hydrological routing will lead to water depths greater than excess runoff rates alone

​

Point 2 requires considering ice shelf stresses, damage, etc.

• This information can be gleaned from observations and calculated within ice sheet models.

​

Earth system model outputs

Firn model emulation

Evolution of firn air depletion, potential surface water coverage, and localized runoff

Predicts firn air content (FAC) using emulator of IMAU FDM

​

Forcing data:

• Surface melt: annual sum
• Rainfall: annual sum
• Snow accumulation: annual sum
• T2m: annual mean
• 10 m wind speed: annual mean
• Melt / Accumulation ratio (MOA): annual mean

Statistical downscaling

SMB FG-provided fields (2 km):

• Surface melt
• T2m: 2-m air temperature
• Snowfall
• Total precipitation

Estimate lake depths following Grau et al., (2025):

• Provides fields of mean lake depth and lake-covered area for each grid cell, based on ice roughness

Runoff to lake area & depth

In-ISM implementation:

• Groups apply own collapse/calving approach driven by runoff availability (post-FAC depletion)
• Couple to existing hydrofracture / damage or stress-based criteria internal to ISM

ISM-defined collapse

Path A

Path B

Firn-focused surface hydrological conditions leading to collapse paths:

• Path A: Excess melt/runoff rates for ISMs to use in own scheme
• Path B: Routing of excess melt into lakes; combined with stress criteria
• Path C: Excess melt/runoff for criteria threshold exceedance (most simple)

Bias corrections

Bias correction using RACMO2.4p1 forced by ERA5 (1995-2014; 11 km)

• T2m: additive mean bias correction
• Surface melt: hybrid multiplicative/additive scheme
• Rainfall: additive + bias corrected t2m-based phase partitioning
• Snowfall: multiplicative, then combined with re-partitioned rainfall

​

sfcWind: native ESM field regridded to 11 km

Combine lake depths + stress state:

• Use lake depth/coverage with ISM-provided stress to diagnose crevasse propagation / hydrofracture

Example stress options:

• Lai et al. (2020) critical resistive stress criterion
• Principle stresses (tension / compression thresholds)
• Nye zero / fracture stability style criteria
• Other community-adopted stress/damage triggers

Stress-aware hydrofracture

Impose collapse following runoff/FAC-based criteria, e.g.,:

• Cumulative excess melt
• Duration/persistence: consecutive years of FAC depletion/high melt; multi-year exceedance
• Optional coupling with damage or stress-based criteria internal to ISM

​

In-ISM implementation: impose ice shelf removal where criteria met.

Prescribed collapse

Path C

ISMIP7 protocol

Why bias correct?

​

Future firn depends on firn today!

​

Top row: not bias corrected

• Native ESM fields statistically downscaled by SMB FG
• Likely large-scale lake coverage by 2100 (all ice shelves!)

​

Bottom row: bias corrected

• ESM outputs corrected using RACMO2.4p1 forced by ERA5; 1995-2014 baseline
• Melt parameterized using bias corrected air temperature
• Far less potential for firn air depletion and lake coverage over 21^st century

CESM2-WACCM SSP5-8.5: 21^st century

Melt/Accumulation ratio (MOA) > 0.7

End-of-centuries emulated firn air content (FAC)

using uncorrected CESM2-WACCM SSP5-8.5 outputs from SMB focus group

Why bias correct?

Result: large-scale FAC depletion by 2100, no FAC by 2200

Why bias correct?

Result: isolated FAC depletion by 2100; expanded depletion by 2200, 2300

End-of-centuries emulated firn air content (FAC)

using bias-corrected CESM2-WACCM SSP5-8.5 outputs from SMB focus group

uncorrected

corrected

difference

FAC evolution�and impacts of �bias correction

Plots show fraction of ice shelf where FAC <= 2m (top plots) and mean ice shelf FAC (lower plots), using uncorrected or bias-corrected CESM2-WACCM SSP5-8.5 data.

​

Vertical lines represent periods where mean ice shelf FAC < 2m has been exceeded for 10 years.

​

With few exceptions (e.g., Wilkins, George VI), bias correction significantly delays FAC depletion and likely meltwater expansion.

​

FAC depletion timing

CESM2-WACCM SSP5-8.5

Years at which FAC < 2m for a decade

​

Antarctic Peninsula:

• ~21^st century

Dronning Maud Land:

• 21st – mid-22^nd century

Wilkes Land:

• ~mid-22^nd century

West Antarctica:

• ~mid-23^rd century

Ross / Ronne-Filchner:

• Sustained FAC, except near grounding zones where downslope winds likely enhance melt

​

Following FAC depletion, excess melt rates can be used to estimate ice shelf removal following one of the protocol pathways.

Ice shelf collapse – including stress conditions (Path B)

• In collapse experiments, stress-based ice shelf vulnerability to hydrofracture should be considered where/when possible
• Under the ISMIP6 collapse forcing, five of 11 ice sheet models had collapse forcing applications where more than one third of the collapsed area was not vulnerable to hydrofracture (Reynolds and Nowicki, accepted)

​

• Also - note that there is growing evidence that ice shelf collapse can occur in the absence of meltwater and hydrofracture
• BUT no sufficient consensus or synthesis of those observations exists yet to constructively parameterize collapse in the absence of surface melt
• Note that such a parameterization would likely involve changes in the ice shelf stress field induced by thinning, expansion of fractures/damage, flow changes, etc.

​

​

Recommended approach to account for these considerations:

• In collapse experiments, ice shelf stress conditions should be considered (Path B above)

​

​

Ice shelf collapse – including stress conditions (Path B) - Examples

Collapse implementation Path B would include the consideration of stress conditions in the ice shelf that enable/discourage crevasse propagation and collapse

• Many ways to compute stress and vulnerability to hydrofracture: dependent on model approach
• One example: considering second principal stress (shown: work by Maya Fields, UM graduate student, AGU 2025; using BISICLES model)

Ice shelf collapse – including stress conditions (Path B) - Examples

Collapse implementation Path B would include the consideration of stress conditions in the ice shelf that enable/discourage crevasse propagation and collapse

• Many ways to compute stress and vulnerability to hydrofracture: dependent on model approach
• Second example: considering resistive stress (shown: Lai et al. (2020) stress pre-conditioning criterion)

In regions where Rxx > R*xx (critical resistive stress): → surface fractures unstable, vulnerable to hydrofracture

​

In regions where Rxx < R*xx: → surface fractures stable, not vulnerable to hydrofracture

Antarctic ice shelf calving

• “There is a dizzying array of conflicting and seemingly contradictory observational studies in different environments that point at different controls on calving” – Bassis et al. (2023)

ISMIP 6 - Antarctic ice shelf calving

• ISMIP6 – retreat left open to modelling groups:
• 6/16 participants used fixed fronts;
• 4/16 used eigen calving;
• 3/16 used minimum thickness;
• 3/16 used divergence and accumulated damage

Since ISMIP6 - lack of community consensus remains: We do not specify any specific calving law or parameterization and leave it open to modelling groups

→ note that including *a* calving approach is better than not accounting for calving at all

Work since ISMIP6 - Wilner et al (2023):

• EC and VM laws most successful
• VM law better for ice shelves extending past their embayment
• EC works better for confined ice shelves
• Difference in extent is mostly in the low buttressing “safety band” (Furst et al., 2016) of shelves

Terminus position error at 10 sample ice shelves for four common calving laws (Wilner et al (2023))

“High-end” retreat (e.g., MICI) proposed approach

• ISMIP6: processes included in response of tributary ice streams left to modeling groups; allowed inclusion of ice cliff failure in collapse experiments
• No group opted to included MICI
• “High-end retreat” scenarios contribute significant uncertainties
• Proposed: observationally-based parameterization to account for stability of cliffs
• Can be implemented in collapse experiments (above) when cliffs are exposed
• Can also be implemented more generally (outside of collapse experiments) to tall glacier cliffs exposed by calving

“High-end” retreat (e.g., MICI) proposed approach

“High-end” retreat (e.g., MICI) proposed approach

Observed rates of ice-cliff calving during retreat of cliffs above 90 m threshold

Needell, Walker, and Bassis (submitted)

Observed: 16 glaciers with >90 m cliffs between 2019-2023; those with recorded calving rates are colored in purple

Observed calving rates during retreat plotted against associated cliff height observations for the selected glaciers; contrasted against existing parameterizations for ice-cliff calving (MICI)

​

→ Observed calving rates are much faster than existing parameterizations would predict based solely on cliff heights

“High-end” retreat (e.g., MICI) proposed approach

• Many (complex) factors may play a role in unstable retreat of tall cliffs
• e.g., surface slope, damage/crevassing, existence of mèlange, presence of bedrock highs, confinement
• Working towards an observationally-constrained cliff calving study/parameterization involving more complexities
• e.g., fitting Crawford et al. (2021) model to observations (previous slide) including lower cliff thresholds

• For ISMIP7: Focus on vulnerability to unstable/runaway retreat
• Observations show that cliffs experiencing stresses on the order of the yield strength of ice, retreating into thicker ice upstream do not undergo runaway retreat as the MICI hypothesis would suggest.
• SO: examine the forces that lead to calving-cliff height change

Needell, Walker, and Bassis (submitted)

“High-end” retreat (e.g., MICI) proposed approach

Data sources and equations used to determine vulnerability to “MICI”/unstable retreat, using data from Tuttulikassaap Sermia as an example.

Components of parameterization (left) are boxed, with cliff-thickening component parameters in purple, and dynamic-thinning component parameters in green

Questions?