Published October 5, 2026 | Version v2

Climate Variability and West Antarctic Ice Sheet Collapse

  • 1. ROR icon Alfred-Wegener-Institut Helmholtz-Zentrum für Polar- und Meeresforschung

Description

Climate Variability and West Antarctic Ice Sheet Collapse — data, model configuration and figure scripts

Authors: Javier Blasco¹, Jan Swierczek-Jereczek²·³, Marisa Montoya²·³, Jorge Alvarez-Solas³, Alexander Robinson¹

¹ Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany ² Department of Earth Physics and Astrophysics, Complutense University of Madrid, Madrid, Spain ³ Geosciences Institute, CSIC–UCM, Madrid, Spain

Contact: Javier Blasco (jablasco@ucm.es)

   
Model Yelmo ice-sheet model v1.14, run through YelmoX (esm experiment), coupled to FastIsostasy
Code version git tag ais_climvar_grl (see Model code and versions)
Domain Antarctic Ice Sheet (AIS), ANT-16KM grid (16 km)
Reference climate 1990–2019 (RACMO2.3p2 forced by ERA5; ORAS5)
Simulated period 1020–2020 CE control (1000 yr, no warming), then 2020–3500 CE warming scenarios
Associated manuscript Manuscript in preparation

1. Overview

This record contains everything needed to reproduce the results and figures of the associated manuscript:

  • Processed model output (1D_data/, 2D_data/): time-series statistics and spatial snapshots from an ensemble of Antarctic ice-sheet simulations designed to quantify the role of climate variability in the timing and magnitude of West Antarctic Ice Sheet (WAIS) collapse.
  • Input and observational data (ice_data/): the topography, velocity, basin, climate-forcing and ocean-observation files on the ANT-16KM grid that are read by the figure scripts.
  • Model configuration: the namelists, launch scripts and git tags/commits needed to rerun the simulations.
  • Figure scripts: a Julia notebook and a Python notebook that generate all main and supplementary figures from the data in this record.

The ensemble applies five levels of atmospheric warming (ΔT = 4, 6, 8, 10, 12 K) combined with stochastic climate variability resampled from the detrended 1990–2019 reference period (100 members per scenario). A parallel set of no-variability (NCV) runs provides the deterministic baseline for each scenario. A sensitivity experiment isolating oceanic (OCN) and atmospheric (ATM) variability is included for ΔT = 10 K, together with runs forced by the historical sequence of variability and by an amplified (×4) Amundsen Sea variability.

2. Contents

README.md
git_logs                                 git tags / commit hashes of the model code
yelmo_esm_Antarctica_1990-2019.nml       main YelmoX namelist
esm_ant.nml                              forcing-file and CMIP-output definitions
opt_ant_esm.sh                           launch script: present-day spin-up (optimization)
1pct_ant_esm.sh                          launch script: control + warming experiments
ManuscriptFigures.ipynb                  Julia notebook: Figures 1–4, S1–S4, S7–S9
MooringFigs.ipynb                        Python notebook: Figures S5–S6
1D_data/                                 CSV time series
2D_data/                                 NetCDF spatial fields
ice_data/                                input and observational data on ANT-16KM

3. Model code and versions

The simulations were run with the following code versions (also listed in git_logs):

Component Repository Version
Yelmo https://github.com/palma-ice/yelmo tag ais_climvar_grl (ea908bde707e71fa3f9aa7a4797f128a495a1ed5)
YelmoX (esm experiment) https://github.com/palma-ice/yelmox tag ais_climvar_grl (0dec53591b4de20677df53271710052d272bb020)
FastIsostasy https://github.com/palma-ice/FastIsostasy commit 27c741a0ce173f9e3fd80bdc8914cde10e0f92f5

4. Reproducing the simulations

4.1 Setup

  1. Clone YelmoX, Yelmo and FastIsostasy at the versions above and build the esm executable following the YelmoX documentation.
  2. Copy the namelists into the YelmoX directory at the locations the scripts expect:
    • yelmo_esm_Antarctica_1990-2019.nml → par/climtip/
    • esm_ant.nml → input/esm/
  3. Make the input data available to YelmoX. The namelists reference inputs as ice_data/{domain}/{grid_name}/... (i.e. ice_data/Antarctica/ANT-16KM/...) and isostasy_data/.... See Section 5 for how these map onto the ice_data/ folder of this record.
  4. The launch scripts were written for the AWI albedo HPC cluster. jobrun, the queue option -q 48h, and the absolute paths (/albedo/home/jablas001/..., in particular path_restart) must be adapted to your system.

4.2 Step 1 — present-day spin-up (opt_ant_esm.sh)

A 15 kyr spin-up under the constant 1990–2019 reference climate (ctrl.run_step=spinup, spinup.clim_var=none). The basal friction coefficient and a basin-wise ocean thermal-forcing correction are optimized (spinup.equil_method=opt) so that the simulated ice sheet matches present-day observations. The run writes the restart files (yelmo_restart.nc, marine_shelf.nc, isos_restart.nc, bsl_restart.nc) used by all subsequent experiments.

4.3 Step 2 — control and warming experiments (1pct_ant_esm.sh)

Transient runs from 1020 to 3500 CE restarted from the spin-up (ctrl.run_step=transient). The first 1000 years (1020–2020 CE) are a control period without warming; from 2020 CE a spatially uniform 1pctCO2-type warming ramp is applied until the target ΔT is reached, after which the climate is held constant.

The type of climate variability is selected with transient.clim_var. The script contains one launch line per experiment type; uncomment the one(s) you need and set path_restart to your spin-up output:

transient.clim_var Output folder Experiment
none noclimvar/ Deterministic, no variability (NCV baseline)
rand, transient.clim_seed=1…100 noclimvar-climvar/{seed}/ Stochastic ensemble, 100 members, random resampling of the 1990–2019 anomalies
hist noclimvar-hist*/ Variability applied in its historical (1990–2019) order

Both atmospheric (near-surface temperature, precipitation) and oceanic (temperature, salinity) variability are taken from the detrended 1990–2019 fields defined in esm_ant.nml (gcm_*_var groups).

5. ice_data/ — input and observational data

All gridded files are on the ANT-16KM grid (see Section 7 for the grid description). The folder is flat: a file referenced in the namelists as ice_data/Antarctica/ANT-16KM/<subfolder>/<file> is stored here as ice_data/<file>.

File Content Variables used Source
ANT-16KM_TOPO-BedMachine.nc Present-day ice thickness, bed and surface elevation, mask H_ice, z_bed, z_srf, mask BedMachine Antarctica (Morlighem et al., 2020)
ANT-16KM_VEL-R11-2.nc Observed surface velocity uxy_srf Rignot et al. (2011)
ANT-16KM_BASINS-nasa.nc Drainage / ocean basins (19 basins) basin_reese Reese et al. (2018)
ANT-16KM_ERA5-3H_RACMO2.3p2_1990-2019_annual_DETRENDED.nc Detrended annual near-surface temperature and precipitation, 1990–2019 t2m_dt (K), precip_dt RACMO2.3p2 forced by ERA5
ANT-16KM_ORAS5_1990-2019_DETRENDED.nc Detrended annual ocean potential temperature and salinity, 1990–2019 thetao_dt (°C), so_dt (PSU), depth ORAS5 (Zuo et al., 2019)
ANT-16KM_ORAS5_1990-2019_DETRENDED_timmean.nc Time mean of the file above thetao_dt ORAS5
ANT-16KM_ORAS5_1979-2024_thetao.nc Annual ocean potential temperature, 1979–2024 (not detrended) thetao (°C) ORAS5
OceMoor1_MIPkitA_PIGN.nc Mooring, Pine Island Glacier North thetao, lev, lat, lon MISOMIP2 MIPkit (Amundsen Sea)
OceMoor2_MIPkitA_PIGS.nc Mooring, Pine Island Glacier South as above MISOMIP2 MIPkit (Amundsen Sea)
OceMoor5_MIPkitA_MIDSHELF.nc Mooring, mid-shelf trough as above MISOMIP2 MIPkit (Amundsen Sea)
DotsonCTDsLADCP_2000_2016.nc CTD/LADCP profiles at the Dotson ice-shelf front, 2000–2016 T, P, Lat, Lon, Time Jenkins et al. (2018)

Third-party datasets are redistributed here in regridded form for reproducibility only; they remain subject to the terms of their original providers, which should be cited when these files are reused.

6. 1D_data/ — CSV files

Years are stored as floating-point values (e.g. 2020.0). Scenario files cover 2020–3500 CE; control files (ctrl_*) cover the 1020–2020 CE control period.

{region} ∈ {wais, eais, ais}; {dT} ∈ {dT4, dT6, dT8, dT10, dT12}.

6.1 Ensemble statistics (with climate variability)

Columns: year, median, q25, q75, min, max (statistics across the 100-member ensemble).

File pattern Variable Unit
slc_{region}_{dT}.csv Sea-level contribution (SLC) m
ag_{region}_{dT}.csv Grounded-ice area 10⁶ km²

Example: slc_wais_dT10.csv — WAIS sea-level contribution for ΔT = 10 K.

6.2 Deterministic no-variability runs (NCV baseline)

Columns: year, slc (SLC files) or year, area (area files).

File pattern Variable Unit
slc_{region}_{dT}_ncv.csv Sea-level contribution m
ag_{region}_{dT}_ncv.csv Grounded-ice area 10⁶ km²

6.3 Control period (1020–2020 CE)

Same variables, units and columns as in 6.1 and 6.2, for the control period preceding the warming ramp.

File Description
ctrl_slc_ais.csv, ctrl_slc_wais.csv Ensemble statistics of SLC
ctrl_ag_ais.csv, ctrl_ag_wais.csv Ensemble statistics of grounded-ice area
ctrl_slc_ais_ncv.csv, ctrl_slc_wais_ncv.csv NCV sea-level contribution
ctrl_ag_ais_ncv.csv, ctrl_ag_wais_ncv.csv NCV grounded-ice area
ctrl_forcing_var_members.csv Variability trajectories of the first 10 members (columns as in forcing_var_members.csv)

6.4 Forcing time series

File Columns Description
forcing_atm_ncv.csv, forcing_ocn_ncv.csv year, dT4, dT6, dT8, dT10, dT12 Deterministic atmospheric and oceanic warming (K) for each scenario
forcing_atm_var_{dT}.csv, forcing_ocn_var_{dT}.csv year, median, q25, q75, min, max Statistics of the stochastic variability component across the ensemble (K)
forcing_var_members.csv year, atm_m1…atm_m10, ocn_m1…ocn_m10 Variability trajectories of the first 10 members (K)

6.5 WAIS rate of volume change

Units: mm sea-level equivalent per year (conversion from ice-volume change: −0.00247 mm per km³ of ice).

File Description Columns
dvidt_wais_ncv.csv NCV rate for all scenarios year, dT4, dT6, dT8, dT10, dT12
dvidt_wais_envelope_{dT}.csv Per-timestep maximum across the ensemble year, dv_max
dvidt_wais_peaks_{dT}.csv Per-member peak rate and its year member_id, peak_year, peak_rate
dvidt_wais_envelope_ocn_dT10.csv OCN-only ensemble envelope year, dv_max
dvidt_wais_envelope_atm_dT10.csv ATM-only ensemble envelope year, dv_max
dvidt_wais_peaks_ocn_dT10.csv OCN-only per-member peak years member_id, peak_year
dvidt_wais_peaks_atm_dT10.csv ATM-only per-member peak years member_id, peak_year
dVidt_wais_{dT}_ncv.csv Single-run rate, deterministic (dT6–dT12) year, dVidt
dVidt_wais_{dT}_hist.csv Single-run rate, historical variability (dT6–dT12) year, dVidt
dVidt_wais_{dT}_x4.csv Single-run rate, historical variability with ×4 amplitude in the Amundsen Sea (dT6, dT12) year, dVidt

Note that file names are case-sensitive (dvidt_… for ensemble products, dVidt_… for single runs).

6.6 Peak sea-level-rise rate distribution

slr_accel_dist_{region}_{dT}.csv — per-member peak sea-level-rise rate within the 2500–3500 CE window. Columns: member_id, peak_year, peak_rate_mm_yr.

7. 2D_data/ — NetCDF files

All files use the ANT-16KM polar stereographic grid (EPSG:3031): dimensions xc (381) × yc (381), grid spacing 16 km, coordinates in km.

7.1 Figure 3 — ice thickness anomaly and grounding at the NCV peak year

fig3_{dT}.nc, one file per scenario (dT6, dT8, dT10, dT12).

Variable Description Unit
mean_dH_anom Ensemble-mean ice thickness minus NCV ice thickness (H̄ᵥ − Hₙᵥ) m
std_dH Ensemble standard deviation of the ice thickness anomaly m
mean_fgrnd Fraction of ensemble members in which the cell is grounded %
fgrnd_ncv NCV grounded mask (0/1) —
mask_frnt NCV ice mask (1 = ice, 0 = open ocean; ice thickness > 1 m), used to draw the ice front —

Global attributes: scenario, peak_year (year of the NCV peak rate of volume change, used as snapshot time), snap_year.

7.2 Figure 4 — OCN/ATM sensitivity at fixed years (ΔT = 10 K)

fig4_dT10_yr{year}.nc, one file per year (2800, 2900, 2950, 3000).

Variable Description Unit
mean_fgrnd_atm_ocn Grounded fraction, ATM+OCN variability ensemble %
mean_fgrnd_ocn Grounded fraction, OCN-only variability ensemble %
mean_fgrnd_atm Grounded fraction, ATM-only variability ensemble %
fgrnd_ncv NCV grounded mask (0/1) —
mask_frnt NCV ice mask (1 = ice, 0 = open ocean) —

Global attributes: dT, snap_year.

Note on grounded fraction: values are the percentage of ensemble members (0–100 %) in which a grid cell is grounded at the snapshot year. A value of 100 % (grey in the figures) means that all members are still grounded there.

7.3 Supplementary fields

File Description Variables
spinup.nc Output of the present-day spin-up; the last time step is the initial state of all experiments H_ice, H_ice_pd_err (m); uxy_s, uxy_s_pd_err (m yr⁻¹); z_bed (m); cb_ref; bmb (m yr⁻¹); mask_bed; tf_corr (K); time
dto_basins.nc Oceanic temperature variability applied in the model dto_var(xc, yc, time) (K), time
2Dsm_wais_{dT}_hist.nc Time-dependent fields, historical-variability run (dT6, dT12) H_ice (m), mask_bed, time
2Dsm_wais_{dT}_x4.nc Time-dependent fields, ×4 Amundsen variability run (dT6, dT12) H_ice (m), mask_bed, time

8. Reproducing the figures

In both notebooks, set the two paths in the first cells:

  • path_data — the root folder of this record (the one containing 1D_data/, 2D_data/ and ice_data/);
  • path_figures — the folder where the PDF figures are written.

8.1 ManuscriptFigures.ipynb (Julia)

Run the first four cells (packages, paths, colour maps, common fields) before any figure section.

Figure Output file Input
Fig. 1 1_climatologies.pdf ice_data/: RACMO2.3p2 and ORAS5 detrended files, BedMachine, basins
Fig. 2 2_wais_1d.pdf forcing_var_members.csv, forcing_{atm,ocn}_ncv.csv, slc_wais_{dT}[_ncv].csv, ag_wais_{dT}[_ncv].csv
Fig. 3 3_dwais.pdf dvidt_wais_ncv.csv, dvidt_wais_envelope_{dT}.csv, dvidt_wais_peaks_{dT}.csv, fig3_{dT}.nc
Fig. 4 4_sensitivity_dT10.pdf dvidt_wais_ncv.csv, dvidt_wais_{envelope,peaks}[_ocn,_atm]_dT10.csv, fig4_dT10_yr{year}.nc
Fig. S1 SM1_pd-spinup.pdf spinup.nc, BedMachine, observed velocity
Fig. S2 SM2_tf_corr_map.pdf spinup.nc (tf_corr), basins
Fig. S3 SM3_ctrl.pdf ctrl_*.csv
Fig. S4 SM4_dto_basins.pdf dto_basins.nc, basins
Fig. S7 SM7_hist_vs_histx4.pdf dVidt_wais_{dT}_{hist,x4}.csv, 2Dsm_wais_{dT}_{hist,x4}.nc
Fig. S8 SM8_hist_vs_deterministic.pdf dVidt_wais_{dT}_{ncv,hist}.csv
Fig. S9 SM9_eais_1d.pdf as Fig. 2 with region = "eais"

Requirements: Julia 1.11 with CairoMakie, GeoMakie, NaturalEarth, NCDatasets, CSV, DataFrames, JLD2, LaTeXStrings, ColorSchemes, Colors, DataStructures, Optim, Distributions, and the helper functions in IcePlots.jl (make_axis_ice2D, gencol_vel, preprocess_ice2D_variable). Figures use the "CMU Serif" font.

8.2 MooringFigs.ipynb (Python)

Figure Output file Input (all in ice_data/)
Fig. S5 SM5_pig_moorings.pdf OceMoor{1,2,5}_MIPkitA_*.nc, ORAS5 files, basins, BedMachine
Fig. S6 SM6_dotson_moorings.pdf DotsonCTDsLADCP_2000_2016.nc, ORAS5 files, basins, BedMachine

Both compare observed ocean temperature (profiles down to 1000 m and the 500–800 m average) with ORAS5 at the nearest valid grid cell.

Requirements: Python 3 with xarray, numpy, matplotlib, pyproj and a NetCDF backend (netCDF4).

9. Ensemble design

Parameter Value
Ice-sheet model Yelmo v1.14, ANT-16KM (16 km), DIVA solver
Initial state 15 kyr optimization spin-up under the 1990–2019 climate
Control period 1020–2020 CE (1000 yr), no warming
Warming scenarios ΔT_atm = 4, 6, 8, 10, 12 K, reached along a 1pctCO2-type ramp starting in 2020 CE and held constant afterwards; ΔT_ocn = 0.25 × ΔT_atm
Ensemble size 100 members per scenario (random seeds 1–100)
Variability source Random resampling of detrended 1990–2019 anomalies: RACMO2.3p2/ERA5 (atmosphere) and ORAS5 (ocean)
Baseline One deterministic no-variability (NCV) run per scenario
Sensitivity experiments OCN-only and ATM-only variability at ΔT = 10 K (100 members each); historical-order variability; ×4 Amundsen Sea variability
Basal melt Quadratic non-local parameterization with basin-wise thermal-forcing correction from the spin-up
Glacial isostatic adjustment FastIsostasy (LV-ELRA)

10. License and citation

Creative Commons Attribution 4.0 International (CC BY 4.0) for the model output, configuration files and scripts produced by the authors. Third-party datasets in ice_data/ remain subject to their original terms (see Section 5).

Please cite the associated manuscript and this Zenodo record when using these data.

Files

ClimVar_dataset.zip

Files (2.3 GB)

Name Size
md5:e24271bb31d5a1d1d1a63aab43e35194
2.3 GB Preview Download