Iceland Glacier Projections — Delivery Overview

OGGM hindcast and CMIP5/CMIP6 projections for Landsvirkjun

Author

Fabien Maussion & Patrick Schmitt

Published

July 7, 2026

What this is

This delivery contains state-of-the-art OGGM hindcast and future projections for all 572 Icelandic glaciers, run to 2100 and 2300 under the CMIP5 and CMIP6 climate model ensembles, together with the notebooks used to select, validate, and analyse them. The projections are intended to support your team’s work on how projected changes in the Atlantic Meridional Overturning Circulation (AMOC) may influence Icelandic glacier evolution — see Landsvirkjun consultancy (google doc) for the original scope of this work.

Configuration used

Two climate forcings were calibrated and validated: CARRA and ERA5 (1991-2026). CARRA (trapezoidal bed shape, lambda=14) is the delivered configuration used throughout the main projections. ERA5 (lambda=8) was selected as the best-performing configuration for this climate forcing and is kept alongside it for comparison — the two configurations differ only modestly in their projected volume loss. The full reasoning and validation evidence behind this choice is in notebook 00, below.

Contents & how to read this delivery

Each notebook below is linked two ways: an nbviewer link (read directly in your browser, no setup needed) and the underlying .py script (a plain, # %%-commented Python file — the notebook .ipynb next to it was generated from this script and is the exact same content, just executed). Either can be opened directly in Jupyter if you want to re-run or adapt the analysis.

00 — Experiment selection & validation (.py script) — the calibration sweep (climate forcing × bed shape × mass-balance model) validated against glacier area, Hugonnet mass change, DEM-differenced volume change, and SMB maps, ending in the decision to ship CARRA, lambda=14.
01 — Comparison with reference models (.py script) — how our Iceland-specific calibration compares to three independent published projections (OGGM standard, GloGEM, PyGEM-OGGM) for the same glaciers.
02 — Projections analysis (.py script) — the main deliverable. Total Iceland glacier volume to 2100 and 2300, the same projections re-grouped by global temperature level and by AMOC index, and total glacier runoff by AMOC index (following the same method as the OGGM-EDU tutorial on glacier water resources). This is the notebook to start from if you want an initial analysis of the projections themselves, or a template to build your own.
03 — AMOC index — two standalone CLI scripts (amoc_index_timeseries.py, amoc_index_classify.py, not notebooks) that build the AMOC-strength classification used in notebook 02, plus amoc_index_table_README.txt documenting the full methodology and the already-computed output tables. See Data below for what is and isn’t shipped here.

Downloading this delivery

Everything in this README — notebooks, scripts, figures, and data — is hosted at:

https://cluster.klima.uni-bremen.de/~fmaussion/share/iceland_projections/

You can browse it directly in a browser, or download the whole delivery as a single archive:

https://cluster.klima.uni-bremen.de/~fmaussion/share/iceland_projections.tar.gz

wget https://cluster.klima.uni-bremen.de/~fmaussion/share/iceland_projections.tar.gz
tar -xzf iceland_projections.tar.gz

The archive is about 4.1 GB (compressed).

Key results

Total Iceland glacier volume, 2000-2100, by scenario (notebook 02)

Total Iceland glacier volume, 2000-2100, grouped by AMOC index (notebook 02)

Total Iceland glacier runoff, grouped by AMOC index — note the “peak water” divergence between groups (notebook 02)

Comparison with published reference models, CMIP6 (notebook 01)

Historical mass change vs Hugonnet et al. observations, shortlisted configurations (notebook 00)

More figures are in the figures/ folder, and each notebook produces the ones relevant to its own analysis.

Data

All data referenced by these notebooks ships alongside them, organised under data/:

data/
  historical/     — CARRA and ERA5 historical OGGM runs (1991-2026)
  projections/    — merged CMIP5/CMIP6 projection NetCDFs (2100 and 2300), CARRA only
  supporting/     — observations (Hugonnet, SMB maps), reference-model CSVs,
                    global-temperature-level tables, and the AMOC index's own
                    supporting files (subpolar-gyre mask, Weijer et al. streamfunction)

Two things are not shipped, to keep this delivery a manageable size:

The full calibration sweep behind notebook 00 (~15 configurations across bed shape and mass-balance model choices) — only the shortlisted configurations’ data is included. The full sweep’s data can be made available on request.
The raw CMIP6/CMIP5 GCM tas files used by amoc_index_timeseries.py to build the AMOC index from scratch (73+ files, too large for this delivery). The already-computed output (i_tas_timeseries.csv, amoc_index_table.csv) is included, so amoc_index_classify.py and notebook 02 run standalone. To regenerate the index from scratch, the GCM files are publicly available at:
- CMIP6: https://cluster.klima.uni-bremen.de/~oggm/cmip6/
- CMIP5: https://cluster.klima.uni-bremen.de/~oggm/cmip5-ng/

Running this delivery

The notebooks use only relative paths (data/, figures/, ../03_amoc_index/), so they must be run from within this folder. A standard scientific Python environment is enough: xarray, pandas, numpy, matplotlib, pyyaml, and jupytext if you want to convert the .py scripts back to notebooks yourself (jupytext --to notebook 02_projections_analysis.py). The AMOC index scripts additionally need regionmask, scipy, and cf_xarray/xesmf-adjacent packages only if re-run from raw GCM data.

About this delivery

These notebooks and scripts were built with the assistance of Claude (Anthropic’s AI coding assistant), under our direction and review. If your team wants to extend this analysis — different scenario groupings, per-ice-cap breakdowns, other hydro variables (melt vs. precipitation split, peak-water timing per ice cap), or anything else the underlying NetCDF data supports — Claude Code can be pointed at this project directory to pick up the analysis directly; the notebooks and this README are written to give it (and any future collaborator, human or otherwise) enough context to continue from here.