Antarctic GlaDS emulation

Tim Hill, 2026 (tim_hill_2@sfu.ca) | https://github.com/timghill/antarctic-glads

Description

The project structure is:

• utils/: shared code for setting up experiments and analyzing outputs
• issm/: individual directories for GlaDS-ISSM model runs, including stored GlaDS ensembles and ISSM solutions
• analysis/: emulator fitting, evaluation, and all analysis
  • analysis/glads/: analysis of the GlaDS ensemble
  • analysis/mean/: analysis for the mean of the perturbed-parameter ensemble
  • analysis/parameters_full/: RF including all features + parameters
  • analysis/parameters_reduced/: simplified RF including 4 features + parameters
• manuscript/: final manuscript figures
• data/: raw data (BedMachine, ice velocities, and other published datasets not reproduced here)
• examples/: notebook showing how to read the trained random forest and make predictions

Each directory has a README file to describe the contents.

Installation

The analysis source code has been tested against python 3.11.5. Package requirements are listed in requirements.txt, and it is recommended to use a virtual environment to manage versions. For example

virtualenv --python 3.11 pyenv/
source pyenv/bin/activate
pip install -r requirements.txt

To install the code for this project on your python path, install in editable (-e) mode with pip from the project root directory:

pip install -e .

Reproducing analysis

Main manuscript figures are made from the following scripts:

1. parameters_reduced/plot_hexbin.py
2. parameters_full/plot_ensemble_spread.py
3. mean/plot_hyperparameter_sensitivity.py
4. mean/plot_hyperparameter_sensitivity.py
5. parameters_reduced/map_delta.py
6. mean/flowlines.py
7. parameters_reduced/future_flowlines.py

Appendix figures:

• A1: analysis/groundingline_statistics.py
• B1: issm/plot_lcurve.py
• C1: analysis/mean/future_flowlines.py
• D1: analysis/mean/plot_heatmap.py
• D2: analysis/mean/plot_trees.py
• D3: analysis/mean/compute_shap.py
• D4: analysis/RF.py