# PyGEM Projections of 21st-Century Glacier Change in Alaska's National Parks

## Facts

- Provider: David Rounce, Brandon Tober, Doug Brinkerhoff, Michael Loso, Patrick Schmitt, Fabien Maussion (Carnegie Mellon University)
- License: CC-BY-4.0 (https://kilthub.cmu.edu/articles/dataset/Data_and_code_for_Projections_of_21st_century_glacier_change_in_Alaska_s_national_parks_calibrated_against_spatially-distributed_elevation_change_observations_/32319708)
- Size: 26 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/CMU-OSPO/Data_and_code_for_Projections_of_21st_century_glacier_change_in_Alaska_s_national_parks
- Access: public, readable without an account
- DOI: https://doi.org/10.1184/R1/32319708.v2
- Category: Environmental Science
- Source: https://doi.org/10.1184/R1/32319708.v2
- Page: https://scigantic.com/archive/f61e3b4a-fbf7-570d-981f-c3bc76416ee6
- Keywords: Environmental Science, Parquet, Hugging Face

## About

Data and code behind projections of glacier mass loss in Alaska's national parks to 2100, from a Bayesian inference framework that calibrates the Python Glacier Evolution Model (PyGEM) against spatially distributed surface elevation change from 1994-2021 airborne laser altimetry (Operation IceBridge Alaska) for 185 glaciers covering over half of the region's glacier area. Calibration cuts parameter uncertainty by 5-18% and hindcast elevation-change error by 0.18 m per year. Projections give 43 +/- 17% to 67 +/- 27% mass loss depending on emissions scenario, with median losses of 57-99% for Lake Clark, Denali, Kenai Fjords and Glacier Bay against 28-51% for Wrangell-St. Elias. The Hugging Face copy (CMU Open Source Programs Office) holds the KiltHub original_files: RGI outlines, OGGM glacier directories, calibration chains, emulator, simulation outputs (one archive is 8.3 GB), elevation-change data, code and SLURM scripts, about 14.6 GB of zip archives, plus a small Parquet table of regional mass and area loss relative to 2015. Licensed CC BY 4.0 on KiltHub; cite Rounce et al. 2026, doi:10.1184/R1/32319708.v2.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("CMU-OSPO/Data_and_code_for_Projections_of_21st_century_glacier_change_in_Alaska_s_national_parks")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `regional_mass_area_loss_relto2015/train-00000-of-00001.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/f61e3b4a-fbf7-570d-981f-c3bc76416ee6
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/f61e3b4a-fbf7-570d-981f-c3bc76416ee6
- Over MCP: `get_schema_card` with id `f61e3b4a-fbf7-570d-981f-c3bc76416ee6` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Environmental Science datasets](https://scigantic.com/topics/environmental-science)
