Framework for assimilating Remote-sensing Observations for Surface mass balance Tuning

FROST is a data assimilation framework tailored for glacier modeling. It couples the 3D glacier model IGM with an Ensemble Kalman Filter to calibrate glacier-specific surface mass balance parameters using remote sensing observations. The method is derivative-free, and scalable. It also provides uncertainty estimates alongside calibrated results.

---

🏗️ Installation

1. Clone the repository

   git clone [email protected]:FAU-glacier-systems/FROST.git

2. Create a virtual environment with conda

   cd FROST
   conda env create -f environment.yml
   conda activate frost_env

---

🚀 Pipeline for Calibration

1. Duplicate the experiments/test_default folder and rename it to your custom experiment name, e.g., experiments/my_run. Adapt the config.yml to your target glacier and desired setup e.g. rgi_id and experiment_name

2. Download elevation change product and adapt the path in config.yml e.g. 'data/raw/hugonnet/11_rgi60_2000-01-01_2020-01-01' https://www.sedoo.fr/theia-publication-products/?uuid=c428c5b9-df8f-4f86-9b75-e04c778e29b9

3. Run the pipeline

   python frost_pipeline.py --config experiments/<experiment-name>/pipeline_config.yml 

   A overview of the pipeline is shown below:

4. View the results:

• Calibration Results data/results/<experiment-name>/glaciers/<rgi-id>/calibration_results.json

• Monitoring Images data/results/<experiment-name>/glaciers/monitor/status.png

• Example

---

🏛️ Architecture

A schematic overview of the FROST calibration workflow:

---

📎 Reference

If you use FROST, please cite:

Herrmann, O. et al. (2025) ‘A Kalman filter-based framework for assimilating remote sensing observations into a surface mass balance model’, Annals of Glaciology, 66, p. e23. doi:10.1017/aog.2025.10020.