https://github.com/kenzaxtazi/bcm4rcm
Robust Bayesian committee machines for regional climate model ensemble learning
https://github.com/kenzaxtazi/bcm4rcm
Last synced: 4 months ago
JSON representation
Robust Bayesian committee machines for regional climate model ensemble learning
- Host: GitHub
- URL: https://github.com/kenzaxtazi/bcm4rcm
- Owner: kenzaxtazi
- Created: 2023-10-09T19:00:40.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-02-20T15:42:46.000Z (4 months ago)
- Last Synced: 2025-02-20T16:27:46.765Z (4 months ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 41.1 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# bcm4rcm
Robust Bayesian committee machines for regional climate model ensemble learning.
Tazi K., Kim S.W.P., Girona-Mata M., & Turner R.E. (2025). Refined climatologies of future precipitation over High Mountain Asia using probabilistic ensemble learning. [arXiv preprint arXiv:2501.15690](https://arxiv.org/abs/2501.15690)
## Code
### experiments
* `bcm_preds`: scripts and visualisations for the robust Bayesian committee machine (BCM)
* `ensemble_learning`: scripts and visualisations for the Wassertein ditance calculations and weight optimisations
* `moe_preds`: scripts and visualisations for the mixture of experts (MoE) and equally weighted mixture model (EW) sampling
* `validation`: scripts and visualisations for the MoE validation
### models
* `guepard_baselines`: updated BCM code from the `guepard` library
### plots
* `hma_map`: map of study area with with subregions
### utils
* `areal_plots`: functions to help plot the maps
* `process_data`: functions to process raw RCM data to make files over desired time periods
## Data
The CORDEX-WAS data are available through the [Earth System Federation Grid nodes](https://esgf-metagrid.cloud.dkrz.de) and the APHRODITE data through [APHRODITE's Water Resources website](http://aphrodite.st.hirosaki-u.ac.jp/download/). The data for the final results analysed in the paper are found on [Zenondo](https://doi.org/10.5281/zenodo.14837272)