mk-prism

mk-prism converts yearly MK-PRISM South Korea climate archives into GeoTIFF products.

The script in main.py unpacks MK-PRISM NetCDF archives from dataset/, exports yearly rasters in EPSG:4326, and reprojects them to EPSG:32652 at 30 m resolution.

Current State

This repository currently contains:

• main.py: the conversion pipeline
• config.yaml: runtime paths, CRS settings, dimension names, and export options
• environment.yml: Conda environment definition
• dataset/ with four MK-PRISM yearly archive files for 2000-2019:
  • MKPRISM_MKPRISMv21_skorea_TA_gridraw_yearly_2000_2019_nc.tar.gz
  • MKPRISM_MKPRISMv21_skorea_TAMIN_gridraw_yearly_2000_2019_nc.tar.gz
  • MKPRISM_MKPRISMv21_skorea_TAMAX_gridraw_yearly_2000_2019_nc.tar.gz
  • MKPRISM_MKPRISMv21_skorea_RN_gridraw_yearly_2000_2019_nc.tar.gz

The repository does not currently include extracted .nc files or generated GeoTIFF outputs. Those are created when the pipeline runs.

Runtime Outputs

When you run the converter, it creates and populates:

• dataset/raw_data/: original tar archives after extraction
• processed_epsg4326/: yearly GeoTIFF outputs in EPSG:4326
• processed_utm52n_30m/: yearly GeoTIFF outputs in EPSG:32652 with 30 m resolution

Variables

The bundled MK-PRISM archives cover four variables:

• TA: mean temperature
• TAMIN: minimum temperature
• TAMAX: maximum temperature
• RN: precipitation

MK-PRISM data download:

• https://climate.go.kr/atlas/ana/cdd

Environment

For public use, prefer the portable environment file:

conda env create -f environment-public.yml
conda activate mk-prism

The existing environment.yml is better treated as a fuller local export. It can still be useful for reproducing the original setup, but it is more platform-specific than most public users need.

If you want to use the exported environment instead, run:

conda env create -f environment.yml
conda activate aibio_310

If you keep environment.yml in the public repo, avoid reintroducing a machine-specific prefix: line.

Usage

Run the full conversion pipeline:

python main.py --config config.yaml

Run only a subset of years:

python main.py --config config.yaml --from 2010 --to 2015

Processing Steps

The pipeline does the following:

• scans dataset/ for *.tar* and *.nc files
• safely extracts .nc members from tar archives
• moves processed tar files into dataset/raw_data/
• reads the configured time dimension (time by default)
• writes yearly GeoTIFF files in EPSG:4326
• reprojects each raster to EPSG:32652
• uses bilinear resampling at 30 m resolution
• writes float32 outputs and fills missing values with -99
• writes tiled, deflate-compressed GeoTIFF files with BIGTIFF=YES

Output Naming

Example geographic output:

processed_epsg4326/MKPRISM_MKPRISMv21_skorea_TA_yearly_2010_2015/MKPRISMv21_skorea_TA_2015_4326_1e-2.tif

Example projected output:

processed_utm52n_30m/MKPRISM_MKPRISMv21_skorea_TA_yearly_2010_2015/MKPRISMv21_skorea_TA_2015_32652_30m.tif

Notes

• Paths in config.yaml are resolved relative to main.py, not the shell working directory.
• Files are filtered by the MKPRISM_MKPRISMv21_skorea_ prefix configured in config.yaml.
• If extracted .nc files already exist in dataset/, the script skips re-extraction and moves matching tar files into dataset/raw_data/.

License

Code and documentation in this repository are licensed under the MIT License. See LICENSE.

The contents of dataset/ are not automatically covered by the MIT License unless you explicitly have the right to release those files under MIT. If you plan to publish this repository, confirm the redistribution terms for the dataset archives or document a separate data license.

Public Release Checklist

Before pushing this repository publicly, verify the following:

• confirm that you have the right to redistribute the files in dataset/
• if the dataset archives are public, check their internal metadata for usernames, internal paths, or processing history you do not want to expose
• prefer environment-public.yml for public setup instructions
• keep generated outputs out of version control via .gitignore