ausclimateservice / agcd Goto Github PK

Validation of the Climate Analysis Forecast Ensemble dataset (CAFE) often involves comparison against the Australian Gridded Climate Data dataset (AGCD).

This repository contains details of the code, data processing steps, and software environment used to process the original AGCD dataset (maintained by the Bureau of Meteorology and hosted at /g/data/zv2/agcd/) into a format consistent with the CAFE dataset (maintained by CSIRO and hosted at /g/data/xv83/dcfp/). This essentially involves converting the original AGCD netCDF files to Zarr format and then regridding to the CAFE spatial grid. The processed files are stored at:
/g/data/ia39/AGCD/post-processed/data/

The following Python scripts contain the code used to process the AGCD data:

• nc_to_zarr.py: converts the netCDF AGCD data files to zarr format
• regrid_agcd_to_cafe.py: regrids AGCD data to the CAFE grid

The Makefile contains the instructions to build and run the sequence of commands that were used to execute the Python scripts and produce each data file:

agcd_v2_precip_total_r005_monthly_1900-2020.zarr.zip

• make zarr TIMESCALE=monthly
• Converts the version 2 monthly AGCD data (0.05 degree grid) to zarr format

agcd_v2_precip_total_cafe-grid_monthly_1900-2020.zarr.zip

• make regrid TIMESCALE=monthly
• Regrids the monthly AGCD data (0.05 degree grid) to the CAFE grid

agcd_v1_precip_total_r005_daily_1900-2020.zarr.zip

• make zarr TIMESCALE=daily
• Converts the version 2 daily AGCD data (0.05 degree grid) to zarr format

agcd_v1_precip_total_cafe-grid_daily_1900-2020.zarr.zip

• make regrid TIMESCALE=daily
• Regrids the daily AGCD data (0.05 degree grid) to the CAFE grid

To find out exactly what commands were executed, run any of those make commands with the -n (dry run) and -B (force make) options (e.g. make -n -B make zarr TIMESCALE=monthly) or open any of the data files and look at the global history attribute.

The following Python libraries are dependencies for running the code: xarray, netCDF4, dask, xesmf, gitpython and cmdline_provenance.

When the Makefile executes a Python command it uses the conda environment at /g/data/e14/dbi599/miniconda3/envs/agcd, which has all the dependencies (and their dependencies) installed. The complete list of packages installed in that environment is shown in environment.yml.

If you'd like to recreate that exact software environment (you don't need to in order to run the Makefile), run the following at the command line:

$ conda env create -f environment.yml

Questions or comments are welcome at the GitHub repostory associated with the code:
https://github.com/AusClimateService/agcd/issues