NCI Data Collection Intake Catalogue

Usage

To access the catalogue and the data you need to be part of the ia39 project.

To load the catalogue

import intake
cat = intake.open_catalog('/g/data/ia39/aus-ref-clim-data-nci/acs-replica-intake/catalogue.yaml') 

NB. if you are using an intake-esm version >= 2022.9.18 as the class esm_datastore has changed we need to pass different arguments to it. Hence we defined a separate catalogue file:

import intake
cat = intake.open_catalog('/g/data/ia39/aus-ref-clim-data-nci/acs-replica-intake/catalogue_new.yaml')

If you are using the CLEX-CMS conda environments, you can use this shortcut

import intake

cat = intake.cat.acsreplica

To load these conda environments you need to be part of the hh5 project.

module use /g/data/hh5/public/modules
module load conda/analysis3

Individual netcdf datasets are catalogued using intake-esm.

Admin

This catalogue exists on Gadi's NCI filesystem under /g/data/ia39/aus-ref-clim-data-nci/acs-replica-intake/catalogue.yaml

Use git pull to download changes from Github

Most of the catalogue csv listings themselves need to be generated, as most datasets are regularly updated. This may be done by running make within the directory

The intake data package is under the directory package/, it simply provides an intake entry point pointing to the catalogue directory. As long as this doesn't change there is no need to re-install the catalogue, even if the csv listings get updated and/or new datasets are added.

Contributing

If you would like to contribute you can clone the repository, create a new branch to work on. Once you added a new dataset push your new branch and create a amerge request.

Adding a new dataset listing

To add a new dataset you need:

• add the main dataset metadata to the catalogue.yaml file
• create a directory for the dataset
• add here:
  • a catalogue.csv.xz file listing all files, good option for stable or "smaller" collections
  • or a ingest.yaml file to generate one on the fly

The catalogue.yaml file lists the main metadata for each dataset:

cmip6_etccdi:
        description: |
            Replica of the Climate extreme indices and heat stress indicators derived from CMIP6 ...
        Project: ia39
        Maintained By: CLEX
        Contact: [email protected]
        Documentation: https://github.com/aus-ref-clim-data-nci/CMIP6_ETCCDI
        License: https://cds.climate.copernicus.eu/api/v2/terms/static/cicero-cmip6-indicators-licence.pdf
        Citation: Sandstad, M., Schwingshackl, C., Iles, ...
        References:
                -  https://cds.climate.copernicus.eu/cdsapp#!/dataset/sis-extreme-indices-cmip6?tab=overview
                - ... 
        driver: intake_esm.esm_datastore
        args:
            esmcol_obj: '{{CATALOG_DIR}}/cmip6_etccdi/catalogue.json'

NB. As shown quotes are not needed for the fields, but they should used when they contain :, urls are scanned correctly.

The catalogue.csv.xz file is a compressed csv file that lists every file in the dataset and the corresponding values for each of attributes defined for the dataset. As an example a snippet of the catalogue for the cmip6_etccdi dataset:

path,index_type,base,frequency,experiment,model,ensemble,variable,date_range
/g/data/ia39/aus-ref-clim-data-nci/cmip6-etccdi/data/v1-0/etccdi/base_independent/yr/ssp370/ACCESS-CM2/txxETCCDI_yr_ACCESS-CM2_ssp370_r1i1p1f1_no-base_v20191108_2015-2100_v1-0.nc,etccdi,base_independent,yr,ssp370,ACCESS-CM2,r1i1p1f1,txx,2015-2100
/g/data/ia39/aus-ref-clim-data-nci/cmip6-etccdi/data/v1-0/etccdi/base_independent/yr/ssp370/ACCESS-CM2/r10mmETCCDI_yr_ACCESS-CM2_ssp370_r1i1p1f1_no-base_v20191108_2015-2100_v1-0.nc,etccdi,base_independent,yr,ssp370,ACCESS-CM2,r1i1p1f1,r10mm,2015-2100
...

As mentioned above, you can provide this list but especially if you're dealing with big datasets or datasets that are regularly updated it is easier to provide a ingest.yaml file to generate this on the fly. Examples of ingest.yaml are available in the existing dataset directories.

The ingest.yaml is composed of 4 main sections:

• The main dataset metadata, as added to the main catalogue.yaml
• find section where you define the main path and any "find" options

find:
    paths:
        - /g/data/ia39/aus-ref-clim-data-nci/gpcp/data
    options: -not -type d -not -path */fx/* -name *.nc

• drs section where you can use Python regular expression to retrieve attributes from the directory structure and filenames

#   /g/data/ia39/aus-ref-clim-data-nci/gpcp/data/mon/v2-3/2021/gpcp_v02r03_monthly_d202106_c20210907.nc
drs: |
    ^(?P<path>/g/data/ia39/aus-ref-clim-data-nci/gpcp/data
    /(?P<frequency>[^/]+)
    /(?P<version>[^/]+)
    /(?P<year>[^/]+)
    /gpcp_(.*[^_]_){2}(?P<date>[^_]+)
    _(.*[^\.]+)
    .nc)

We provided a simple code test_drs.py to test your DRS regex separately.

• assets section, here the path is added as an attribute column, all the others are automatically generated based on the DRS pattern. This is also where the attributes used to groupby and aggregate the data are defined.

assets:
    column_name: path
    format: netcdf
aggregation_control:
    # Name of the variable in the file
    variable_column_name: variable
    # Grouping keys are made of these columns, joined by '.'
    groupby_attrs:
        - realm
        - frequency
        - variable
        - version
    aggregations:
        # Join along the existing time dimension
        - type: join_existing
          attribute_name: date_range
          options:
              dim: time

Sometimes you need to do some further processing of the attributes you can derived from the DRS in that case you can do that by coding in a postprocess.py script. You add this script and any input files that you might need in the dataset directory, this will be automatically run after the ingest.yaml. There are examples both in the gpcc and frogs datasets.

More information is available in the official intake-esm documentation.