How to run the PCMDI Metrics Package (PMP)

Owned by Chris Golaz
Last updated: Mar 21, 2017 by Qi Tang

The page illustrates how to run the PCMDI metrics package (PMP) on ACME model output. Please be aware that PMP is still experimental and under development. Currently, PMP is only supported on crunchy.llnl.gov.

Prerequisites

  1. Account on crunchy.llnl.gov
  2. Climatology files generated by ncclimo (Generate, Regrid, and Split Climatologies (climo files) with ncclimo). Sample files are available on crunchy for the step-by-step guide below.

Step-by-step guide

Get ACME PMP utility scripts included in ACME PreAndPostProcessingScripts repository

For illustration, we clone the repository under ~/ACME_code (Note: need to add the SSH key to GitHub account before cloning)

mkdir -p ~/ACME_code

cd ~/ACME_code

git clone git@github.com:ACME-Climate/PreAndPostProcessingScripts.git

Set-up working directory and environment

Create working directory and copy utility scripts

mkdir ~/mytest

cd ~/mytest

cp ~/ACME_code/PreAndPostProcessingScripts/create_model_diagnostics/pcmdi_metrics_package/* .

Load conda environment

export PATH=/usr/local/anaconda2/bin:${PATH}

source activate latest

Pre-process climatology files

Note: PMP uses the time variables in the climatology files to determine which months they are for, assuming the file with smallest time is January and so on. This causes problem when the climatology files are created using ncclimo with -a scd option. Since scd means seasonally continuous December (The first month used will be Dec of the year before the start year), the December climatology file will have the smallest time variable. cdscan command in the next step will misplace the input for one month (e.g., the December file for January, etc.).

check_climo.py will check the time variables in the January and December climatology files. If the December file has a smaller time value, it will be added by 365.

python check_climo.py $input_path

Use cdscan to create xml describing input files

Note: Input files are monthly climatology. Don't put seasonal and annual climo files in the input directory.

mkdir xmls

cdscan -x xmls/alpha6_1850_0070-0099.xml /work/golaz1/ACME_simulations/20160520.A_WCYCL1850.ne30_oEC.edison.alpha6_01/pp/clim_rgr/0070-0099/*.nc

The output file be a file named xmls/alpha6_1850_0070-0099.xml that describes the input netcdf files.

Process climatologies to be consistent with CMIP data structure

mkdir data

python transform_acme_clims_to_CMIP5_structure.py

The result will be a number of individual netcdf files in the sub-directory data/alpha6_1850_0070-0099:

ls data/alpha6_1850_0070-0099/

pr.nc  prw.nc  psl.nc  rlutcs.nc  rlut.nc  rsutcs.nc  rsut.nc  ta.nc  tas.nc  tauu.nc  tauv.nc  ua.nc  va.nc  zg.nc

Compute derived fields (currently only CRE fields)

python compute_acme_derived_vars.py -b data

Two additional netcdf files for LW and SW cloud radiative effects will be created under data/alpha6_1850_0070-0099:

ls data/alpha6_1850_0070-0099/
pr.nc  prw.nc  psl.nc  rltcre.nc  rlutcs.nc  rlut.nc  rstcre.nc  rsutcs.nc  rsut.nc  ta.nc  tas.nc  tauu.nc  tauv.nc  ua.nc  va.nc  zg.nc

Run PMP mean climate statistics

mkdir results

pcmdi_metrics_driver.py -p acme_pmp_parameterfile.py

This is the core of PMP. It will compute climate statistics comparing the model output with various observational datasets. The statistics are stored in json files located under the subdirectory results/pmp:

ls results/pmp/*.json

results/pmp/pr_2.5x2.5_esmf_linear_metrics.json      results/pmp/ta-850_2.5x2.5_esmf_linear_metrics.json

results/pmp/prw_2.5x2.5_esmf_linear_metrics.json     results/pmp/tas_2.5x2.5_esmf_linear_metrics.json

results/pmp/psl_2.5x2.5_esmf_linear_metrics.json     results/pmp/ua-200_2.5x2.5_esmf_linear_metrics.json

results/pmp/rltcre_2.5x2.5_esmf_linear_metrics.json  results/pmp/ua-850_2.5x2.5_esmf_linear_metrics.json

results/pmp/rlut_2.5x2.5_esmf_linear_metrics.json    results/pmp/va-200_2.5x2.5_esmf_linear_metrics.json

results/pmp/rstcre_2.5x2.5_esmf_linear_metrics.json  results/pmp/va-850_2.5x2.5_esmf_linear_metrics.json

results/pmp/ta-200_2.5x2.5_esmf_linear_metrics.json  results/pmp/zg-500_2.5x2.5_esmf_linear_metrics.json

There is one json file for every variable. Additionally, re-gridded netcdf files are under the subdirectory interpolated-output.

Create portrait plots

mkdir plots

python portrait_plots.py

The resulting output should be a png image file under the subdirectory plots:

ls plots

acme_pmp-control_tests_rms_GLB.png

This image should look like

To generate portrait plots for other regions (tropics, nothern extra-tropics, souther extra-tropics), edit portrait_plots.py and change definition of the region from global to another option and then rerun the script.

reg = 'GLB'

#reg = 'NHEX'

#reg = 'SHEX'

#reg = 'TROPICS'


Questions?

Chris Golaz may be able to help.

Acknowledgements

Many thanks to Peter Gleckler and Charles Doutriaux (Unlicensed) for their help getting me started with PMP.