Skip to main content

Parallel Python Averager for Climate Data

Project description

August 19, 2016




A package used for computing averages from climate model output.

Authors: Sheri Mickelson, Kevin Paul, and John Dennis
Version: 0.9.10
Copyright: Contained within LICENSE.txt
Comments and feedback:

What the PyAverager Can Do

The PyAverager can create the climatology files needed by the AMWG, OMWG, Land, and Ice CESM diagnostic packages (the full list of averages is defined within the ‘Specification’ section). It is able to compute averages from previously generated averages (such as monthly averages for season averages) or from scratch. The PyAverager can operate on monthly time-slice and time-series files that exist in the same directory.


The PyAverager depends on the PyNIO, mpi4py, and PyTools packages to be installed on your system. PyNIO is needed for NetCDF file I/O. mpi4py and PyTools are needed for the parallel communication, though it is possible to run the PyAverager in serial mode without mpi4py.

If you are not running on a CESM supported machine, installation information can be found at:
(git clone
(svn co

CESM Supported Machine Information (steps to take to get these packages into your path):

Add the following to the top of your bsub script or execute them on the command line and then type module save to always keep them in your environment:
module load python
module load all-python-libs
module load asaptools
(If asaptools fails to load, it has since been added to all-python-libs module)

How To Install ASAP PyTools
- git clone pytools
- cd into pytools
- python install --user

Building and Installing the PyAverager

Check out the source code (3 options):
- git clone
- pip install pyAverager --user

If checking out the code via svn or git, you must install the pyAverager (the pip method will install the package for you).
o $ cd PyAverager
o $ python install --user

Make sure the install location is added to your $PYTHONPATH
o You can also type python to get the interactive terminal and then type
from pyaverager import PyAverager, specification
You will get an error if it is not in your path

To install documentation, run:
o $ doxygen Doxyfile
The documentation will be created in the apidocs directory.

Running the Examples

Running the examples on Yellowstone:
(For other machines, you will need to create a queue submission script similar to examples/ runAvg_mpi.csh)

o $ cd examples
o Open runAvg_mpi.csh for editing
o Set the correct project number to run under
o Select an example to run (control_*.py)
o Edit the control_*.py script you would like to run
(See the Specifier section below for more details on editing the control script)
o Run
o $ bsub < runAvg_mpi.csh


The PyAverager is a python library that is referenced from another python script. In order to run the PyAverager, you need to specify parameters so the average knows what types of averages to compute, input/output locations, and any averaging options you would like to add. The example directory contains several files that you can use as templates. You can copy one of these scripts and modify the top section to fit your data.

CESM naming conventions that the PyAverager follows by default:

Slice: $CASE.$comp.$stream.$year-$

Series: $CASE.$comp.$stream.$var.$year1$month1-$year2$

(If your file names do not match this pattern, you will need to pass the file_pattern variable to the specifier)

Types of Averages

The table below lists the types of averages the PyAverager can compute.
Average Option Description Output Name Can be Weighted? Can Be Created As a Dependency?
ya Yearly Average $CASE.$ Yes No
tavg Ocn average across years tavg.$Year1-$ Yes Yes
annall Land model, annual averages
concat together $ Yes Yes
moc Ocn MOC file $ Yes Yes
hor.meanyr Ocn hor.mean year file $REG_hor.meanyr.$ Yes Yes
hor.meanConcat Ocn, concat of hor.meayr $REG_hor_mean_hor.meanConcat.
$CASE.$Year1-$ Yes Yes
mocm Ocn MOCM $ No No
ann Annual Average $ Yes Yes
djf Winter Average $ Yes Yes
mam Spring Average $ Yes Yes
jja Summer Average $ Yes Yes
son Fall Average $ Yes Yes
jan January Average $ Yes Yes
feb February Avg $ Yes Yes
mar March Average $ Yes Yes
apr April Average $ Yes Yes
may May Average $ Yes Yes
jun June Average $ Yes Yes
jul July Average $ Yes Yes
aug August Average $ Yes Yes
sep Sept Average $ Yes Yes
oct Oct Average $ Yes Yes
nov Nov Average $ Yes Yes
dec Dec Average $ Yes Yes
mavg Concat of all monthly averages mavg.$Year1-$ No Yes
mons Lnd, concat of monthly averages $ No Yes
jfm Ice Winter Avg $ Not Now Yes
fm Ice Feb & Mar Avg $ Not Now Yes
amj Ice Spring Avg $ Not Now Yes
jas Ice Summer Avg $ Not Now Yes
ond Ice Fall Avg $ Not Now Yes
on Ice Oct & Nov Avg $ Not Now Yes
preproc Ice pre proc file ice_vol$CASE_$Year1-$ Not Now Yes

Can be created as a dependency? option:
The averages that are listed in the above table as being able to create averages as dependencies have the ability to use previously calculated averages to calculate a new average. To use this option, append dep_ in front of the average name (ie, dep_jja). Without dep_, a jja average would loop over all June, July, and August values within the year ranges and create an average. With dep_, the PyAverager will create and output a June average, July average, and August average. Then it will open these average files and average these values to create the jja average file. In most cases, it is faster to run with dep_, but it should be pointed out that the answers between using and not using the dep_ option will differ due to order of operation.

Specifier Arguements
See examples/ for how to set all available options to send to the create_specifier function.

Variables that must be passed to the specification.create_specifier class:
o in_directory: directory where the input data is located
o out_directory: directory where the output will be produced
o prefix: the case name, plus component name (ie. b40.20th.track1.1deg.006.cam2.h0)
o suffix: the end of the input file names (usually nc)
o date_pattern: 'yyyymm-yyyymm'
o avg_list: a list of averages to compute DEFAULT = Empty List
Format: ['ya:1850','mavg:1850:1890'] ya is the only average to take one year. All other averages expect a start year and end year separated by a colon. The available average choices are listed in the above table.

Variables that are mandatory for the Ice and Ocean Diags:
o The following variables are used by the Ocean Model Diags for the hor.meanConcat file creatation:
o mean_diff_rms_obs_dir: directory that contains the obervartion files needed to calculate the hor.mean.Concat file (Ocean Model).
o region_obs_file_suffix: the suffix of region obs files found in the mean_diff_rms_obs_dir directory
o region_nc_var: variable name the contains the region mask information (Ocean Model)
o regions: regions to create files for (ie[1:'Sou',2:'Pac']) region int that corresponds to the region_mask, region name.
o region_wgt_var: variable name that contains the region weight info
o obs_file: observation file (contains the region_nc_var and region_wgt_var)
o obs_dir: directory where the obs_file is located in

o The following variables are used by the Ice Model Diags for the Pre_Proc file:
o ice_obs_file: a netCDF file that contains area/weight information
o ncl_location: the location of the ncl script used to create the reg_file (usually provided with this source code in pyaverager/ directory
o reg_file: the name of the netcdf file that contains the region mask information. If it does not exist, it will be created for you.

Optional variables that can be passed to the specification.create_specifier class:
o ncformat: either 'netcdf4c' (netcdf4 compressed (lev=1)), 'netcdf4' (netcdf classic), and 'netcdf' (netcdf3 classic) DEFAULT = 'netcdf4c'
o file_pattern: needed for non-cesm data
For file name:
Use: ['$var','_','$prefix','_','$m_id','_','$date_pattern','.','$suffix']
o hist_type: either slice or series DEFAULT = 'slice'
o m_id: experiment/or other unique id (can be used to id ensemble members)
o weighted: Boolean to weight averages (when available, see about table) DEFAULT = False
o split: Are the files split between lat coordinates (used in cice series files)
o split_files: strings differentiating the different pieces DEFAULT = null
o split_orig_size: list of lat/lon names and their original full size DEFAULT = null
o varlist: ['a','list','of','vars','to','avg'] DEFAULT = Full list
o clobber: If a user specified average exists on disk, delete if set to true. DEFAULT=False
o serial: run in serial or parallel mode DEFAULT=FALSE (parallel mode)
o main_comm: the simple_comm object. If one isnt passed in, one will be initialized for you.

To generate the obs file needed for the Ocean hor.meandiff calculation:
(or yellowstone users can copy/use files in /glade/p/work/mickelso/PyAvg-OMWG-obs/obs)
- From omwg obs_data: and SALT
- POP history file
- Copy to
- ncks -A -v SALT
- ncrename -O -d X,nlon -d Y,nlat -d depth,z_t
- ncatted -a _FillValue,TEMP,c,f,-99.
- ncatted -a _FillValue,SALT,c,f,-99.
- ncatted -a missing_value,TLAT,d,,
- ncatted -a missing_value,TLONG,d,,
- ncatted -a _FillValue,TLAT,d,,
- ncatted -a _FillValue,TLONG,d,,
- ncatted -a _FillValue,TAREA,c,f,-99.
- ncatted -a _FillValue,TAREA,m,f,1.0e36
- ncatted -a _FillValue,TAREA,m,f,-99
- ncatted -a _FillValue,,m,f,-99
- ncatted -a _FillValue,TAREA,m,f,-99
- ncatted -a _FillValue,TAREA,o,f,-99
- ncatted -a _FillValue,REGION_MASK,o,i,99

To generate the regional obs files needed for the Ocean hor.meandiff calculation:
(or yellowstone users can copy/use files in /glade/p/work/mickelso/PyAvg-OMWG-obs/obs)
- From omwg obs_data: and SALT
- POP history file
- Copy to
- ncks -A -v SALT
- ncrename -O -d X,nlon -d Y,nlat -d depth,z_t
- ncatted -a _FillValue,TEMP,c,f,-99.
- ncatted -a _FillValue,SALT,c,f,-99.
- ncatted -a missing_value,TLAT,d,,
- ncatted -a missing_value,TLONG,d,,
- ncatted -a _FillValue,TLAT,d,,
- ncatted -a _FillValue,TLONG,d,,
- For regions in the table below:
o ncwa -m REGION_MASK -T eq -M <reg_number> -w TAREA -a nlon,nlat -v TEMP,SALT <reg>
o <reg_number> Table
Sou Pac Ind Atl Lab Gin Arc Hud
1 2 3 6 8 9 10 11

- For Glo:
o ncwa -m REGION_MASK -T gt -M 0 -w TAREA -a nlon,nlat -v TEMP,SALT

PyAverager Error Codes
errors 1-19: average list errors
1: Listed average is not in the know average list
2: Average cannot be created with dependencies
3: Average must list only one year
4: Average must have a start year and an end year
5: Date ranges are inconsistent and cannot run this average with dependencies

errors 20-39: input file problems
20: Cannot find the file (triggered in three different checks points)
21: Missing files to calculate DJF. You need either the previous December or the January and February from last year+1
22: Time series files are split, but the dates between them are not contiguous (triggered in two different checks points)
23: A date was found within two different time series files. Not sure which to use.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for pyAverager, version 0.9.10
Filename, size File type Python version Upload date Hashes
Filename, size PyAverager-0.9.10.tar.gz (37.5 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page