PPUG_running_averager

Running post processing scripts to create climatology and average files

Note: The [comp] option corresponds to 'atm', 'ice', 'lnd' and 'ocn'.

Synopsis

$PP_CASE_PATH/[comp]_averages

Compute the component averages and climatologies where comp is one of atm, ice, lnd, or ocn. Works with either history time slice or variable time series data. If using variable time series data, then the timeseries process must complete first. Uses the env_postprocess.xml and env_diags_[comp].xml settings to define behavior. The $PP_CASE_PATH/logs/[comp]_averages.log contains the stderr and stdout from the job submission. To submit the script to compute the averages on yellowstone use commands:

bsub < atm_averages

bsub < ice_averages

bsub < lnd_averages

bsub < ocn_averages

Details

The [comp]_averages bash wrapper scripts call the corresponding [comp]_avg_generator.py python script in parallel which reads the env_diags_[comp].xml file, defines the necessary input specifier �class option to the pyAverager tool, and calls the averager to generate climatologies and averages using either history time slice files or single variable time series files as input.

The climatologies generation adhere to the CESM Model File Naming Conventions.

Users may customize the behavior of the climatologies and averages generation by modifying the XML input values in the env_diags_[comp].xml. See Configuring the post processing case files for details.

After customizing the env_diags_[comp].xml file, you can submit the [comp]_averages bash wrapper script to the queue. The $PP_CASE_PATH/logs/[comp]_averages.log file contains stderr and stdout output from the job. Each [comp]_averages job is independent of any others and so can be submitted to the queue concurrently with the exception of the timeseries generation.

Dependencies

The XML settings for each component allow for input history slice files or single variable time series files for either the model or control averages. If single variable time series files are specified as the input, then the timeseries generation needs to complete successfully first.