Testing Timeseries Input with E3SM Diagnostics

timeseries_input_e3sm_diags.md

Testing Timeseries Input with E3SM Diagnostics on ACME1/AIMS4

This is a tentative guide. It will eventually be put on the official documentation site here.

You guys can probably figure out how to run this on Edison. Just use module load, create you env, and adjust the paths in the py accordingly.

Installation

1. Follow these instructions under Environment for development here, up to and including step 4 only. Don’t do step 5.

2. Remove cdp, we need the latest from master.

   conda remove cdp --force
   

3. Install the latest cdp from master.

   git clone https://github.com/CDAT/cdp.git
   cd cdp
   python setup.py install 
   

4. Go to the climo branch of e3sm_diags and install the latest code there.

   git clone https://github.com/E3SM-Project/acme_diags.git
   cd acme_diags
   git checkout climo
   python setup.py install 
   

Running the diagnostics

5. Create and run the following py parameters file.

   Remember to change results_dir and num_workers accordingly.

   For the current timeseries files in test_data_path, the timeseries specific variables you should only change are test_start_time_slice and test_end_time_slice.

   reference_data_path = '/p/cscratch/acme/data/obs_for_acme_diags/'
   # The below test_data_path is where the climo model data is stored.
   # test_data_path = '/p/cscratch/acme/data/test_model_data_for_acme_diags/'
   # New path of where the timeseries files are located.
   test_data_path = '/p/user_pub/work/E3SM/1_0/piControl/1deg_atm_60-30km_ocean/atmos/129x256/time-series/mon/ens1/v1/'
   short_test_name = 'some_test'
   
   # New parameters for timeseries input.
   test_timeseries_input = True
   test_start_yr = '0001'
   test_end_yr = '0500'
   # You can change these to get difference slices.
   test_start_time_slice = '0001'
   test_end_time_slice = '0012'
   
   # sets = ['zonal_mean_xy', 'zonal_mean_2d', 'lat_lon', 'polar', 'cosp_histogram']
   sets = ['lat_lon']
   
   seasons = ['ANN']
   backend = 'mpl'
   
   results_dir = '/var/www/acme/acme-diags/zhang40/dataset/lat_lon_timeseries_input_08_21'
   
   multiprocessing = True
   num_workers = 48

6. Your results should look something like this.

Adjusting the time slicing stuff

In the climo branch, the Dataset class is located here:

acme_diags/acme_diags/driver/utils/dataset.py

The function _get_var_from_timeseries_file() on line 431 is where the time slicing occurs. Feel free to change or don't even use the get_start_end_time_slice() function. This is what gets the start and end slices.

Remember you can always use one of the diagnostics in this or other cfg files to make testing faster.

Email me if you have any issues, need help, or have any feedback.