Debug checklist

Problem:  Model dies almost immediately.  May not have an error message.

  1. Check to be sure you are assigning the correct mappings in env_conf. This is a common mistake.   The atm to ocean STATE vars are the ONLY ones that use bilinear mapping. if you accidentally give the bilinear mapping (or area average mapping) the wrong assignment, the model will crash right away. E.g.
    <!--"atm to ocn flux mapping file for fluxes (char) " -->
    <entry id="MAP_A2OF_FILE"   value="map_T31_to_gx3pt_aave_da.nc"  />
    <!--"atm to ocn state mapping file for states (char) " -->
    <entry id="MAP_A2OS_FILE"   value="map_T31_to_gx3pt_bilin_da.nc"  />
    <!--"ocn to atm mapping file for fluxes (char) " -->
    <entry id="MAP_O2AF_FILE"   value="map_gx3pt_to_T31_aave_da.nc"  />
    <!--"ocn to atm mapping file for states (char) " -->
    <entry id="MAP_O2AS_FILE"   value="map_gx3pt_to_T31_aave_da.nc"  />
  2. if you have created new mapping files (particularly important for Deep Time simulations) run scrip_test (section 5.1.3 in the Paleo Tech Note).  This tool creates spatial maps in netCDF format from the files created by scrip.   These files are easy to look at using ncview and can often be used to identify problems.
  3. If you get any time steps in at all, write out an instantaneous cpl history file at every time step to look at the various variables.  Sometimes the atm/ocn grid ratios at the poles are too large (or too small).
  4. if the model is crashing at timestep 0, and you have correctly set the mapping files (#1),  try setting info_debug in env_run.xml to the highest value.  This will write extra coupler STDOUT information to the log file which might help identify a problem.