Commits on Mar 13, 2023

1. ice_history_write: fix initial condition metadata under 'hist_avg' …

   When writing averaged history outputs (hist_avg=.true.), this setting
   also affects the initial condition. Even if the actual data variables
   written to the initial condition are not averaged (they are taken
   more or less directly from the restart or the hard-coded defaults,
   modulo aggregation over categories), their attributes ('cell_method' and
   'time_rep') imply they are averaged, and the 'bound' attribute of the
   'time' variable refers to the 'time_bounds' variable.
   
   Make the metadata of the initial condition more correct by:
   - not writing the 'time_bounds' variable (and the corresponding 'd2' dimension)
   - not writing the 'bounds' attribute of the 'time' variable
   - not writing the 'cell_method' attributes of each variable
   - writing the 'time_rep' attribute of each variable as 'instantaneous'
   instead of 'averaged'.
   
   Do this by checking 'write_ic' at all places where we check for the
   value of 'hist_avg' to write the above variables and attributes in each
   of the 3 IO backends (binary, netcdf, pio2).

   

   phil-blain committed Mar 13, 2023
   Configuration menu

   • View commit details
   • Copy full SHA for 9f4504d
   • Browse repository at this point

   Copy the full SHA

   9f4504d View commit details

   Browse the repository at this point in the history

2. drivers/{nemo_concepts,standalone}: write initial condition at initia… …

   …l time
   
   In CICE_InitMod::cice_init, we call ice_calendar::advance_timestep
   before writing the initial condition, such that the 'time' variable in
   the initial condition is not zero; it has a value of 1*dt (the model
   time step). The initial condition filename also reflects this, since
   'msec' (model seconds) also has a value of 1*dt and is used in
   ice_history_shared::construct_filename. This leads to the initial
   condition filename not corresponding to the model initialization
   date/time but rather 1*dt later.
   
   Since we call 'accum_hist' after initializing the forcing, any forcing
   field written to the initial condition has values corresponding to
   msec=dt, whereas the ice state corresponds to msec=0, leading to an
   inconsistency.
   
   Fix that by calling 'accum_hist' to write the initial condition _before_
   calling 'advance_timestep'. Since we now call 'accum_hist' before
   initializing the forcing, any forcing field written to the initial
   condition have its default, hard-coded value, instead of its value at
   time=dt. An improvement would be to read the forcing at time=dt, write
   the initial condition, advance the time step, and read the forcing
   again, but let's not complicate things too much for now.

   

   phil-blain committed Mar 13, 2023
   Configuration menu

   • View commit details
   • Copy full SHA for d72cdf9
   • Browse repository at this point

   Copy the full SHA

   d72cdf9 View commit details

   Browse the repository at this point in the history