Restarting the model

Hi,

I'm trying to restart UFS from restart files.
Is there any documentation on this? As I can't seem to find any.

In input.nml I have changed

warm_start=t

Then linking the files from RESTART to INPUT

When re-run it overwrites the output files (for example atmf000.nc) and doesn't seem to be continuing from the last restart file.

Any help would be great!
Thanks

Hi Tim,

that's a little more involved unfortunately.

1. Move/link the files in the RESTART directory to the INPUT directory (note that any timestamp information must be removed from the name, they should have exactly the same names as cold start data has).

2. Toggle the following options in input.nml, for restarts they should then be
 

WARM_START=.true.
NGGPS_IC=.false.
EXTERNAL_IC=.false.
MAKE_NH=.false.
MOUNTAIN=.true.
NA_INIT=0
nstf_name=?,0,?,?,? ! this means set the second value to 0, leave all others as is

3. Edit model_configure and adjust the forecast length as desired (the value is always relative to the cold start initialization time, not the forecast length from the restart time)

This *should* work, please give it a try.

Hi Dom,

Thanks for the quick response. That did work (kind of)... I got a load of slope warnings and then a lookup table overflow

WARNING from PE 787: diag_manager_mod::send_data_3d: A value for dynamics in field slp (Min: 6.18169E+002, Max: 7.41228E+002) is outside the range [ 8.00000E+002, 1.20000E+003], and not equal to the missing value.

FATAL from PE 808: compute_qs: saturation vapor pressure table overflow, nbad= 1

I'll dig into why I'm getting these warnings and errors. However the restart works!
Thanks!

Per the wiki doc Restarting the Coupled Model, the restart files are:

 

  • coupler.res
  • fv_core.res.nc
  • fv_core.res.tile[1-6].nc f
  • v_srf_wnd.res.tile[1-6].nc
  • fv_tracer.res.tile[1-6].nc
  • phy_data.tile[1-6].nc
  • sfc_data.tile[1-6].nc

 

What's the status of a restart if, for some reason, these files don't exist in the RESTART directory? While it seems that this situation would be an edge case, what's the suggested next step(s)?