neg. dp (m) problem with running HYCOM in a new domain.

[Jeongin Kim]

Hello, 

I'm trying to make a restart file from a climatology (cold start) using a similar method to the GOMb0.08 example.

Since I'm setting the new domain (North Atlantic Ocean), I don't have the existing restart file. 

So, I'm trying to shoot a one-day run without forcing to make a restart file, and then use this to an archv2restart.f so that I can convert an archive file (subregioned GLBb 1994_01 file) to a restart file.

However, I'm getting multiple neg. dp (m) flags, and those locations are quite far from the boundaries, so I want to believe it is not a problem of masking or relaxation.

Could you let me know the clue of these error flags, and give a hint to resolve this?

Also, I'm a bit unsure that I can change LIMITS to LIMITI to avoid HYCOM to find restart_in.[ab] at here.

setenv YOF ""
#setenv YOF `echo ${Y01} | awk '{printf("%04.0f", $1+1900)}'`
C
C --- local input files.
C
if (-e    ${D}/../${E}y${Y01}${A}.limits) then
  /bin/cp ${D}/../${E}y${Y01}${A}.limits limits
else
#  use "LIMITI"  when starting a run after day zero.
#  use "LIMITS9" (say) for a 9-day run
#  echo "LIMITS" | awk -f ${D}/../${E}.awk y01=${Y01} ab=${A} >! limits
  echo "LIMITI" | awk -f ${D}/../${E}.awk y01=${Y01} ab=${A} >! limits
endif

I use the same grid resolution of GLBb0.08 and GOMb0.08, and run without any forcing, tidal forcing, etc.

For your information, I've attached the figure of the original domain and my masked domain, relaxation mask file, csh and log file, and blkdat.input file.

I'm idm, jdm are 1475, 953, and running with the configuration of src_2.3.01_relo_mpi,  with 24MPI.

Let me know if you need any other information to figure this out.

Many thanks,

Jeongin

[Alan Wallcraft]

The time steps from GOMb0.08 won't work for the Atlantic because the grid spacing at 50N (say) is smaller than at 30N.

For the global model with assimilation, we use:

 180.0    'baclin' = baroclinic time step (seconds), int. divisor of 86400
 6.428571         'batrop' = barotropic time step (seconds), int. div. of baclin/2

These are smaller than you need, and it is particularly important to get batrop as large as possible.

Run the batrop script for your domain and set batrop at about 75% of the cfl minimum.  For example, GOMb008 has min cfl of 23 seconds and batrop is 18.75 (81%).  If you set it too large HYCOM will blow up immediately.  This is probably what is happening to you.

The baclin time step is less sensitive, but you want it as large a practical to reduce run time.  I'm sure 240 will work in your case and 300 or 450 might also work.  These are all divisors of 1 hour, which is necessary if you use 1 hour forcing or output.

Alan.

[Jeongin Kim]

I appreciate you so much, Alan!

The problem has been resolved by changing batrop and baclin.

Best, 

Jeongin

2025년 10월 20일 월요일 오전 7시 37분 32초 UTC-4에 Alan Wallcraft님이 작성: