Hello,
I’m having trouble with the floats in my ROMS run — the output file flt.nc is entirely filled with NaNs (including at the first time step). I have debug and rerun the model several times and the issue persists.
What I have checked so far:
Lfloats == T in my roms.in file.
NFLOATS matches the number of positions in floats.in.
Not using biological behavior — FBIONAM = behavior_oyster.in is commented out.
FLOATS and FLOAT_VWALK CPP options are both activated.
POS lines in floats.in are properly formatted. Example:
POS = G, C, T, N, Ft0, Fx0, Fy0, Fz0, Fdt, Fdx, Fdy, Fdz
1 1 1 1 0.0d0 55.250d0 -4.560d0 -10.0d0 0.0d0 0.0d0 0.0d0 0.0d0
Float release time (Ft0) is at the start of the simulation, so they shouldn’t have time to leave the domain before being recorded.
The model runs much slower with floats activated, so I believe they are being processed during the run.
Additional observation:
In one earlier test, when I ran the model in serial mode, I did get some non-NaN values in the flt.nc file. In my current runs (which are in parallel/MPI mode), all values are NaN. Could this point to a partial problem related to parallel execution or domain decomposition for floats?
My question:
Since the flt.nc file has NaNs even for the initial positions, I suspect the floats may not be initializing properly. What should I check next? Are there specific log messages in ocean.out that indicate failed float initialization?
Any suggestions on what to inspect or possible reasons for this would be appreciated.
Thanks,
Hsin-Yi Chen