search

NOAA-EMC / WW3

WAVEWATCH III
Other
265 stars 542 forks source link

Triad calculation failing for cold starts (TR1) #1041

Open ukmo-ccbunney opened 1 year ago

ukmo-ccbunney commented 1 year ago

Describe the bug When using the TR1 flag to enable non-linear triad interactions, the w3str1md.F90 is performing a divide by zero when starting from a coldstart.

This is due to EMEAN being zero at coldstart, therefore causing problems at this line: https://github.com/NOAA-EMC/WW3/blob/02cb72f7b15d7384c507813a3a6569265f588491/model/src/w3str1md.F90#L431

Perhaps easily solved by adding a test on EMEAN after summing the spectrum between sections 3 and 4? Something like:

! 3.  Add tail beyond discrete spectrum
!
    EBAND  = EB(NK) / DDEN(NK)
    EMEAN  = EMEAN  + EBAND * FTE
    SIGM01 = SIGM01 + EBAND * FTF
!
! 3.1 [NEW] Check for zero energy
!
    IF(EMEAN .EQ. 0.0) THEN
      S(:) = 0
      D(:) = 0
      RETURN
    END IF
!
! 4.  Final processing
!
    SIGM01 = SIGM01 / EMEAN

To Reproduce Run a TR1 test case with a cold start. You might need debug flags to catch the divide by zero.

Expected behavior Runtime error something like this (generated by GNU compiler): Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

ukmo-ccbunney commented 1 year ago

Tagging @aronroland for his opinion :)