Issue restarting solutions in fraction of a day

[rsoutelino]

Hi All,

I've recently bumped into an issue when restarting forecast simulations that run 2x daily, e.g. when restart files have the first time record at a fraction of a day (12 h, for example). I've tracked it down and found a possible typo on dateclock.F 430:436, as below.

Code: Select all

!
!  Fractional date number (units=second).
!
      DateNumber(2)=REAL(MyDay,dp)*86400.0_dp+                          &
     &              REAL(MyHour,dp)*360.0_dp+                           &
     &              REAL(MyMinutes,dp)*60.0_dp+                         &
     &              MySeconds

The multiplier for the number of seconds in an hour is missing a "0". This causes the below bit of the code from mod_netcdf.F 5275:5277 to add an inconsistent offset to the time array.

Code: Select all

DO i=1,Asize(1)
    A(i)=(my_Rdate(2)+A(i))-Rdate(2)
END DO

Which crashes the initialization in check_multifile.F. Below are the relevant snippets of ROMS stdout and ncdump.

Code: Select all

 NLM: GET_STATE - Reading state initial conditions,                       2019-01-21 13:12:00.00
                   (Grid 01, t = -0.4500, File: tinyapp2d20190122_00z_rst.nc, Rec=0001, Index=1)
                - free-surface
                   (Min = -3.92467349E-02 Max =  5.34361359E-02)
                - vertically integrated u-momentum component
                   (Min = -5.60959179E-02 Max =  8.09049525E-03)
                - vertically integrated v-momentum component
                   (Min = -6.47255244E-02 Max =  3.94585012E-03)

 CHECK_MULTIFILE - Error while processing Forcing multi-files:
                   data does not include initialization time = 2019-01-21 13:12:00.00

Code: Select all

ncdump -v ocean_time -t tinyapp2d20190122_00z_rst.nc

		:CPP_options = "ROMS2D_NO-TIDE_NO-OBC, ANA_FSOBC, ANA_M2OBC, ASSUMED_SHAPE, ATM_PRESS, !BOUNDARY_A !COLLECT_ALL..., CURVGRID, DOUBLE_PRECISION, GLS_MIXING, KANTHA_CLAYSON, MASKING, MPI, NONLINEAR, NO_LBC_ATT, N2S2_HORAVG, PERFECT_RESTART, PROFILE, K_GSCHEME, RADIATION_2D, REDUCE_ALLGATHER, !RST_SINGLE, SPHERICAL, STATIONS, UV_ADV, UV_COR, UV_LDRAG, UV_VIS2" ;
data:

 ocean_time = "2019-01-22", "2019-01-22 12", "2019-01-23", "2019-01-23 12", 
    "2019-01-24", "2019-01-24 12" ;

Because hour=0 cancels it out, this doesn't affect simulations that start flush in "0h", which are largely the most common case for hydrodynamical applications I guess.

I've applied this fix locally here and tested it, it works fine.

Cheers,
Rafael

[arango]

Indeed, good catch Thank you for reporting this bug. I will update the code soon.