defined an allocatable ncid array instead of fixed length array to avoid ww3_ounf to crash when opening a new netcdf with a out of range ncid value

Need to check previous svn issues that explain differences in mod_def files for some test cases (ww3_tp2.7, ww3_ts4 and mww3_test_07).

A module based on the generalized kinetic equation (Gramstad and Stiassnie 2013), which takes into accout both resonant and quasi-resonant four wave interactions, will be implemented as NL5 in WW3. The modified kinetic equation by Janssen (2003) will also be included in the NL5 framework.

NCEP operations require that ounf output to a fixed-name file. Currently a quick-fix is used in ww3_ounf.ftn:

! For NCEP application, requires fixed netcdf file name
FNAMENC='ww3.gridded.nc'

This issue is a request for a feature to allow users choosing that in ounf.inp.

Triads do not alter the results in the right direction and magnitude (2019). Looking into things I did some rewrite and found that the explicit part of the source term was off by a factor of 2. Moreover, reallocation would cause issues and run-time penalty. I rewrote following the WWM implementation

The Lake Erie SMC propgagtion only regtest (ww3_tp2.10) is going unstable (at least on our Cray system). Output fields after the initial timestep are full of "NaN", "Inf" or unrealistically large values. This is causing problems for comparison of regtest suites.

A requirement for operations, is going to be the ability to restart WW3 from a checkpoint. This means selecting the correct restart file and then continuing to write output files from that point.

A proposed solution is to add an option to not only label the restart files with dates such as
restart.YYYYMMDDHHMMSS.ww3 (or some form of YYYYMMDDHH) and to explore the possibility of having output such as out_pnt.ww3 also be for each time (out_pnt.YYYYMMDDHH.ww3) so that you would not have to over-write anything in the binary file or cycle through to find out in initialization where you should be in the binary file.

We could also consider adding an option to put the output in folders, since this potentially could increase the number of files which is of a concern in coupled models where all models have lots of output files and it could be an issue for the file system.

@aliabdolali @ajhenrique

NCEP R&D machine has moved from MOAB to slurm batch queueing. An update is being developed by @ajhenrique.

The subroutine TAU_WAVE_ATMOS dynamically allocates six working arrays every call but does not release them. The following should probably be added a the end of the subroutine:

  DEALLOCATE(SIG10Hz,CINV10Hz,DSII10Hz)
  DEALLOCATE(SDENSX10Hz,SDENSY10Hz,UCINV10Hz)

Comparing compile and run times when running a regression test is fundamental to see impacts of changes to model results, runtime and to compilation metrics. This issue deals with adding counters for both in the script run_test.

I had an error runinng ww3_shel, it complained about a problem reading ww3_shel.inp

Diving in the code, seems this can be fixed by commenting an extra READ in line 1303

From
READ (NDSI,,END=2001,ERR=2002) IPRT, PRTFRM
!/DEBUGINIT write(740+IAPROC,), 'Before reading IPRT'
!/DEBUGINIT write(740+IAPROC,), 'Before read 2002, case 10'
READ (NDSI,) IPRT, PRTFRM
!/DEBUGINIT write(740+IAPROC,*), ' After read 2002, case 10'

to
READ (NDSI,,END=2001,ERR=2002) IPRT, PRTFRM
!/DEBUGINIT write(740+IAPROC,), 'Before reading IPRT'
!/DEBUGINIT write(740+IAPROC,), 'Before read 2002, case 10'
!!! READ (NDSI,) IPRT, PRTFRM
!/DEBUGINIT write(740+IAPROC,*), ' After read 2002, case 10'

We can see that in ww3_shel.inp there are options for reading the flags for mud and ice, but in multi there are no options for reading input files with ice and mud for each grid.

Add an option to keep all build information (mod/obj_*, tmp, exe) in a different directory than model. One advantage to adding this feature would be that it would allow for multiple builds to occur simultaneously and could help speed up regression tests significantly.

In line 1876- 1878 (wmupds subroutine)

          IF ( FLAGLL ) THEN
              IF ( ICLOSE.NE.ICLOSE_NONE ) JJ = 1 + MOD(J-1,NXI)
              IF ( JJ.LT.1 .OR. JJ.GT. NXI ) CYCLE

This leads to a crash if no closure is used as JJ is not defined before being called.

I think it should be like this instead, copying what is done in line 1835

          IF ( FLAGLL .AND. ICLOSE.NE.ICLOSE_NONE ) THEN
              JJ =  1 + MOD(J-1,NXI)
              IF ( JJ.LT.1 .OR. JJ.GT. NXI ) CYCLE

It might have been reported before, maybe in the mailing list, but the version here still has the error.

After the successful transition of the master and develop branches from NOAA VLab to GitHub, a few branches need to be cleaned up or transferred across to forks or institutional repos.

A multitude of comp/link scripts for specific applications/machines, some deprecated, populates the model/bin directory. @mickaelaccensi has whipped up a set of template comp/link files with and env-setting script that is lined up to replace individual comp/link files, inc association with the w3_setup script. The work will require collaboration between users to make sure the new approach works across platforms. An idea is to include modulefiles use in compilation across machines to help users and developers in their initial model setup.

Regtests: matrix scripts run a series of regtests, and are the basis to develop our code management gold standard. @aliabdolali has started documenting regtesting using matrix scripts, @JessicaMeixner-NOAA will add wiki page.

I am working on a feature branch fb_rtd2 which is an extension to the /RTD switch. It provides a method to output boundary conditions to a rotated lat/lon grid. This fulfills our need to create one-way nesting to a 9-nmi grid around Greenland. I will await Chris Bunney and colleagues at UKMO to give it a critical look and expect to send a pull request before end of august.
https://github.com/CarstenHansen/WW3/tree/fb_rtd2

Restructure ww3_ounf code to use alternative, user defined varaible name and metadata packages, e.g. CF-1.7 compliant data and variable names used by Copernicus Marine Environment Monitoring System and Met Office ServiceHub

WCOR switch is well implemented but missing in make_makefile.sh

Change manual, guide and w3initmd to reflect new version number (7.00)

@ajhenrique has created a temporary fix to allow usage of ounf in NCEP's GLWU system. A more consistent solution for this problem will be provided under issue #23.

@flampouris

In the manual for The Restart File Processor (ww3_uprsrt) section, there are two places where it says:
UPD0C:: ELIMINATED from version x.xx
UPD1 :: ELIMINATED from version x.xx

what version is x.xx?

Some link scripts need to be bugfixed for libs= statement that removes netcdf libs= settings. This occurs in link.theia but may occur in other scripts.

DB1 module was originally coded using the monocramatic wave dissipation dimensional variables. It was updated to use the spectral form, but some coefficients were no updated accordingly which resulted in a dimensionally-incorrect version. @aronroland @ajhenrique and @thesser1 are working to fix this inconsistency.

Runs of the regtests matrix have indicated that test ww3_tp1.6 is not bit for bit comparable in independent runs of a given branch. Differences are seen in out_grd.ww3 and out_ont.ww3 files. This should be investigated, as pointed out by @ajhenrique, @aliabdolali and @JessicaMeixner-NOAA .

Hello WAVEWATCH III Developers,

Wednesday is out first bi-monthly WAVEWATCH III Developers meeting since the transition to GitHub. The focus of this meeting will be on the transition to GitHub itself, a description of our new authoritative and trusted repositories, new milestones for v7, and recent developments.

Highlights also include the presentation "Met Office WAVEWATCH III profiling and scale-ability activities" by Dr Andrew Saulter (UKMO), and an open forum where developers can discuss items of interest.

When: Wednesday, June 5th 2019, 10:30am EDT

Presentation material will be posted via google hangouts:
https://meet.google.com/egi-wopu-sqn

For audio, we will use the teleconference line: 1-877-953-0315
Passcode: 1262920#

Let's go modeling!

Henrique,

For the WAVEWATCH III Development Group

@12cm103
@ajhenrique
@aliabdolali
@bauj0001
@brandaodeco
@breichl
@CarstenHansen
@Clarenburg
@emezac63
@erdc
@FCOO
@flampouris
@hellais
@hkershaw-brown
@huangynj
@j-gonzalezlopez
@jessicaborgesp
@JessicaMeixner-NOAA
@JoaoGRRodrigues
@lileipku00
@mickaelaccensi
@randobucci
@rickyspaceguy
@sahammhd
@seglion
@tiagofaro
@tomdurrant
@ukmo-ccbunney
@ukmo-waves
@umr-lops
@AndresSepulveda
@wavespotter
@XianwuXue-NOAA
@xuexianwu
@yuemaaa
@yuwana1989
@zhongtouwang2019

Add in wave spectrum update method, following Lionello et al (1992), to ww3_uprstr.ftn

Re-structured SMC propagation code for improved scaleability in hybrid MPI-OpenMP mode

The following hotfixes will be merged to NOAA-EMC/WW3 master and develop:

• Bugfix in ww3_ounf.ftn: NBIPART check against NOSWLL. See #2.
• Updating link files for grib compile, change to ww3_uprstr for clean compile with OMPH OMPG flags,
• Option for writing to fixed netcdf gridded output file with NCO switch (more appropriate fix in v7, see #8).
• DB1 depth-breaking_bugfix: added proper scaling coefficients for radian freq. [radHz]. See #7.

The update affects the following downstream forks:
@ajhenrique ajhenrique / WW3
@aliabdolali aliabdolali / WW3
@erdc erdc / WW3
@flampouris flampouris / WW3
@hellais hellais / WW3
@huangynj huangynj / WW3
@j-gonzalezlopez j-gonzalezlopez / WW3
@JessicaMeixner-NOAA JessicaMeixner-NOAA / WW3
@lileipku00 lileipku00 / WW3
@rickyspaceguy rickyspaceguy / WW3
@sahammhd sahammhd / WW3
@tomdurrant tomdurrant / WW3
@ukmo-waves ukmo-waves / WW3
@umr-lops umr-lops / WW3
@wavespotter wavespotter / WW3
@xuexianwu xuexianwu / WW3
@yuemaaa yuemaaa / WW3
@yuwana1989 yuwana1989 / WW3
@zhongtouwang2019 zhongtouwang2019 / WW3

I'm not sure if it is a bad use from my part or an issue in the code

line 1873 when interpolating for a 2-component field the following formula is used for A3

A3(IX,IY) = SQRT ( &
RD11(IX,IY) * ( XC(IX21(IX,IY),IY21(IX,IY))**2 &
+ YC(IX21(IX,IY),IY21(IX,IY))**2 ) &
+ RD21(IX,IY) * ( XC(IX22(IX,IY),IY21(IX,IY))**2 &
+ YC(IX22(IX,IY),IY21(IX,IY))**2 ) &
+ RD12(IX,IY) * ( XC(IX21(IX,IY),IY22(IX,IY))**2 &
+ YC(IX21(IX,IY),IY22(IX,IY))**2 ) &
+ RD22(IX,IY) * ( XC(IX22(IX,IY),IY22(IX,IY))**2 &
+ YC(IX22(IX,IY),IY22(IX,IY))**2 ) )

I have an issue in my case where the number inside the square root is negative, from a negative value of RD22.
Probably an issue from having a 1/12 degree regular grid, I suspect that the different rounding in my forcing grid and forcing file is giving a very small negative value of RD22 at some point when computing the weights.

Maybe a way around it could be a 'transfer as it is' option in the nml input file, like in the inp version?

@MicahDavid reported

I was receiving the following error using OpenMPI:

At line 394 of file w3iogomd.F90
Fortran runtime error: Unit number is negative and unit was not already opened with OPEN(NEWUNIT=...)

Making the fix to w3iogomd.ftn on line 404 alleviated this error:
FROM:
IF (IFJ .LT. NOGE(IFI)) WRITE(NDSEN,1007) IFI
TO:
IF (NDSEN .GE. 0 .AND. IFJ .LT. NOGE(IFI)) WRITE(NDSEN,1007) IFI

The profile output gmon.out and time_count.txt should be removed from the regtests/bin/matrix.comp script so that they are not included in regression test comparisons.

when creating new points file from input file, add relative to points file instead of filename

@ajhenrique will work on this, tagging up with EMC admins to find out more.

Hi,

I am receiving the following error in my simulation, when using ww3_shel

....
Type 6 : Partitioned wave field data
-----------------------------------------
From : 1968/06/06 00:00:00 UTC
To : 2001/01/02 00:00:00 UTC
Interval : 01:00:00

        output dates out of run dates : Track point output deactivated
        output dates out of run dates : Nesting data deactivated
        output dates out of run dates : Partitioned wave field data deactivated
   Wave model ...

corrupted size vs. prev_size

Program received signal SIGABRT: Process abort signal.

The same configuration ran ok in my laptop. Any idea what it could be?

Attached are the output, switch and ww3_shel.inp files

error.log
ww3_shel.inp.txt

switch.txt

error in flag status to read time with new CF convention when month and day are written on only one character : YYYY-M-D

Currently ounf creates date-tagged output files. Some applications require fixed output file names. @ukmo-ccbunney and @mickaelaccensi will collaborate towards a fixed-file option.

@ajhenrique working with @JessicaMeixner-NOAA to iron out inconsistencies.

The W3PSMCMD module uses OMP directives on various loop constructs, but these are not protected by one of the OMPx compile switches.

The !/OMPG switch should be used as these loops parallelize well in both hybrid and pure OMP mode.

OUTSIDE_BOUNDARY is defined over the full nodes NODES indexes whereas TMPSTA and ZBIN are defined over the clean up list of nodes NX

A bug in ww3_ounf is causing the netcdf gridded post-processing wrapper to fail. The line

IF(NBIPART .GT. NOSWLL) THEN

needs to be changed to

IF(NBIPART .GT. NOSWLL+1) THEN

this reflects the fact NOSWLL sets the number of swell partitions in addition to the windsea partition thus a total of NOSWLL+1 partitions.

@aliabdolali @ajhenrique @ukmo-ccbunney

Hello WW3 Developers

Tomorrow we will have our Bi-Monthly Call to discuss the development of WAVEWATCH III.

During this meeting, we will have a presentation about the unstructured grid implicit scheme implementation in WAVEWATCH III. We will also present a summary of recent developments included in the develop branch and updates on master branch changes. The meeting will end with an open forum where all developers can express their ideas and suggestions as well as ask any questions relevant to the development community.

This meeting will be hosted by NCEP and led by Ali Abdolali. He will add an agenda soon.

Cheers

Henrique

Presentation Title: Improving Efficiency and Accuracy of WAVEWATCH III on Unstructured Grids
By Ali Abdolali - NOAA/EMC

Join hangouts meeting: meet.google.com/pwq-vnuq-ady

@ajkammerer93
@aliabdolali
@bauj0001
@benoitp-cmc
@brackins-john
@brandaodeco
@breichl
@CarstenHansen
@chrishalbert
@Clarenburg
@djcrowley4
@djtitze
@dongming007
@emezac63
@erdc
@MathieuDutSik
@fabio-p-piccoli
@FCOO
@flampouris
@fmlokt
@fradscnr
@GabySalgado
@graig-sutherland
@hellais
@hkershaw-brown
@huangynj
@j-gonzalezlopez
@JeffOskamp
@jessicaborgesp
@JessicaMeixner-NOAA
@JoaoGRRodrigues
@kayo-ide
@lauryrenac
@lileipku00
@ling2019
@Miazhi
@MicahDavid
@mickaelaccensi
@mithundeb
@mmuscato
@QingxiangLiu
@randobucci
@RichardMGorman
@rickyspaceguy
@RobertoPadilla-NOAA
@sahammhd
@SallyDillon12
@seglion
@tiagofaro
@tomdurrant
@ukmo-ccbunney
@ukmo-waves
@umr-lops
@AndresSepulveda
@wavespotter
@XianwuXue-NOAA
@xuexianwu
@xychengso
@YoungjaeYu
@yuemaaa
@yuwana1989
@zhongtouwang2019

Extending the current IC5 module by including another two dissipative ice models:

1. the viscoelastic Robinson & Palmer ice model (i.e., the RP model in Mosig et al. (2015))
2. the Order 3 power law proposed by Meylan et al. (2018)

Andy Saulter with @ukmo-ccbunney will work to add these enhancements.

crash with gnu compiler

use a generic template for all the reading/writing