Enable no local timesteping #356

willem4 · 2025-11-04T15:38:17Z

What was done

minimum transport time step dtmin_transp set now computed in setdtorg()
if transportlocaltimestep=0 then reduce the current flow time step dts by the dtmin_transp if smaller than the current flow time step.

Evidence of the work done

Video/figures
<add video/figures if applicable>
Clear from the issue description
Test in issue is correct now

Tests

Tests updated
e02_f022_c107
Not applicable

Documentation

Documentation updated
Created issue: UNST-9393
Not applicable

Issue link

…oes local timesteps but in all cells.

… local time step

FlorisBuwaldaDeltares

I don't understand the whole local timestepping business but apart from 1 suggestion and 1 question the code looks good. If @V-Chavarrias agrees with the implementation then go ahead and merge

FlorisBuwaldaDeltares · 2025-11-05T08:49:40Z

...dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/get_jaupdatehorflux.f90

-      integer, intent(in) :: nsubsteps !< number of substeps
      integer, intent(in) :: limtyp !< limited higher-order upwind (>0) or first-order upwind (0)
      integer, dimension(Ndx), intent(in) :: jaupdate !< cell updated (1) or not (0)
      integer, dimension(Lnx), intent(out) :: jaupdatehorflux !< update horizontal flux (1) or not (0)


Suggested change

integer, dimension(Lnx), intent(out) :: jaupdatehorflux !< update horizontal flux (1) or not (0)

integer, dimension(Lnx), intent(out) :: jaupdatehorflux !< horizontal flux on link needs updating (1) or not (0)

FlorisBuwaldaDeltares · 2025-11-05T08:55:43Z

...dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/get_jaupdatehorflux.f90


      if (timon) call timstrt("get_jaupdatehorflux", ithndl)

-      if (nsubsteps == 1) then


it is not clear to me why this check (and the corresponding input argument) could be removed. It is also not clear to me why they were replaced by a check on limtyp.

If you only have one substep, you need to update the fluxes.

Originally, this was done in subroutine get_jaupdatehorflux (this if-clause). However, I guess that somebody found it was better to bypass the calling altogether and made the same if-clause in update_constituents:

if (nsubsteps > 1) then call get_jaupdatehorflux(nsubsteps, limtyp, jaupdate, jaupdatehorflux) end if

Hence the one inside get_jaupdatehorflux is redundant.

V-Chavarrias

Thanks!

V-Chavarrias · 2025-11-05T19:33:34Z

src/engines_gpl/dflowfm/packages/dflowfm_kernel/src/dflowfm_data/unstruc_model.f90

      end if

      call prop_set(prop_ptr, 'numerics', 'TransportAutoTimestepdiff', jatransportautotimestepdiff, 'Auto Timestepdiff in Transport, 0 : lim diff, no lim Dt_tr, 1 : no lim diff, lim Dt_tr, 2: no lim diff, no lim Dt_tr, 3=implicit (only 2D)')
+      call prop_set(prop_ptr, 'numerics', 'transportLocalTimeStep', ja_transport_local_time_step, '0=no, 1=yes (default), 2=yes but all cells use the same local time step. Flag for local time stepping in the transport module')


Remember to send Priska the new input flags

V-Chavarrias · 2025-11-06T05:16:15Z

src/engines_gpl/dflowfm/packages/dflowfm_kernel/src/dflowfm_data/unstruc_model.f90

      call prop_set(prop_ptr, 'Time', 'dtInit', dt_init, 'Initial computation timestep (s)')

-      call prop_set(prop_ptr, 'Time', 'timeStepAnalysis', ja_time_step_analysis, '0=no, 1=see file *.steps')
+      call prop_set(prop_ptr, 'Time', 'timeStepAnalysis', ja_time_step_analysis, '0=no, 1=yes (default)')


Default is 0.

I created UNST-9399 to explain in manual.

V-Chavarrias · 2025-11-06T05:33:30Z

...dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/get_jaupdatehorflux.f90


      if (timon) call timstrt("get_jaupdatehorflux", ithndl)

-      if (nsubsteps == 1) then


If you only have one substep, you need to update the fluxes.

Originally, this was done in subroutine get_jaupdatehorflux (this if-clause). However, I guess that somebody found it was better to bypass the calling altogether and made the same if-clause in update_constituents:

if (nsubsteps > 1) then call get_jaupdatehorflux(nsubsteps, limtyp, jaupdate, jaupdatehorflux) end if

Hence the one inside get_jaupdatehorflux is redundant.

V-Chavarrias · 2025-11-06T05:40:06Z

...gpl/dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/get_ndeltasteps.f90

-!         write(6,*) dtmin
-!      end if
-
+         if (ja_transport_local_time_step==2) then


Could this be moved right after nsubsteps is computed (new line 71) and prevent the loop on ndxi? E.g.:

else logtwo = log(2.0_dp) nsubsteps = max(1, 2**int(log(dts / dtmin) / logtwo + 0.9999_dp)) if (ja_transport_local_time_step==2) then ndeltasteps = nsubsteps numnonglobal = ndxi else dtmin = dts / nsubsteps ! get number of substeps do kk = 1, Ndxi dt = dtmax(kk) if (dt < dts) then ndeltasteps(kk) = min(2**int(log(dt / dtmin) / logtwo), nsubsteps) numnonglobal = numnonglobal + 1 else ndeltasteps(kk) = nsubsteps end if end do ! fictitious boundary cells do LL = Lnxi + 1, Lnx ndeltasteps(ln(1, LL)) = ndeltasteps(ln(2, LL)) end do endif

V-Chavarrias · 2025-11-06T05:48:37Z

...s_gpl/dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/ini_transport.f90

+      nsubsteps = imiss
+      ndeltasteps = imiss
+      jaupdatehorflux = imiss
+      numnonglobal = imiss


For me to remember; nothing to be done with this comment.

imiss is -999. This is filled in alloc_transport:

call realloc(ndeltasteps, Ndx, keepExisting=.false., fill=1) call realloc(jaupdate, Ndx, keepExisting=.false., fill=1) call realloc(jaupdatehorflux, Lnx, keepExisting=.false., fill=1)

and numnonglobal in get_ndeltasteps, and it is always set.

V-Chavarrias · 2025-11-06T05:54:21Z

...dflowfm/packages/dflowfm_kernel/src/dflowfm_kernel/compute_transport/update_constituents.f90

-      jaupdate = 1
-
+      jaupdate = 1 ! always update the first step
+      jaupdatehorflux=1 ! always update the first step


It is not really needed. jaupdatehorflux is used inside comp_fluxhor3D but there there is a check on the number of substimesteps:

if (nsubsteps > 1) then if (jaupdatehorflux(LL) == 0) then cycle else

…ping-in-transport

…nsport' of https://github.com/Deltares/Delft3D into fm/bugfix/UNST-9391-enable-no-local-timestepping-in-transport

…ping-in-transport

willem4 added 6 commits November 4, 2025 14:47

enable no local timesteping

7b9856a

typo

4f2caed

Update setdtorg.f90

59c9766

move to m_flowparameters

441cf9d

Update dimr_dflowfm_all_but_validation_cases.xml

c4f726e

Update setdtorg.f90

1a983df

willem4 requested a review from FlorisBuwaldaDeltares November 4, 2025 15:41

Update setdtorg.f90

42ed97d

willem4 marked this pull request as draft November 5, 2025 03:22

willem4 added 3 commits November 5, 2025 05:07

revert original code for ja_transport_local_time_step==1

1f99a92

change implementation for localtimestepping=0, added option 2 which d…

8935593

…oes local timesteps but in all cells.

update local timestep logic such that jaupdatehorflux=1 at each first…

0b32900

… local time step

FlorisBuwaldaDeltares reviewed Nov 5, 2025

View reviewed changes

V-Chavarrias reviewed Nov 6, 2025

View reviewed changes

willem4 added 16 commits November 11, 2025 08:42

reorder jasourcesink application, set tsd=dmiss_neg

f604736

disable gallapatti temporarily

27d0b43

disable openmp loops temporarily

39facea

set explicit sink in the explicit source term

dd654ce

Set computed flux to 0.0 temporarily

0b2f716

dt_loc = dts

1c0d092

Update get_ndeltasteps.f90

a80a679

Update update_constituents.f90

f76dfdc

write SourSink choice to dia

23be8dd

reactivate Galappatti

2cef4f6

revert explicit sedimentation

f84cf0e

Merge branch 'main' into fm/bugfix/UNST-9391-enable-no-local-timestep…

6ad755f

…ping-in-transport

add variable to set source to explicit

0cdc756

deactivate Galappatti, add sink_implicit_factor

79eb3ad

Merge branch 'fm/bugfix/UNST-9391-enable-no-local-timestepping-in-tra…

ff529e5

…nsport' of https://github.com/Deltares/Delft3D into fm/bugfix/UNST-9391-enable-no-local-timestepping-in-transport

switch off Galappatti

bd612f7

willem4 added 11 commits November 13, 2025 15:28

changes required for unused variable error

97ee363

add theta factors for sediment source and sinks

f4bcaca

update implicit part every sub timestep

75d591a

update - not conservative yet

5365fcf

Merge branch 'main' into fm/bugfix/UNST-9391-enable-no-local-timestep…

27edc6c

…ping-in-transport

update to flux

c6196d0

fix build

e5b779f

Update fm_erosed.f90

970777b

Merge branch 'main' into fm/bugfix/UNST-9391-enable-no-local-timestep…

6ad9c61

…ping-in-transport

remove dt_flux

7fdafc9

Merge branch 'main' into fm/bugfix/UNST-9391-enable-no-local-timestep…

8a4a45b

…ping-in-transport

	integer, dimension(Lnx), intent(out) :: jaupdatehorflux !< update horizontal flux (1) or not (0)
	integer, dimension(Lnx), intent(out) :: jaupdatehorflux !< horizontal flux on link needs updating (1) or not (0)


		if (timon) call timstrt("get_jaupdatehorflux", ithndl)

		if (nsubsteps == 1) then

Enable no local timesteping #356

Are you sure you want to change the base?

Enable no local timesteping #356

Uh oh!

Conversation

willem4 commented Nov 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What was done

Evidence of the work done

Tests

Documentation

Issue link

Uh oh!

FlorisBuwaldaDeltares left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

V-Chavarrias left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

willem4 commented Nov 4, 2025 •

edited

Loading