SEDMORPH parameters

Sediment modeling collaborators: issues, applications, information exchange

Moderators: arango, robertson, rsignell

Post Reply
Message
Author
nganju
Posts: 82
Joined: Mon Aug 16, 2004 8:47 pm
Location: U.S. Geological Survey, Woods Hole
Contact:

SEDMORPH parameters

#1 Unread post by nganju »

When using the SEDMORPH option, what is an appropriate "mud_morph_fac"? Does a value of 10, for example, mean the bed changes are scaled up by 10 at every time-step? Thanks in advance...

jcwarner
Posts: 1172
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

morph factor

#2 Unread post by jcwarner »

excert from
"Development of a three-dimensional, regional, coupled wave, current, and sediment-transport model"
Warner, Sherwood, Signell, Harris, and Arango
submitted to Computers and Geosci.

.......
The bed model accounts for changes in sea floor elevation resulting from convergence or divergence in sediment fluxes. These morphological changes can have significant influence on flow and transport when they are larger than a few percent of the water depth. The morphological changes are accounted for by equating the bottom boundary condition of the vertical velocity to the rate of change of elevation of the sea floor. This method is completely mass conserving and retains tracer constancy preservation.
A morphological scale factor is also provided to allow an increased rate of morphological change as would be useful for simulating trends over time periods longer than typical simulations. Methodologies for morphological updating are described by Roelvink (2006).
For bedload transport, the scale factor is multiplied against the bedload transport rates. For suspended load transport the scale factor is applied by multiplying it to the exchange of sediment (erosive or depositional flux) at the bed-water interface. The magnitude of sediment concentrations in the water column are not modified – just the exchange rate on and off the bed. For both bedload and suspended load, sediment is limited in availability as described previously, based on the true amount of sediment mass (not multiplied by the scale factor). This morphological scale factor method works well for systems with unlimited sediment in the seabed. However, it can generate extra sediment in systems with limited supplies of bed sediment. This occurs when the amount of sediment to be eroded is limited by the amount available: application of the morphological scale factor places more sediment in suspension than is actually present on the bed and subsequent deposition places this new mass in the bed.

glesser

#3 Unread post by glesser »

However, it can generate extra sediment in systems with limited supplies of bed sediment. This occurs when the amount of sediment to be eroded is limited by the amount available: application of the morphological scale factor places more sediment in suspension than is actually present on the bed and subsequent deposition places this new mass in the bed.

Can you please give an example of how this occurs? As far as I am aware, even with limited sediment availability, morphological acceleration factors only cause mass errors if a) the morfac is CHANGED while sediment is in suspension or b) the amount of sediment in suspension is different at the start and end of the "control" interval. I don't understand why a simple limitation of sediment in the bed should lead to mass errors. Many thanks.

csherwood
Posts: 39
Joined: Fri Apr 02, 2004 4:46 pm
Location: USGS, Woods Hole, USA

#4 Unread post by csherwood »

Here is my understanding of the problem: the morph factors (in ROMS and D3D) act on the erosion or deposition rates...so at the end of a time step, they deposit or erode X times the amount calulated by the model (call that F for flux). This causes bed elevation changes that are X times greater, so the coupling between morphology -> flow -> sediment transport -> morphology is basically on steroids.

Deposition is not a problem. But if the amount of material available for erosion is limited (by the stratigraphy or total thickness of the bed) and that limit is enforced, the amount eroded may be less than X times the calculated amount. So erosion at one spot may be <X*F, but deposition elsewhere may be =X*F. If a closed system with sediment transport keeps bumping into the erosion limit and the morph factor is in place, sediment mass can increase.

The moral of the story is that mixing stratigraphic limits on sediment availability and morph factors is a risky modeling behavior.
Chris Sherwood, USGS
1 508 457 2269

glesser

#5 Unread post by glesser »

Hi Chris

A coupla points to think about:

1) In D3D I never "enforced the limit". i.e. in principle the bed sediment quantity could have gone negative - but I never saw this happen due to the "soft landing" we implemented when bed thicknesses became small. This prevents (this) problem from occurring. Note: I suspect that this behaviour may have changed in D3D with the more recent stratigraphy bed model and so the behaviour you describe may now occur. If it does, then I expect there are creative solutions that could be implemented to solve it.

2) Using a morfac of 1 (i.e. not accelerating the bottom changes) doesn't solve the problem you describe either. If your numerical scheme is such that transport can "bump into the erosion limit" - most likely because it is implicit - then using high morfacs only makes this MORE LIKELY to occur, it doesn't fundamentally cause the problem (try putting X=1 into your own equations). Under this scenario the only safe morfac to use is zero! which really isn't that much fun. An explicit numerical scheme would work regardless of morfac used, but I imagine this would have other problems...

In summary, I still don't think you can blame this observed loss of continuity on morfac. Morfac is just showing up shortcomings in the numerical scheme...

Is that fair?


Giles

jcwarner
Posts: 1172
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

#6 Unread post by jcwarner »

Giles-

I am not sure what is in D3D, and do not know what a 'soft landing is."

For ROMS, you are not correct in your interpretation of the morph fac method. Don't try to think of this as anything complicated - it is very simple. If the morph fac is zero, then no sed will be suspended or deposited. If morph fac = 1, then this is 'normal conditions.' The amount of bed elevation change is based on the actual amount of sediment resuspended, and the amount of sed added to the water column is also based on the actual amount of sed resuspended. So there are no issues here. The amount of sed that is resuspended is limited by the amount available on the bed (+ any depositing). So the bed will never go negative. For other morph factors, it is as described above.

These concepts of morph fac need to be applied very carefully. There is not a lot of user experience on these issues.

glesser

#7 Unread post by glesser »

OK, forget the "soft landing", it sounds like you have found your own method.

However, if you have been able to implement a bottom b/c that "The amount of sed that is resuspended is limited by the amount available on the bed (+ any depositing)" then why can't you implement a scheme where...

The amount of sed that is resuspended is limited by [the amount available on the bed (+ any depositing * morfac)] / morfac ? Assuming morfac is a known constant.

As this would appear to solve the problem that you and Chris describe using morfacs in ROMS.

csherwood
Posts: 39
Joined: Fri Apr 02, 2004 4:46 pm
Location: USGS, Woods Hole, USA

#8 Unread post by csherwood »

That doesnt work if stratigraphy or limited sediment is enforced. Imagine a cell with only a little available sediment in the top layer...say 1 mm...(for real geologic reasons, not numerical reasons). A big resuspension event strips the bed and puts everything is suspension...and morph factor is 10, so it would like to remove 10 mm of sediment, but the next layer is too coarse to move, so the amount eroded is still 1mm...but when things calm down and it is redeposited, no limit applies, so 10 mm deposits...presto, more sediment.

So, morph factor is fine as long as you dont enforce limits on erosion, but if you do, it can distort things.

Do you know of a case using D3D where a morph factor is used in conjunction with stratigraphy or a hard limit on initial sediment availability?
Chris Sherwood, USGS
1 508 457 2269

rsignell
Posts: 124
Joined: Fri Apr 25, 2003 9:22 pm
Location: USGS

#9 Unread post by rsignell »

I first realized this problem with MORPHFAC and limited sediment during Delft3D runs where I started growing large sand banks on either side of a tidal headland where the sand had been completely eroded at the headland tip. I found I was generating huge amounts of sand (beach replenishment folks would have been quite happy!) because of exactly the mechanism Chris describes.

I had a MORPHFAC of 100. If 1 mm was available on the bed, and the true physical processes wanted to remove that 1 mm of sediment, then that 1 mm of sediment goes up into suspension and is advected downstream. MORPHFAC then wants to erode the bed by 100 mm at that location, but that is not possible, as there is was only 1 mm there. So the bed erodes by only 1 mm. Downstream, when the 1 mm of sediment deposits, MORPHFAC happily puts 100 mm of sediment on the bed.

As John mentioned, this is NOT a problem if MORPHFAC=1, only if it exceeds 1.

This potential problem exists in both ROMS and Delft3D, and in any other model that uses this MORPHFAC approach.

-Rich

bjagers
Posts: 5
Joined: Tue Nov 28, 2006 5:55 pm
Location: Deltares

#10 Unread post by bjagers »

For the time being I tend to agree with Giles, but that opinion might be affected by the way in which things got implemented in Delft3D.


Step 1: determine sediment transport capacity (reference concentration) not taking into account limited sediment availability.
Step 2a: reduce bedload by factor representing thickness of sediment layer (proximity factor for running out of all sediment).
Step 2b: reduce sediment transport rate and concentration by fractional bed composition.
Step 3: do advection diffusion step.
Step 4: determine bed load gradients and deposition/entrainment rate and consequently determine mass flux.
Step 5: update bed composition (possibily multiplied by morphological factor).


Step 2a and 2b reduce the likelihood of eroding sediment that is not available but these options do not prevent it. Step 2a concerns Giles' "soft landing" but only works if you run out of sediment at the same time as you run out of sediment fraction, i.e. uniform sediment, well mixed transport or last sediment fraction being eroded. Step 2b works only if you run out of one sediment fraction, or phrased more correctly if you don't run out of at least one sediment fraction.


If the transport gradient and net entrainment result in a net erosion rate of F m3/s and assume there is V m3 of sediment in the bed available, then the time step is limited to V/F seconds. If a morphological factor of X is applied then the time step is limited to V/(X*F) seconds which is a smaller time step. So, the same problem can occur for X=1, but is less prominent and thus less likely to occur.


As Giles already cited John indicated: "The amount of sed that is resuspended is limited by the amount available on the bed (+ any depositing)." In Delft3D the deposition term is an implicit term whereas the resuspension term is explicit. This makes the resuspension trivial if you limit it to the available sediment only, but difficult to limit it to the amount of sediment available plus the sediment *going to be deposited*. How does ROMS get around this point? By implementing the deposition also as an explicit term or some other way?


By the way we are currently fixing this by not allowing any removal of the depleted sediment fraction until the shortage has been replenished; at least this guarantees that the error does not accumulate. This is probably one of the "creative solutions" Giles was referring to.


However, I have a more fundamental concern about the stratigraphy module and a morphological factor in tidal areas. The morphological factor is not used as a time accellerator but as a time shifter. Effectively all low tide events are grouped together and all high tide events are grouped together. However, that does not make sense if you are tracking the history of the bed composition. Eroding 10 cm and redepositing 10 cm in some dynamic equilibrium leads to less mixing then eroding 1 m and redepositing 1 m if a morphological factor of 10 is used. The morphological factor used in the tidal sense is messing with the timeline of events. In rivers we just use the morphological factor to speed up morphological time without reordering events; think of a bigger morphological time step than the flow time step. Acklowedging the additional risk of sediment depletion as addressed in the discussion, this latter approach would be acceptable.


I know that Guy and others have applied the stratigraphy module together with time-varying morphological factor for Capitol Lake. Also the combination of Dano's mormerge tool to continuously combine multiple parallel scenarios into one averaged bed level change together with stratigraphy module has been proposed and discussed, but as far as I know that kind of simulation has not yet been carried out (neither in Delft nor in US).


Rich, were you referring to the long-term simulation of the headlands. That problem was associated with a bug in the correction for the suspended load in the bed load region below the reference height. That problem got solved in the end, wasn't it? Or are you referring to continued problems?


Best regards,

Bert

rsignell
Posts: 124
Joined: Fri Apr 25, 2003 9:22 pm
Location: USGS

#11 Unread post by rsignell »

Bert,

Yes, it now appears my statements above about Delft3D and MORFAC could be in error, and that the non-conservative nature of the sediment I experienced was likely related to another bug that was fixed (although I still need to find time to go back and check).

After some offline discussion with you and Giles Lesser, I now understand that Delft3D limits the flux of sediment to the water column when there is insufficient sediment on the bed. To reuse the example I gave previously, in Delft3D, if erosion wants to remove 1 mm of sediment, MORFAC is 100, and only 1 mm of sediment is on the bed, that 1 mm is removed, but then only 0.01 mm of sediment is fluxed into the water column so that when it is deposited again, the 1 mm returns to the bed.

Thus in Delft3D, sediment is conserved no matter what MORFAC is, but not without some undesirable consequences:
  • Sediment concentrations in these conditions will be artificially low, which could be a problem in sediment stratified regimes. But MORFAC is often used in sandy sites where sediment stratification is less of a problem.

    MORFAC isn't actually acting as designed (to accelerate morphological development) in these conditions. But most likely the eroded areas are not contributing that much to the morphological development anyway, and MORFAC is still accelerating the development where there is sufficient sediment.
Overall, it's probably better to limit the flux this way and have these minor sins rather than the major sin of not conserving sediment. Currently, we don't limit the flux to the water column in ROMS. Should we? I think we probably should, but it would be nice to hear some more discussion.

-Rich

njarry
Posts: 7
Joined: Tue Jan 08, 2008 3:57 pm
Location: ACRI-IN

Re: SEDMORPH parameters

#12 Unread post by njarry »

Hi everyone,

I'm working on the lake_signell test case (run OK) and plotting my results with NCARG.

I would like to visualize the bathymetry and its evolution through the time, so I activated the SED_MORPH option.
ccnt < ccnt.in is OK but when SED_MORPH option is activated, bathymetry does not belong to the plottable fields so I can't visualize it.

Does anyone have an idea ?

Thank you

jcwarner
Posts: 1172
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

Re: SEDMORPH parameters

#13 Unread post by jcwarner »

when sed morph is on, the variable 'bath' contains the changing bathymetry.

njarry
Posts: 7
Joined: Tue Jan 08, 2008 3:57 pm
Location: ACRI-IN

Re: SEDMORPH parameters

#14 Unread post by njarry »

thank you for your help

I know about the variable "bath" but it's not a plottable field either

jcwarner
Posts: 1172
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

Re: SEDMORPH parameters

#15 Unread post by jcwarner »

so are you saying that the var is in the netcdf file, but the NCARG plotting package that you are using does not allow that var to be visualized?

njarry
Posts: 7
Joined: Tue Jan 08, 2008 3:57 pm
Location: ACRI-IN

Re: SEDMORPH parameters

#16 Unread post by njarry »

In the ocean_lake_signell.in, I checked that the logical switch to activate writing of "bath" into HISTORY output file is T ( Hout(idBath)== T )

So I think the variable bath is in my output netcdf file (ocean_his.nc).

Then, in the ccnt.in, I put the field ID to plot to 59 (bathymetry at RHO points) but when I try to issue the ccnt < ccnt.in, an error message tell me that the variable "h" is not a plottable field.

And there isn't any plottable field concerning bathymetry

jcwarner
Posts: 1172
Joined: Wed Dec 31, 2003 6:16 pm
Location: USGS, USA

Re: SEDMORPH parameters

#17 Unread post by jcwarner »

can i suggest that you put this post in the
ROMS Tools and Techniques
section, as it seems more to be an issue with the plotting package.
thanks.

Post Reply