Erroneous FLOW Results Adjacent to Boundary  DFlow Flexible Mesh  Delft3D
intro story DFlow FM
DFlow Flexible MeshDFlow Flexible Mesh (DFlow FM) is the new software engine for hydrodynamical simulations on unstructured grids in 1D2D3D. Together with the familiar curvilinear meshes from Delft3D 4, the unstructured grid can consist of triangles, pentagons (etc.) and 1D channel networks, all in one single mesh. It combines proven technology from the hydrodynamic engines of Delft3D 4 and SOBEK 2 and adds flexible administration, resulting in:
An overview of the current developments can be found here. The DFlow FM  team would be delighted if you would participate in discussions on the generation of meshes, the specification of boundary conditions, the running of computations, and all kinds of other relevant topics. Feel free to share your smart questions and/or brilliant solutions!
=======================================================  Sub groups

Message Boards
Erroneous FLOW Results Adjacent to Boundary
John McDowell, modified 1 Year ago.
Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsDear all,
I've set up a coupled FLOW and WAVE model for my area of interest in North America. The model is relatively large (900x300x1) but coarse with a structured grid of 200m cells over the entire domain. The model runs through fine, but produces warnings regarding to high flow speeds.
These abnormally high flow speeds always occur adjacent to the only open boundary, as narrow jets of high flow, that then propagate into my wider model, resulting in bad data. Please see the attached image, which shows a small portion of my domain. The open boundary is horizontal along the bottom of the picture (South), and is meant to be representative of opening onto the Atlantic ocean.
My boundaries are astronomically forced from TPXO 7.2, and there is no physical feature in the bathymetry from GEBCO that would produce such a response. An example of the boundary conditions applied in FLOW is below:
South1 Z A 3 1 3 1 1000.0 South1A South1B
As you can see, I have set the Alpha reflection parameter to 1000 (default, but have varied it and observed similar results), and the boundaries are water level defined. Wind is applied in my model, but only relatively low speeds with gradual direction changes. I am using a time step of 0.1.
I do not know whether it is due to the coupled WAVE element of the model, where I have applied a single uniform boundary with relatively conservative conditions as follows:
[Boundary]
Name = South
Definition = orientation
Orientation = south
SpectrumSpec = parametric
SpShapeType = jonswap
PeriodType = peak
DirSpreadType = power
PeakEnhanceFac = 3.3
WaveHeight = 1
Period = 5
Direction = 180
DirSpreading = 4
Does anyone have any ideas what will be causing these strange boundary conditions? Is there anything in the model setup that I have overlooked?
Thanks very much,
John
Attachments:
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsHi John.
A couple comments:
Your reflection parameter can be set to 0 for now, I'm not sure why it's defaulting to 1000.
Your timestep can be larger than that, maybe 2 minutes or more at that grid resolution.
The effect you are observing is not uncommon. Try using Chezy rather than Manning for models where depth is large. 55 is a good value to start with. If that still doesn't work, you could also try a thin band of high friction (say Chezy 10? See what works) along that boundary.
Get the FLOW model running happily first before you try to bring in waves. One step at a time!!
Good luck.
Phil
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsHi Phil,
I've tried running just the FLOW model through now, setting Alpha to 0 and increasing the time step to 2 minutes.
The alpha parameter seems to impact the results very little, perhaps slightly reduced jets magnitude due to the boundary being less reflective.
The time step has made the model run much faster, but I am still seeing strange boundary effects.
I'll try altering the bottom friction, some of my model is in excess of 200m deep so perhaps suitable for Chezy (however, the strange boundary jets are seen in shallow water).
May I enquire as to the exact syntax of changing the Chezy parameter to 55? As far as I can see, the only parameters I can alter regarding Roughness is:
Roumet= #C#
Ccofu = 2.0000000e02
Ccofv = 2.0000000e02
These values default to very low, I assume I should not be changing them to 55? Do I alter a value in the boundary file?
Thanks again for your time,
John
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsThat's correct  the timestep won't remove the boundary effects, just run your model faster.
Your model is already set to Chezy (Roumet= #C#) but the values in place are suitable for Manning, not Chezy. Chezy should be somewhere between approx 30 and 60, with 50 or 55 being typical. As configured, you have way too much friction in the model. No need to alter a boundary file.
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsHi Phil,
Thanks for the explanation. I've now done as you suggested but get similarly strange boundary results. I suspect it relates to the time step or the grid size...
It's my understanding that the time step can also impact the results? My models unfortunately don't run through completely with the larger time steps you recommend. Does this seem correct? I get messages such as:
*** MESSAGE Courant number for Uvelocity in Upoint equals 1.01 for (m,n,k) = (1,220,1), at nst = 2755. Advised time step: 0.999E01 minutes.
Decreasing the time step to the recommended values seems to increase the time before the high flow velocities occurs (see attached for my most recent run). But after several simulation hours, the boundary condition returns. The flow in the jet is heading into the model, and now is so focused that it causes the error:
*** WARNING Velocity change too high > 5.00 m/s (per 0.5 DT) after 2863 timesteps in the following points:
(m,n,k) = (37,212,1), v0, v1, abs(v0v1): 0.713E+01 , 0.129E+01 , 0.584E+01
(m,n,k) = (38,213,1), v0, v1, abs(v0v1): 0.678E+01 , 0.986E+00 , 0.579E+01
(m,n,k) = (42,220,1), v0, v1, abs(v0v1): 0.110E+02 , 0.164E+02 , 0.538E+01
(m,n,k) = (42,221,1), v0, v1, abs(v0v1): 0.142E+02 , 0.198E+02 , 0.565E+01
(m,n,k) = (47,208,1), v0, v1, abs(v0v1): 0.755E+01 , 0.135E+02 , 0.591E+01
(m,n,k) = (48,208,1), v0, v1, abs(v0v1): 0.195E+01 , 0.446E+01 , 0.641E+01
(m,n,k) = (49,209,1), v0, v1, abs(v0v1): 0.118E+02 , 0.181E+02 , 0.627E+01
(m,n,k) = (50,209,1), v0, v1, abs(v0v1): 0.235E+01 , 0.415E+01 , 0.650E+01
(m,n,k) = (51,210,1), v0, v1, abs(v0v1): 0.491E+01 , 0.140E+02 , 0.910E+01
(m,n,k) = (53,211,1), v0, v1, abs(v0v1): 0.269E+01 , 0.882E+01 , 0.613E+01
*** ERROR Water level change too high > 25.00 m (per 0.5 DT) after 2863 timesteps in the following points:
(m,n) = (73,218), s0, s1, abs(s0s1): 0.383E+02 , 0.889E+01 , 0.295E+02
From searching for this error on the forums, it seems the solution is to reduce the timestep or increase the grid size...
Do you have any thoughts on this? Is it possible that simply the number of grid cells is too high? It is under the recommended amount of 300,000, and we saw vast models into the millions at the Delft Software Open days, so I would be very confused (and disappointed at the limitation) if this were the case.
Thanks again for your help and advice!
John
Attachments:
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsHi Phil.
Thanks for your speedy reply  hope you're well mate!
I'm implementing your suggestions now, I'll post any updates when I have them.
Thanks again,
John
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsOnce your timestep is small enough to satisfy the courant criterion, then decreasing it further is unlikely to help.
Your grid is quite coarse, so maing it coarser won't help.
Did you try the thin band ( a few cells wide) of high friction (say Chezy 10) all along the boundary?
Given that some of your other "default" settings have been unusual, have you checked viscosity? Should be around 50 for this scale model.
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsHi Phil,
I think I've settled the time step issue, it seems to be most stable around 0.2 minutes. Slower than ideal, but at least the model runs through.
The viscosity was an excellent shout, I've been reading about it this afternoon, and didn't realise I needed to scale it to my grid size (seems obvious now). This has had a huge improvement to the fidelity of the model.
However, I am running the simulation for a relatively long period of time (days to months) and over time the same strange boundary conditions seem to build up again. I hope you don't mind, but I've attached my .mdf file to this message. If you have a free moment, could you give it a quick look over/compare it to your larger models to make sure I have not overlooked anything, or incorrectly assumed that other default values are appropriate?
Thank you again for your continued help, pints on me next time we're in Delft.
Cheers,
John
P.S. I've not yet tried the high friction thin boundary technique, but am loathed to do so unless it is the last option. I'm trying to keep things as quick and simple as they can be, and having to tailor a friction boundary seems (to a Delft novice) like it will be a lengthy/fiddly process!
Attachments:
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsI'll have look at the .mdf.
It takes 2 minutes in Quickin to generate a spatially varying friction. I reckon it will sort your issue.
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsYou could also try it without the wind .... as i said, keep it simple at first. That lower wind drag coefficient is perhaps a little high, default is 6.3e4.
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsI've tried with, and without wind and it doesn't seem to make a huge difference...the only reason I included it this time is because at some point I would like to try to couple the model with waves. Thanks, I'll alter the drag coefficient now. I wish there was an easy to use resource for approximating all these coefficients!
The FLOW manual has lots of good information on appropriate Time Steps (meeting Courant criteria), Alpha Parameters and many others. But in the example of Horizontal Eddy Diffusivity it simply states "dependent on grid size", and I can't see a guide section that covers what to set the bed friction at, or even which method is most appropriate to the situation.
I guess this all comes with time! And I suppose the implication is not to use the model until you've learned what the parameters do!
I'm currently having a play with QuickIn to generate a spatially varying friction file.
Thanks again,
John
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsAny joy John? :)
John McDowell, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsPhil  Yes, thanks! The spatially varying friction did the trick!
To anyone that may experience these type of boundary errors in the future, these were the solutions that positively impacted upon my model:
1. Ensure that your viscosity and diffusivity terms are appropriately scaled to your grid size/length scale of your model. This goes a long way to suppressing the extent that the erroneous boundary jets propagated into the wider model
2. Increasing the grid size/decreasing the time step gives the model more time to resolve the flow, and so can prevent the jets from occurring for a time, as the energy can be allowed dissipate. However, for lengthy runs the jets still build up over time, despite utilising a very small time step of 0.1 for 200m grid cells.
3. The solution that had the most positive influence on the boundary conditions was varying the bed friction at the boundaries. This completely removed the boundary jets, and as far as I can tell, has had very little impact on the hydrodynamics of the rest of the model (no significant amount of energy appears to be lost). This can be applied by:
a) Loading the computational grid into QUICKIN
b) Selecting the cells immediately adjacent to the boundary with a polygon
c) Exporting as a depth file with the values inside the polygon altered to be of higher friction
d) Editing the extension from .dep to .rgh
e) Setting the model to read the .rgh file, rather uniform friction
All credit to Phil Shepperd for these suggestions, each very much appreciated.
I feel like I've learned a lot this week!
Thanks again,
John
arun svv, modified 9 Months ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 11 Join Date: 4/22/15 Recent PostsJohn McDowell:Phil  Yes, thanks! The spatially varying friction did the trick!
To anyone that may experience these type of boundary errors in the future, these were the solutions that positively impacted upon my model:
1. Ensure that your viscosity and diffusivity terms are appropriately scaled to your grid size/length scale of your model. This goes a long way to suppressing the extent that the erroneous boundary jets propagated into the wider model
2. Increasing the grid size/decreasing the time step gives the model more time to resolve the flow, and so can prevent the jets from occurring for a time, as the energy can be allowed dissipate. However, for lengthy runs the jets still build up over time, despite utilising a very small time step of 0.1 for 200m grid cells.
3. The solution that had the most positive influence on the boundary conditions was varying the bed friction at the boundaries. This completely removed the boundary jets, and as far as I can tell, has had very little impact on the hydrodynamics of the rest of the model (no significant amount of energy appears to be lost). This can be applied by:
a) Loading the computational grid into QUICKIN
b) Selecting the cells immediately adjacent to the boundary with a polygon
c) Exporting as a depth file with the values inside the polygon altered to be of higher friction
d) Editing the extension from .dep to .rgh
e) Setting the model to read the .rgh file, rather uniform friction
All credit to Phil Shepperd for these suggestions, each very much appreciated.
I feel like I've learned a lot this week!
Thanks again,
John
Excellent John,
You solved my problem also. In fact, I'm also facing a similar issue in my domain along one stretch of Indian coast. I will try to implement your solution. I've another problem. The currents are reflecting at the boundaries and reentering into the domain causing eddies like features, which I think not realistic. Can you help me in resolving that issue? The problem is enhancing in my smaller domain with morphology on. A lot of bathymetric changes are observing along the boundaries. Your replies really matter. Thanks in advance.
regards
Arun
John McDowell, modified 9 Months ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Youngling Posts: 13 Join Date: 9/6/17 Recent PostsHi Arun,
Yes, the eddy like features you are describing sound very similar to my issue  did you have any luck implementing the friction varying solution?
First off, just to check that the Alpha reflection coefficient at your boundary is set appropriately? Setting it too high can prevent waves or components of the flow from leaving the model.
I think essentially these type of issues occur due to the model being forced by the astronomic boundary, but having too much energy to dissipate. Therefore the flow aligns itself such that some of the energy can leave the model (through unrealistic jets). Especially with morphology, which is driven by flow velocities, any unrealistic increase is only likely to remove more sediment and further deepen the channel and propagate the unrealistic flow regime.
My advice would be:
 Try to increase the bed friction locally near the boundaries, this will dissipate a lot of energy.
 Try a few runs with varying values for the horizontal/vertical eddy diffusivity term  you don't want these large scale eddies to dissipate too quickly, as that is unrealistic. You want them to continue to be a sink for energy (not contributing to the flow speeds).
 Specifically for you, is it possible to alter your morphology grid such that the sediment near the boundary is highly cohesive/hard to erode? Your boundary should be far from your area of interest, so it shouldn't affect your results perceptibly. It will however prevent the channel widening/narrowing at the sensitive boundary vicinity due to any unrealistic flows.
Let me know if any of these options work for you? Model set up is something that interests me a lot, so I'm always interested in hearing the successes/failures people have had.
Cheers,
John
Phil Shepperd, modified 1 Year ago.
RE: Erroneous FLOW Results Adjacent to Boundary
Padawan Posts: 44 Join Date: 12/4/18 Recent PostsExcellent. Now that you don't have unrealistically high flow speeds, I reckon you can nudge that timestep up quite a bit. As a check, load your grid and bathy into quickin, set the timestep and look at the Courant map it generates. Cheers John. Phil.