Vertical mixing in a 3D model with a temperature gradient - D-Flow Flexible Mesh - Delft3D
intro story D-Flow FM
D-Flow Flexible MeshD-Flow Flexible Mesh (D-Flow FM) is the new software engine for hydrodynamical simulations on unstructured grids in 1D-2D-3D. 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 D-Flow 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
Vertical mixing in a 3D model with a temperature gradient
Ben Williams, modified 7 Years ago.
Vertical mixing in a 3D model with a temperature gradient
Jedi Knight Posts: 114 Join Date: 3/23/11 Recent Posts 00
Hi all,
I'm running a 3D model with a reasonably uniform temperature gradient. The same temperature gradient is specified as a uniform starting condition and also at the boundaries.
The problem I am having is 'artificially high' mixing. CTD data I have for a location shows that the temperature gradient persists almost all the time. However the model seems to be over-egging the mixing so that over time the water column at the calibration location (about 6km from the boundary) is entirely mixed, even though the water coming in at the boundary still has the temperature gradient applied.
Has anyone got any tips for controlling the degree of vertical mixing in Delft3D?
Model details: 100m water column, dx=dy= ~20 --> ~150m, sigma layer model with 10 layers applied uniformly. Only a very mild bed slope in the model going from 100m depth to about 10m depth over roughly 5km. Current speed at the boundary is about 1m/s, applied uniformly over the water column.
I've tried k-E, k-L and constant eddy viscosity models. I've also tried miniscule background eddy viscosity and diffusivity values.
Thanks,
Ben
I'm running a 3D model with a reasonably uniform temperature gradient. The same temperature gradient is specified as a uniform starting condition and also at the boundaries.
The problem I am having is 'artificially high' mixing. CTD data I have for a location shows that the temperature gradient persists almost all the time. However the model seems to be over-egging the mixing so that over time the water column at the calibration location (about 6km from the boundary) is entirely mixed, even though the water coming in at the boundary still has the temperature gradient applied.
Has anyone got any tips for controlling the degree of vertical mixing in Delft3D?
Model details: 100m water column, dx=dy= ~20 --> ~150m, sigma layer model with 10 layers applied uniformly. Only a very mild bed slope in the model going from 100m depth to about 10m depth over roughly 5km. Current speed at the boundary is about 1m/s, applied uniformly over the water column.
I've tried k-E, k-L and constant eddy viscosity models. I've also tried miniscule background eddy viscosity and diffusivity values.
Thanks,
Ben
Adri Mourits, modified 7 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Yoda Posts: 1221 Join Date: 1/3/11 Recent Posts 00
Hi Ben,
May be it helps to switch on "anti creep", by adding the following line to your mdf-file:
Sigcor=#Y#
Or in the FLOW-GUI -> Numerical Parameters -> "Correction for sigma-coordinates"
Regards,
Adri
May be it helps to switch on "anti creep", by adding the following line to your mdf-file:
Sigcor=#Y#
Or in the FLOW-GUI -> Numerical Parameters -> "Correction for sigma-coordinates"
Regards,
Adri
Fernando Barreto, modified 3 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Padawan Posts: 74 Join Date: 6/21/12 Recent Posts 00
I'm a facing the same problem with a anomalous cooling of my surface sigma layers by the bottom layers. Did you find a solution for your problem?
Regards,
Fernando Barreto
Regards,
Fernando Barreto
Ben Williams, modified 3 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Jedi Knight Posts: 114 Join Date: 3/23/11 Recent Posts 00
Hi Fernando,
After some months of pain I did find a solution. There were multiple problems that I was dealing with.
My suggestion to you is :
1) Forget using the z layer model with parallel. There are simply too many problems with it.
2) If a temperature gradient exists in either the horizontal or vertical, the courant number for baroclinic terms requires a very fine time step to satisfy relative to barotrophic terms. I had to coarsen my horizontal grid by factor 5 to achieve realistic time step. Dx=dy=150m.
3) it's really hard to stop layers below surface interfering with surface layer. Try describing your gradient with at least 5 layers, bearing in mind the limitation of each row only being 1. 2 or 0.8 that of adjacent cells . If your tidal range is greater than surface layer thickness this is also a problem that is hard to beat.
Might help if you post more details of your model setup
Ben
After some months of pain I did find a solution. There were multiple problems that I was dealing with.
My suggestion to you is :
1) Forget using the z layer model with parallel. There are simply too many problems with it.
2) If a temperature gradient exists in either the horizontal or vertical, the courant number for baroclinic terms requires a very fine time step to satisfy relative to barotrophic terms. I had to coarsen my horizontal grid by factor 5 to achieve realistic time step. Dx=dy=150m.
3) it's really hard to stop layers below surface interfering with surface layer. Try describing your gradient with at least 5 layers, bearing in mind the limitation of each row only being 1. 2 or 0.8 that of adjacent cells . If your tidal range is greater than surface layer thickness this is also a problem that is hard to beat.
Might help if you post more details of your model setup
Ben
FP
Frank Platzek, modified 3 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Youngling Posts: 3 Join Date: 3/24/11 Recent Posts 00
Dear Ben, dear Fernando,
I am not sure whether your questions are still open or whether you have managed to resolve them by adapting your models.
I do have a number of questions concerning your posts and would be very greatful if you could answer them. It might be that in answering some of these questions, we are able to improve the performance of your models.
1) Are you running sigma- or z-layer models?
2) What time steps are you aiming for (in combination with a certain horizontal/vertical resolution)? The Delft3D z-layer module (with default settings) is explicit in the momentum and constituent/tracer advection and therefore is subject to a Courant condition based on the flow velocity (in contrast to the sigma-layer version, which has a much less strict stability condition due to the Cyclic ADI advection scheme).
3) Are you running in parallel?
Related to question 3, I was a bit surprised to read that Ben suggested not to use the parallel z-layer version of Delft3D. Which version of Delft3D are you using (which revision number)? The parallel z-layer module of Delft3D was realized just a few years a go and in the past year a number of changes have been made to improve the robustness of the parallel code. If you are using a new version of Delft3D, I would be very interested to hear what kind of problems have you encountered. We have had a number of recent applications of the parallel z-layer code, where very satisfactory results were obtained.
Kind regards,
Frank
I am not sure whether your questions are still open or whether you have managed to resolve them by adapting your models.
I do have a number of questions concerning your posts and would be very greatful if you could answer them. It might be that in answering some of these questions, we are able to improve the performance of your models.
1) Are you running sigma- or z-layer models?
2) What time steps are you aiming for (in combination with a certain horizontal/vertical resolution)? The Delft3D z-layer module (with default settings) is explicit in the momentum and constituent/tracer advection and therefore is subject to a Courant condition based on the flow velocity (in contrast to the sigma-layer version, which has a much less strict stability condition due to the Cyclic ADI advection scheme).
3) Are you running in parallel?
Related to question 3, I was a bit surprised to read that Ben suggested not to use the parallel z-layer version of Delft3D. Which version of Delft3D are you using (which revision number)? The parallel z-layer module of Delft3D was realized just a few years a go and in the past year a number of changes have been made to improve the robustness of the parallel code. If you are using a new version of Delft3D, I would be very interested to hear what kind of problems have you encountered. We have had a number of recent applications of the parallel z-layer code, where very satisfactory results were obtained.
Kind regards,
Frank
Ben Williams, modified 3 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Jedi Knight Posts: 114 Join Date: 3/23/11 Recent Posts 00
Hi Frank,
Are you asking the question in the official capacity of Deltares? If so, let me know your email address and I will start a dialogue with you offline.
I think there remains improvements to be made in the z model with parallel, particularly in artificial mixing along MPI boundaries. I used several tagged versions, the most recent dating from June 2016. I recognise that several improvements have been made since then but it was a difficult site with multiple numerically challenging aspects.
Lets see if we can work together on this.
Regards,
Ben
Are you asking the question in the official capacity of Deltares? If so, let me know your email address and I will start a dialogue with you offline.
I think there remains improvements to be made in the z model with parallel, particularly in artificial mixing along MPI boundaries. I used several tagged versions, the most recent dating from June 2016. I recognise that several improvements have been made since then but it was a difficult site with multiple numerically challenging aspects.
Lets see if we can work together on this.
Regards,
Ben
Fernando Barreto, modified 3 Years ago.
RE: Vertical mixing in a 3D model with a temperature gradient
Padawan Posts: 74 Join Date: 6/21/12 Recent Posts 00
Hi Ben,
thanks a lot for your reply.
Actually I was trying to use Deft3D in a case with a deep bathymetric field (max depth around 1200). Testing different configurations of viscosity/diffusion values I managed to control the field for the serial Z case. For the sigma case I'd not have success (it mixes a lot vertically, cooling surface layers and generating strong baroclinic jets).
I will try your suggestions, thank you
Regards,
Fernando T. C. Barreto
thanks a lot for your reply.
Actually I was trying to use Deft3D in a case with a deep bathymetric field (max depth around 1200). Testing different configurations of viscosity/diffusion values I managed to control the field for the serial Z case. For the sigma case I'd not have success (it mixes a lot vertically, cooling surface layers and generating strong baroclinic jets).
I will try your suggestions, thank you
Regards,
Fernando T. C. Barreto