Hi all --
I was just wondering what the status is of wetting/drying capabilities for ROMS. Is that built into any of the newer versions yet? I noticed on a post back in Aug 05 that it was something that was being worked on.
thanks
Raz
ROMS wetting & drying
Current status of wetting/drying-
Currently we have a version of ROMS that includes a rudimentary method for wetting and drying. The formulation is based upon the concept of a 'critical depth' (Dcrit) criteria. As the model progresses, the total depth (h+zeta) in each cell is compared to Dcrit. If h+zeta<Dcrit then we impose a 'flux blocking' algorithm that prevents transport out of that cell. Water can flow into any cell at any time, but if the total depth is less than Dcrit the water can not flow out. Cells become rewet as water flows back in from adjacent cells.
The method has been used successfully in several simple geometries, standard test cases, and several more complicated configurations. The method is mass conserving and can give stable results.
This method is simple, but it also has problems. It relies on a criteria to fail before action is taken. As soon as the total depth falls below Dcrit the velocities are immediately shut off. This can create small 'shocks' that are not friendly. Also, a small thickness of water (user defined, typ ~5 cm) is left on each cell while it is dry. What happens to the tracers in those cells? Does temp increase due to solar flux? What about a dissolved tracer - this wet/dry method could actually lead to artificial increased dispersion of that tracer. We need a different method.
What have we tried: We tried many other methods, including:
- I looked at the Oey paper in OM - they take similar approach but they calculate Dcrit criteria at u/v points, I had used rho points. I think his method could lead to small oscillations in the water level. But essentially it is cell face blocking as before.
- We looked at the Burchard et al method in GETM. They use a formalism based on shallow water theory that essentially shows a balance between the barotropic pressure gradient and the bottom stress. They artificially increase the bottom stress to prevent transport out of the cells. This can cause delay of re-wetting the cells.
Proposal:
I spent a few days down at Hernan's working on this. We called Sasha and he suggested a few alternatives. We need to spend some time to investigate several options, but they include an approach based on an upstream scheme for the free surface with higher order corrections (maybe something like a flux-limiter method). To implement this methodology, we felt that step2d should be revised to be consistent with the version in UCLA's code. This will allow better management of the limiter to accommodate the time stepping. We are looking into this deeper as time permits. It is high on the list.
Currently we have a version of ROMS that includes a rudimentary method for wetting and drying. The formulation is based upon the concept of a 'critical depth' (Dcrit) criteria. As the model progresses, the total depth (h+zeta) in each cell is compared to Dcrit. If h+zeta<Dcrit then we impose a 'flux blocking' algorithm that prevents transport out of that cell. Water can flow into any cell at any time, but if the total depth is less than Dcrit the water can not flow out. Cells become rewet as water flows back in from adjacent cells.
The method has been used successfully in several simple geometries, standard test cases, and several more complicated configurations. The method is mass conserving and can give stable results.
This method is simple, but it also has problems. It relies on a criteria to fail before action is taken. As soon as the total depth falls below Dcrit the velocities are immediately shut off. This can create small 'shocks' that are not friendly. Also, a small thickness of water (user defined, typ ~5 cm) is left on each cell while it is dry. What happens to the tracers in those cells? Does temp increase due to solar flux? What about a dissolved tracer - this wet/dry method could actually lead to artificial increased dispersion of that tracer. We need a different method.
What have we tried: We tried many other methods, including:
- I looked at the Oey paper in OM - they take similar approach but they calculate Dcrit criteria at u/v points, I had used rho points. I think his method could lead to small oscillations in the water level. But essentially it is cell face blocking as before.
- We looked at the Burchard et al method in GETM. They use a formalism based on shallow water theory that essentially shows a balance between the barotropic pressure gradient and the bottom stress. They artificially increase the bottom stress to prevent transport out of the cells. This can cause delay of re-wetting the cells.
Proposal:
I spent a few days down at Hernan's working on this. We called Sasha and he suggested a few alternatives. We need to spend some time to investigate several options, but they include an approach based on an upstream scheme for the free surface with higher order corrections (maybe something like a flux-limiter method). To implement this methodology, we felt that step2d should be revised to be consistent with the version in UCLA's code. This will allow better management of the limiter to accommodate the time stepping. We are looking into this deeper as time permits. It is high on the list.