I'm busy building a 2d only Hec model for a mining pit, which requires that I use a precipitation approach (net precipitation after losses which i will apply uniformily over various homogenous 2d areas of the pit and surrounds). The available lidar has an average pixel density of about 1 point every 2x2m. The pit itself is characterized by various benches and roads with bunding (informal earthen walls) along many of them. I consequently need to retain as much detail as possible to assess flow pathways and so am using a 2x2m cell size. The use of breaklines has occured to me in order to reduce cell size but the bunding is so dense that it would look ridiculous and take that much time as well.
What I need help with is on the choice of diffusion vs. full momentum.
At present I am using a small test area (160,000 cells) and can get a mostly converging run with diffusion using all default parameters and a computation interval of 20sec. But when trying full momentum (default parameters), the model starts diverging from the start, even when i choose a computation interval as low as 0.1sec.
Will diffusion suffice - for a initially dry pit, with near vertical walls becoming water falls during the simulation? It may have to if i can't get the full momentum working?
Thanks for any feedback.
....on a related albeit less important note - when looking at ras mapper results, the animation shows pulsing waves of flow in what is expected to be a more sustained flow. Any thoughts on that?
The diffusion wave is going to be more stable for the water falls on the vertical wall.
There isn't really enough info to say whether full momentum is needed, but if the velocities (ignoring the water falls over the vertical walls) are generally pretty low and it is generally overland and shallow flow, than diffusion wave would probably be fine. If you have fast water flowing down a steep road that makes sharp turns, for instance, momentum would, in theory, do better, but as long as it is pretty shallow I doubt it would make much difference (and there would probably be larger sources of error/uncertainty in the model).
And if you can't get the full momentum to run, diffusion is, obviously, going to give you a better answer than nothing.
If you have a 2m cell then the velocities need to be around 2 m/s or less for a 1 second time step or you would start to get interesting results. The 20 second time step is to large and probably breaks the Courant Criteria. You can get a smaller timestep than 0.1 seconds which can be done in the Unsteady Plan Options for 2D meshes using a multiplier.
The pulsing waves is due to too high of a timestep.
The diffusion wave solver is very stable and will run with high timesteps, but give poor results.
Thanks Jarvus and Cameron - those responses are very welcome.
I will keep an eye on velocities once i have the model finalized and will run it with full momentum if possible, by either reducing time step, cell size or reducing velocity (through more representative Mannings N values). Otherwise diffusion - to at least get a result (with the necessary caveats).