BASEMENTBasic Simulation Environment for computation of environmental flow and natural hazard simulationLaboratory of Hydraulics, Hydrology and Glaciology (VAW)ETH Zurich | ![]() |
You are not logged in.
Dear all, I have noticed that, in certain cases, the temporal propagation of a flood wave depends on the output time step (TIME > out): if the output time step is smaller, the flood wave propagates faster. Apart from this parameter, both models are identical (computational grid, CFL number, etc.); the results should therefore be identical at the same time step. Has this also been noticed in other use cases? A minimal working example (based on the Flaz tutorial) could be provided.
Offline
Hi Daniel
I noticed this as well in a model some time ago. In short: Yes the output timestep affects the solution. The reason is that the timestep just before your exact output (lets say at t = 10 s) is adjusted (reduced) to match exactly the 10 s timestamp. So for many output timesteps (also affected by tracking_timestep in FLOOD) many small timesteps might get introduced to match your output pattern. However, in most cases this effect should be rather small.
Hope this helps!
Best, Lukas
Offline
Hello Lukas, thank you for your prompt reply! The numerical time step is approximately 0.1–0.2 s, and the flow velocity is approximately 2–3 m/s. The output time step varies between 6 and 600 s. At an output time step of 6 s, the front of the wave is approx. 20-25 m ahead of the front at an output time step of 600 s. In light of your explanation, this seems (too) much to me; I would rather expect a difference of max. 1 m...
Offline
Hi Daniel
What is roughly the triangle size in the mesh? What are your initial conditions, wet or dry?
Offline
The triangle size is variable and about 2-10 m2, the initial condition is dry. The inflow is a source (type "total", sink "exact").
Offline
is there a difference between wet and dry initial conditions (in case it is a river bed)? Can you maybe reduce the influence by changing (increasing) the minimal_water_depth?
Offline
I suspect that the problem is related to the inflow source. When I add the inflow via a stringdef at the upstream boundary of the model (with "uniform_in"), the wave propagation is identical at output time steps 6 and 600 s (i.e. the wave front is always at the same location at 10, 20, 30, ... min, regardless the output time step). However, if I add the inflow via a source over an area with a specific material index, the wave propagation depends on the output time step.
Offline
This effect is really surprising and should be investigated using a minimal test case.
Offline