How to deal with the blow-up at boundaries due to high wind?

[jayasankar]

Hello All,

My simulations are getting terminated due to blow-up at boundaries, To be specific node like (1,2),(2,2),(1,4) etc...

The wind velocities are very high in this region, around 12±5 m/s.
Grid Resolution is 0.2 to 0.25 degrees, I have tried time steps 1800 sec and 1200 sec.

Thank you, Jayasankar

[kate]

I would keep trying shorter timesteps to see if that solves the problem or if there's something else going on. If you go for half the timestep and the model runs for twice as many steps, there's something else going on.

[jayasankar]

Hi Kate,

thank you for the suggestion ,

I am trying with time step....

Could you give me an idea about time steps used in various model in general?

Is it normal to use time steps of 600 sec for a model with 0.25 degree grid resolution and maximum depth 5000m.

Also, I'm kind of thinking about editing wind velocities by reducing them along strips of say 2 degrees, along the boundary, would that be an acceptable practice?

Thank you, Jayasankar

[kate]

Is it normal to use time steps of 600 sec for a model with 0.25 degree grid resolution and maximum depth 5000m.

It depends on more than that, including your vertical grid structure. We have a domain at 11 km resolution in which we started with 42 vertical levels, then went to 60, then back to 50. The timestep for the latter is now 360, with an option to go shorter as needed. Many of my domains have a default timestep (say 100 sec) and a commented out shorter step (80 sec) to get through the rough spots.

Are there other models for which you could use a longer timestep? I expect so, including Sasha's version of the code.

I have been too lazy to fuss with the winds in the manner you suggest. Easier to just use a short timestep.

[jayasankar]

Hi,

I am trying not to compromise on the speed, hence i thought about reducing the wind velocities along the boundaries.

I am wondering if anyone had tried this way before, or if this would make noticeable variations in the model predictions in the entire domain.

Thank you.