Input_sounding + Sounding_type=ConstantDensity does not work!

[Adiram97]

Hello,

I am trying to run DNS with Boussinesq equations by providing a profile for velocity and temperature in the form of input_sounding. But when I try sounding_type=constantDensity, it gives me an error saying "Arrays are hard-wired upto 256 high only"

I would just like to give a profile. So if there is issue with non-ideal sounding_type, I can even shift to some other format to provide a profile to start the simulation. But I am not sure if there is any other format available to provide a profile

I can even shift to init_type=ConstantDensity instead of Input_sounding. But is there a way to provide a velocity profile if the init_type is constant density?

Thank you very much

[asalmgren]

init_type is used to define the base state -- so I would suggest keeping that as ConstantDensity but then use the files in Source/Prob -- specifically you can use ERF*UserDefined.H -- to set the velocity and theta initial values. Then set erf.prob_name = UserDefined in your inputs file.

[Adiram97]

hi,
Thank you for your response
I am still trying to understand how to define Velocities in ERF_InitCustomPertVels_UserDefined.H. Also because, I need the functionalities of abl problem type. So I am still trying to understand how to define profiles.

I'll update you once I have something

[asalmgren]

@Adiram97 -- can we close this issue or is there still an outstanding problem?

[Adiram97]

@asalmgren

Hello,
Actually I did not manage to solve. It is because I would like to use all the capabilities of prob type abl (So that I have block refinement, IBM etc) while wanting to define initial profiles for velocities. So I do not know how to incorporate both by adding it in Userdefined.H

For the time being what I have been doing is to use an old version of the code from November last year or so to initialize with input_sounding for 1 timestep such that it adds initial perturbations. Then using the generated chk and restart from the field for velocities, adding profile for temperature with the newer version of the code.

If you have better solution/suggestion, please let me know

Thank you

[asalmgren]

Hmm. prob_name only affects the initial perturbations, it doesn't impact anything else. I'll be out of the office for a few days but let's see if we can sort this out next week. in Exec/ERF_Prob.cpp we have the mapping from prob_name to the perturbation file - so you can see that if prob_name is ABL then the state perturbations are initialized in Prob/ERF_InitCustomPert_ABL.H and the velocity perturbations are intiialized in Prob/ERF_InitCustomPertVels_ABL.H. So one thing you could do as a test is to leave prob_type as ABL but edit the Prob/ERF_InitCustomPertVels_ABL.H file to do what you want. Can you try that and see if that's what you want?

[Adiram97]

Hi,

Unfortunately I do not have access to my cluster for a few days. So I'll try this out on Monday

[Adiram97]

Hi @asalmgren

So I finally have access and I checked a bit. So I see that U_0 and V_0 is a scalar. So basically what I would like to do is to give intial profiles of velocity so as a vector. So I tried changing to vector such that I read array from input file but I have been having some issues with implementation as there's a loop over gpus before the perturbation implementation and std::Vector is not seen by gpus. So I am simply trying to understand how to declare it better

[asalmgren]

@Adiram97 -- can you remind me why input_sounding type doesn't work? That does allow you to input different values of U_0,V_0 at different z.

[Adiram97]

@asalmgren So, with input_sounding, if I do sounding_type= ConstantDensity, it gives me an error saying "Arrays are hardwired upto 256 only"

[asalmgren]

How many cells in the vertical would you like to have?

[Adiram97]

I have various test cases and it depends on the problem. Somewhere it is around 500 and in other cases it is around 1000

[asalmgren]

@Adiram97 -- I changed the test on 256 so it only applies when using terrain_fitted coordinates, which I think you aren't. See #3207. I've merged that PR -- can you give it a try?

[Adiram97]

Hi @asalmgren , thank you so much for this. I'll pull and test in a bit.

But, I do use stretched mesh and if I understand correctly, stretch mesh somehow activates the terrain_fitted coordinates.
So I hope that is not an issue

[asalmgren]

Good point about stretched mesh -- you're right that capability wasn't there -- I've added it in #3217. Can you try this PR and see if everything works ok? (I"ve only compiled, not run it.)

[Adiram97]

Hi, It is not compiling for me actually
I am getting the following error from Source/IO/ERF_Plotfile.cpp
..//Source/IO/ERF_Plotfile.cpp(1814): error: explicit capture matches default
auto f = =,this {
I tried to remove "this" and compile and it compiles successfully. But the simulation does not run giving a Segfault