netcdf turbulence reading rule in amrwind

[QB-Lin]

In AMR-Wind, I used the netcdf file to provide turbulence inflow

ABL.bndry_file = ../netcdf/bndry_file.nc # bndry_file.nc #bndry_file.native
ABL.bndry_io_mode = 1 # 0 = write, 1 = read, -1 = no use
ABL.bndry_planes = xlo
ABL.bndry_output_start_time = 0.0
ABL.bndry_var_names = velocity # temperature
ABL.bndry_output_format = netcdf # netcdf or native

I set time step (dt) =1s in the time variable of my netcdf file, but the actual time step in amrwind is 0.2 s

Amrwind read netcdf according to the number of time step or the actual time value ?

I assume that at 10th time step, the actual time in amrwind simulation is 2 s

If Amrwind read netcdf according to the number of time step, the netcdf turbulence data at the 10 th time step (10 s) is extracted.

If Amrwind read netcdf according to the actual time value, the netcdf turbulence data at the 2nd time step (2 s) is extracted. If there is no t = 2 s data, amrwind will make linear intepolation.

Which is correct?

[marchdf]

It uses actual time and linearly interpolates the planes between the time steps.

[QB-Lin]

Thanks. I have another 3 questions about the netcdf turbulence file setting.

Q1:

In my amr wind simulation, my lo corner is not at the origin of global coordinate
for example
"
geometry.prob_lo = -400 -200 0 # Lo corner coordinates
geometry.prob_hi = 600 200 300 # Hi corner coordinates
"
I set "xlo/level_0/lo" = [-200;0], and set "xlo/level_0/hi" = [200;300]. The amr-wind could not read the netcdf data
The error is as below:
"
Reading input NetCDF file: ../CDRFG/bndry_file.nc
terminate called after throwing an instance of 'std::runtime_error'
what(): Assertion `nc_dat == los' failed, file "/usr/local/1_software/AMR-Wind3.5.0/amr-wind/wind_energy/ABLBoundaryPlane.cpp", line 806
"
When I set "xlo/level_0/lo" = [0;0], and set "xlo/level_0/hi" = [400;300], the amr-wind could read the netcdf data. (I still set geometry.prob_lo = -400 -200 0, and set geometry.prob_hi = 600 200 300 ).
I made this change because I saw the netcdf generated from ABL makes this setting.
In this way, "xlo/level_0/hi" = "xlo/level_0/lengths".
Is this setting correct?

Q2:

In the netcdf generated from ABL, I saw "xlo/normal" =0, "xlo/side" =0, "xlo/perpendicular" =[1; 2]. I set these same values in my DIY netcdf. Is this correct?
Assume x, y, and z means 0, 1, and 2, respectively. Why is "side" set as 0?

Q3:

In the example: https://exawind.github.io/amr-wind/user/abl_bndry_io.html#inflow-file-structure-netcdf
I saw that 'xlo/level_0/nx' is 48.
What is the physical meaning of nx for the turbulence inflow?
From my perspective, xlo only had one layer of grid cell.
I could understand ny and nz. They mean the cell number of xlo surface at level_0 in the y and z direction, respectively.

[marchdf]

Hi,

Q1: this confuses me. I would have assumed your first approach would be the correct one. And looking at the code, it still feels like the correct one. I am worried there's a floating point comparison issue. One way to know for sure is to run your case and see if you get the inflow you expect.

Q2: This is correct. "Why is "side" set as 0?" -> this is an internal thing that is tracked by the netdf writer. Sides are 0 to 6 for each face of a cube. 0 means xlo.

Q3: You can ignore nx on the xlo face. We write all of them out for each plane for consistency and simpler code (fewer conditionals).

[github-actions]

This issue is stale because it has been open 30 days with no activity.