Using refined boundary plane data in a simulation

[lawrenceccheung]

Bug description

I'm seeing a likely box array mismatch error when refining the boundary plane data generated in one simulation to be used in another kynema-sgf simulation. The segfault error I'm seeing is:

[flight1:19282] *** Process received signal ***
[flight1:19282] Signal: Aborted (6)
[flight1:19282] Signal code:  (-6)
  what():  Assertion `amrex::match(hdr.m_ba,mf.boxArray()) && hdr.m_ncomp == mf.nComp()' failed, file "/projects/kynema/builds/flight/kynema_flight_exfix_2026-05-28/kynema-manager/stage/ndeveld/spack-stage-kynema-sgf-main-rdgreyluw6t6yopfi64jjpg6cbxhu7go/spack-src/submods/amrex/Src/Base/AMReX_VisMF.cpp", line 1637

Note that my ultimate goal is to use boundary plane data generated from an ERF simulation at different resolution and apply it to a kynema-sgf simulation. However, I see the same error with the ERF boundary planes, and I created a kynema-sgf only test case which can be quickly used to debug the process

Steps to reproduce

1. The initial boundary plane data is used using a modified GABLS input file here: abl.1.inp. Note that this simulation is run with domain parameters

geometry.prob_lo        =   0.       0.     0.  # Lo corner coordinates
geometry.prob_hi        =   400.   400.   400.  # Hi corner coordinates
amr.n_cell              = 128 128 128 # Grid cells at coarsest AMRlevel
amr.max_level           = 0           # Max AMR level in hierarchy

2. Use the refine_native_boundary_plane.py to refine the boundary planes.

3. Use the refined boundary planes in the abl.restart.inp input file, which has twice the grid resolution as the original case:

geometry.prob_lo        =   0.       0.     0.  # Lo corner coordinates
geometry.prob_hi        =   400.   400.   400.  # Hi corner coordinates
amr.n_cell              = 256 256 256 
amr.max_level           = 0           # Max AMR level in hierarchy

Steps to reproduce the behavior:

1. Compiler used

   • GCC
   • Intel LLVM
   • oneapi (Intel)
   • nvcc (NVIDIA)
   • rocm (AMD)
   • with MPI
   • other:

2. Operating system

   • Linux
   • OSX
   • Windows
   • other (do tell ;)):

3. Hardware:

   • CPU
   • GPU

4. Machine details ():
   Using SNL flight machine.

5. Input file attachments
   Input files can be downloaded from this gist: https://gist.github.com/lawrenceccheung/d77a7c4d54ff31516d61c315c55735d1

6. Error (paste or attach):
   See error output above.

7. If this is a segfault, a stack trace from a debug build (paste or attach):

Expected behavior

Kynema-SGF information

==============================================================================
                Kynema-SGF (https://github.com/kynema/kynema-sgf)

  Kynema-SGF version :: 043f87d-DIRTY
  Kynema-SGF Git SHA :: 043f87d33caaaaf69da98fc2fb98485a667b0864-DIRTY
  AMReX version    :: 26.05-41-gba3cb0c68650

  Exec. time       :: Sun Jun  7 15:15:51 2026
  Build time       :: May 28 2026 19:46:23
  C++ compiler     :: IntelLLVM 2025.3.0

  MPI              :: ON    (Num. ranks = 448)
  GPU              :: OFF
  OpenMP           :: OFF

  Enabled third-party libraries: 
    NetCDF    4.8.1
    HYPRE     3.1.0

           This software is released under the BSD 3-clause license.           
 See https://github.com/Kynema/kynema-sgf/blob/development/LICENSE for details. 
------------------------------------------------------------------------------

Additional context

[lawrenceccheung]

Tagging @mchurchf in case he is also thinking about doing ERF/AMR-Wind coupling via boundary planes.

[lawrenceccheung]

Update: the other utility generate_native_boundary_plane.py is working fine with any resolution, so we can probably use that to achieve what I want.

[github-actions]

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

[mbkuhn]

@lawrenceccheung should we address this issue, or is your workaround good enough?