filament-instability-redux

Setup for simulation of a mixed layer filament using Oceananigans. Setup to easily change reference conditions, initialisation method, cooling etc. Hopefully this is useful later. Simulations and post processing done on SciNet Mist, Niagara systems respectively.

Setup

Environment used on Mist (PowerPC) in $SCRATCH/.julia-mist

  [a9b6321e] Atomix v0.1.0 ⚲
  [052768ef] CUDA v5.3.5 ⚲
  [033835bb] JLD2 v0.5.10
  [9e8cae18] Oceananigans v0.93.3 ⚲
  [6fe1bfb0] OffsetArrays v1.15.0
  [276daf66] SpecialFunctions v2.5.0

Environment used on Niagara in $SCRATCH/.julia-niagara

  [a9b6321e] Atomix v0.1.0 ⚲
  [6a3955dd] ImageFiltering v0.7.9
  [033835bb] JLD2 v0.5.10
  [9e8cae18] Oceananigans v0.93.3 ⚲
  [d0ccf422] Oceanostics v0.14.5
  [6fe1bfb0] OffsetArrays v1.15.0
  [276daf66] SpecialFunctions v2.5.0

Publication

To recreate simulations for publication, submit jobs as follows

Mist (simulations)

scripts/preinit/m2.sh -> scripts/Ri/Ri00.sh
scripts/preinit/m2-Ri01.sh -> scripts/Ri/Ri01.sh
scripts/preinit/m2-Ri02.sh -> scripts/Ri/Ri02.sh
scripts/preinit/m4.sh -> scripts/amplitude/m4.sh
scripts/preinit/m6.sh -> scripts/amplitude/m6.sh
scripts/preinit/m8.sh -> scripts/amplitude/m8.sh

Niagara (post-process)

scripts/post-process/pp-Ri00.sh
scripts/post-process/pp-Ri01.sh
scripts/post-process/pp-Ri02.sh
scripts/post-process/pp-m4.sh
scripts/post-process/pp-m6.sh
scripts/post-process/pp-m8.sh

Figures

fig/paper_figures.jl creates all figures and videos included in the paper or as supplementary material. Functions that create figures are individually documented in their respective source files, located in fig.

Usage

If you would like to modify the simulations, there are a few options. Running a simulation is done with
julia [JULIA OPTIONS] -- path/to/simulation.jl output_folder stop_time save_interval Ro Ri α β Nx Ny Nz Q ampl init_time init_rate preinit preinit_path
A description of each of the arguments:

Basic arguments

• output_folder Path to folder for simulation output
• stop_time Simulation stop time in units of 1/f
• save_interval Time between saved snapshots
• Ro Minimum Rossby number of reference state
• Ri Minimum Richardson number of reference state
• α Ratio between filament separation and frontal jet width
• β Aspect ratio, mixed layer depth relative to filament separation
• Nx Number of across-front grid cells
• Ny Number of down-front grid cells (size of domain will be such that horizontal grid is isotropic)
• Nz Number of vertical grid cells (3/4 of which are in the mixed layer)

Additional arguments

The remaining arguments have additional behaviour:

• Q Surface cooling (flux of buoyancy upwards)
  • If zero, then the model uses a SmagorinskyLilly closure and CenteredSecondOrder advection.
  • If non-zero, then the model uses no closure and implicit WENO advection of order 9.
• ampl Level of noise
  • If preinit_path is empty, then ampl is the maximum absolute value of the symmetric uniform distribution that initialises the u and w fields.
  • If preinit_path is set, then ampl is a prefactor that multiplies the fluctuation fields from the pre-initialisation.
• init_time The amount of time to restore the reference buoyancy profile for
• init_rate The rate of buoyancy relaxation to the reference state during the initialisation
• preinit Whether this simulation is a pre-initialisation
  • If true, the simulation runs without a filament.
  • If false, the simulation runs as normal.
• preinit_path Path to pre-initialisation
  • If empty, the initial condition is some uniformly distributed noise with maximum absolute value ampl applied to the u and w fields.
  • If set, the initial condition is the reference state plus the fluctuations in the final saved timestep in preinit_path, multiplied by ampl.