Sounding ingestion

[willoughby-convective]

One of the great uses of Cloud Model 1 and WRF (ideal mode) is being able to spin-up an atmospheric profile from a sounding. Could you add this to Breeze?

[willoughby-convective]

One of the great uses of Cloud Model 1 and WRF in ideal mode is the ability to spin up a three-dimensional simulation from a single vertical profile of the atmosphere. You hand the model a sounding (surface pressure and temperature, then
height or pressure plus potential temperature, water vapor mixing ratio, and the two horizontal wind components at each level), it builds a horizontally homogeneous base state in hydrostatic balance, and you start integrating. It's the
standard way to study idealized convection, boundary layer evolution, squall lines, supercells, sea breezes, and any other process where you want a clean, reproducible thermodynamic and kinematic environment rather than a full reanalysis or
forecast initial condition. The Weisman and Klemp 1982 sounding is the canonical example, but the same workflow is used with observed soundings, modified soundings, and parameter sweeps over CAPE, shear, and moisture.

I'd love to be able to do the same thing in Breeze.jl. My use case is running large eddy simulations of deep moist convection in idealized environments, where the controlled variable is the input sounding and I want to compare runs across
many profiles without standing up a mesoscale parent domain. Today I prototype these in Cloud Model 1, but I'd prefer to stay in Julia and take advantage of Breeze's WENO advection, anisotropic minimum dissipation subgrid scale model,
adaptive timestepping, and RRTMGP radiation.

A rough sketch of what an API could look like:

using Breeze

Read a sounding from a file (Cloud Model 1 input_sounding format,
or any of the other community formats)
sounding = read_sounding("input_sounding"; format = :cm1)

Initialize a model from the sounding. The constructor handles:
- hydrostatic base state (pressure, density, Exner)
- interpolation of theta, qv, u, v to the model grid
- optional warm bubble or random potential temperature perturbations
to break horizontal symmetry
- default surface boundary conditions consistent with the sounding
model = AtmosphereModel(
grid;
initial_condition = sounding,
perturbation = WarmBubble(amplitude = 2.0, radius = 10_000, center = (0, 0, 1_400)),
microphysics = Morrison2Moment(),
radiation = RRTMGP(),
sgs = AnisotropicMinimumDissipation(),
)

[glwagner]

Are you able to use set! with the sounding profiles that you have? If you need to interpolate from the sounding grid to the model grid, you can use interpolate!. I'd be happy to help with this! What are the variables that you have in your sounding data? If it is a small file, you can share here and we can try to put some code together.

[glwagner]

If there are standard formats, then it may be possible to automate the process with something like

sounding = Sounding(filename, format)
set!(model, sounding)

however, given the tools available in Oceananigans, automation may not be essential. We can first figure out how to write this out in a script, and then consider packaging the code we come up with into a source code utility, if it seems like a good idea.