Project: Ocean Only Configuration - The Bare Minimum

This project is all about how we create the most basic ocean_only domain to shake things out and get spinning as fast as possible. This is one way of testing everything before we get into our research questions.

Download this project by either:

Running the CrocoDash CLI command: crocodash template --machine derecho --notebook crocodash.projects.ocean_only
Copying the file from your CrocoDash checkout: gallery/crocodash/projects/ocean_only.ipynb

Grid Generation¶

The first thing we are going to do is think about how to get a domain. Since we want to get something fast, I might look at some scaling data we’ve put online: https://crocodile-cesm.github.io/SeaSloth/mom6_scaling.html

Given how fast the scaling 100x100 domain is, I think we should go with that. If we provide ~40 cores for the ocean, it’ll run wicked fast.

We’ll pick a standard vertical grid.

from CrocoDash.grid import Grid
from CrocoDash.vgrid import VGrid

grid = Grid(
  resolution = 0.05, # in degrees
  xstart = 278.0, # min longitude in [0, 360]
  lenx = 5.0, # longitude extent in degrees
  ystart = 7.0, # min latitude in [-90, 90]
  leny = 5.0, # latitude extent in degrees
  name = "basic_grid",
)


vgrid  = VGrid.hyperbolic(
    nk = 75, # number of vertical levels
    depth = 2000,
    ratio=20.0 # target ratio of top to bottom layer thicknesses
)

Now that we have our grids, we need to setup our bathymetry. Normally, we’d use the Topo.set_from_dataset function to go from a bathymetry dataset to grid bathymetry. However, we’re trying to move with purpose, so we can use Topo helper functions to create our own fake fun little domain.

from CrocoDash.topo import Topo

topo = Topo(
    grid = grid,
    min_depth = 9.5, # in meters
)

topo.set_bowl(max_depth = 2000, dedge = 200)

topo.depth.plot()

Great! So I’m going to create my case now, and this step will hardly ever change!

from pathlib import Path
from CrocoDash.case import Case

# CESM case (experiment) name
casename = "basic_ocean"

# CESM source root (Update this path accordingly!!!)
cesmroot ="<CESM>"

# Place where all your input files go 
inputdir = Path("<inputdir>") / casename
    
# CESM case directory
caseroot = Path("<casedir>") / casename


case = Case(
    cesmroot = cesmroot,
    caseroot = caseroot,
    inputdir = inputdir,
    ocn_grid = grid,
    ocn_vgrid = vgrid,
    ocn_topo = topo,
    project = "<PROJECT>", # Update this with your project code
    override = True,
    machine = "derecho",
    compset = "CR_JRA"
)

Now, the last step is to generate forcings, which is done as a two step configure and process. Since it can take a long time to get raw data from outside sources, we’re going to use a fake data generator that comes with CrocoDash.

I also think we can run for 60 days. That doesn’t sound too bad.

case.configure_forcings(
    date_range = ["2019-01-01 00:00:00", "2019-03-03 00:00:00"],
    product_name="reference_ocean",
    function_name="get_reference_ocean_data"
)

case.process_forcings()

Cool! Once this is done, we have to ./case.build and ./case.submit in our case directory. Before that though, here’s some things I’d xmlchange to make the domain run faster.

I’d set the NTASKS = 10, which sets every component to 10 cores, and makes the data atmosphere regridding very fast. (All components share these 10 cores)
I’d set the NTASKS_OCN = 40 because of the scaling data we saw: https://crocodile-cesm.github.io/SeaSloth/mom6_scaling.html
I’d set the ROOTPE = 0, which ensures that all components share the same cores. If I set ROOTPE_OCN=10, it would skip the first 10 cores and run on the next 40 cores.
I’d set the JOB_WALLCLOCK_TIME = 00:30:00 --subgroup case.run, which ensures the queue picks up our job realllly fast.

And there we go! A fake running ocean model! Isn’t that awesome!