Ocean and Ice Modeling

Review

Overview

Teaching: min
Exercises: min
Questions

  • What have we learned so far?

  • Where are we going?

Objectives

Midway through the semester is a good time to pause and review what we have learned so far and what we have left.

What have we learned so far?

Where are we going?

Atmospheric Modeling

  1. Dynamics (Equations of motion) Assumptions: hydrostatic Linearize, discretize, solve numerically Predict: thickness(w),u,v,q,T

  2. Physics (Sub grid-scale Parameterized)
    • Clouds
    • Radiation
    • Convection
    • Microphysics
    • Boundary layer
    • Fluxes
    • Gravity waves
  3. Boundary Forcing Information exchnaged with or prescried by other components of the climate system (e.g. soil moisture, sea surface temperature)

CESM Quickstart

A review of our CESM quickstart procedure

Create a newcase (this script is located in CIMEROOT/scripts)

./create_newcase --case CASEROOT --res RESOLUTION --compset COMPSET --project UGMU0032

Setup the case (run from your case directory)

./case.setup

Make your changes to the namelists, .xml files, and/or source codes

Build the case

qcmd -- ./case.build

Submit the run

./case.submit

CESM Compsets we have used so far

Today we will work with

Key Points

Ocean and Sea Ice Modeling

Overview

Teaching: min
Exercises: min
Questions

  • How to Ocean and Ice Models work?

Objectives

Just like the atmosphere and land models are typically run together, the ocean and sea ice models are also typically run together. We will learn about what ocean and sea ice models predict, what they explicitly resolve, what they paramaterize, and some of the challenges of ocean and sea ice modeling.

Ocean Models

The CESM2 ocean component is called POP (Parallel Ocean Program). The next version of CESM (CESM3) will use a different ocean model called MOM (Modular Ocean Model). Regardless of the exact model used, the fundamentals of ocean modeling are similar and consist of dynamics and physics. They may also consist of biological and chemical processes as well, but we will not discuss this part.

Dynamics

Parameterized Physics

Some Challenges in Ocean Modeling

Key Points

Sea Ice Modeling

Overview

Teaching: min
Exercises: min
Questions

  • How Does Sea Ice Modeling work?

Objectives

Why is it important to model sea ice?

Sea Ice Models

The CESM2 sea ice component is called Community Ice CodE (CICE). Sea ice models have two primary components, some models have a 3rd component:

Dynamics

Force balance to determine the motion of the sea ice Wind stress, water stress, internal ice stress, Coriolis, and stress associated with sea surface slope Resistant to converence and shear

Thermodynamics and Radiative Properties

Ice Thickness Distribution

Sub-gridscale parameterization to represent the spatial hterogeneity in ice

Key Points

CESM G compset

Overview

Teaching: min
Exercises: min
Questions

  • How to run Ocean and Ice models?

Objectives

We will setup and run 4 ocean/sea-ice experiments to learn how to work with these components.

Ocean/ice active compsets start with G.

Just like the F compsets needed SST data as the boundary forcing, G compsets require boundary forcing from the atmosphere. The standard atmospheric forcing provided to a G compset is called CORE (Coordinated Ocean-ice Reference Experiments), version 2, atmopsheric datasets (Large and Yeager 2009)

Our Experiments

Control Case: default settings, 1 year__

Setup, build, and run a control case for 1 year. Use the following:

Overflow Parameterization Case

Create a case exactly the same as the control case, but with the overflow parameterization turned off.

This means we will not attempt to parameterize the mixing associated with dense water sinking in the Denmark Strait, Faroe Bank Channel, and Weddell and Ross Seas.

Modifying parameterizations is done in the namelists.

After we build the model or run ./preview_namelists, the namelists that the model uses are located in:

/glade/scratch/USERNAME/CASENAME/run/xxx_in

Remember, when we want to make a change to a namelists, we use user_nl_xxx to override the default behavior. We can always run ./preview_namelists to check our namelists get resolved properly.

To create an exact replica of another case, you can use create_clone rather than create_newcase, as follows, replacing NEWCASE and OLDCASE with the correct case names:

./create_clone --case ~/cases/NEWCASE --clone ~/cases/OLDCASE

Setup and build the case.

Modify the user_nl_pop namelist as follows:

overflows_on = .false.

overflows_interactive = .false.

Run the case for 1 year.

Where do you expect the overflow parameterization to matter?

What variables would you look at in your output?

Change the snow albedo on sea ice case

Create a clone of your control simulation.

Modify user_nl_cice as follows:

r_snw = 2.00

This is a tuning parameter that specifies the number of standard deviations away from the base optical properties in the shortwave radiative transfer code. The default value is -2.00. It is used in the equation: rsnw_nonmelt = 500 - r_snw * 250 (in microns).

Higher values of r_snw -> lower rsnw_nonmelt -> higher albedo.

Run the model for 1-year.

What do you think will happen by increasing the albedo of the ice?

How can you check?

Some variables to look at:

albedo to convince yourself this worked as expected

ice fraction to see how changing albedo impacted the ice

Increase the zonal wind stress in the ocean case

Create a clone of your control simulation Run case.setup

This exercise demonstrates how to change the source code (i.e. Fortran) of the model. The original source code resides in:

/glade/work/USERNAME/cesm2.1.3/

Similar to namelists, we don’t change the original, we make changes in an alternate location, then modify it to override the default. Changes to source code go in SourceMods/src.xxx

Also, source code changes must be included in the build, so we make the changes before building the model.

Copy the file /glade/work/USERNAME/cesm2.1.3/components/pop/source/forcing_coupled.F90 to CASEDIR/SourceMods/src.pop/

cp /glade/work/USERNAME/cesm2.1.3/components/pop/source/forcing_coupled.F90 ~/cases/gwindstress/SourceMods/src.pop/

Now let’s look at the code.

Find the subroutine rotate_wind_stress

This code rotates the wind stress vector to map it onto the correct location in the model local grid coordinates. We can think of the first component (array index 1) as the zonal direction (on the model grid) and the second component (array index 2) as the meridional component on the model local grid coordinates.
Add a line of code to increase the first component by 25%.

Build and run the model for 1-year.

Be sure to check that the model built successfully before you run it. If you had a typo, you will likely get a build error and your model will not run because it did not build.

Where do you think there will be the biggest impacts of increased zonal wind stress?

How can you check the impact?

What variables would you look at?

Looking at our output

Once you have submitted your model runs, they will run relatively quickly if the queue is not too busy. In the meantime, you can look at the output of my model runs to explore the results of these experiments.

Experiment 1 (control): /glade/scratch/cstan/archive/gcontrol/ocn/hist/

Experiment 2 (overflow): /glade/scratch/cstan/archive/goverflow/ocn/hist/

Experiment 3 (snow albedo): /glade/scratch/cstan/archive/gsnowalb/ice/hist

Experiment 4 (wind stress): /glade/scratch/cstan/archive/gwindstress/ocn/hist

How would we read this data?

To launch the Jupyter notebook on the NCAR computers

  1. Log in to the Production NCAR JupyterHub
  2. Start a server
import xarray as xr

path='/glade/scratch/cstan/archive/gcontrol/ocn/hist/'
files='gcontrol.pop.h.0001-*.nc'
ds=xr.open_mfdataset(path+files,combine='nested',
                    concat_dim='time')
ds

We can look at annual mean values by taking the mean in time.

ds_amean=ds.mean(dim='time')

Key Points