cubedsphere 0.1

Library for post processing of MITgcm cubed sphere data

cubedsphere

Library for post processing of MITgcm cubed sphere data

Capabilities:

• regrid cubed sphere datasets using xESMF and xgcm
• open datasets created by the mnc package
• open datasets using xmitgcm (needs current PR #98)
• plot original cubed sphere data
• some more small utilities
• more to come...

Note:

Work in progress! This library is a collection of tools that I found useful to use for the interpretation of cubed sphere data.

Installation:

Clone this repo:

git clone https://github.com/AaronDavidSchneider/cubedsphere.git
cd cubedsphere

Create conda environment:

conda create -n mitgcm

Activate environment:

conda activate mitgcm

Install dependencies:

conda install -c conda-forge xesmf esmpy xgcm matplotlib

Install cubedsphere:

pip install -e .

You can now import the cubedsphere package from everywhere on your system

Example Usage

See examples/example.py. The following plots have been created using data from tutorial_held_suarez_cs.

import matplotlib.pyplot as plt
import cubedsphere as cs
import cubedsphere.const as c

# Specify directory where the output files can be found
outdir = "/Volumes/SCRATCH/sim_output/xmitgcm_test/nc_test"

# open Dataset
ds = cs.open_mnc_dataset(outdir, 276480)

# regrid dataset
regridder = cs.Regridder(ds, 5, 4, reuse_weights=False, filename="weights", concat_mode=False)
# Note: once weights were created, we can also reuse files by using reuse_weights=True (saves time).
ds_reg = regridder.regrid()

Only takes few seconds!

# do some basic plotting to demonstrate the dataset
# determine which timestep and Z to use:
isel_dict = {c.time:0, c.Z:0}

# do some basic plotting to demonstrate the dataset
fig = plt.figure(figsize=(8,6), constrained_layout=True)
ds_reg[c.T].isel(**isel_dict).plot(vmin=260,vmax=312, add_colorbar=False)
U, V = ds_reg["U"].isel(**isel_dict).values, ds_reg["V"].isel(**isel_dict).values
cs.overplot_wind(ds_reg, U, V)
plt.gca().set_aspect('equal')
plt.savefig("../docs/temp_reg.png")
plt.show()

# Now also plotting theta without regridding (on the original grid):
fig = plt.figure(figsize=(8,6), constrained_layout=True)
cs.plotCS(ds[c.T].isel(**isel_dict), ds, mask_size=5, vmin=260, vmax=312)
plt.gca().set_aspect('equal')
plt.savefig("../docs/temp_direct.png")
plt.show()

Tests with xmitgcm:

See examples/example_xmitgcm.py

Since we do not have the grid information, we need to fallback to the nearest_s2d interpolation with concat mode. Concat mode means that instead of regridding each face individually and summing up the results, we first concatenate the ds along the X dimension and regrid on the flattened dataset afterwards.

We can fix this and keep using conservative regridding by using the grid files from the mnc dataset:

In the future we will use xmitgcm.utils.get_grid_from_input function instead.

ToDo:

Postprocessing:

• interface xmitgcm to enable the use of .meta and .data files -> added wrapper
• how do we expand lon_b and lat_b from left to outer for xmitgcm wrapper? -> either nc file or soon with xmitgcm.utils.get_grid_from_input

Testing:

• compare results with matlab scripts

Interface:

• which values should be hardcoded? -> done in const.py
• special tools needed for exorad?

future Ideas:

Regridding tools:

• use ESMPy as an alternative to xESMF (requires 6 processors)

Credits

Many of the methods come from: https://github.com/JiaweiZhuang/cubedsphere

I would especially like to thank @rabernat for providing xgcm and @JiaweiZhuang for providing xESMF.