Meteorological plotting utilities
Meteorological data are often stored in various formats (GRIB, netcdf, Zarr...), with various tools to read them. Sometimes these tools include plotting functions (e.g. EPyGrAM), sometimes they don't (e.g. cfgrib)
Metplotlib aims to be source-agnostic: you load the data the way you prefer, metplotlib does the plots
Many of the existing tools are specific to an institution or a computing platform. Some dependencies may not be public, or the installation may complex, sometimes requiring super-user rights.
Metplotlib aims to be based on a limited set of widely used dependencies, namely: Numpy, Matplotlib, Cartopy.
With uv (recommended):
uv add metplotlib@git+https://github.com/ThomasRieutord/metplotlibWith pip:
pip install metplotlib@git+https://github.com/ThomasRieutord/metplotlibIf you wish to contribute, please see the guidelines in the Contributing section.
Once you have cloned the metplotlib package locally, activate your favorite environment, go to the package root directory (where there is the pyproject.toml) and run
git clone [email protected]:ThomasRieutord/metplotlib.git
cd metplotlib
pip install -e .The environment must be activated each time you want to use the package. To check the installation:
python tests/import_test.py
python -i tests/two_vars_test.pyFor example, to make a plot with the mean sea level pressure isolines over 2-m temperature color levels:
from metplotlib import plots
mslp, t2m, lon, lat = ...replace with your favorite way of loading data...
fig, ax = plots.twovar_plot(mslp, t2m, lons=lon, lats=lat, cl_varfamily = "temp")
fig.show()
Same example with wind speed instead of 2-m temperature:
from metplotlib import plots
mslp, wind, lon, lat = ...replace with your favorite way of loading data...
fig, ax = plots.twovar_plot(mslp, wind, lons=lon, lats=lat, cl_varfamily = "wind")
fig.show()
One more example, for ensemble forecasts at a specific location represented by quantiles time series:
from metplotlib import plots
temp = ...ndarray of shape (n_mbr, n_ldt)...
leadtimes = ...ndarray of shape (n_ldt,)...
fig, ax = plots.quantiles(temp, x=leadtimes, xlabel="Leadtime (hours)", ylabel="2m temperature", title = f"Forecast ...")
fig.show()
To compare two variables in two different states:
from metplotlib import plots
mslp0, mslp1, wind0, wind1, lon, lat = ...replace with your favorite way of loading data...
fig, ax = plots.twovar_comparison(mslp0, mslp1, wind0, wind1, lons=lon, lats=lat)
fig.show()
To plot data into a different projections than that initially provided:
import numpy as np
import cartopy.crs as ccrs
from metplotlib import plots
lambert_proj_params = {
"a": 6367470,
"b": 6367470,
"lat_0": 53.5,
"lat_1": 53.5,
"lat_2": 53.5,
"lon_0": 5.0,
"proj": "lcc",
}
# Figure CRS: what the figure will look like
fig_crs = ccrs.LambertConformal(
central_longitude=lambert_proj_params["lon_0"],
central_latitude=lambert_proj_params["lat_0"],
standard_parallels=(lambert_proj_params["lat_1"], lambert_proj_params["lat_2"]),
)
# Data CRS: how the data is stored
data_crs = ccrs.PlateCarree()
nx, ny = (245, 265)
lon, lat = np.meshgrid(
np.linspace(-20, 3, nx),
np.linspace(45, 60, ny),
)
t2m = 30 * np.cos(np.pi*lat/180) + np.sin(20*np.pi*lon/180)
mslp = 1015 + 10*(np.sin(20*np.pi*lon/180) + np.cos(25*np.pi*lat/180))
fig, ax = plots.twovar_plot(mslp, t2m, lons=lon, lats=lat, cl_varfamily = "temp", figcrs=fig_crs, datcrs = data_crs)
fig.show()
...still to be done...
Contributions are welcome! Please open issues and merge requests with your contributions. Contributors are asked to keep their code clean and easy to read.
We recommend the use of uv as package manager. The installation in dev mode is as follows:
git clone [email protected]:ThomasRieutord/metplotlib.git
cd metplotlib
uv syncMake sure you include tests in your contribution. At least provide instructions in your PR on how to review your contributions.
All tests written in the tests folder must pass without error.
To launch them locally, use the following:
uv run pytest tests/.We recommend to use ruff as linter and formatter. It is already included in Metplotlib's optional dependencies. To lint and format your code, just run the following commands (and fix whatever is needed):
uv run ruff format .
uv run ruff check .MIT licence