Pythie is a Python package to postprocess weather forecasts. Presently, it contains the modules:
-
MBM: a module to make member-by-member ensemble forecasts postprocessing based on past (re)forecasts. It implements the methods described in:
- Bert Van Schaeybroeck and Stéphane Vannitsem. Ensemble post-processing using member-by-member approaches: theoretical aspects. Quarterly Journal of the Royal Meteorological Society, 141 (688):807–818, 2015. [link].
-
To be continued...
The present release v0.1alpha of Pythie is a preliminary version for the v0.1 release. The release of a Python package is planned for the v0.1 release.
Pythie is presently compatible with Linux and Mac OS.
It is not compatible with Windows for the moment, but a Windows compatible version will be released soon.
The easiest way to run Pythie for the moment is to use an appropriate environment created through Anaconda.
First install Anaconda and clone the repository:
git clone https://github.com/Climdyn/pythie.git
Then install and activate the Python3 Anaconda environment:
conda env create -f environment.yml
conda activate pythie
The latter instruction will install a minimal environment for Pythie that is needed for the postprocessing to take place. If you want to benefit from the advanced functionalities (plotting, notebooks, etc...), you have to install the extra packages:
conda env update -n pythie -f extra_environment.ymlPlease note that these extra packages are also needed to build the documentation.
To build the documentation, please run (with the conda environment activated):
cd documentation
make html
You may need to install make if it is not already present on your system. Once built, the documentation is available here.
The documentation is also available online on read the docs: https://pythie.readthedocs.io
You may also want to run the tests provided inside the documentation:
make doctest
In case of a test failure, please report it on the issue page of the Pythie GitHub repository.
Warning: You need the extra packages to be installed in order to be able to build the documentation.
Examples are provided in the References pages of the documentation, and through the use of Jupyter notebooks. The latter can be found in the notebooks folder.
These notebooks needs the download of a dataset stored on Zenodo [link to the data]. This dataset can be downloaded by running the following script:
bash -i download_sample_data.sh
in the root folder of Pythie. The data are placed by the download script in the ./data folder.
From there, running
conda activate pythie
cd notebooks
jupyter-notebook
will lead you to your favorite browser where you can load and run the examples.
Warnings: Beware of your internet connection usage! The size of the example dataset is roughly 2 gigabytes!
Pythie needs mainly:
Check the yaml file environment.yml for the full list of dependencies.
For various usages, Pythie can optionally need:
- IPython
- Jupyter for notebooks
- matplotlib for plotting
- iris to manipulate NetCDF files
- pandas to manipulate CSV and dataframes
- ffmpeg to make movies and gif ;-)
Check the yaml file extra_environment.yml for the full list of extra dependencies.
- Better interfacing with libraries such as: pandas, iris, xarray.
- Release of a Python package for the version v0.1
- A more involved test framework
Writing better interfaces to other libraries is a priority so if you think you can help us, please contact the main authors.
In addition, if you have made changes that you think will be useful to others, please feel free to suggest these as a pull request on the Pythie GitHub repository.
A review of your pull request will follow with possibly suggestions of changes before merging it in the master branch. Please consider the following guidelines before submitting:
- Before submitting a pull request, double check that the branch to be merged contains only changes you wish to add to the master branch. This will save time in reviewing the code.
- Please document the new functionalities in the documentation. Code addition without documentation addition will not be accepted. The documentation is done with sphinx and follows the Numpy conventions. Please take a look to the actual code to get an idea about how to document the code.
- The team presently maintaining Pythie is not working full-time on it, so please be patient as the review of the submission may take some time.
Articles and newsletters related to Pythie:
- Demaeyer, J., Vannitsem, S. & B. Van Schaeybroeck, 2021: Statistical post-processing of ensemble forecasts at the Belgian met service, ECMWF Newsletter No. 166, 21–25. [link]
- Vannitsem, S. & J. Demaeyer, 2020: Statistical postprocessing of ECMWF forecasts at the Belgian met service, ECMWF Newsletter No. 164, 4–5. [link]
The Pythie developers would like to thank the following people:
- Bert Van Schaeybroeck from RMIB for his help in debugging parts of the code.
- Susanne Theis from DWD for communicating us the existence of the CDC portal.