The pySTEPS initiative is a community that develops and maintains an easy to use, modular, free and open source python framework for short-term ensemble prediction systems.
The focus is on probabilistic nowcasting of radar precipitation fields, but pySTEPS is designed to allow a wider range of uses.
The pySTEPS package needs the following dependencies
- numpy
- scipy
- opencv
- pillow
- pyproj
- matplotlib (for examples)
Additionally, the following packages can be installed to provide better computational efficiency:
- dask and toolz (for code parallelisation)
- pyfftw (for faster FFT computations)
We recommend that you create a conda environment using the available environment.yml file to install all the necessary dependencies:
conda env create -f environment.yml
The latest version can be installed manually by downloading the sources using:
git clone https://github.com/pySTEPS/pysteps
To install using pip run:
pip install ./pysteps
Or, to install it using setup.py run (global installation):
python setup.py install
For user installation:
python setup.py install --user
If you want to install the package in a specific directory run:
python setup.py install --prefix=/path/to/local/dir
IMPORTANT: All the dependencies need to be already installed!