This repository contains the machine learning code used in the paper: Exploiting radar polarimetry for nowcasting thunderstorm hazards using deep learning, submitted to Natural Hazards and Earth System Sciences.

Installation

You need NumPy, Scipy, Matplotlib, Seaborn, Tensorflow (2.6 used in development), Numba, Dask and NetCDF4 for Python.

Clone the repository, then, in the main directory, run

$ python setup.py develop

(if you plan to modify the code) or

$ python setup.py install

if you just want to run it.

Downloading data

The dataset for the polarimetric variables and quality indices, pretrained models and results can be found at the following Zenodo repository: https://doi.org/10.5281/zenodo.7760740. Follow the instructions there on where to place the data. The dataset for radar can be found at the following Zenodo repository: https://zenodo.org/record/6802292

Pretrained models

The pretrained models are available at https://doi.org/10.5281/zenodo.7760740. Unzip the files models-*.zip found there to the runs/run* directory and results.zip to the runs/run*/results directory..

Running

Go to the scripts directory and start an interactive shell. There, you can find training.py that contains the script you need for training and plots_sources.py that produces the plots from the paper.