The project aims to create an open-source weather model using data from Météo-France and other sources like ERA5. The project focuses on downloading and processing various weather datasets. The overall goal is to build a dataset that can be used for machine learning to predict the weather.
All code element of the data part are in the meteodata folder.
The first element of the project will be to handle the different data sources. We currently have mainly 3 sources of data:
-
radar data coming from meteo france API (BUFR files and HDF5 files) Those are images files that we can convert to numpy arrays
-
Ground station data coming from meteo france API (CSV files) Those are CSV files that we can convert to pandas dataframes
-
ERA5 data coming from ECMWF API (NetCDF files) Those are NetCDF files that we can convert to xarray dataarrays
-
(POSSIBLY) satellite data coming from EUMETSAT API (NetCDF files) But currently I'm not sure of the availability of this data
The key idea is to use machine learning to predict the weather a few hours / days in advance. So we need to download the data every day to create a database (no existant yet).