Benblob688/SWIFT_RF_Rainfall

Scripts and notebooks to tune, train, verify and apply random forest models for MSG-based rainfall retrieval over Africa. README.md # SWIFT_RF_Rainfall Python scripts to tune, train, verify and apply random forest (RF) models for MSG-based rainfall retrieval over Africa. [insert image of final published product output working here] # Prerequisites This package has been developed and tested entirely on the JASMIN compute cluster, within the SWIFT Group Workspace (GWS). If operating outside of the SWIFT GWS on JASMIN, there are a number of other dependancies that are required for these scripts to function. A Python environment file from `conda` is included which is not OS-specific. Setting up a new Python environment for this package is recommended. If you already recieve real-time SEVIRI images or NWCSAF CRR products through a satellite dish, you will have these dependencies already. Completely new users of NWCSAF products will have to set these up and gain a license from EUMETSAT. - Spinning Enhanced Visible and InfraRed Imager (SEVIRI) satellite files in `HRIT` format. These can be recieved directly with a satellite dish (see Roberts et al. 2021) or by downloading them from EUMETSEAT under "By format > HRIT" where the format is explained (a file for each channel plus EPI and PRO files demarking the start and end of an observing window) as well as links to "Get Access". Either method requires a login/license with EUMETSAT. - xRITDecompress should be installed to be able to decompress the `HRIT` files. SEVIRI data in NetCDF and other formats could be supported, but the scripts given here must then be modified by the user. At some point the satellite images are all in `numpy` arrays which is a common overlap point in the scripts for anyone who wishes to modify the scripts to support other data formats. # Summary There are 6 steps to creating a custom version of the algorithm published in Pickering et al., 2022. | | Step | _Description_ | | -- | ---------------- | ------------- | | 0. | Model Parameters | _high-level settings used commonly throughout the process_ | …