This repo contains the codes of the Liquid Water Path (LWP) and Integrated Water Vapor (IWV, sometimes also refered to as "Precipitable Water Vapor" or PWV) retrieval presented in Billault-Roux and Berne, 2021 (https://doi.org/10.5194/amt-14-2749-2021).
This uses as input measurements from the cloud radar/radiometer "WProf" (Küchler et al. 2017, https://doi.org/10.1175/JTECH-D-17-0019.1), namely the 89-GHz brightness temperature (TB).
Optional inputs are information on the geographical location, surface atmospheric conditions, and reanalysis products (ERA5, https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5).
A neural network is used for this retrieval, implemented through the keras python library (Chollet 2015, https://keras.io). The requirements for running the retrieval are listed in requirements.txt (Note: more packages might be needed for fully re-training the model: this was not tested).
The user is most likely interested by the code contained in the juypter notebook implementation_lwp_retrieval.ipynb which illustrates how to implement the retrieval using real data.
The repo additionally contains the following directories and files:
- The "dataset_creation" directory contains scripts to generate the training dataset (fetch radiosonde profile, run forward model...). This uses the radiative transfer model PAMTRA (https://pamtra.readthedocs.io/)
- The "parameters" directory contains the neural network parameters + values to use for normalization (tree of subdirectories which are browsed through depending on the chosen input features).
- The "training_scripts" directory contains a few scripts that were used during training of the model.
- "download_ERA5_data.py" is a script for the download of reanalysis data, which should be adapted by the
- ICEGENESIS_Jan_ERA5_for_LWP.nc: example of ERA5 data in NetCDF format
- tools.py: a few useful functions
Questions should be directed at anne-claire.billault-roux@epfl.ch