This repository contains the code for the paper "Distribution-Guided Ensemble Postprocessing for S2S Precipitation Forecasts: A Seamless Pathway using Deep Generative Models". The code is published for open research purposes.
The proposed model's key training code is located in scripts/3.train/plmodules.py:
- CvaeGanModule_v2: Our proposed deep generative model for distribution-guided ensemble postprocessing
- DetermModule: Used for MOS (Model Output Statistics) baseline experiments
Key requirement: PyTorch Lightning version 1.9.5
Other packages (PyTorch, NumPy, xarray, etc.) can be installed with standard versions without specific requirements.