Machine learning and data assimilation with the Lorenz '96 system
This repository contains code for investigating machine learning and data assimilation with the Lorenz '96 system, a toy weather model, inspired by Dueben and Bauer.
aicontains scripts for training a neural net to emulate the Lorenz '96 system.
assimilationcontains scripts for running data assimilation with the numerical model and the trained model.
forecastcontains scripts for comparing "weather forecasts" run with the numerical model and the trained model.
numerical_modelcontains FORTRAN90 code for running the numerical model. This is compiled with f2py so it can be called from Python.
- Set up conda environment. Run
conda activate lorenz96_machine_learning.
source setup.shfrom the root directory. This will build the numerical model and add it to the Python path.