Distribution-Informed Prediction

Code for calibrating geophysical predictions under constrained probabilistic distributions.

Project layout

• pyqg/: Online hybrid emulator of quasi-geostrophic turbulence; generate data with data.py, train with train_neural_network.py, run simulations with cali_online.py, then inspect pyqg_result.ipynb.
• toy_example_pco2/: Toy example and air–sea pCO2 application; see notebook flow and data notes in its README.

Environment

• Python 3.10–3.11
• Install dependencies: pip install -r requirements.txt
• GPU is recommended for training-heavy scripts if available.

Getting started

1. Create/activate a virtual environment.
2. Install requirements (above).
3. Follow the per-folder README instructions to prepare data, run training, and reproduce figures/tables.