stratospheric-transport

Proof-of-concept: Physics-informed neural network for stratospheric transport

General description

This repository provides a neural network for stratospheric transport studies. The model is designed to predict residence time from age-of-air observations using physics-informed or supervised boundaries. Data workflows for retrieval of renanalysis tropopause-derived features or retrieval of age-of-air training data are part of the repository. Three approaches to learn time patterns (seasonal, multi-annual, decadal) can be selected in addition.