Recovering the parameters underlying the Lorenz-96 chaotic dynamics

TL;DR

This paper investigates three deep learning algorithms to objectively infer parameters of the Lorenz-96 model, a simplified climate system, aiming to improve climate model parameterization and reduce uncertainty.

Contribution

It introduces and compares three deep network architectures for data-driven parameter inference in a chaotic climate-like model.

Findings

Convolutional networks outperform fully-connected networks in parameter recovery.

All models successfully recover underlying Lorenz-96 parameters.

Deep learning approaches can enhance climate model parameter estimation.

Abstract

Climate projections suffer from uncertain equilibrium climate sensitivity. The reason behind this uncertainty is the resolution of global climate models, which is too coarse to resolve key processes such as clouds and convection. These processes are approximated using heuristics in a process called parameterization. The selection of these parameters can be subjective, leading to significant uncertainties in the way clouds are represented in global climate models. Here, we explore three deep network algorithms to infer these parameters in an objective and data-driven way. We compare the performance of a fully-connected network, a one-dimensional and, a two-dimensional convolutional networks to recover the underlying parameters of the Lorenz-96 model, a non-linear dynamical system that has similar behavior to the climate system.