Defining error accumulation in ML atmospheric simulators

TL;DR

This paper introduces a clear definition and metric for error accumulation in ML atmospheric simulators, distinguishing between fixable model errors and inherent system chaos, and demonstrates improved forecasting performance.

Contribution

It provides the first formal definition and measurement of error accumulation in autoregressive ML atmospheric models, along with a regularization method to mitigate it.

Findings

Improved RMSE and spread/skill metrics in atmospheric forecasts.

Regularization based on the new error accumulation definition enhances model performance.

Demonstrated effectiveness on real-world weather prediction tasks.

Abstract

Machine learning (ML) has recently shown significant promise in modelling atmospheric systems, such as the weather. Many of these ML models are autoregressive, and error accumulation in their forecasts is a key problem. However, there is no clear definition of what `error accumulation' actually entails. In this paper, we propose a definition and an associated metric to measure it. Our definition distinguishes between errors which are due to model deficiencies, which we may hope to fix, and those due to the intrinsic properties of atmospheric systems (chaos, unobserved variables), which are not fixable. We illustrate the usefulness of this definition by proposing a simple regularization loss penalty inspired by it. This approach shows performance improvements (according to RMSE and spread/skill) in a selection of atmospheric systems, including the real-world weather prediction task.