Calibrated Forecasts of Quasi-Periodic Climate Processes with Deep Echo State Networks and Penalized Quantile Regression

TL;DR

This paper demonstrates how deep echo state networks combined with penalized quantile regression can improve the forecasting and uncertainty calibration of quasi-periodic, ocean-driven climate processes, which are challenging for physical models.

Contribution

It introduces a data-driven approach using deep echo state networks and ensemble-based calibration to enhance predictability and uncertainty quantification of complex climate events.

Findings

Effective forecasting of multi-year ocean-driven climate events.

Calibrated uncertainty estimates through ensemble methods.

Applicable to various environmental prediction problems.

Abstract

Among the most relevant processes in the Earth system for human habitability are quasi-periodic, ocean-driven multi-year events whose dynamics are currently incompletely characterized by physical models, and hence poorly predictable. This work aims at showing how 1) data-driven, stochastic machine learning approaches provide an affordable yet flexible means to forecast these processes; 2) the associated uncertainty can be properly calibrated with fast ensemble-based approaches. While the methodology introduced and discussed in this work pertains to synoptic scale events, the principle of augmenting incomplete or highly sensitive physical systems with data-driven models to improve predictability is far more general and can be extended to environmental problems of any scale in time or space.