Dynamics of real-time forecasting failure and recovery due to data gaps

TL;DR

This paper investigates how data gaps impact real-time forecasting systems, revealing that restart time is crucial and that accuracy does not fully recover even after data is restored, based on a Lorenz model study.

Contribution

It provides the first systematic analysis of forecasting failure and recovery dynamics caused by data gaps using a Lorenz model-based system.

Findings

Restart time significantly affects recovery speed.

Forecast accuracy does not fully recover after data gaps.

Data gaps can cause persistent degradation in forecast quality.

Abstract

Real-time forecasting is important to the society. It uses continuous data streams to update forecasts for sustained accuracy. But the data source is vulnerable to attacks or accidents and the dynamics of forecasting failure and recovery due to data gaps is poorly understood. As the first systematic study, a Lorenz model-based forecasting system was disrupted with data gaps of various lengths and timing. The restart time of data assimilation is found to be the most important factor. The forecasting accuracy is found not returning to the original even long after the data assimilation recovery.