Data-Driven Prediction of Thresholded Time Series of Rainfall and SOC models

TL;DR

This paper compares the predictability of thresholded events in a SOC sandpile model and rainfall data, revealing how scaling behavior affects prediction accuracy and ROC curve characteristics.

Contribution

It introduces a scaling theory for event predictability in SOC models and analyzes differences in rainfall data, advancing understanding of extreme event prediction.

Findings

Scaling of quiet-time distributions enhances predictability in SOC models.

ROC curves can be extrapolated for extreme events in SOC systems.

Rainfall data shows non-scaling behavior, complicating prediction for high thresholds.

Abstract

We study the occurrence of events, subject to threshold, in a representative SOC sandpile model and in high-resolution rainfall data. The predictability in both systems is analyzed by means of a decision variable sensitive to event clustering, and the quality of the predictions is evaluated by the receiver operating characteristics (ROC) method. In the case of the SOC sandpile model, the scaling of quiet-time distributions with increasing threshold leads to increased predictability of extreme events. A scaling theory allows us to understand all the details of the prediction procedure and to extrapolate the shape of the ROC curves for the most extreme events. For rainfall data, the quiet-time distributions do not scale for high thresholds, which means that the corresponding ROC curves cannot be straightforwardly related to those for lower thresholds.