Climatological Analysis: A Fast Non-linear Signal Filtering Approach

Abstract

This paper proposes a new approach to detect, model and analyze climatological signals solely based on the largest eigenvalue of a signal derived matrix, used as a measure of innovation. The resulting process offers a substantial reduction in computational complexity with respect to presently applied methods. Furthermore, the approach yields much precise and detailed result. A case study illustrates the application of the proposed approach to the identification and analysis of a Nino event, tracing the interactions between different regions in the planet and pointing out perturbation paths otherwise unknown, allowing the prediction of the phenomena months before the climatological development.