Optimization of a Causal Decomposition Algorithm

Abstract

Empirical Mode Decomposition (EMD) is a data-driven signal analysis technique. Its wide acceptance has led to diverse expansions, improvements and applications. The so-called Causal Decomposition (CD) application establishes and quantifies causality between signals, along their different oscillation modes. In the present work, we present our own version of the decomposition algorithm, improving features such as the alignment between the respective signal modes and the precision of the final result. In order to quantify the improvement rate, the new algorithm is applied to the same series studied with previous versions of the method. Those series correspond to the El Niño – Southern Oscillation (ENSO) phenomenon, precipitation in the cities of Buenos Aires (Argentina) and Junín (Buenos Aires, Argentina). The causality detected is, in many cases, indicative of the presence of drivers that underlie the data.