Abstract

We present a novel approach of analysing–visualising time series of a geophysical variable and we characterise its abrupt transitions in comparison to benchmark time series produced with model dynamical systems: a mathematical model (stochastic resonance) and a climate model of intermediate complexity (2D meridional ocean circulation with an atmospheric forcing). The method combines a genetic segmentation algorithm that uses ordinal regression and clusters the different segments of the time series around centroids located in a six-dimensional(6D) space of statistical metrics. After detecting statistical similarities it helps compare the type of transition observed in the time series to three separate studied types: a) noise transition, b) subcritical bifurcation crossing and c) transition to a limit cycle. The proposed method complements the causality analysis of a record of abrupt transitions in a geophysical system.