Spatiotemporal Pattern Extraction with Data-Driven Koopman Operators for Convectively Coupled Equatorial Waves

Joanna Slawinska, Rutgers University

We study spatiotemporal patterns of convective organization using a recently developed technique for feature extraction and mode decomposition of spatiotemporal data generated by ergodic dynamical systems. The method relies on constructing low-dimensional representations (feature maps) of spatiotemporal signals using eigenfunctions of the Koopman operator governing the evolution of observables in ergodic dynamical systems. This operator is estimated from time-ordered data through a Galerkin scheme applied to basis functions computed via the diffusion maps algorithm. We apply this technique to brightness temperature data from the CLAUS archive and extract a multiscale hierarchy of spatiotemporal patterns on timescales spanning years to days. In particular, we detect for the first time without prefiltering the input data traveling waves on temporal and spatial scales characteristic of convectively coupled equatorial waves (CCEWs).We discuss the salient properties of waves in this hierarchy and find that the activity of certain types of CCEWs is modulated by lower-frequency signals.