Nonlinear time series analysis of coastal temperatures and El Niño–Southern Oscillation events in the eastern South Pacific

Abstract. We carry out a nonlinear time series analysis motivated by dynamical systems theory to investigate the links between temperatures on the eastern South Pacific coast, influenced by the Humboldt Current System, and El Niño–Southern Oscillation (ENSO) events. To this aim, we use a set of 16 oceanic and atmospheric temperature time series from Chilean coastal stations distributed between 18 and 45∘ S. The spectral analysis indicates periodicities that can be related to both internal and external forcing, involving not only ENSO, but also the Pacific Decadal Oscillation, the Southern Annual Mode, the Quasi-Biennial Oscillation and the lunar nodal cycle. The asymptotic neural network test for chaos based on the largest global Lyapunov exponent indicates that the temperature dynamics along the Chilean coast is not chaotic. We use local Lyapunov exponents to characterize the short-term stability of the series. Using a cross-entropy test, we find that two stations in northern Chile, one oceanic (Iquique) and one atmospheric (Arica), present a significant positive cross-dependence between local Lyapunov exponents and ENSO. Iquique is the station that presents the greater number of regional characteristics and correlates with ENSO differently from the rest. The unique large-scale study area, combined with time series from hitherto unused sources (Chilean naval records), reveals the nonlinear dynamics of climate variability in Chile.