Forecasting ENSO with a smooth transition autoregressive model

Abstract

This study examines the benefits of nonlinear time series modelling to improve forecast accuracy of the El Niño Southern Oscillation (ENSO) phenomenon. The paper adopts a smooth transition autoregressive (STAR) modelling framework to assess the potentially regime-dependent dynamics of sea surface temperature anomaly. The results reveal STAR-type nonlinearities in ENSO dynamics, resulting in superior out-of-sample forecast performance of STAR over the linear autoregressive models. The advantage of nonlinear models is especially apparent in the short-and intermediate-term forecasts. These results are of interest to researchers and policy makers in the fields of climate dynamics, agricultural production, and environmental management.

A novel approach for discovering stochastic models behind data applied to El Niño–Southern Oscillation

“…Such re-sampling aims to preserve the overall pdf of and , while erasing all self-correlation and cross-correlation at all lags. An independent noise assumption is commonly employed in physically-based simple ENSO models 4 , 28 , 39 . We perform a forward integration of the model for 100,000 monthly time steps starting from neutral conditions in January and using Euler’s method.…”

Bispectral analysis of nonlinear interaction, predictability and stochastic modelling with application to ENSO

“…A number of stochastic univariate models of ENSO have been fitted such as: smooth transition autoregressive (STAR) models (Hall et al, 2001;Ubilava and Helmers, 2013), autoregressive conditional heteroscedasticity type models (ARCH) (Ahn and Kim, 2005), and threshold AR models (De Gooijer, 2017). This section aims at fitting a minimal univariate stochastic model for El Niño 3.4. index driven by a multiplicative Gaussian delayed noise, able to reproduce the observed empirical spectrum, skewness and bispectrum as well as assess the impact it has on predictability, compared to benchmark linear models.…”

Chaotic signature of climate extremes

“…Rejection of chaos in the time series of ENSO has been disputed based on analysis carried out using Largest Lyapunov Exponent (Kawamura et al 1998), Correlation Dimension (Kawamura et al 1998), close return plots (Ahn and Kim 2005) and determinism (Binder and Wilches 2012). The theory that ENSO time series is a stochastic system rather than a chaotic one has led to the development of stochastic approach for ENSO (Ubilava and Helmers 2013;Hall et al 2001). Several studies have identify chaos in the dynamics of ENSO using tools such as bred vector (Tang and Deng 2010), dynamical models (Vallis 1986;Chang et al 1996;Tziperman et al 1994), false nearest neighbour and correlation dimension (Chang et al 1996;Tsonis 2009), Lyapunov Exponent (Tsonis 2009) and nonlinear prediction error (Elsner and Tsonis 1993).…”