Modelling the El Nino Southern Oscillation

ENSO is a global phenomenon that occurs in two distinct phases, widely believed to be the main source of inter-annual climate variability, with the potential to cause catastrophic climatic conditions of flood and drought. The flow-on effect of these conditions are far reaching; weather sensitive industries including agriculture, commercial fishing, construction, and tourism are disrupted, making ENSO-related economic losses estimated in the billions, with the potential for damage now in the trillions. Even higher are the unquantifiable losses incurred by ENSO: human losses as well as those suffered by communities without the benefit of precautionary notices are often the hardest hit. A mechanism to provide long lead times used to develop strategies for dealing with the anticipated impacts of ENSO are necessary to mitigate the losses otherwise incurred.

This research focuses on probabilistic forecasting of the ENSO signal, using a Bayesian-based modelling and forecasting approach. Research in this area has led to the develop of a Bayesian Binary Tree (BBT) model, possessing the distinctive feature in that it is tailored to model ENSO when exhibiting two distinct phases. We have discovered that signal persistence is important when forecasting more than a few months in advance, and that using probabilistic forecasting provides a simple yet comparable alternative to the more complex dynamic forecasting models, in terms of forecast quality. Current work in this area is focused on extending the BBT to a ternary state model which incorporates three distinct phases of the ENSO phenomenon.

This research involves collaboration with an overseas partner.

Contact: Belinda Chiera

Key publications

BA Chiera, JA Filar, DS Zachary & AH Gordon, 'Comparative Forecasting and a Test for Persistence in the El Nino Southern Oscillation', Uncertainty and Environmental Decision Making, International Series in Operations Research and Management, 2010.

Areas of study and research

+ Click to minimise