12-month El Niño forecast on elninoprediction.nl

Peter Nooteboom, UU, CSS
Peter Nooteboom

The Centre for Complex Systems Studies has recently launched the website elninoprediction.nl. The website presents El Niño forecasts for up to one year in advance. This is a major improvement on the conventional predictions, which are only reliable for up to six months in the future. The prediction model was developed by the Centre’s members Peter Nooteboom, Qingyi Feng and Prof. Henk Dijkstra, in collaboration with the complexity institute in Mallorca. The predictions on the website are based on complexity research combining network theory and machine learning.

Anticipating on natural disasters

During an El Niño event, sea surface temperature increase in the Eastern Pacific. This occurs once every 4 to 6 years during a Northern Hemisphere winter, and is associated with natural disasters such as droughts in Africa or floods in California. A good El Niño prediction gives us an opportunity to anticipate for the possible disasters that are associated with these higher sea surface temperatures. Predictions far in advance of an event could save lives and reduce the economic costs of the disasters.

Network theory and machine learning

So far, El Niño has been accurately predicted up to six months in advance (also known as the ‘spring-predictability barrier’). The new prediction model makes use of a machine learning approach. This is an artificial neural network which recognises patterns in the system at an early stage that could lead to an El Niño event a year later. Network theory is used to feed the right information into the ‘machine’. For the winter of 2017/2018, the one-year prediction was quite good: it forecast a strong La Niña event (the opposite of El Niño), and a year later a moderate La Niña event did indeed occur.

More information about the prediction method on www.elninoprediction.nl

Publication

Nooteboom, P. D., Feng, Q. Y., López, C., Hernández-García, E., and Dijkstra, H. A.: Using network theory and machine learning to predict El Niño, Earth Syst. Dynam., 9, 969-983, https://doi.org/10.5194/esd-9-969-2018, 2018.