Hurricane Patricia in 2015 was given a boost by El Ni単o UPI / Alamy Stock Photo
Artificial intelligence is learning how to predict El Ni単o climate cycles. The hope is that the technology could be used to improve climate predictions and give policy-makers more time to prepare.
El Ni単o can cause severe weather and devastating damage. A phase of the El Ni単o-Southern Oscillation, it occurs when water warms over the tropical Pacific Ocean, shifting east and increasing rainfall and cyclones over the Americas while pulling rain away from Indonesia and Australia.
Strong El Ni単o events are associated with intense storms and flooding in some areas, and drought and fires in others.
Advertisement
Yoo-Geun Ham at Chonnam National University in South Korea and his colleagues built an AI that can predict the development of an El Ni単o system up to 18 months before it occurs.
They trained the system using historical global oceanic temperature data from 1871 to 1973, as well as nearly 3000 simulations of El Ni単o events between 1961 and 2005 that had been generated by existing forecasting models. The data included information about the surface sea temperature, as well as ocean heat content the average heat in the top 300 metres of the ocean.
Ham and his colleagues tested the AI on data from 1984 to 2017 and found that it was more accurate than existing climate forecasts, which can only reliably predict events up to a year in advance.
The AI correctly predicted 24 out of 34 climate events as either El Ni単o, La Ni単a or normal 18 months ahead of time, says Ham. In comparison, a leading existing model predicted fewer than 20 correctly.
The algorithm also learned to recognise precursors to an El Ni単o event, such as the Indian Ocean dipole, an irregular warming and then cooling of the eastern part of the Indian Ocean.
Nature,
Topics:
