In a recent publication in Nature, scientists have been looking into how artificial intelligence could forecast El Niño/Southern Oscillation, or what they call ENSO. Here, they discuss the already known information that variations in ENSO are due to a wide gradient in climate conditions, as well as in ecosystems, over various regions. The research was aimed to use the technology to improve weather forecasts, giving both the government and residents adequate time to prepare.
Researchers from Chonnam National University, South Korea claim to have programmed an AI that could predict an El Niño at most 18 months before it occurs.
They used statistical data on global oceanic temperatures from 1871 to 1973 and 3000 simulations from 1961 to 2005 generated from existing technology. The data included other information like surface sea temperature and ocean heat content.
The team tested their program to predict some occurrences from 1984 to 2017, which resulted in 24 out of 34 correct predictions-that is a 74 percent accuracy. This is a higher success rate compared with an existing model, which produced only 20 correct predictions-56 percent. Again compared to this existing model, the new AI program could predict 1.5 times the capability of the existing.
A retired climate scientist from Canada, William Hsieh, who was not involved in the study, explained how the program could help overcome what lacks in the current system. "The use of the climate models to create extra training data is a clever way around the shortcomings of other approaches," he said. "It appears enough of an advance that it should be deployed for real forecasts."
On the other hand, climate scientist and El Niño modeling expert from the International Research Institute for Climate and Society at Columbia University, Stephen Zebiak, talks about its practicability, considering the response time of people, particularly the government. "The kind of lead time that is actionable is probably less than a year," he said, "because decision-makers are unlikely to take action further in advance."
The team has already begun to look into forecasts up to 2021 while also looking into whether their program could also be used as basis to predict the occurrence of La Niña-the opposite of El Niño-and into the Indian Ocean Dipole pattern. They are also planning to extend the forecast period further than 18 months.