A group of researchers from the Potsdam Institute for Climate Impact Research, Beijing Normal University and Justus-Liebig-Universität Giessen has found a way to predict El Niño events up to a year before they occur. In their paper published in the Proceedings of the National Academy of Sciences, the group describes their complexity-based approach to better predicting the seemingly random weather events. ...
Once they found that pattern, the researchers went analyzed yearly surface temperature data from 1984 to 2018 to make predictions about El Niño events in the past. They report that their method correctly predicted nine out of 10 El Niño events (and had three false positives.) Additionally, they found that the higher the disorder the previous year, the stronger the following El Niño event. The researchers conclude that it is now possible to predict El Niño events up to a year in advance with reasonable accuracy.
Some of the work has come from an Israeli university, it seems, so the research got a lot of reporting in the Jewish press. For example:
“This novel climate network approach is very promising for improving El Niño prediction,” said Prof. Shlomo Havlin, an Israel Prize-winning physicist from Bar-Ilan University who was involved in developing the algorithm.As El Nino is usually associated with less rainfall and higher temperatures in Australia, the 2020 prediction is really not good news.
“Conventional methods are unable to make a reliable El Niño forecast more than six months in advance. With our method, we have roughly doubled the previous warning time,” stressed JLU physicist Armin Bunde, who initiated the development of the algorithm together with his former PhD student Josef Ludescher.
Hans Joachim Schellnhuber, Director Emeritus of Climate Impact Research, explained: “This clever combination of measured data and mathematics gives us unique insights – and we make these available to the people affected.”
He pointed out that the prediction method does not offer one hundred percent certainty: “The probability of El Niño in 2020 is around 80%. But that’s pretty significant.”
2 comments:
El Nino is the result of solar activity. So its unlikely that they will be able to predict the El Nino without being able to predict solar activity. Theodore Landschiedt already could predict El Nino before he died in 2004. He'd use planetary orbits to deduce how far the sun was from the centre of gravity of the solar system and then he'd interpolate how it would behave at that lead to all manner of successful predictions.
Now it might be if you programmed all past data into a neural network program the computer could figure it all out even without explicitly putting in the planetary data or the predictions of the solar data. But that seems a bit of a stretch so their model should fail.
One missed year and three false positives when you already knew the answer? So thats a bad model. Its doesn't backtest and no successful predictions so far. You have "predictions" about the past, but you need to be able to predict the future. So zero runs on the board for future predictions and they even failed for predicting the past. Its actually quite hard to fail to predict the past. As I'm sure you can agree.
So this is a dud model. Its not going anywhere.
Post a Comment