Unless you're putting money down today the model doesn't mean anything.
Predicting the past isn't the same thing as predicting the future. You can train a model or compose an algorithm that fits existing data but that doesn't necessarily mean it has predictive power far into the future.
ENSO events are inherently chaotic, one of those pendulums that are known to swing back and forth but difficult to predict exactly when.
In times well past there was a certain degree of divination that looked at accumulated patterns and made guesses that worked more often than not in the short term (less than a year in advance.
Then came "black box" non physics based methods:
Recent advancements in artificial intelligence (AI) have pushed these boundaries, achieving accurate predictions up to 16–18 months in advance. However, the "black box" nature of AI models has precluded attribution of this accuracy to specific physical processes.
Not being able to explain the source of the predictability in the AI models results in low confidence that these predictions will be successful for future events as the Earth continues to warm, changing the currents in the oceans and atmosphere.
and now:
"Unlike the 'black box' nature of AI models, our XRO model offers a transparent view into the mechanisms of the equatorial Pacific recharge-discharge physics and its interactions with other climate patterns outside of tropical Pacific," explained Fei-Fei Jin, the corresponding author and professor of atmospheric sciences in SOEST.
"The initial states of the extratropical Pacific, tropical Indian Ocean, and Atlantic enhance ENSO predictability in distinct seasons. For the first time, we are able to robustly quantify their impact on ENSO predictability, thus deepening our knowledge of ENSO physics and its sources of predictability."
Still getting 18 month predictions, but now with a delicious senses of likely understanding why.
Yet another example of people blaming chaos when it’s not the relevant limiting factor. If chaotic behavior manifested at a sufficient level on the 18 month time scale to preclude forecasting El Niño, neither the back of AI models nor the new improved physics-based model would be capable of making such predictions.
Chaos comes in degrees, people still catch tumbling tennis rackets despite the Dzhanibekov Effect | intermediate axis theorem.
Not every horseshoe map is folded to infinity and beyond. Creeks flow predictably enough and yet still contain turbulent vortexes.
Dynamic systems aren't an all or nothing proposition wrt chaos, there are often contained embeddings an portions that are difficult to predict but not impossible subject to error bounds.
We are talking about large scale systems on long timelines that have massive momentum and energy. Predicting something 18 months out isn't hard (Florida will have a hurricane in the next 18 months).
Predicting "will crops get enough water in the mid west in any given year" is impossible because of "chaos" because of complexity.
This result is not surprising to any one who pays attention to how the math for these systems works.
Precisely the point. Chaos wasn't impactful on the relevant timescale for large scale dynamics like El Nino, and shouldn't be blamed for our previous inability to predict beyond 12 months.
We use a mixture of physical and statistical models to forecast ENSO, including physical models of varying degrees of complexity.
A long-standing challenge has been for these models to exhibit reliable skill at lead-times beyond ~6-12 months, especially when forecasts are initialized in the early Spring. Data-driven models ("AI" or "machine learning") seemed to break through this barrier a bit, but what's notable about this paper is that a physical model with a new underlying mechanism seems to catch up a bit.
That hasn't been my experience these days, however I do use meteograms which are far better at describing the weather conditions than the basic weather spiel stating highs and lows and whether it will rain or not.
That's not true anymore it depends on number crunching ability and willingness to spend money. Most countries are not willing to spend the money for granular details for a small towns or cities weather
I'm not sure if this will improve long-term weather forecasts, like this [1], where an el nino was predicted but it still rained more than expected.
[1] Farmers frustrated after destocking following BOM's incorrect El Niño forecast: https://www.abc.net.au/news/rural/2024-01-22/farmers-frustra...
Unless you're putting money down today the model doesn't mean anything.
Predicting the past isn't the same thing as predicting the future. You can train a model or compose an algorithm that fits existing data but that doesn't necessarily mean it has predictive power far into the future.
Wait, what were they basing the previous model on?!
ENSO events are inherently chaotic, one of those pendulums that are known to swing back and forth but difficult to predict exactly when.
In times well past there was a certain degree of divination that looked at accumulated patterns and made guesses that worked more often than not in the short term (less than a year in advance.
Then came "black box" non physics based methods:
and now: Still getting 18 month predictions, but now with a delicious senses of likely understanding why.Yet another example of people blaming chaos when it’s not the relevant limiting factor. If chaotic behavior manifested at a sufficient level on the 18 month time scale to preclude forecasting El Niño, neither the back of AI models nor the new improved physics-based model would be capable of making such predictions.
Chaos comes in degrees, people still catch tumbling tennis rackets despite the Dzhanibekov Effect | intermediate axis theorem.
Not every horseshoe map is folded to infinity and beyond. Creeks flow predictably enough and yet still contain turbulent vortexes.
Dynamic systems aren't an all or nothing proposition wrt chaos, there are often contained embeddings an portions that are difficult to predict but not impossible subject to error bounds.
> Yet another example of people blaming chaos
This isn't how this works, at all.
We are talking about large scale systems on long timelines that have massive momentum and energy. Predicting something 18 months out isn't hard (Florida will have a hurricane in the next 18 months).
Predicting "will crops get enough water in the mid west in any given year" is impossible because of "chaos" because of complexity.
This result is not surprising to any one who pays attention to how the math for these systems works.
Precisely the point. Chaos wasn't impactful on the relevant timescale for large scale dynamics like El Nino, and shouldn't be blamed for our previous inability to predict beyond 12 months.
We use a mixture of physical and statistical models to forecast ENSO, including physical models of varying degrees of complexity.
A long-standing challenge has been for these models to exhibit reliable skill at lead-times beyond ~6-12 months, especially when forecasts are initialized in the early Spring. Data-driven models ("AI" or "machine learning") seemed to break through this barrier a bit, but what's notable about this paper is that a physical model with a new underlying mechanism seems to catch up a bit.
It appears to be an arms race between physics based models vs machine learning models.
Current SOTA (which this is apparently matching) is AI based.
“Global climate model”
Meanwhile it's still basically a coin flip if it's going to rain tomorrow.
Better than a coin flip is tomorrow’s weather is going to be roughly like today’s weather.
That hasn't been my experience these days, however I do use meteograms which are far better at describing the weather conditions than the basic weather spiel stating highs and lows and whether it will rain or not.
That's not true anymore it depends on number crunching ability and willingness to spend money. Most countries are not willing to spend the money for granular details for a small towns or cities weather