Why can data models predict the future climate but not the weather?
Why are weather forecasts often wrong but modelling of the future climate highly accurate? I would presume a more complex system over a longer period would be harder to predict.
In: 0
Who said their climate prediction are highly accurate? And weather prediction is pretty damn accurate. But it’s kinda like sports. I can predict the top teams each year with pretty good accuracy. But I can’t mail down how my hits one player will get in one specific game
> I would presume a more complex system over a longer period would be harder to predict.
You are not wrong, and in part you are onto why it’s harder to predict the weather.
**Weather** is what conditions of the atmosphere are over a short period of time. **Climate** is how the atmosphere “behaves” over relatively long periods of time.
You can think of weather as the very specific details, and Climate as the broad average.
So is it going to rain tomorrow in my town for 30 minutes, is much harder than saying rain is likely in the next month, or the average temperature will rise by 2 degrees.
Hope that makes sense
Let’s make an analog.
You are able to save on average 2000 a month. That means in 10 years you should have 240.000 on your bankaccount.
This doesn’t mean you will be saving 2k each month. Some months it is 500 and others it is 3k.
In other words we see a trendline which should mean x to the weather. But that doesn’t corrolate directly for the weather we will be having on a daily basis.
I think the predictions people are looking for regarding weather are far more precise.
For example the prediction about climate that comes to my mind is “the average temperature of earth will rise”.
(Stuff like a rising water level are direct conclusions of this one very vague prediction.
But if you are talking about weather you want to know what will the temperature be next Tuesday at exactly 4 pm in this very specific city.
Another thing to keep in my mind with weather you get a new chance to check the predictions everyday.
While with climate it needs a few years to get an answer.
Because complex systems regress towards a median, in accordance with a model’s predictions, especially over long intervals, whereas short-term predictions are subject to greater volatility. In other words, it’s much easier to say that 1,000 cars will go down the road, than it is to say that car X will take the next off-ramp.
YMMV I guess but the weather forecasts I have seen in the past several years have been extremely accurate.
Weather is a product of climate. So while minute weather patterns are too chaotic to predict, the overall contitions which set up these weather patterns are less chaotic and more predictable.
Considering how chaotic weather is, we actually do a really good job with weather prediction. It takes some of our best computers though
(1) Predictions about future climate tend to be very general (things like “average worldwide temperature increase of X degrees”) while weather predictions tend to be a lot more precise (“You’ll see 3-5 inches of rain in this city between 9am and 1pm tomorrow.”) The more precise the prediction, the easier it is to miss.
(2) We don’t really yet know if predictions about the climate are accurate because they’re largely predictions of events that are still to come.
Imagine tracking a molecule of air as it moves around inside a balloon: its speed, trajectory, spin, etc. Now try to predict how that molecule will behave as it moves through that balloon as it bounces off of the rubber and into other air molecules.
I hope you can understand that this would be incredibly, INCREDIBLY hard to do with accuracy given how many variables you’d need to track, and how small variations can throw off your estimates very rapidly.
BUT… what if instead of tracking the behavior of individual molecules, we track the behavior of all those trillions of air molecules as a whole: temperature, pressure, volume, etc. This is much easier, because the individual behaviors of all those particles will average out. This is how we get the ideal gas law: PV = nRT.
Another way to think of it is to consider predicting the results of a coin flip: how many times will it land on heads? You’ll only be able to get it right 50% of the time. But what if we consider 10 coins, and consider the results as a whole: how many times will the heads come up between 40% and 60% of the time? That’s a lot easier, and can be mathematically tracked along what’s called a [normal distribution](https://www.fourmilab.ch/rpkp/experiments/statistics.html). Furthermore, when you increase that number to 100 coins, or 1000 coins, that normal distribution gets narrower and narrower, because as you add more coins the system “averages out” more and more.
The larger a system is, the more accurately you can gauge its “average state.” Tomorrow’s weather is a much smaller system than the overall climate (which can be roughly seen as an average of a region’s weather patterns over a long period of time). The idiosyncrasies of daily changes to the weather effectively average out over a long period and is easier to predict, and trends are easier to observe.
Latest Answers