Messy Models

Hello and welcome to the continuation of my previous blog post: She's a Modeller.
If you haven't read that and you're not doing a PhD in atmospheric physics, you will need to in order to understand what I'm gonna say today.
Climat modelling sounds pretty cool, doesn't it?
The very notion of us being able to predict the weather is just mind-shattering.
BUT, there's always a "but".
Climate models are imperfect, and we will never be able to create a carbon copy of our Earth.
As a matter of fact, even if we get the perfect climate model, we still wouldn't be able predict the weather perfectly.
There are three big problems in climate modelling, two of them are kinda related and the last one is the absolute FATHER of "Why-You-Suck-At-Climate-Modelling".

I've already talked about discretisation last week (splitting the globe up into tiny boxes).
Since our computers need to calculate the atmospheric conditions in each box, one at a time, this can be extremely time-consuming.
But we are always trying to get those boxes in higher resolutions (smaller boxes) so that the Earth looks more "smooth" and less "blocky" (Figure 1).
This creates a sort of paradox: we want more (smaller) boxes, but that'll mean it takes even longer for our computers to finish calculating the whole thing.
So, as much as want to, we can't make those boxes too small.
And big boxes have big consequences!
So that's our first dilemma.

Figure 1. As supercomputer processing speeds get faster, we can afford to increase the resolution of our model. That means we can have more, smaller boxes; rather than a few big, bulky ones. This enables us to make a more accurate map of the Earth as well as areas of elevations (such as moutains).
This makes our model more realistic and Earth-like. Taken from Climate Modelling 101.

Now, on to our second problem.
If we have big boxes (more than 100 km wide), we can't represent smaller processes.
CLOUDS are an excellent example of that. We don't know whether they help reduce or induce the greenhouse effect.
But we do know they play a major role in Earth's climate.
But we don't see a full 100 km cloud do we?
That's the problem. We can't make a whole box full of clouds because that will be unrealistic and create horrible climate predictions for us.
Not only that, but we don't even know what its function is (greenhouse effect - yay or nay?).
So us scientists need to rely on (if you haven't guessed it yet) MATHS to try to represent the clouds with an mathematical equation.
This process is called parameterisation.
Don't even try pronouncing it, even I can't spell it right.
Even with parameterisation, we can't get it exactly right, simply because... we don't actually know what clouds do, exactly.
That is a major flaw. We are trying to replicate something we know very little about.
You know, normally, climate models can help advance our understanding of these clouds.
But they're just TOO.DAMN.SMALL.
And if we make our boxes small enough (around 1 km?) to realistically portray these clouds, it'll take infinity years to finish our modelling due to the sheer amount of boxes we get.
See where I'm getting at here?
Discretisation and parameterisation are both limitations to our climate model.
I hope I've made it all understandable up till this point.
The final problem will be a whole lot technical, and wayyyy more interesting.
I promise!

THE REASON YOU SUCK AT MODELLING: CHAOS

Okay, so this is our final challenge, AND our biggest one.

Unlike discretisation and parameterisation, this is not a "technical" problem.

This is all about that bass The Chaos Theory.

Sounds cool, huh?

What the Chaos Theory says is that if you put some numbers (temperature, pressure etc) into a model that's complex, non-linear and dynamic (basically, our atmosphere system), bad things happen.
The guy who came up with the Chaos Theory, Edward Lorenz was playing around with a similar model back in the 1960s.
He put a few numbers in and let the system run for a while. Came back and recorded his results.
Then, he did the test again, only this time, he rounded off his numbers before typing them into the system.
When he came back from his coffee, the results were INSANELY (much emphasis) different from his first test.
So, this is because the system is chaotic. We can't predict what goes on with it.
Not only that, we have systematic errors going on with our measurements.
Some measuring devices can only give you a certain amount of accuracy a.k.a decimal places, where as perfect modelling might need you to be really accurate and have 20 decimals places!
(28.3950358395030128394 degrees celsius, anyone?)
And what if I don't have a 20-decimal thermometer?
Well then, sucks to be you.
Your predicted weather will look nothing like the actual weather.
Therefore, modelling and predicting the climate is nigh impossible and I've just wasted your time in these two blog posts. =)
Thank you very much.

Figure 2. The Chaos Theory is also popular known as the Butterfly Effect, where a butterfly can flap its wings in Brazil and that flap will cause a cyclone to appear in Texas two years later. The butterfly's flap symbolises the small changes in measurements we input into our model. Whether it flapped its wings or not can make all the difference in a cyclone appearing (symbolising tiny changes in the beginning can result in massive differences after long periods). Taken from Mr. Lovenstein.

I'm just trolling.
The Chaos Theory only kicks in when you make the model predict the climate over a long period of time.
Right now, we can predict up to 5-7 days accurately.
10-14 days if we're really pushing it.
But after that, you get the system going haywire!
THAT is the reason why you only get weather forecasts a week in advance.
Any more ahead and the weather station can't guarantee you whether the predictions are true.
But doesn't this sound a little contradictory?
We admit that we can't predict the weather past 2 weeks tops.
And yet we claim to be able to predict how Earth surface temperature will go up by *insert number* degrees in 100 years if we keep emitting greenhouse gases.
So... what's the deal with that?

Well, you'll just have to stick around for the final part of this trilogy!
*Whoopie!*

Listening to Father Figure - Glee Cast

Still modelling,

TK