UK United Kingdom

A chaotic beast, probably: wacky weather and climate forecasting

“Prediction is very difficult. Especially about the future” – so said Neils Bohr, the Danish physicist and 1922 Nobel Prize winner. And you know what? I think the bloke was onto something there, especially…

A precise weather forecast, months in advance, will never be possible. Matt Smith

“Prediction is very difficult. Especially about the future” – so said Neils Bohr, the Danish physicist and 1922 Nobel Prize winner.

And you know what? I think the bloke was onto something there, especially when it comes to seasonal climate outlooks.

In the past 12 months we’ve seen – as Today Tonight likes to describe it when they interview myself and my colleagues at the Australian Bureau of Meteorology – truly wild and wacky weather.

But no-one tells you this wackiness is a combination of weather and climate. And that, despite all the calls for a precise weather forecast for everyone’s town/farm/dam many months in advance, it will never be possible.


Well, first of all, let’s get the weather and climate thing sorted. As the saying goes: climate is what you expect, weather is what you get. And indeed that’s pretty much it – climate is just the averaging of weather over time.

But when it comes to climate forecasts – in this case we’re talking about seasonal forecasts – there are some fundamental differences between what you can do for weather and what you can do for climate.


Predicting weather and predicting climate are actually quite separate beasts. Sure, they both use stonking big computers and telephone-book-thick reams of Fortran code (yes, Fortran, for those sniggering up the back – it’s still the fastest floating point arithmetic language). But the divergence is in how the forecasts are made.

Climate is all about forces. Not “F=ma” type force, but rather a push towards one state (such as wetter and cooler over Australia) or another. Think of the sun – that’s your classic climate driver.

As it moves from overhead in the northern hemisphere to overhead in the southern hemisphere it “forces” a change in the climate. Hence the seasons we get each and every year. In fact, you could say the calendar on your wall is your very own climate model – how cool is that?

So when it comes to producing a seasonal climate outlook, the important things to look at are the climate drivers. The biggies for myself and colleagues are the oceans, as they store a massive amount of heat (and “evaporable” moisture), and will in turn force the atmosphere to respond one way or another.

So, for a climate forecast, we ask: “Is the Pacific Ocean warmer or cooler than normal?” In other words, do we have El Niño, neutral or La Niña? Are the sea surface temperatures around Australia warmer or cooler? What’s the state of the Indian Ocean? All these factors (and more) will try and force the atmosphere to alter slightly from its mean state.

These drivers and their strength are known as the “boundary conditions”, because they set the limits of what’s possible over the season ahead.

Cyclone Yasi NASA

Look at what happened in the 2010 wet season. Late in the year the Coral Sea was very warm – typical of a La Niña – which nudged the region considered “favourable” for tropical cyclone development closer to the Queensland coast. The result? Tropical Cyclones Tasha, Anthony, then the biggie, Yasi all did their darndest to ruin the sunshine state’s summer.

To highlight what a boundary condition can do, tropical cyclones have only ever crossed the Queensland coast multiple times in the one wet season during La Niña events.

So what about weather forecasts?

Weather forecasts

With a weather forecast, while it’s important to have the boundary conditions about right, the most important things are your initial values. This is the data you plug into the model from every single spot you can get your hands on.

Manual observations, satellites, floating buoys, aircraft, ships, automatic weather stations, ocean gliders, radars, weather balloons – all of it goes in to paint an extremely detailed picture of the three-dimensional current state of the atmosphere and its weather at a precise time.


Now, the weather being a chaotic beast, any little errors in these observations have the potential to become big ones – it’s like the butterfly flapping its wings in the Amazon causing a cyclone over India a week later.

OK, that’s stretching it a bit, but fundamentally that’s what chaos says can happen: a cascade of ever-increasing errors/impacts.

Hence if you have your physics right (just as for a climate outlook) and you have really good initial conditions, the “chaos” (or the explosion of the errors) is minimised for a few days and we can get really good forecasts on small time and space scales.

But this will only last a few days. Typical accuracy is about a week, or approximately the time it takes to cycle from one weather pattern to the next. Beyond that, the atmosphere “forgets” what happened a week ago and starts to head down its own merry path.

So if it’s a billion-to-one to predict exactly what will happen on a particular day in a few months time, how can we trust a seasonal forecast? Well, a colleague of mine came up with a great analogy.

Take one hammer. Take one vase. Bang ‘em together. What will you get?

The drover’s dog could tell you the answer is a smashed vase. (That’s the climate-forecasty bit; the known forces/drivers/boundary conditions have given a very predictable, general, result: an ex-vase.)


But Professor Bohr’s wisest owl couldn’t be expected to tell you where every single shard of pottery will land. That’s nigh-on impossible (well, unless you knew the exact location and force on every particle in the universe – then it’d be a doddle). Not being able to predict where every piece will land doesn’t discount the fact the hammer is the natural enemy of the vase.

Similarly, just because you can’t predict exactly where rain will fall on a particular Thursday in five weeks' time (or even five days' time), that doesn’t mean we can’t predict what will generally happen that month or season.

Playing with probabilities

Any decent climate outlook worth its Fortran code will be given as a probability.


Research shows us that, at best, only about 70% of our year-to-year seasonal climate is predictable; the other 30% is chaotic random (weathery) stuff. (It must be one of the few fields where you expect to be “wrong” occasionally and just have to cop it.)

A typical climate forecast will calculate there’s a 60% chance of more rain than normal in a general region next season.

Does this help?

Well, we know that hedging one way or another, over time, will let you win in the end.

Zdenko Zivkovic

Think of it this way – a European roulette wheel has one “0”, 18 red and 18 black slots. If you only bet one colour and the payout is double for a win, the casino would still win in the end because they have the 0. (In reality, when odds and payouts are taken into account, a casino has about a 2.7% edge).

The casino knows it will always make a profit if the game is played often enough.

Climate forecasts give you far better odds that 2.7% and, used wisely over a period of time, will give an edge and ultimately a win overall.

And in a changing climate, an edge is what we all need.

Which leads me to proclaim the following:

“Prediction is very difficult, especially about the future, but at least we have stonking great computers to sway the odds.” – Andrew Watkins, 2012. (Don’t think that will win me a Nobel Prize though.)

Articles also by This Author

Sign in to Favourite

Join the conversation

11 Comments sorted by

  1. Stephen Prowse

    logged in via LinkedIn

    I would like to congratulate the author for explaining the difference between weather and climate without once mentioning climate change or global warming.

  2. Tim Scanlon

    Author and Scientist

    Thanks for the article Andrew. Good to have this all spelled out very everyone.

    It would be great to have more of the BOM information presented like this. I know the people I work with appreciate the explanation to contextualise climate and weather.

  3. Matt Stevens

    Senior Research Fellow/Statistician/PhD

    Great article, but o mention climate change - how about the climate change modellers giving us some probabilities when predicting temperature change? This would inform the debate and ensure that dogmatic language is not used, which has happened way too much in the past and lla it has done is fed into the skeptics arguments.

    1. Roger Jones

      Professorial Research Fellow at Victoria University

      In reply to Matt Stevens

      Dear Matt,

      given that CSIRO/BoM projections are framed in terms of probabilistically-defined ranges, IPCC projections are given as ranges with judged uncertainties and the Climate Commission's communications are broadly consistent with those, could you please point us to the sources of these dogmatic predictions. Thanks!

    2. Matt Stevens

      Senior Research Fellow/Statistician/PhD

      In reply to Roger Jones

      Hi Roger, ok, maybe not dogmatic predictions...though the predictions certainly didn't match the last 10 years of no temperature increase. Clearly there are other factors affecting temperature changes that the models do not account for. I have a modelling background (though many years ago), and do understand that these are verty complex models. However, if variables are not included that place constraints on the model, then the results, well are unreliable...I do believe that the climate is increasing and that CO2 is an important variable, but I also think that there is many unknowns, and it is the job of scientists doing the work to address the concerns of the general public when questioned. Further, I do worry that the general public (at least in the USA and Australia to a lesser degree) often do not trust scientists and governments need to do more in ensuring all citizens are educated in the sciences.

    3. Roger Jones

      Professorial Research Fellow at Victoria University

      In reply to Matt Stevens

      Matt, put that way - you've uncovered a really interesting question and it can't really be answered by better probabilities in the way you were anticipating in your first comment. You'll appreciate some of the difficulties being a statistician.
      Climate projections are provided using a signal to noise model where the statistics strip out the central line/curve from the noise of climate variability. The $64 question is to whether all of the variability we observe is natural, or whether the climate…

      Read more
    4. Matt Stevens

      Senior Research Fellow/Statistician/PhD

      In reply to Roger Jones

      Thanks Roger. It certainly is complex and clearly there are many things that do get debated in the scientists world. In terms of the non-linear relationship between CO2 and temperature, it clearly is not linear. I came across something about sun spots as a driver of climate change, which makes intuitive sense I would think to most people. It also appears from past records that the sun activity was the main driver of ups and downs in climate (corresponds exceedingly well with warmer periods and ice age periods. From what I know, the 11 year sun cycle is modelled in the IPCC models, but do they also model changes in sun temperature (whether using minimas or maximas or an average)?

    5. Dale Bloom


      In reply to Matt Stevens

      Thanks for this article Roger.

      The energy of the planet includes thermal heat energy in the oceans, atmosphere and land masses, kinetic energy in moving ocean currents and wind, latent heat energy contained in cloud cover and even chemical energy bound up in plant growth from photosynthesis. One type of energy can be converted into another, and to develop a model for weather, an energy audit would have to be included, which makes it extremely complicated. I am uncertain if we have the equipment…

      Read more
    6. Andrew B. Watkins

      Manager of Climate Prediction Services at Australian Bureau of Meteorology

      In reply to Dale Bloom

      G'day Dale,

      The climate models used today are well advanced from the old energy balance models - they are physics based (dynamical), models, essentially solving the primitive equations for momentum, energy and mass. The degree to which these models can simulate the global climate is, frankly, incredible, and really are one of the most amazing pieces of human ingenuity of the past half century (Saturn V rockets included - but then again, I look into a cars engine bay and go "hey, these inanimate…

      Read more
    7. Andrew B. Watkins

      Manager of Climate Prediction Services at Australian Bureau of Meteorology

      In reply to Matt Stevens

      G'day Matt,

      The IPCC doesn't have any models at all - rather they collect information from models run by universities, research centres, government agencies etc etc. The IPCC is really just an intelligent portal of others research.
      In terms of the sun, the models do take into account solar changes explicitly (the sun has been 'cooling' for about 4 decades now). Some really neat work by Australia's own Julie Arblaster and Jerry Meehl from US National Center for Atmospheric Research show that you can only model modern changes in the climate closely by including volcanoes, solar strength, aerosols, ozone and greenhouse gases. Sadly they weren't able to model the observed global climate trend using natural forcings alone.

    8. Dale Bloom


      In reply to Andrew B. Watkins

      Thanks Andrew, and sorry I mistook your name in a previous post. The latest models seem very interesting, although I had been thinking of weather as a manifestation of energy, or different forms of energy produce different forms of weather.

      My background automatically made me think of the planet as being a closed system, and energy in has to produce energy out or energy being stored somehow.

      Thanks again.