Predicting the Weather

Brian Golding takes us through how a weather forecast is put together, and how much we can trust it.
16 May 2013

Interview with 

Brian Golding, Met Office


In the northern hemisphere, summer is supposedly on its way, but we haven't seen very much evidence of that yet here in Cambridge. Looking on the sunny side though, we decided to find out how we forecast the weather, so we'll know if it's going to be arriving anytime soon. Dominic Ford spoke to Brian Golding at the UK's forecasting centre, the MET Office...

Brian - Fundamentally, the weather of the whole world is driven by heat coming from the sun and if that heat warmed up the Earth uniformly then we wouldn't have any weather. But because some place has absorbed more heat than others, forests that are absorbing heat, oceans that are reflecting it, those variations in the heating then communicate themselves to the atmosphere, they result in differences in density and different densities then drive the wind and the whole circulation of the atmosphere.

Dominic - So, I guess in trying to work out what the weather is going to do, you're basically simulating those processes to work out what's going to happen next.

Brian - Yes, that's exactly what we do. The mathematical equations that describe those processes are quite well-known. Interestingly though, if you solve those equations, they diverge into uncertain solutions. I like to say the atmosphere has a very bad memory like me. So, we always have to refresh the information about what the current state of the atmosphere is, before we can start the forecast and that's why we have lots of observations from all around the world, and satellites looking down on the Earth, sensing what the state of the atmosphere is, so that we can get to the best starting point for forecast.

Dominic - And I guess it's interesting that people have been interested in the weather for hundreds, if not, thousands of years and physicists understand how gas behaves quite well. So, why is it so very difficult to solve these equations and workout what the weather is going to do tomorrow?

Brian - Yeah, it's a very interesting problem. It comes down to the fact that the atmosphere will respond to these variations in heating on a whole spectrum of scales. So, at the very small scale, you'll get little eddies which may turn into big eddies, turn into thunderstorms, perhaps with tornadoes. And then at the other end of the spectrum, you've got these large depressions that track across the globe. And in between, you've got a whole variety of scales of fronts and clouds, and when those interact with each other then it becomes very difficult to make a precise prediction.

Dominic - I know in the last year or so, the MET Office has moved over to using probabilities. So, you talk about a 50% chance of rain rather than to then say, it will or won't rain. What do these probabilities mean exactly?

Brian - If we talk about the probability of rain then what we're saying is that at the location that the forecast is for and the period that it's for then there is a certain chance of having rain. The source of that uncertainty might come from a number of areas. It might be that it's going to rain for a part of the time, but we don't know whether the time that it's going to be raining is during the time that the forecast is for or not, or it might be that it's going to rain somewhere in the country, but we don't know what the chances are that it's going to rain in that particular location. Or it may be that the way in which the weather develops might result in rain or it might not result in rain at all. There are all these different sources of uncertainty.

Dominic - Now, these physical processes that you're modelling to know what the weather is going to do, are they the same all around the globe or rather specific physical phenomena that you have to worry about in specific geographic regions?

Brian - Fundamentally, they're the same everywhere, but they do come in different mixtures in different parts of the world. So, the equations are the same, but in developing the models, we have to optimise the performance by comparing it with what actually happens in well-studied situations. And that optimisation has to be done for different parts of the world.

Dominic - So, how do you go about doing that?

Brian - The main way is large international field experiments where scientists choose a location and they set up observing facilities, often involving aircraft flights to sample the atmosphere in great detail. They will involve the satellite observations that are available, surface observing equipment and at the same time, then they will run our forecast models for those same locations and carry out very detailed comparisons of what the forecast model produces and what's observed. By doing that, they can learn what the shortcomings are of the models, improve some of the details that go in the equations, parameters that relate to vegetation for instance or to the interaction with topography. And they might change those and look at how that alters the forecast and whether it makes it closer to the detailed analysis.

Dominic - So, does the same apply locally here?

Brian - Indeed. We had a detailed field experiment in the Welsh boarders a few years ago, looking at the way in which the atmosphere behaves on a clear night in winter. As the atmosphere cools down at night then the wind tends to develop down the slopes of the valleys, bringing cold air down into the bottom of the valley. We did detailed observations of how that happened and we compared it with the way in which the model developed, and one of the things that it told us for instance is that you really need to have a very fine resolution model in order to get that right, even finer than our finest models. So, we have a 1.5 kilometre model at the moment. We would really like to go down to perhaps 200 or 300 metres when we can afford it because we know it would produce a better forecast in those circumstances.

Dominic - And what will that take?

Brian - It would require a computer with a factor of at least a hundred greater power than we've currently got.


Add a comment