The Science of Weather Prediction

Image Credit: Unsplash: Fabian Albert

Vanshika Dhyani discussed the Science of Weather Prediction with Professor Conor Sweeney, President of the Irish Meteorological Society

Weather prediction is an act of calculation based on observational data— collected from studying the atmospheric conditions of a specific place at a given time—that provides information about future advancements in weather conditions.    

The modern age of weather forecasting began in the late 1830s with the invention of the electric telegraph that allowed information about the weather to travel hundreds of kilometres instantaneously. 

Francis Beaufort—a hydrographer for the British Navy, and Robert FitzRoy—his protégé laid the foundation of the science of forecasting. Beaufort devised a scale called the  Beaufort wind force scale to relate wind activity with conditions at sea or on land. 

FitzRoy compiled weather statistics and used the telegraph to correspond a weather forecast. Readings from barometers near the coast were studied and compared with temperature and wind observations to compose a forecast. 

Imprecision and inaccuracy are a dilemma with forecasting that stem from an inadequate understanding of meteorological conditions because the weather is a chaotic system.

"Chaos is a branch of mathematics and what chaos theory says is that if you have a chaotic system, then where that system ends up depends really really critically on exactly how that system started," explained Professor Conor Sweeney from UCD School of Mathematics and Statistics.

The advent of computers transformed the science of weather prediction. With the help of a computer, the density, pressure, and potential temperature scalar fields and the velocity vector field of the atmosphere could be determined and interpreted to obtain a forecast.

Today, a computer model provides meteorological information by solving a set of primitive equations and the ideal gas law to predict the weather.  Primitive equations are nonlinear partial differential equations and cannot be solved analytically thus Numerical weather prediction (NWP) models can only reach approximate solutions.

"Weather forecasting is one of the great, big data problems. To solve the weather for anywhere in the world we need to know the weather everywhere in the world. Computer model is the whole, globe of the earth and it needs to solve what happens everywhere around the earth to figure out what's going to happen tomorrow and the next day, and the next week."

“The weather doesn't care what happens on the ground. Weather is what happens in the sky.”

"In Ireland, the people you see on television are full meteorologists. They come from Met Éire and the Irish Weather Service. So they've done their degrees in Physics and Maths, and they've done their training in meteorology." Professor Sweeney explained that since computer models are still an approximation of the weather conditions, we need meteorologists to interpret and combine the computer model output with what they know about Ireland and its seasons. Thus concluding that there is a need for human input on forecasting. 

But for how long?

"At the moment, if you open up the weather app on your phone you'll get the same forecast as someone down the street but in the future, it could become more localised."

He believes that weather prediction "is a real opportunity for machine learning because the computers don't always get it right and a lot of times humans have to adjust for that, adapt for that." 

"If phones start recording what temperatures they sense that could all feed into this machine learning algorithm and then we can have personalised forecasts" he continued.

It is essential to know the condition at the higher parts of the atmosphere to be able to forecast how the weather is going to change on Earth's surface. This is why weather prediction turned a new chapter with TIROS-1—the first successful weather satellite launched by NASA in 1960.Atmospheric conditions are monitored by geostationary and polar-orbiting satellites.

"The satellites really made a massive difference in forecast skill" he told the University Observer. “The weather doesn't care what happens on the ground. Weather is what happens in the sky.”

"Our forecast got much better. It particularly made improvements in areas where there weren't many observations," he said, still discussing the subject of satellites.

"When the satellites came along we had data everywhere. And that helped our forecast skill get better. It is still advancing all the time, we've newer instruments, we've got new satellites, we've more powerful supercomputers which means we can run at a higher resolution. And critically, we can run more forecasts. "

"They[meteorologists] realised quickly that there is no point in having a really, really really highly accurate model. Because if it is off we don't know how far it's off. Actually what they do now is they run loads of models." Forecasting is chaotic and very uncertain which makes predicting weather nearly impossible. 

Then how do we predict what the weather is going to look like?

Weather is predicted using probabilistic forecasting where the same calculation, is made multiple times to minimize errors and uncertainties in the estimate. "The European centre run 52 different forecasts every time they run a forecast. And that's really important because, on one hand, the forecast has to be less detailed because you've just got the one computer so if you're going to run 50 of them, you've got 50 times less computer power. But if all 50 models agree that this is what's going to happen the next day, then you can say 'I'm very confident that the temperature is going to be in this span.' If the computer models all start disagreeing, after a day and a half or two days then you can say 'okay, there are a lot of uncertainties in the atmosphere, over this area.'" Professor Sweeney explained.

"There is no good having a single forecast, that's what we call deterministic forecasting, what we really need to have is a whole load of forecasts which is probabilistic forecasting and by using those we get a much better idea of what's going to happen."

The strides made in the field of forecasting have helped predict severe weather conditions, which in turn supported national weather services issue warnings ahead of time to prepare citizens. With the introduction of machine learning in forecasting alongside advances in science and technology, the future of weather predictions looks bright. 

[Fun fact: Beaufort (born and raised in Ireland) trained FitzRoy, commanded the ship HMS Beagle on its second voyage that carried Charles Darwin (a recent graduate). The journey to the Island of Galapagos  led to the theory of evolution recorded in 'The Origin of Species']