Why humans are still essential for your daily weather forecast
Weather forecasting is a success story. And a huge business. While the daily forecasts heavily rely on computer simulations, the added value of the human forecaster is still unmatched. A recent scientific study shows why.
Weather generates millions in tourism revenue when it’s good, and destroys as much when it’s bad. It’s one of those few things nobody can change, but everybody is concerned about. So weather forecasting is important, and national agencies as well as private businesses around the world are charged to improve those forecasts continuously.
At its beginnings, forecasting was done by drawing weather maps based on atmospheric measurements, and then trying to find similarities among the many maps they have drawn over the previous years. If there was a similar map, the idea was that something similar would also happen this time. So the forecast was based on what happened a while ago, assuming that things would repeat reasonably often and in a reasonably similar way. That was ok for the next day, but after that, those forecasts weren’t any good. After all, the father of chaos theory, Edward Norton Lorenz, was a meteorologist.
By the beginning of the 20th century, people started having better ideas. By that time, the main mathematical equations which describe the behaviour of the weather were already known. But there were two problems: The equations only describe changes to the weather, meaning that they are only useful if an initial state (i.e. the exact state of the atmosphere and ocean right now) is known — the first problem was that observations were too sparse for this initial state to be known. The second problem was that even if a forecaster knew the initial state, the equations were so complicated that computing them would take way too long. In 1922, L.F. Richardson computed that 64,000 humans would be needed to compute the weather over the globe fast enough to just keep up (let alone being faster to be able to forecast). It took him two years to compute a 6-hour forecast he made to illustrate his method. Nevertheless, what Richardson described as method should many years later become almost exactly what we make our computers do for our forecasts.
Perhaps some day in the dim future it will be possible to advance the computations faster than the weather advances and at a cost less than the saving to mankind due to the information gained. But that is a dream. [L.F. Richardson, 1922]
Of course, by now Richardson’s dream is reality — has been for quite some time. As computers become more powerful and weather models more complex, one might think that the human forecaster might become (or already be) obsolete. It is true that there are efforts to completely automate weather forecasting, and some of the commercially available weather apps and products do just that.
But one of the main results of our recent paper is that even the most advanced weather models with the highest possible resolutions can still be surprisingly far from the truth. The good news is that they are ok for big, important events. But they aren’t that good for normal, everyday weather. And this is where the value of the human forecaster comes in. Through her, human knowledge and experience manages to extract reliable information from a soup of contradicting data. Because after all, daily weather forecasts are really quite good — and that’s mainly thanks to human intervention.
So in the end, humans still beat machines. Daily. Repeatedly. Thank you unknown forecasters who live somewhere in my weather app. I hope you have a sunny day.
Dr. Martin Jucker is a lecturer in climate dynamics at the University of New South Wales, Sydney, Australia. He has also written an op-ed for Iceberg about the stratospheric influence on weather in the Southern Hemisphere.
