It is 10.5 degrees on Jacob Schipper’s farm in Flevoland. The wind is from the north, it is not raining. Barometric pressure: 1,039 hectopascals. At least that is what Schipper’s weather station says, a pole with measuring equipment that is a few meters from his house.
“As a farmer you are quite busy with the weather,” says Schipper, with his hands deep in the pockets of his blue overalls. For example, he uses the measurements to determine when to irrigate the land. “If a shower of 25 millimeters has just fallen, then I know I can wait a few days.” He prefers to do that measurement himself, because a rain cloud passing over the official KNMI station at Lelystad Airport can leave his own country dry just like that.
Schipper is not the only one using the data. His weather station also continuously sends them via WiFi to the Met Office, the British counterpart of the KNMI. It collects weather data from all kinds of countries, such as Belgium, Sweden and New Zealand, and makes it available to the Netherlands as well. On the map of WOW-NL, the name of the Dutch variant of the project, there are hundreds of stations. They bear names such as Meteo Tuitjenhorn, Mooi Weer Boxtel and Schipper Weather, Schipper’s station in the middle of the Flevopolder.
Photos Diewertje Bravenboer
Slippery roads
With the data from all amateur meteorologists, the KNMI can improve its weather forecasts, says Jan Willem Noteboom, research coordinator at the weather institute. “If, for example, frost moves over the Netherlands, it is important to know where the zero-degree line is. From there the roads become slippery. Our own forty weather stations sometimes overlook cold areas that WOW-NL’s five hundred stations can see.”
Scientific studies have already shown that amateur weather stations have added value, not only in the Netherlands. Weather authorities in a handful of European countries are experimenting with the data. Norwegian meteorologists measured for example, that they can make temperature predictions a third more accurate, from an average deviation of 1.24 to 0.84 degrees Celsius. Large errors of more than three degrees decreased by almost 70 percent, depending on the number of weather stations in an area.
Floods in Limburg
Dutch researchers saw that hobby measurements can help to better estimate wind speeds, although some weather stations are too close to buildings to say anything meaningful. Local gusts that missed the official stations were picked up by WOW-NL.
The KNMI wants to use all the data to improve its weather warnings in particular. According to Noteboom, there was room for improvement during the floods in Limburg in the summer of 2021. “Our own weather stations could not properly see the very local and extremely heavy precipitation, and the precipitation radar cannot properly see through a wall of rain. ” The devices in the backyards of Limburgers were the only ones that showed how much it had really rained.
At the moment, weather warnings are still issued per province, but according to the KNMI, this should also be possible per zip code. The institute is therefore working on a new warning center that can warn more accurately, of which the measurements from WOW-NL must become a part.
Read about the rain of July 2021: The Netherlands is more often confronted with extreme rainfall
The KNMI is not that far yet. Noteboom: “We sometimes compare WOW-NL to our own measurements, but they are not yet a permanent part of our services.”
The reason is simple: some weather stations are less accurate than others. Schipper ordered it from China for about a hundred euros. “After a while, the wind sensor runs a bit stiffer, and when the sun shines on it, it thinks it’s warmer. Or sometimes it hasn’t rained for a week, but he still measures some precipitation because morning dew is dripping into it.” Moreover, the wind is not at all from the north, as the wind vane indicates.
Some outliers are easy to spot, says Noteboom. “It happens that people have the thermometer inside. If it is 5 degrees outside, and you see a measurement of 25 degrees somewhere, then you already know that it is wrong.” But in other cases, outliers are justified. “A thunderstorm can drop the temperature in a short time.”
:strip_icc()/s3/static.nrc.nl/bvhw/files/2023/05/data100906734-dcc081.jpg|//images.nrc.nl/QQXAjAkWxA4zFjfhh2mtamOhGWk=/5760x/smart/filters:no_upscale():strip_icc()/s3/static.nrc.nl/bvhw/files/2023/05/data100906734-dcc081.jpg)
Jacob Schipper at his station, with sensors for wind speed, wind direction and precipitation. Photo Dieuwertje Bravenboer
Weird sightings
The hobbyists’ measurements will never be perfectly accurate, but when are they accurate enough? And how does the KNMI know whether that level has been reached? This is where the meteorologists in De Bilt got stuck, almost eight years after the launch of the project.
It is a problem that more crowdsourcing projects encounter, says Liesbeth Gijsel, coordinator of a Belgian platform for citizen science. “Weird observations are always coming in, and sometimes people even deliberately send in false results.”
Throughout Europe, Gijsel sees an increase in research that non-scientists can participate in, especially because everyone now has a smartphone. Different projects handle incorrect measurements differently, for example by checking outliers manually. The general rule: the more measurements, the better. With a larger amount of data, outliers are easier to pick out, and it is less bad if they are included anyway.
Noteboom is aware of this and is therefore looking for even more weather data. For example, there are also weather stations on advertising pillars along the highway, municipalities carry out weather measurements and modern cars are also eligible. “Car manufacturers fill them completely with sensors.”
Trees and population density
However large the mountain of data may be, it is still necessary to correct inaccuracies. How exactly, that is what meteorologists and data scientists are figuring out. KNMI researchers think for example, to have found a way to combine data about the landscape, such as trees and population density, with temperature measurements, making them more reliable. Other KNMI employees came up with a statistical model to adjust measurements of thermometers that get too hot because they are exposed to full sun. And yet another group found out that rain measurements around Amsterdam became much more accurate if 12 percent of the least reliable figures were deleted.
Despite all the improvements, the use of amateur weather stations is still in its infancy, says Noteboom. “It could be years before they become a permanent part of the KNMI models.”
Jacob Schipper will not lose any sleep over it. “My brother connected the station to WOW-NL. I had already forgotten that I was participating in it.”