Will it be 42 degrees next week or will the mercury remain at a ‘mean’ degree or 25. It is not often that the weather models disagree so much. Why are they suddenly measuring such different temperatures?

Samsung or iPhone

With multiple calculations for the beginning of next week, the American weather model comes to temperatures above 40 degrees, while the European calculations keep it at much lower temperatures of around 25 to 30 degrees. Meteorologist Peter Kuipers Munneke explains to Scientias.nl: “An explanation for the differences between the models lies in the fact that certain processes in such a model are calculated slightly differently and that both models also have a different resolution, for example with regard to the formation and dissolution of clouds and the evaporation of moisture. In the somewhat longer term, this contributes to uncertainty in expectations. Compare it to a Samsung and an iPhone. You can take a photo with both devices, but one photo is sharper than the other and due to different lens choices, the proportions and colors can result in a slightly different photo.”

Small shifts, big fluctuations

It is these small differences that sometimes lead to large temperature fluctuations. In addition to a main calculation, a weather model consists of about fifty measurements of small shifts (the so-called plume forecast, see box). It is precisely in some of those measurements that the American model reaches record temperatures. The fact that the models draw such different conclusions does not mean that they are not reliable. Kuipers Munneke: “Weather models have improved enormously in recent decades. For the short term of one to three days they are very reliable. For the medium term of three to eight days, the models have become very good at indicating how certain the expectation is over the past ten years. And we see that in the current situation: the forecast for the weekend and the days after (six to nine days ahead) is very uncertain.”

weather forecast. Source: KNMI

Climate change

How uncertain a forecast is also depends on the weather type and the expected air currents. “The main reason for the uncertainty is that warm and cooler air will be close to each other,” explains the NOS weatherman. “For France we can be sure that it will be warm this weekend and for Denmark it will remain cooler. But where the exact limit is is a matter of a few hundred kilometers and that is actually a detail in the period of a week in advance. With major consequences for expectations.”

However, global warming has no effect on the reliability of weather forecasts. “The reliability of the weather models continues to improve over time and you can safely say that the improvements due to more computing power and better formulas are faster than the unpredictability of the atmosphere itself.”

the plume

When it comes to the longer-term weather forecast, the plume quickly comes up. KNMI calculates this plume forecast using the Ensemble Prediction System (EPS). The name plume refers to the lines in the graph that, due to their variation, somewhat resemble a plume of smoke. If the plume of smoke fans out far, the differences between the calculations are large and so is the uncertainty of the prediction. Suppose that two calculations result in a temperature of 42 degrees, then that means that the probability of this is about 5 percent.

To arrive at a weather forecast, the main calculation of the ECMF (the European weather model), the so-called operational forecast, is used, supplemented with 51 forecasts of a lower resolution. A small variation of the starting situation is always taken into account. This simulates the uncertainties of the weather. Think of a slightly different air pressure or temperature. Each of those 52 calculations comes with its own prediction in terms of temperature, precipitation, air pressure and wind. The calculations are performed every twelve hours. This creates a continuously updated weather picture, including uncertainties.

Because no matter how advanced the weather models are, it is still difficult to predict the weather more than a week in advance.