Open the dashboard

For teachers · Worksheet 2 of 5

Why the models disagree: the 51 scenarios

A tiny error at the start can end up as very different weather a few days later. That is why weather centres do not calculate a single forecast, but many.

Aim

To understand that the atmosphere is a chaotic system, what an ensemble forecast is and how to read an uncertainty band that widens as the days go by.

Several lines leave together from the same point and spread apart towards the right: this is how the scenarios diverge as the days go by. todayin 7 days

Activity

  1. 10 min
    The paper experiment. Drop a sheet of paper five times from the same height and mark where it lands. However hard you try, it never comes out the same: a tiny difference when you let go changes its whole path. The same happens with the atmosphere.
  2. 10 min
    The band of the 51 scenarios. In the Enthusiast view, find the “Probability · ECMWF ENS” card. The temperature band runs from the coldest 10 % of scenarios to the warmest 10 %. Measure its width at the beginning and at the end of the chart.
    Open the Enthusiast view
  3. 10 min
    Six different models. In the same view, in “Model comparison” and in the meteogram, see how far the six models spread apart. Do they spread apart at the same time as the band?
  4. 10 min
    Debate. To organise a school trip five days from now, what is more useful to you: a single number or a range with its probability? Each group argues for one side.

Questions

  1. What is a scenario (or “member”) of an ensemble of forecasts?
  2. Why does the ECMWF calculate 51 scenarios instead of just one?
  3. How does the width of the band change as the days go by? What does it mean?
  4. If 40 of the 51 scenarios give rain, how would you say it in a weather forecast?
  5. How is the six models disagreeing different from the 51 scenarios disagreeing?

Answers for teachers
  1. It is a complete run of the same model that starts from a slightly different initial state, within the margin of error of the observations (and with small variations in its physics).
  2. Because the atmosphere is chaotic: small errors grow over time (this is the “butterfly effect” described by Edward Lorenz in 1963). The ECMWF ensemble (ENS) has one control run and 50 perturbed runs: 51. By counting how many scenarios do something, we estimate its probability.
  3. It widens: uncertainty grows with the lead time. One or two days ahead the band is usually narrow; a week ahead, much wider.
  4. 40 out of 51 is about 78 %: “rain likely, about 8 in 10 chances”. The dashboard puts it like this: “X in 10 scenarios”.
  5. The six models differ in how they are built (resolution, physics, how they use the observations). The 51 scenarios differ in where they start, with the same model. Both measure uncertainty, and the dashboard uses both.