I recently switched Météo-Parapente to the AROME forecast from Météo-France. It was both an urgent economic necessity, but also – and above all – an improvement. At least, that's what I believe.

(I'm not talking about the temporary disappearance of certain data (precipitation, thermals, wind at altitude). That's temporary. It's just that this information requires a little more work, and that for the moment I'm on the block.)

But since I switched to AROME, I hear a lot of people say that these new forecasts are completely fanciful.

Personally, that surprises me. From a scientific point of view, AROME forecasts should be much more reliable than previous forecasts. In particular, because of the better quality input data.

So I'm trying to figure out where the problem is:

• AROME, who would be wrong?
• The files made available, which would be too interpolated?
• The “Météo-France” brand, which would give the impression of a bad forecast?
• My system, in which there would be a bug?
• The user, who would misinterpret this new data?

For now, I'm missing a few clues. All I was told was that the forecasts weren't good. But without knowing neither the place, nor the day, nor the real conditions, that does not help us much. Help me !

Aiguebelette, August 28

Searching the forums, I came across this:

As this message had the merit of specifying a date, a place and the conditions, I conducted my investigation...

So I looked for Aiguebelette:

Which led me to the following predictions:

Indeed, on the right, ceiling announced around 600 meters.

But on the map on the left, I see that the red cursor is placed at the bottom of the relief.

By unchecking the interpolation box (at the bottom of the map), we see the "pixels" of the model appear. Here, the right forecast therefore corresponds to the "landing" pixel.

The colors of the map guide us. You can clearly see that the ceiling is low above the two lakes. The relief is more colorful.

So I clicked on a neighboring pixel, this time on the relief:

There, the announced ceiling exceeds 1500 meters. The forecast has nothing to do with the previous one. However, the cursor has moved very little.

A beginning of explanation?

For now, I don't have enough real feedback. This unique case does not make it possible to understand what is happening. However, some assumptions can be made.

My first hypothesis is that the AROME model works at a much finer level than before. Before, we had an average forecast, fairly consistent from one point to another. Now, we have too many details and therefore we get lost. Personally, I don't find it shocking to have 1500 of plaf on a relief and very little down at the edge of the lake. So that would mean that our old approach is no longer suitable. Either we have to point more precisely, or we have to reinvent a new way of presenting all that.

My second hypothesis is linked to the previous one: the AROME model works at a much finer level than before. But Météo-France does not provide us with the raw outputs of AROME. The files we are given are interpolated on a different grid. So we have more details than before, but the interpolation completely destroys the information by mixing everything up a bit haphazardly. Basically, we make noise. Here I am speculating a bit. I wish someone would prove me wrong.

And now ?

For now, I'll leave the site as it is. Time to see if there is something wrong and, if so, what...

If we realize that the AROME forecasts are not usable, we will set up a plan of attack. I'm not against backtracking, if that makes sense.

In the meantime, I'm counting on you to compare the forecasts to reality, and try to see if it sticks or not.