September emptied the villages, not the passes: the first numbers

How we made the comparison, and why it matters

The September days we have in full are Tuesday the 1st, Wednesday the 2nd and Thursday the 3rd. Three weekdays in a row, with no weekend inside them.

Comparing them with the whole of August, Saturdays and Sundays included, would have made September look emptier than it is. So we kept only August's weekdays, and the comparison below is weekdays against weekdays.

That is not a cosmetic caveat. At the Sella pass, August goes from 21 completely full days out of 23, which is the figure we published in the guide to the Dolomites in September and which includes weekends, to 15 out of 19 if you look at weekdays alone. Same car park, two different questions.

The occupancy below is the hourly average between 8 and 18, on the days our plausibility check considers valid.

• Monte Pana, 340 spaces: August 59%, first three days of September 15%
• Plan de Gralba, 400 spaces: August 28%, September 6%
• Sciuz, 82 spaces: August 65%, September 55%
• Sella pass, 199 spaces: August 44%, September 46%

The valley floor really did empty out

Monte Pana is the clearest case: forty-four points down. In the three September days it never dropped below 215 free spaces out of 340, which means it was never even a third full. On August weekdays it had reached zero on 7 days out of 20.

Plan de Gralba goes from 28% to 6%, and the low point of the three days is 343 free spaces out of 400. It is a large car park that in September, on weekdays, sits practically empty.

Sciuz falls the least, ten points, and stays the busiest of the three: it is the smallest, 82 spaces, and the closest to the lifts. It never filled either: the low point across the three days is 19 free spaces.

The picture on the valley floor is consistent: in September, on a Tuesday, Wednesday or Thursday, you will find a space.

The pass did not

The Sella pass is the only one of the four that does not fall. It rises by two points, which with a three-day sample simply means it stayed where it was.

And it still fills. On Tuesday the 1st it did not reach zero, but it came close: 14 free spaces out of 199 at nine in the morning. On Wednesday the 2nd it was completely full for two hours, from nine to ten. On Thursday the 3rd for four hours, from eight to eleven.

The sequence of the three days gets worse rather than better, and the hour it fills moves earlier: nine first, then eight. We are not building a trend on it, because three points are not a series, but the direction should be said.

The one thing that changes from August is the timing. On August weekdays the car park reached zero at 7:28 on average. Across the three September days, at 8:30. That is an extra hour of margin, not the freedom to arrive whenever you like.

• 1 September: low of 14 free spaces out of 199, never completely full
• 2 September: full from 9 to 10
• 3 September: full from 8 to 11
• Average hour of first filling: 7:28 on August weekdays, 8:30 across these three days

A hypothesis, and we say it as one

The explanation that comes to mind is the school calendar. The people parking on the valley floor, at Monte Pana or Plan de Gralba, are largely those bringing children, taking the lifts, walking out from the village: when schools reopen, that demand disappears almost overnight.

The people stopping at the Sella pass are there for the pass: photography, a walk from the top, motorbikes, the four-pass loop. That is demand that does not follow the school calendar, and it faces a car park of 199 spaces, which is small.

With three days we cannot prove it, and we are not presenting it as proven. It is the hypothesis we will test at the end of the month, when the days will be thirty and the weekends will be in there.

The useful part: after two the pass is empty

On all three days the Sella pass returns to 199 free spaces, that is completely empty, by 14 or 15. On the Thursday, the busiest of the three, it already had 173 free spaces at 14.

It confirms what we measured in August and wrote about in the midday window: the pressure is all in the morning, and it drains fast. In September it drains even earlier.

If you can choose your hour, the early afternoon is when the most contested car park in the Dolomites is literally free.

What this comparison is worth, and what it is not

It is worth three days. They are all weekdays, they are consecutive, and they all sit in the same week: whatever was particular about that week, the weather first of all, is inside our numbers and we cannot separate it out.

It contains no weekend, and a weekend in the Dolomites is another thing entirely. The gap between the two August figures for the Sella pass, 21 days out of 23 with weekends and 15 out of 19 without, gives the measure of how much they weigh.

And it says nothing about the rest of the month. The first three days of September are not September: they are the first three days of September.

The real assessment will come at the end of the month, with the thirty days and the weekends in it. How we keep these measurements, and what happens when a reading does not hold, is on the page about the data we check.

Three sensors out of seven said nothing

We follow seven car parks. The numbers above use four, and it is worth saying why the other three are out.

The Gardena pass still does not work. It reports 500 free spaces out of 500 for all twenty-four hours, every single day, since the first day we started recording it. That is not an empty car park: it is a sensor that does not count, and our check discards it.

The Centralpark in Ortisei is the trickiest case, because it looks alive: it sends around fourteen hundred readings a day. Except the value is always the same, so of those fourteen hundred not one is any use.

The Mont Sëuc garage does not declare its capacity. We know how many free spaces it has moment by moment, but without the total we cannot turn it into a percentage, and a percentage is what you need to compare it with the others.

Four out of seven is the real basis of this piece. We would rather say so than pad the sample with readings that do not hold.

And why 3 September nearly did not exist

One last thing, because it bears on how reliable all the rest is. Until this morning 3 September was not in our archive, and neither was the 4th.

The reason is a mistake of ours. The data collection lives inside the same script that publishes the articles, and it sits after the check that decides whether it is the right hour to publish. Today that check fired at 10:56 against an hour set at 11:00: the script exited four minutes too early, and on its way out it took the measurement with it. No article, and no data.

We recovered the days because the source keeps a few days of history, and re-reading it this morning brought back thirty-five records. Had we waited a week, they would have been lost.

We have separated the two: the data is now collected regardless, including on days when nothing goes out. It is the same class of fault as June's flat sensor: silent, green on every dashboard, and visible only by going to check whether the number is actually there.