What this site does that a flat-country site cannot

Three things about Switzerland break the assumptions behind an ordinary point forecast. Each of them changes a number by several degrees, and each is handled explicitly here.

1. The model does not know where the ground is

Even at 1 km, model orography misses real terrain badly. Measured against the station list:

A raw nearest-gridpoint value is answering a question about the wrong altitude. So every temperature field on the maps is moved onto one common 1 km reference terrain before it is drawn, which is what makes flipping between models show the forecast changing rather than the orography.

2. The textbook lapse rate is wrong-signed in a cold pool

Everyone corrects temperature for altitude at 6.5 K/km. On a winter Plateau day under Hochnebel, or in a snow-covered valley at dawn, the valley floor is colder than the slope above it, and that correction makes the forecast worse rather than better. Moving a value from 900 m down to a 450 m valley floor through a 5 K inversion:

So the station product corrects along the model's own vertical stratification, which reproduces inversions instead of erasing them. The Inversion map shows where the lid is.

3. A summit is not a valley with a different number on it

Which correction is right depends on whether a site sits inside the boundary layer or above it, and the two answers differ by several kelvin. Measured against ICON-CH1 over 23 Swiss stations:

A peak like the Jungfraujoch stands in the free atmosphere, so it takes the free-air temperature from the model's pressure-level profile. A valley keeps the model's own screen-level field, because free air knows nothing about the cold pool sitting in it overnight. Jungfraujoch is the clearest case: raw +0.7 K, lapse-corrected −6.2 K, free air −0.4 K.

The snow line is a wet-bulb problem

Rain turns to snow at the wet-bulb zero, not the dry-bulb zero: falling precipitation evaporates, cools the layer it falls through, and drags snow several hundred metres below the 0 °C isotherm. MeteoSwiss computes this properly and ships it as SNOWLMT, which is what the Snow line map draws.

One trap worth stating, because it is the kind of number that looks right. In a dry subsiding column the wet-bulb zero can sit more than 1,500 m below the dry-bulb zero — over Zurich on 9 September 2026, humidity fell from 84 % at 2.6 km to 4 % at 3.5 km, putting the wet-bulb 1 °C level at 1,930 m against a freezing level of 3,650 m. That is arithmetically correct and meteorologically empty, because no precipitation exists in that column. A snow line only means something where something is falling.

Föhn, and why it needs three things at once

The Föhn map is the product of three conditions: a pressure difference across the Alps, wind at ridge height blowing across the crest, and terrain that actually descends downwind. All three have to be true together, which is what stops it lighting up on any windy day. It fires for south föhn into Altdorf and Chur, and for north föhn into Ticino.