Skip to content

Validation (Val-Lab)

The Val-Lab tab shows how the model compares with independent observations. It appears where satellite or station data exist for the region. Opening it shows a chooser with two areas: Satellite and Stations. ← All validation tools returns to the chooser.

Satellite

Compares modelled snow cover with MODIS fractional snow-covered area (fSCA).

  • A side-by-side view: model on one side, satellite on the other, with a ▶ button to step through the days.
  • An fSCA / Binary toggle. fSCA compares the fraction of each pixel covered by snow. Binary compares snow/no snow, which is less sensitive to how MODIS estimates partial cover.

What to look for: the snow line and its movement up and down slopes during the season. Gaps where cloud hid the ground are filled from neighbouring days in the MODIS product, so a single day can be off; trends over weeks are reliable.

Stations

Compares the model with station measurements (temperature, precipitation, snow depth…, as available).

  • Map mode: Cluster, Point, 90 m, Cluster+lapse. These are different ways of reading the model at the station:
    • Cluster: the value of the terrain unit containing the station.
    • Point: a run done specifically for the station's own elevation and exposure.
    • 90 m: the value from the 90 m grid pixel.
    • Cluster+lapse: the unit's value corrected to the station's elevation.
  • View: Map (bias at each station) or Bias structure (bias against elevation, season…).

Clicking a station opens the station panel with the two time series, Downscaled and Observed, and summary scores.

A station is a point; a unit is an area

Even a perfect model would not match a station exactly. A valley station sits in a unit whose mean elevation may be hundreds of metres higher, and wind and snow vary over metres. Large differences at a few stations are expected; a consistent bias across many stations is meaningful.