Loading verification data…
Observation locations & per-point scores
Bias & MAE by lead time
RMSE by lead time
Observation vs. model (scatter)
Spatial variogram — temperature variability vs. distance
Empirical semivariogram γ(h) = ½·mean[(Ti−Tj)²] computed from all observation pairs in each lead-time window. Observations (dashed) are the reference.
About this page
Each surface observation is paired with the nearest model grid point (max 0.5° distance accepted). AROME Arctic 2.5 km (00 Z and 12 Z, hourly steps, 66 h), IFS HRES 0.1° (00 Z and 12 Z, hourly steps, 72 h), AIFS 0.1° (00 Z and 12 Z, 6-hourly steps, 72 h), IFS experimental 0.1° (00 Z and 12 Z, hourly steps, 72 h, expver 80). ECMWF models downloaded on 0.1° regular grid (closest to the IFS native O1280 ≈ 9 km resolution). BIAS = mean(model−obs) — positive means model too high. MAE = mean absolute error. RMSE = root-mean-square error. Statistics updated daily.