The bold line is the accuracy-weighted consensus. The shaded band shows where the models disagree — narrow band, easy forecast; wide band, pack for both. Tap the chart to jump to an hour.
Tap any hour to preview it above. Rain % = share of models forecasting rain.
In Copenhagen the models rarely miss the temperature — they miss the rain. So each model is scored above all on its rain hit rate (did it correctly call rain or no rain, hour by hour, over the past 48 hours here), plus its temperature error. Rain accuracy counts for 60% of the vote.
How it works: for hours that already happened, yesterday's forecast from each model is checked against analysed conditions over Copenhagen. Models that have been sharp lately earn a bigger vote; models that drifted get politely outvoted. That's why this consensus tends to beat any single source.