Which model is most accurate for Europe?
-
The default for Europe is ECMWF but NEMS and ICON have higher resolution. Is there some statistical comparison of these models, i.e. what these models predicted versus actual observations from weather stations across Europe?
Not sure how should I choose between the models, I assume the default is the best in most aspects?
-
@fh In my personal experience, both ECMWF and ICON EU are very accurate. Maybe ICON is better cause it has a slightly better resolution. I don't use NEMS that much, because it lacks some computing power to compute certain perimeters.
-
@theradarguy
@fh
However NEMS is much more accurate in mountainous region, in particular for temperature forecast, where other models fail.
You can do your own observation: Select « Reported temp. » or « Reported wind » and choose a weather station close to your location. Then you can compare which model show the closest values.
I know comparison of ECMWF model with few models including GFS in north extra tropical hemisphere, but not with ICON nor NEMS.
-
@fh
My opinion:
ECMWF is better in forecasting rain and wind (but overestimates wind gusts).
ICON (because of better resolution) forecasts better the temperature, low cloud cover and wind gusts.For further reading:
http://apps.ecmwf.int/wmolcdnv/scores/surface.mean/tp
https://www.wpc.ncep.noaa.gov/html/model2.shtml#verification
https://www.dwd.de/EN/research/weatherforecasting/num_modelling/05_verification/verification_node.html
-
Looking at the temperature forecast for now at my location (Prague), the observed temperature is 20 °C, EMCWF shows 26 °C, the other models 23 °C.
And looking at the peak Snežka (1602 m), EMCWF shows 20 °C, observed temperature is 10 °C. GFS shows 23 °C (lol).
Weather.com is almost spot on for both locations, I wonder what model they use.
-
@fh Weather.com uses a combination of weather models that make it more accurate. Which is the ideal way to forecast weather. The more models you have to compare, the more accurate the forecast will be.
-
@TheRadarGuy I'd guess that weather.com also perform local adjustments to the numerical prediction forecast(s) based on terrain, nearby meteorological stations, satellite data, etc.
It would make sense because in one cell of a 9 km grid, the elevation and therefore ground temperature can vary wildly.
BTW, I found that Weather Underground (owned by owners of weather.com) use their own model for US locations with hourly updates: https://www.wunderground.com/about/data. Dark Sky also use custom model: https://status.darksky.net/2017/06/14/hourly-updates.html.
