Understanding the Compare Forecast Feature in Windy.com
-
@Korina I like the detail in HRRR
-
I use MBLUE mostly as it seems to be more accurate in West Sussex UK for temperature, rain, sun & wind. Probably as it has the nearest meter.
-
I use ecmwf
-
@Korina
ECMWF is by far the most accurate in my area -
@Korina MBLUE seems to work best for Minnesota
-
I
m a ocean sailor from Argentine, sailing high latitude (more than 50 degree). I always compare the four models !!! Its more accurate and make more safety our navigation...
Recomendation, always compare all models to plan a route
Josè -
@Korina Tracy from Texas here & we're melting. A combination of Basic and GFS helps me pray for rain & a change in the jet stream.
-
@Korina NAM and HRRR seem to be more accurate where I live.
-
@Korina In Belgium I use ECMWF and ICON. Both are very good.
-
@Korina i use Ecmwf raraly metoblue when in mountians or near the lakes
-
@Korina I prefer the Mixed models view. The more information, the better, as far as I’m concerned. For different seasons, and different conditions, I prefer various models.
-
@Korina In Belgium I use ECMWF and ICON. Both are very good.
-
@Korina I use the ECMWF model the most for sailing in the Med and western Atlantic
-
-
As a sailor, I always used ECMWF. This is one we can lean on especially that hurricane season here in Atlantic-Caribbean area already started. 👍♥
-
This post is deleted! -
I'm in the UK so for me the UKV model is the best predictor in the short term, although I do use the others to get a feel for the the general trends beyond the scope of UKV.
-
Here in Greece Macedonia The ECMWF 9km model has been the most reliable model.
-
ICON-D2 seems to work the best for me!
-
For one reason or another peoples selected model preference is often bias or at the very least subjective ….
Suggestion: Have a verification feature. The user selects their location and the app verifies the different model results over selected time periods … day, week, month. Start with something simple, max and min temps for example and build from there. The info would tell the user which model is actually performing the best and display model biases. This could be added as a premium feature.
