GFS (NOAA) Forecast Inaccuracies
The last few weeks, the GFS (NOAA) forecast has not appeared accurate for my location (48.29/-122.64?48.082). The high temps are always about 10 degrees below all the other forecasts when displayed in "compare" mode (image 1 below). The GFS temps are also lower than the temps NOAA shows on its own hourly forecast page for the same location (image 2 below). Again, this forecast inconsistency has appeared only in the last few weeks. Prior to this GFS was much more consistent with the other forecasts and the NOAA page.
@needsomecoffee Hello, I suggest trying other models available for your area to compare their results.
@korina I am using Windy's 5-forecast compare, so I am using different models. Just now I have only 4 to compare as the GFS data Windy is getting from NOAA's data appears to be wrong. Not a big deal, but it was not a problem previously. Something appears to be in error re: how Windy is pulling/using the NOAA data. NOAA data, as per my second post, is pretty easy to review on NOAA's site. The data pretty clearly shows Windy's error. Thanks for the reply.
The discrepancy that you have noticed is not observed everywhere. Here in Western Europe GFS is matching quite well to other models even if it is generally 3°C above the others for the max temp.
So, it would be useful that you give your approx. location.
I put the lat/long coordinates for my location (North Whidbey Island, WA) in my original post: 48.29/-122.64.
I just checked the daily high temps for GFS (NOAA) number from Windy 5-fcst compare vs. the NOAA forecast for the same location.
Fri Aug 12: Compare -5F lower
Sat Aug 13: Compare -5F lower
Sun Aug 14: Compare -8F lower
I have seen some other posts where this inconsistency is being noted. It may not be at all locations.
Hope it gets fixed, but this thread does not suggest Windy engineering will investigate. Unless they are grabbing the NOAA number from a non-NOAA site, this does not make sense.
Not too big a loss as GFS is less accurate overall vs. the other 4 models which I tend to rely on more.
I have not found a GFS prediction for this island on a NOAA website. On 2 other weather apps the discrepancy of GFS and other models is approximately the same as that seen on Windy. This is confirms by the Observation vs forecast tool in Windy.
How can we explain this discrepancy? GFS has the lowest resolution of all model used in Windy. The GFS grid has a 22km spacing which means that if a point is on the island, the others are probably offshore. The maximum temperature over the waters is much lower than over the land. Then between grid points the temperature is interpolated and smoothed. So with GFS you cannot get a precise maximum temperature in this island.
You can see the effect of model resolution by comparing GFS 22 km and HRRR 4km
GFS 22 km
Due to higher resolution HRRR shows definitely a better contrast between sea and land than GFS.
As I noted, the GFS/NOAA model in my 5-model comparison is the least accurate so I use it the least. Also, I have the NOAA hourly forecast bookmarked, so it is just one click to get the the "real" NOAA forecast.
Cliff Mass (University of Washington Professor) has a blog where he has described why NOAA's models are less accurate (standard USA stupid stuff).
The weird Windy !== NOAA numbers problem exists for my location. So now my "Compare" is between 4 models not 5 unless this gets fixed. (Again, this problem appeared recently. It did not exist previously.)
unless this gets fixed. (Again, this problem appeared recently. It did not exist previously.)
Windy will not fix the GFS temperature issue on this island. Windy cannot change the GFS resolution.
To be more clear, here are the 4 points of GFS grid as shown by https://spotwx.com/ . One is on the island, the 3 others are over the sea. Interpolating (averaging) the temperatures of these 4 points cannot give a good temperature over the land of the island in the hotter hours in summer.
BTW HRRR is more precise because the 4 points are on the land.
Why does the NOAA give a good temperature with GFS?
Probably because they use the nearest point of the 4, in this case that on the land. They do not average the 4 temperatures.
Why can’t Windy do the same instead of averaging the 4 temperatures?
Because this is the only way to draw a temperature map. The ‘nearest point’ can only be used for a point forecast.
Why some months ago, GFS was providing correct temperatures in Windy?
Probably because at that time of the year, the temperature over the land and that over the sea were not so different as they are in summer.