Why same model (ECMWF) differ in other sites/apps
-
I need an explanation:
Comparing two sites (windy.com and windy.app) for the forecast of absolutely the same spot in the mountain using absolutely the same model - ECMFW, there is great difference in the temperature, snow accumulation etc.
Windy.com says there would be 0.1cm of snow accumulation but windy.app says there will be 9cm. The temperature differ from -10C (windy.app) to -5C (windy.com).
I am talking about a forecast for the same day. Not 6 days from now.
Using the premium version of both sites and once again - the same model!Thanks.
-
Check these 2 points:
- Are the run time (reference time) exactly the same in both apps when you make your observation ?
- Is the resolution of the model the same ? In windy.com the resolution of the ECMWF model is 9km: This is the high resolution version which requires high costs. Many weather apps use the free version of this model with a resolution of around 40 km.
-
In case you don’t know, Windy.com and Windy.app are two different companies. The code behind the two forecast displays will likely be different.
The other thing you need to check is the Reference time for the forecast. This is the time that the predictive model is loaded with current conditions. It can take 6 hours of compute time before the model actually comes out with a new forecast.
Some sites are quicker than others in updating when a new forecast is released. Windy.com is pretty quick so it could be that Windy.app is still displaying the previous forecast update while Windy.com is showing the latest one. It that case, predictions can vary a lot. Alway check the forecasts Reference time, and make sure they agree, if you want to compare them.
-
I am completely aware that windy.com and .app are different things. That's why I am comparing them.
I check and compare them throughout the day a few times. And not on one occasion as well. For different spots as well.
The reference time is almost exactly the same - about an hour difference. The model in windy.app is 14km resolutions as per their site. I don't think that's so great of a difference.
Anyway, it's just frustrating. I am mainly a winter user and the main objective is temp, snow, wind. I just wanted to know where the difference in snow accumulation comes from, because if one of the sites was saying snow 1,00m and the other 1,10m - that's fine, 10% difference not big deal. But between 10cm and 0cm it's quite the opposite. Anyhow, it is what it is.
-
You may try to compare the Snow depth by ECMWF themselves {run time 12z - 01 27)
With the same parameter at same date by windy.com (same run time ):
These 2 maps are identical.
-
Alright, let's try again. Here are two screenshots - windy.app and windy.com showing absolutely the same spot at the same time. One of the forecasts has been updated almost 1 hr ago, and the other one - almost 4hrs ago, hence 3hrs difference (but I must say that there was the same pattern of discrepancy between the two even when the update times were in favour of the other app.)
As you may see: Per windy.app the calculated snowfall is almost TRIPLE as much as per windy.com.
Something is just not right - I cannot wrap my head around how the same forecast model has that much of discrepancy between two different apps. The 3hrs difference in the update time cannot be a legitimate reason. If a forecast model cannot predict the weather for the SAME day why does it exist at all?
-
Hi @Dimitar-Mihaylov, according to this article, windy.app uses some other model from ECWMF (14km resolution, updated twice a day). ECMWF-HRES model on windy.com has 9km resolution and is updated 4 times a day.
-
In fact for the atmosphere parameters windy.app is probably using the free version of IFS which comes with a 25km resolution. It is called “WMO essentials”.
But for sea parameters they use the 14km resolution model.
At ECMWF there is no 14km resolution model for the atmosphere parameters as you can see :
https://confluence.ecmwf.int/plugins/servlet/mobile?contentId=324860211#content/view/324860211
They deceive the users using the 14 km resolution as the general resolution of the ECMWF forecast they show. -
At the altitude of your ski resort, 25km resolution vs 9km resolution that makes a big difference. In addition model updates can be different.
Check the run time of the models, called the Reference time in Windy. -
Well, okay, so we can assume the windy.com forecast should be more accurate.
Would've been great if I could stay at a mountain hut and just make a prospective trail comparing the two forecasts :)
Anyhow.
One more question which is offtopic - as I can see few of you are developers/moderators - do you consider adding a "freezing level" line as in the windy.app interface? That feature I must say is really handy. Thanks.
-
Hi @Dimitar-Mihaylov, already available: https://www.windy.com/-Show---add-more-layers/overlays?deg0,48.546,31.399,5
-
@Filip_K said in Why same model (ECMWF) differ in other sites/apps:
Hi @Dimitar-Mihaylov, already available: https://www.windy.com/-Show---add-more-layers/overlays?deg0,48.546,31.399,5
Oh, great.
Can I add it into the "forecast window" with the wind/temp/snow/rain/etc or I can see it only on the map when I move it around? Because when I choose one of my saved favourite spots the freezing level is not shown.
Thanks again.
-
@Dimitar-Mihaylov
I'm afraid only map and airgram display freezing level.In airgram, freezing level is the white line between blue and green mass. And the left axis is in pressure levels which makes it quite difficult to get actual freezing altitude.
But picker displays such value anywhere:
-
Yes, that's what I thought. It's not really useful in that configuration but hopefully it will get added in the forecast grid.
-
@Dimitar-Mihaylov
What you call the “forecast grid” is probably the Basic meteogram called Basic in Windy. For the freezing level you just need to switch on Airgram, as @Filip_K said.
I don’t know if a customisable meteogram could be made. Probably not easy. -
@idefix37 said in Why same model (ECMWF) differ in other sites/apps:
@Dimitar-Mihaylov
What you call the “forecast grid” is probably the Basic meteogram called Basic in Windy. For the freezing level you just need to switch on Airgram, as @Filip_K said.
I don’t know if a customisable meteogram could be made. Probably not easy.I checked the Airgram but as mentioned before - the reference axis for the freezing level is air pressure not altitude which is not useful, at least not for me.
-
@Dimitar-Mihaylov
Yes, but you must know that the weather model calculations are all based on air pressure levels, not on real elevation.
However it would be useful to display the average elevation corresponding to these pressure levels. As it is displayed just below for clouds but not very detailed:Or you need to make your own pressure / averaged elevation scale yourself.
-
@idefix37 said in Why same model (ECMWF) differ in other sites/apps:
@Dimitar-Mihaylov
Yes, but you must know that the weather model calculations are all based on air pressure levels, not on real elevation.
However it would be useful to display the average elevation corresponding to these pressure levels. As it is displayed just below for clouds but not very detailed:Or you need to make your own pressure / averaged elevation scale yourself.
I understand and know that most of the calculations are based on the air pressure levels but that doesn't change the fact it's not convenient for the normal user.
This is exactly every application's "you had one job" thing - to process the information and present it in a understandable way.
If I wanted to do my own calculations with the raw data then I would not use any app at all.
Anyway - that was just a suggestion for improving the UI.
-
@Dimitar-Mihaylov
I do not ask you to do your own calculations with the raw data, just to imagine thisThis has been already asked to Windy…