The Windy overview of ECMWF says:

ECMWF 9km
Very accurate model provided by the 
European Centre for Medium-Range Weather
 Forecasts. Clear winner compared to other 
forecast models.

That may be true in some categories. For snow (levels and new) it is horrifically false, at least at altitude (mountains or not.)

The following is a typical example:

We live at 7050 ft on high plains in Colorado Springs. Just finished a two day round trip to Monte Vista, CO (7500 ft high plains) and nearby NWRs via La Veta Pass.

There was no snow on the ground at any time (except maybe an inch in some shaded north facing slope spots at the top of the pass.)

ECMWF thought we were seeing a foot or more of snow almost everywhere we went, and no less than a half inch except from ~ Pueblo to Walsenburg. (Ground modeled-reality and forecast)

In our region, ICON is far more accurate for snow, both existing depth and new snow.

SUGGESTIONS:

• Correct the above summary text
• Consider support for different models for different data streams.

Thbanks!

I've used this feature in the past (a few years ago.) It appears that at the very least, the UI has changed significantly since the instructions were written ( https://community.windy.com/topic/9425/how-to-use-the-observation-vs-forecast-feature. )

Anomalies I am observing, and a few questions/suggestions:

A) Select Reported Wind and a model.

• Most stations disappear (ie almost ALL) Even though they have plenty of wind reporting.
• Stations not reporting a current non-zero windspeed disappear. BUG?
• CANNOT select a time frame
• No comparison(s) show up

Much preferred the UI where all models are graphed. Oh well.

Suggestion/Question
B) I would LOVE it if we could compare much more than reported wind speed!

• Temperature of course
• But even more, for where we live (7000 ft in Colorado USA)... precipitation! Both rainfall and SNOW. Who is accurate about snow?

Suggestion/Question
C) Without information re the prediction timeframe, this feature is almost useless. Example: the HRRR model is updated SO frequently for the near future, that of course at TOBS (Time Of Observation) it is "correct". A real example:

• Last night we had ~~ 8.5-9 inches of snow.
• Yesterday noon, HRRR was predicting a total of ~~2.5 inches in the following 24 hours. All other models ranged 4.5 to 9-10 inches.
• Over the 24 hr timeframe, HRRR "predictions" kept changing. What a surprise (not): the "predictions" at TOBS were about right.

Same is true for other models. So...

Suggestion: add a dial to the feature, allowing the user to select how far in advance the models should be evaluated. Perhaps a 24 or 36 hour default?

I've used this feature in the past (a few years ago.) It appears that at the very least, the UI has changed significantly since the instructions were written ( https://community.windy.com/topic/9425/how-to-use-the-observation-vs-forecast-feature. )

Anomalies I am observing, and a few questions/suggestions:

A) Select Reported Wind and a model.

• Most stations disappear (ie almost ALL) Even though they have plenty of wind reporting.
• Stations not reporting a current non-zero windspeed disappear. BUG?
• CANNOT select a time frame
• No comparison(s) show up

Much preferred the UI where all models are graphed. Oh well.

Suggestion/Question
B) I would LOVE it if we could compare much more than reported wind speed!

• Temperature of course
• But even more, for where we live (7000 ft in Colorado USA)... precipitation! Both rainfall and SNOW. Who is accurate about snow?

Suggestion/Question
C) Without information re the prediction timeframe, this feature is almost useless. Example: the HRRR model is updated SO frequently for the near future, that of course at TOBS (Time Of Observation) it is "correct". A real example:

• Last night we had ~~ 8.5-9 inches of snow.
• Yesterday noon, HRRR was predicting a total of ~~2.5 inches in the following 24 hours. All other models ranged 4.5 to 9-10 inches.
• Over the 24 hr timeframe, HRRR "predictions" kept changing. What a surprise (not): the "predictions" at TOBS were about right.

Same is true for other models. So...

Suggestion: add a dial to the feature, allowing the user to select how far in advance the models should be evaluated. Perhaps a 24 or 36 hour default?

In some parts of the world, there are few obvious geologic features to provide context in the background map.

Right now for example, I'm wanting to see the big weather picture in SW USA. It would GREATLY help if I could see the major (interstate) highways at least, if not one or two additional highway road layers.

WIthout that, we only see OSM roads when zoomed all the way in to a local place.

I would LOVE to see more zoom levels with OSM, at least as an option! (maybe OSM-enabled-layer as an option?)

Thanks!

Has anyone else noted just how awful some of the models are, for hurricane forecasts?

I had great interest in Beryl, as a family member lives on the Yucatan peninsula. Three models available: NHC-At, UKM, ECMWF

In general, NHC-At has been pretty good, ECMWF unrealistically low, and UKM truly crazy-low.

On Wed as Beryl approached Jamaica, I noted that just after a forecast update the discrepancy became rather extreme. Current wind speed was ~140mph. The noon forecast, only a few hours later, was literally a drop to half that in the UKM model, and a bit less extreme in ECMWF. Even 3 hours before noon, with wind speed increasing, the UKM forecast only increased to 80mph.

The result: only the NHC-At model gave me any sense that it could produce a believable forecast.

Why are the other two models used at all for this purpose?

The Windy overview of ECMWF says:

ECMWF 9km
Very accurate model provided by the 
European Centre for Medium-Range Weather
 Forecasts. Clear winner compared to other 
forecast models.

That may be true in some categories. For snow (levels and new) it is horrifically false, at least at altitude (mountains or not.)

The following is a typical example:

We live at 7050 ft on high plains in Colorado Springs. Just finished a two day round trip to Monte Vista, CO (7500 ft high plains) and nearby NWRs via La Veta Pass.

There was no snow on the ground at any time (except maybe an inch in some shaded north facing slope spots at the top of the pass.)

ECMWF thought we were seeing a foot or more of snow almost everywhere we went, and no less than a half inch except from ~ Pueblo to Walsenburg. (Ground modeled-reality and forecast)

In our region, ICON is far more accurate for snow, both existing depth and new snow.

SUGGESTIONS:

• Correct the above summary text
• Consider support for different models for different data streams.

Thbanks!

@Ondřej-Šutera Is there a place for feature suggestions?

The reason for this thread is that it would be wonderful to have snow forecast in the embed forecast during cold months 🤠

Is there a limit to how many (time stamped) observations can be uploaded in a single API call? I don't see a reason why it would fail, but the docs don't say one way or another.

I'm planning to write a catch-up function for weewx similar to what is available for other wx online databases :)

@Ondřej-Šutera can I do layers in the forecast?

@Ondřej-Šutera my real question is about the embed URL interface, and how to enable temp isoline.

As you see above, I was told that isolines are not supported even in the API. Clearly they are.

Do you know if temp isoline is, or may be, supported in embed URL's?

@Ondřej-Šutera at 7000 ft in Colorado, there is very little relationship between predicted precipitation and predicted snowfall.

I just confirmed the hard way. Both guesses are correct.

When you click "show key",a new key is generated, and the existing key is no longer valid.