Windy Launches Premium 15-Day Forecast for ECMWF Model using AIFS technology
-
@Polina-Nozdrina
I can't make the payment smoothly. I've never used such a crappy payment system. -
@Victor_windy Hello, could you please provide more info about your device?
-
@浙江省应急 HI, I see you were successful at the end and your account is a Premium one. Do you still need any help?
-
Hi @Suty, it's a Samsung Galaxy S20 5G (SM-G981U1)
-
@gavin-swonnell yeah! Mine too!
-
Is their any data published on the accuracy of this model for forecasting wind, sun, rain, temperatures out to 15 days?
I find your description of the model contradictory. At the beginning it's touting using this model plan for future activities with more accuracy and at the end it's stating forecasting out 15 days is risky and caution should used for planning activities around it. Would be good to know the positive predictive value when it forecasts a windy day on a certain body of water 15 days out (ie what % of the time that it predicts a windy day does it turn out to be correct) and likewise what is the sensitivity (what percent of windy days does it truly pickup 15 days out).
We recognize predicting weather/wind is challenging but it would be helpful to know if the accuracy is a flip of the coin (50%) or more like 80-90%.
Thanks for all you have done with this great app! -
Greate, Perfect application for an AI.
Will also the current prediction be better with the AI ?
Re. The operation of the AI,- will the AI operate with constantly fintuning or will the AI training be updated as software in general, ie in steps as new releases?
For paragliding on a specific spot many times its possible understand how the local whether usually deviate from the forecasted .
IF the AI take in to account the actual weather at a spot, the local variations at this spot ought to learned by the AI.
Correct?
Then the AI also ought to make better forecasts for that specific spot even in the short and medium term forecast,I usually look the predicted weather for a place and a specific day. I then follow how the prediction changes day by day. Then it's possible to see a trend in how the forecasted weather changes. This give me a hint on how the predicted weather for that day and place is most likely to change if changed at all.
- Is this a valid way of interpret the possible the forcasted weather i.e. the possible changes in weather compared to the predicted original prediction?
Is there any way the see the calculated certainty of presented forecasts?
-
@SwairCare said in Windy Launches Premium 15-Day Forecast for ECMWF Model using AIFS technology:
Will also the current prediction be better with the AI ?
The addition of AIFS is just an extension of the current ECMWF model with lower resolution
https://community.windy.com/topic/34378/windy-launches-premium-15-day-forecast-for-ecmwf-model-using-aifs-technology/4For paragliding on a specific spot many times its possible understand how the local whether usually deviate from the forecasted .
IF the AI take in to account the actual weather at a spot, the local variations at this spot ought to learned by the AI.
Correct?
Then the AI also ought to make better forecasts for that specific spot even in the short and medium term forecast,For paragliding, in mountain I guess, you can try Meteoblue which includes AI algorithms. This model is better and more accurate in mountains. But it is only available in Basic forecast table, not as a weather map and not in Airgram / Meteogram.
Is there any way the see the calculated certainty of presented forecasts?
This is not shown in Windy.
-
I idefix37 referenced this topic on
-
First -- thank you developers for integrating this forecast.
Second --
the warnings about how the tail end of these longer forecasts are less reliable has no visual indication on the forecast, nor any way to disable this longer-term, less accurate forecast. That caveat only appears on the announcement of this product.
Likewise, there should be a visual cue that after day ~6 or so, the forecast is not of the same level of accuracy. Perhaps a change in background or hashing to visually cue the less accurate tail end.
EG When NOAA releases its extended hurricane forecasts, you'll notice that the cone gets larger and the central line of the path is disabled by default. This is because there is less certainty.
Personally, I'd rather have a button to turn off this 15-day completely and revert to the old week long forecast. It's hard to read on my iPhone and I know it's less reliable.
Thank you,
-
@atthenius
Man can’t say that the AIFS extension is less accurate, but it has a lower resolution.
See what ECMWF says about this AI model.
https://www.ecmwf.int/en/newsletter/178/news/aifs-new-ecmwf-forecasting-system
Note that its resolution is now 0.25°, same as GFS.Then there is a clear mark in the point forecast table, but not when you are just looking at the map.
And at the end of the point forecast table you can set the forecast period back without the AIFS extension.
You get only 10 days forecast in the table but also on the map.
By the way I don’t understand why it is said 7 days, or I miss something. -
I idefix37 referenced this topic on
