Windy Launches Premium 15-Day Forecast for ECMWF Model using AIFS technology

SwairCare

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?

idefix37

@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/4

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,

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.

atthenius

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,

idefix37

@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.