Track Forecasts for a day-time/location over time
-
Long time Windy subscriber (and user) here.
We're sailors, and it's -super useful- to track the forecast for a particular day & location as it evolves.
Use case:
- I have a regatta in Biscayne Bay, Florida, USA on a Saturday 10 days away.
- "Mark this spot to capture" daily snapshots for the next 10 days of what the wind forecast is for the available models
- Review UI - of how the forecast has evolved day to day - is it stable, is it strobing, is it all over the place?
- After the target date has passed, post-mortem review: how did this model, at this point in time, track against observed weather at a nearby weather station (say, VAKF1 for Biscayne Bay FL).
We might be able to say, based on reviewing the data, things like
- "GFS 48hrs out was spot on"
- "ECMWF was mostly stable over the 10 days, and was close to the observed conditions"
- "ICON was accurate on the timing of the afternoon wind shift, 24hrs out"
We race in locales with stable forecasts, and in locales with... less stable forecasts. We know the date we'll be sailing. Sometimes it's accurate for an East wind, and a West wind means lower accuracy.
I would expect similar things for VFR flights, oceanic crossings – you know you want to fly or sail on this date and time – not just what's the forecast at this point in time, but is it stable, is it reliable?
If any of this is useful, interesting, happy to flesh out the use cases more. Not a trivial feature req. But it seems to me that you have all the underlying bits of data and UI parts.
It'd need additional storage for the requested captures – but that it'd also be a useful paid feature to offer, and a sticky feature too, a reason for customers to stay subscribed.
-
@martin-langhoff said in Track Forecasts for a day-time/location over time:
But it seems to me that you have all the underlying bits of data and UI parts.
First, Windy would have to archive the forecast data from each model, which is not currently done as far as I know.
Windy already offers a comparison of the readings from a weather station and the forecasts from the different models at same time, but not at further lead time.
Have you tried the “Observation vs forecast” feature ?
-
@idefix37 you picked one of my favorite weather stations. :-)
Yes I've seen that feature. It's intriguing as it hints at the kind of feature I would like to see.
The comparison is made against presumably the last / most recent forecast before the observation. So that might be 1hr before the observation. That's not a very interesting forecast point – "is this model accurate 1hr out" is not a high value question. :-) – I'm interested in what the forecast said, say, 24hrs before the observation.
And yes, I'd expect this to involve some additional data retention.
-
@martin-langhoff
Yes, but this feature will highlight if a model shows bias at this location. For example at a weather station by the sea, this feature will show that some models are wrong for temperature prediction because some of their grid points are above the sea, and so the prediction is too low during the daytime and too high during the nighttime.
Of course it does not allow to say if a model is efficient for a 10-day time frame.
