On Tuesday, 21 June 2016 16:42:09 UTC+1, Stephen Davenport wrote:

Indeed. However, I would note that the better apps use a weighted model & ensemble mix.

Is that a plug, Stephen? (This is only said partly in jest as I'd be interested to know how the Meteogroup App works on this front).

Richard

On Tuesday, June 21, 2016 at 2:07:55 PM UTC-4, Norman Lynagh wrote:
Stephen Davenport wrote:

On Monday, June 20, 2016 at 3:00:22 AM UTC-4, Norman Lynagh wrote:


At the time that I wrote that I was looking at the 00z run of the GFS. The
rain that was predicted for Thursday on that run had disappeared on the 06z
run. It just confirms the nonsense of apps basing their predictions on one
run of one model.

========

Indeed. However, I would note that the better apps use a weighted model &
ensemble mix. Even so, expectations are usually too high for a free app.

Stephen.

I bow to your greater knowledge, Stephen :-) However, any app that gives only
categoric predictions for several days in advance has no credibility IMHO, no
matter how the content is generated. The science just isn't good enough yet.


========

I certainly won't disagree that even with those nuances deterministic forecasts more than a couple or few days ahead have, shall we say, er, limited usefulness. : - ) Especially for weather type. Temperatures verify rather better.

Did you see that Accuweather has now extended its 45-day daily forecasts to 90 days? Madness.

Stephen.

On Wednesday, June 22, 2016 at 5:49:49 AM UTC-4, Richard Dixon wrote:
On Tuesday, 21 June 2016 16:42:09 UTC+1, Stephen Davenport wrote:

Indeed. However, I would note that the better apps use a weighted model & ensemble mix.

Is that a plug, Stephen? (This is only said partly in jest as I'd be interested to know how the Meteogroup App works on this front).

Richard

========

I think a few use that sort of approach. But too many just take GFS gridpoints, I think, without even any downscaling.

I'm no expert on the specifics of apps, though.

Stephen.

On Wednesday, June 22, 2016 at 11:57:09 PM UTC+1, Stephen Davenport wrote:
On Tuesday, June 21, 2016 at 2:07:55 PM UTC-4, Norman Lynagh wrote:
Stephen Davenport wrote:

On Monday, June 20, 2016 at 3:00:22 AM UTC-4, Norman Lynagh wrote:


At the time that I wrote that I was looking at the 00z run of the GFS. The
rain that was predicted for Thursday on that run had disappeared on the 06z
run. It just confirms the nonsense of apps basing their predictions on one
run of one model.

========

Indeed. However, I would note that the better apps use a weighted model &
ensemble mix. Even so, expectations are usually too high for a free app.

Stephen.

I bow to your greater knowledge, Stephen :-) However, any app that gives only
categoric predictions for several days in advance has no credibility IMHO, no
matter how the content is generated. The science just isn't good enough yet.


========

I certainly won't disagree that even with those nuances deterministic forecasts more than a couple or few days ahead have, shall we say, er, limited usefulness. : - ) Especially for weather type. Temperatures verify rather better.

Did you see that Accuweather has now extended its 45-day daily forecasts to 90 days? Madness.

Stephen.

Except, of course, on the limited number of times when it is quite possible, with high confidence (80%+) to determine that, even at 10 days hence, they do.