On Oct 2, 8:09*am, Dawlish wrote:
On Oct 1, 5:29*pm, Len Wood wrote:





On Oct 1, 5:24*pm, pete wrote:

What I've done is look at the forecasts and compared them to
each other. *I haven't correlated whether the weather they
forecast actually happened, just how often the forecasts changed
(since a *5 day forecast that changes the very next day can't
be relied on, so isn't any use).

As the weather got more changeable after the summer, so the
accuracy of the forecasts dropped. Compared with August's 50%
likelihood of the next-day forecast being accurate, Spetember's
40% *(i.e. it missed the mark on 3 more days) is disappointing.

The basic problem of having too many categories is even more
pronounced this time, as when I reclassified the 9 types of
weather in Septemer's forecast down to just the basics:
sun, cloud, rain the chances of being correct next day rise
to 77% and the forecast for +4 days goes from a lower than
random chance to a usable 11 days out of 30

* * * * * forecast *forecast on: * *times * num of *days
-- date -- on day * D-1 D-2 D-3 D-4 correct changes ahead
Wed *1 Sep * *S * * *S * S * S * S * *4 * * * 0 * * *4
Thu *2 Sep * *S* * *SI *SI *SI *SI * *0 * * * 1 * * *0 !
Fri *3 Sep * SI * * *S *WC *SI *SI * *2 * * * 3 * * *0
Sat *4 Sep * *S * * SI * S *SI *WC * *1 * * * 4 * * *0
Sun *5 Sep *LRS * * SI *SI * S *SI * *0 * * * 3 * * *0 !
Mon *6 Sep * HR * * HR LRS *SI *SI * *1 * * * 2 * * *1
Tue *7 Sep *HRS * *HRS HRS *HR *SI * *2 * * * 2 * * *2
Wed *8 Sep * *D * * LR LRS HRS HRS * *0 * * * 3 * * *0 !
Thu *9 Sep * SI * * *S *SI *SI HRS * *2 * * * 3 * * *0
Fri 10 Sep * GC * * GC *WC *SI *LR * *1 * * * 3 * * *1
Sat 11 Sep * SI * * LR *SI *LR *HR * *1 * * * 4 * * *0
Sun 12 Sep * *S * * SI *SI * S * S * *2 * * * 2 * * *0
Mon 13 Sep * SI * * SI *SI *SI * S * *3 * * * 1 * * *3
Tue 14 Sep * LR * * LR *LR *HR *LR * *3 * * * 2 * * *2
Wed 15 Sep * SI * * *S * S *SI *SI * *2 * * * 2 * * *0
Wed 15 Sep * SI* * * S * S *SI *SI * *2 * * * 2 * * *0
Fri 17 Sep * *S * * GC *WC *SI * S * *1 * * * 4 * * *0
Sat 18 Sep * SI * * SI *SI *HR *SI * *3 * * * 2 * * *2
Sun 19 Sep * SI* * *LR *LR * S *HR * *0 * * * 3 * * *0 !
Mon 20 Sep * SI* * *SI *SI *SI * S * *3 * * * 1 * * *3
Tue 21 Sep * SI* * *SI * S *LR * S * *1 * * * 3 * * *1
Wed 22 Sep * *S * * *S * S * S LRS * *3 * * * 1 * * *3
Thu 23 Sep *HRS * *HRS HRS LRS LRS * *2 * * * 1 * * *2
Fri 24 Sep *HRS* * HRS *LR HRS * S * *2 * * * 3 * * *1
Sat 25 Sep * *S * * SI LRS *SI *GC * *0 * * * 4 * * *0 !
Sun 26 Sep * SI * * GC *SI LRS *GC * *1 * * * 4 * * *0
Mon 27 Sep * SI* * HRS *GC *GC LRS * *0 * * * 3 * * *0 !
Tue 28 Sep * GC * * SI *SI *WC *SI * *0 * * * 3 * * *0 !
Wed 29 Sep * LR * * HR *HR *LR *GC * *1 * * * 3 * * *0
Thu 30 Sep * SI * * *S *SI *SI * S * *2 * * * 3 * * *0

For last 30 days: avg 2.50 different forecasts per day
On average forecasted 0.83 days ahead
Correct 4 days ahead: 1 times
Correct 3 days ahead: 4 times
Correct 2 days ahead: 8 times
Correct 1 day ahead: 12 times (40.0%)
Number of times the weather was same as yesterday(*): 7 times (23.3%)
Number of days every forecast was wrong(!): 7 times (23.3%)

Key:
* D: * * * * * * Drizzle *1 occurrences, predicted * 0 times under-forecast
*GC: * * * * *Grey Cloud *2 occurrences, predicted * 8 times about right
*HR: * * * * *Heavy Rain *1 occurrences, predicted * 8 times over-forecast
HRS: * Heavy Rain Shower *3 occurrences, predicted *10 times about right
*LR: * * * * *Light Rain *2 occurrences, predicted *12 times over-forecast
LRS: * Light Rain Shower *1 occurrences, predicted * 8 times over-forecast
* S: * * * * * * * Sunny *7 occurrences, predicted *26 times about right
*SI: * * Sunny Intervals 13 occurrences, predicted *43 times about right
*WC: * * * * White Cloud *0 occurrences, predicted * 5 times over-forecast

Sorry, but have I missed something? Forecasts by who? Are you saying
that the UKMO forecasts are crap?- Hide quoted text -

- Show quoted text -

It's really the kind of thing the MetO should be doing for its
customers. The analysis of model accuracy that is produced by NOAA,
doesn't reflect what happens on the ground and analyses like this do.
I'm always grateful to Pete for doing this, as it takes time and
effort.- Hide quoted text -

- Show quoted text -

I am grateful too.
Although the results look a bit alarming.
Must reflect the difficulty in trying to forecast for a specific
location. Something we all know is a tall order.

Len
Wembury