With regard to forecasting accuracy, I think people’s expectations are too high.  They want all the answers at their fingertips and accurate to the n’th degree. They forget we’re dealing with nature and a slight change to one variable can skew the balance.

Originally Posted by: Caz 

I've said it before, forecasts which try to be really accurate are worse than the vague 80s style ones. I've lost count of how many times I've heard people moaning that some app showed a shower of rain at 2 PM, only for it not to happen. The old vague "sunshine and showers in the afternoon" would, paradoxically, be a better forecast in that case.

Similarly there is a "gotcha" with the ever increasing resolution of the models: it increases the risk of a phantom small-scale feature being generated, which then renders everything afterwards inaccurate... it's an object lesson why we shouldn't just look at a single run in isolation.

We see the reverse of this too, most commonly in winter, when a straight easterly 4 days out suddenly develops all sorts of kinks and features nearer the time. The details are seldom right first time, which just underscores the need for ensembles over everything else - and indeed, looking at ensembles of all the models (as Decider does, a tool the MetO use), rather than just one model suite.

 

Originally Posted by: Retron 

Anybody planning a half term holiday to Croatia or Italy next week should keep an eye on the models and forecasts. Looks like torrential rain is likely in places, no doubt with associated flooding. 

Originally Posted by: GezM 

Not sure if anyone here attended the UK Met forum yesterday, but there were some interesting pieces of news. They are intending to drop the concept of an operational run and switch over entirely to an ensemble based system in (IIRC) 2025. The resolution will be increased, and the control run will essentially be treated as the deterministic solution, although (as I understand it) it may be possible for one of the ensemble runs to be picked as the most representative solution. Also, MOGREPS-G will be run to 14 days instead of 7 days, the first week will at 10km resolution and the second week at 20km due to the computing resources required. Instead of getting 4 x 18, there will be 8 x 9 runs through a 24 hour period and this again is being done to optimize CPU cycles. Significant improvements to the physics in the model/s are also being implemented and it was interesting to hear about some of the known issues with the current suite.

Originally Posted by: Brian Gaze 

I've been following these comments with regard the supposed better accuracy of forecasts back in the days of Ceefax and Oracle/Teletext ltd with interest. I followed Ceefax etc from its inception in Sep 1974 to its demise in Oct 2012. My recollection is that it was frequently wrong, and, except under the most static of high pressure systems the weather for the week ahead in newspapers was not worth the paper printed on. I'm sorry but the rose tinted spectacles you are using when trying to assure us that in the golden days of Ceefax the forecast was never wrong and even substantially better than now is incorrect.
Lets look at some facts from when I was working in weather forecasting field.
1972. We moved to an all singing all dancing 10 level model with a 300 km grid.
1982. Upgraded to 15 level 200 km grid, which was later resolved to 150 km grid
1987. The Great Storm of 15/16 October 1987.
We were using the above model for Global Forecasting (15 levels, 1 to 2 runs per day, 150 km grid.) There was also a regional (mainly local to the British Isles finer mesh model which was used experimantally for topographically induced wind fields and local visibilities) The model had placed into it from buoys, ships, aircraft and satellite some 1200 observations over 24 hours. This was processed at around 4 million calculations per second.
And because of the lack of observation in the area of the storms development, plus late arrival of key data, the track and rapid intensification of the storm were missed.
Nowadays I believe (cos I'm now retired and dont get to see all the latest stuff!) that UKV can resolve convective weather to 1.5 KM resolution and soundings through the whole thickness of atmosphere.
The Cray XC40 supercomputer is the one I belive is now in use and has 46000 cores, and top speed of processing of 16 Petaflops (1 petaflop = 10 to the 15 calcs per second.)
Well I could go on and bore everyone, but I think it is safe to say the the assertion that the models were better and more accurate back in the days of good old CEEFAX, are ..well. silly. If you want a good read about Met Office devlopments since the 1950's Try here... it gives tha facts rather tahn rose spectacled opinions. https://www.metoffice.gov.uk/weather/learn-about/how-forecasts-are-made/computer-models/history-of-numerical-weather-prediction 
 

Originally Posted by: Bow Echo 

Huge mishaps in the big rooms at Exeter the last 36hrs.
Major warnings cancelled at very short notice. 
Yellow warnings ‘removed’ (not seen this before) and then Red high alert warnings cut in geographical size by 70%!!!
Definately some sensationalist fired today. 

Originally Posted by: White Meadows 

Huge mishaps in the big rooms at Exeter the last 36hrs.
Major warnings cancelled at very short notice. 
Yellow warnings ‘removed’ (not seen this before) and then Red high alert warnings cut in geographical size by 70%!!!
Definately some sensationalist fired today. 

Originally Posted by: White Meadows 

The red warning was not “cut in size”. A new ted warning, covering a smaller area, was added after the expiry of the original one. 

Originally Posted by: Rob K 

WX temp charts for week 1 show below norm for NE Europe, well above in the Balkans, and close to norm for W Europe. In week 2 the cold weather intensifies and expands W to Scandinavia and S through Russia; any warmth confined to shores of the Med. W Europe still close to norm. Rain on Atlantic coasts for the coming fortnight - Portugal, France, and up to S Norway, not forgetting Britain.

GFS Op - Currently a multi-centred trough from Iceland to Portugal, this wrapping up to leave remnants as 995mb N Sea Tue 24th while a larger Atlantic depression approaches, pushing a broad trough ahead of it, to reach 970mb SW Ireland Fri 27th. This meanders over Britain until Wed 1st when there is a narrow N_S ridge of HP over Britain, sandwiched between LP Denmark (N-lies for E coast) and the remains of TS Tammy (S-lies for W coast). This persists until the end of the week but is finally displaced by LP S of Iceland, SW gales Sat 4th followed by (cold) NW-ly gales Sun 5th.

ECM - similar to GFS; the 'meandering' LP stays nearer or just off W coasts rather than centred over Britain. Yesterday ECM was showing a strong rise of pressure Tue 31st but this seems to have gone.

GEFS - temps varying a little either side of norm (day-night 4C) with good ens agreement to Wed 1st, after which mean remains near norm but with increasing spread balanced on either side. A dry day or two, after which rain in many places at many times, heaviest in w/b Tue 24th, least wet in NE England and SE Scotland (and they need a break!), wettest in S & SW England. 

Originally Posted by: DEW 

It's not been that wet so far this autumn here when compared to many other places in the UK. Not that many will notice here. I'm sure it will change at some stage. Perhaps the GEFS is onto something? What is notable are the large swells that come up the Channel from time to time from Atlantic storm activity. The other day some really powerful deep ones swept in like tsunamis up a beach I was on.  

Originally Posted by: NMA 

Next week it's the turn of the South?

Accumulations of rainfall from ECMWF high res in round terms, each on top of 20mm or so earlier in the week:

S Wales, Gower - 110mm Wed into Thu & another 40mm by Tue 31st
N Devon, Bideford - 120mm Wed into Fri, likewise more to come
S Coast, just SW of the Isle of Wight - 100mm Thu into Sat
Sussex, Brighton - 80mm Thu into Sat
Cap Gris Nez, only just across from Dover - 200mm Thu into Sat

The heavy rain isn't predicted to move far inland, with, say, the line of the M4 only getting 20mm or so on top of the 10mm earlier in the week. But it'll only need a slight shift in the predicted boundaries....

Originally Posted by: DEW