The house that Jack built

This is the purpose-designed house that Jack built.

This is the Cray XC40 supercomputer that cost £97 million, has 1.6 petabytes of memory and runs over 16 trillion calculations per second that sits in the purpose-designed house that Jack built

This is the sophisticated NWP model, that runs on the Cray XC40 supercomputer, that sits in the purpose-built house that Jack built

This is the last night’s forecast produced by the sophisticated NWP model, that runs on the supercomputer, that sits in the purpose-built house that Jack built.

These are the actual temperatures for 04 UTC close to the house that Jack built.

In my reworking of the old nursery rhyme, Jack is of course synonymous with the Met Office, if you’ve not already guessed! What I’m trying to say I suppose is this:

You may have the fastest and the most accurate forecast model in the world, but if you can’t visualise what the model is telling you, then you may be just wasting your time.

These NWP  models generate so much highly detailed forecast data, a lot of which forecasters can’t possibly assimilate, and getting the salient facts about the weather across to the public, such as the overnight minimum temperatures seems to elude them. Compare the forecast temperatures in the graphic from the BBC with the actual temperatures at 04 UTC this morning, and you will notice that they are around 3°C too high in most of Devon. It makes no difference that the Cray XC40 supercomputer is located just 2.26 km to the west of the airport, and even though the mesoscale output from the model may have correctly forecast the temperature at 04 UTC, the BBC graphics are still wrong.

Some may argue that the temperature that the BBC show are for “towns and cities”, but that to me is just a clever get out on their part. There is a problem forecasting extreme temperatures, especially overnight minimum temperatures which has never successfully been resolved, in fact I would say that little effort has ever been expended either by the Met Office, or the company that provides the BBC graphics in doing so. Let’s hope that Meteogroup, in these days of 4K television, come up with an improved way of displaying extreme temperatures which more accurately reflects what the model is  forecasting when they finally take over the service next March.

Ex-tropical storm Rina holds up cold front clearance on Saturday

Figure 1 – Data courtesy of the NHC (approximate GFS positions from the 00 UTC 10 November model run)

I wondered what was holding up the cold front clearance on Saturday across southwestern parts of England (fig 2), but then I noticed on the Berlin Meteorological Institutes website, that the second shallow low the follows behind the low that tracks WNW- ESE across Ireland, Wales and the southeast of England during Saturday morning, was labeled ex-tropical storm Rina (fig 1). The Met Office of course are having none of that because it wouldn’t be correct would it.

Figure 2

Will Rina be the finale of the Atlantic hurricane season?

The last advisory on Rina highlighted what had been yet another rather unusual tropical cyclone in 2017 (fig 3):

Figure 3 – Courtesy of the NHC

Rina lives on in Europe

Interestingly, that rather shallow low which was Rina was when it crosses the UK is forecast to develop into quite a deep low of 991 hPa by 12 UTC on Sunday, as it tracks southeast across the Alps into the northern Adriatic, and quite a significant feature in that part of Europe (fig 4).

Figure 4 – Courtesy of the Met Office

Yellow Warning for rain?

Even more interestingly, so far the Met Office haven’t issued any yellow warnings for heavy rain in southwestern parts during Saturday. I just wondered if the tropical origins of the air might even enhance the rainfall in theses parts.

Graphics speak with forked tongue Kemosabe

Figure 1 – Courtesy of the BBC

There are numerous problems that I see with this approach to forecasting the overnight minimum temperature that David Braine uses quite regularly in the weather forecast on BBC southwest, and last night’s forecast he gave was a case in point (fig 1):

  • The colour contoured temperatures are invariably at odds with the individual spot values.
  • You never know the exact location of any of the spot values, for instance is the 8 for Sennen or the 1 for Sherborne?
  • Are the temperatures from two separate models, and why are they usually so different and misleading?
  • I know that we live on a peninsula down here in the southwest, but why are 6 out of the 8 spot values at coastal site?

I have a theory that the colour contours show the minimum temperature for the overnight period, and the spot values are the forecast temperature, in this case for 05 UTC even though the minimum will occur close to dawn. If what I believe is the case the solution looks simple.

  • Never combine the extreme colour contoured temperatures for a period with spot values for a fixed time.
  • Instead of using spot values for a fixed time pick out temperatures for towns and cities from the colour contoured value.
  • Use slightly higher resolution data and finer contours to highlight the differences between valleys and moorland. In Anticyclonic situations the valleys will be much colder that the hills.
  • Zoom in a lot more, and pan from west to east, at least 75% of the area in the forecast for the southwest is the open sea!

Here for the record are last nights minimum temperatures from 1800 to 0600 UTC (fig 2).

Figure 2

Guardian: BBC extends Met Office weather contract due to delays in switch to rival

The big question that I have in this fiasco that seems to have been dragging on for years is:

Just who will provide the raw NWP data that Meteogroup will use in their new state-of-the-art graphics solution?

If it’s to be Met Office NWP data, all I can imagine is been the price that the Met Office have been demanding for it that’s been the sticking point. It’s well-known that the Met Office are very restrictive in publishing our NWP data on the Internet. I imagine that up until now Meteogroup have made extensive use of free American GFS model data in most of their products, and you can understand if they stump up millions a year to use Met Office model data, how they might like to use it in other commercial services that they have, perhaps for use in other countries across Europe. If the use of Met Office data is restricted for use with only the BBC, that would mean that Meteogroup will have to run dual NWP systems, which might be a bit of headache at times, especially as the two models can differ markedly after T+36.

We will know more next March, that’s if they can stick to this timetable. All that I can say is that there is a lot of expectation riding on this new Meteogroup service, lets hope when it finally does take off, it’s all been worth the effort on the BBC’s part.

A critique of the October news release

I should skip this post now if you don’t like me having a go at what the Met Office push out in their various social media outlets, because I’m just going to give an honest critique of what I think of their latest news release that’s looking back at the weather of October 2017:

Textual content

I can’t really fault what they say about last month’s weather, I’ve written quite a few monthly weather diaries for the UK myself, and know that it’s not easy to condense everything into one and keep it readable, although I do wish they would fully justify it!

Colour contoured charts

I’ve always like the colour contoured climate charts that the Met Office produce. Although they haven’t changed for at least twenty years or more it still looks professional, but I think it’s about time to see what ArcGIS can do.

Climate tables

What let’s the whole article down in my opinion, is the inclusion of the badly formatted HTML tables of climate values (fig 1). This is what’s wrong with them in my opinion:

  • Both tables are left aligned on the page and not centred.
  • The values in the columns are left and not right aligned.
  • The titles in the header are left aligned and not centred.
  • Why not use the word ‘anomaly’ or simply ‘anom’ rather than ‘Diff from average’?
  • Why not combine the two tables into one like this (fig 2):

Figure 2

I’ve generated this graphic from a bespoke application that I use to downloads and visualise the free gridded monthly data that the Met Office make available. Obviously this is a graphic file, but I could have just as easily output HTML. I think you’ll agree that it’s much more professional and more importantly easy to read than the Met Office offering? (If not get an appointment with your optician as soon as you can). The good thing about doing this in software, is that it’ll work just as well for next month, as it would for January 1963. Here’s a screen shot with all the regional values for January 1963 as an example (fig 3).

Figure 3

Overnight fog rather more extensive and earlier than expected

Figure 1

The overnight fog arrived much earlier and was a little more extensive than the Met Office warned of their yellow warning issued yesterday afternoon (fig 1). The warning was for the period starting 02 UTC this morning (fig 2), but there were visibilities of 100 metres or less at Exeter airport from as early as 21 UTC, which must be quite embarrassing for forecasters, when the airport is such a well-known frost hollow and fog trap. In fact by 20 UTC, under an almost full moon you could see the Culm Valley where we live 8 km to the north filled with fog. The main rail line to London and the M5 also run up the Culm valley towards Taunton. Fog did take longer to form further east, but the Met Office had to extend the area at 0720 UTC this morning to cover areas to the west of London. I wonder if the TAF for Exeter airport required amending?

Figure 2 – Courtesy of the Met Office

Dunkeswell in fog at 20 UTC

Dunkeswell have gone into fog at 20 UTC with a temperature of 4.7°C (fig 1). Maybe my adapted Middle Wallop fog TDA had the right idea after all! Certainly the yellow warning for fog that the Met Office issued this afternoon seems to have already gone awry, because the validity time doesn’t start till 02 UTC! Of course technically it may not be true radiation fog, and could well be upslope stratus, but 100 metres visibility is still an F in the Beauforts in my book.

Figure 1

Tonights yellow alert for fog and the adapted Middle Wallop technique

Figure 1 – Courtesy of the Met Office

I was only saying complaining after the last yellow alert for fog that the Met Office issued, that a bit of advanced notice wouldn’t be a bad idea, and lo and behold, at least with this one, they have!

Coincidentally, I’ve recently been working on a Tactical Decision Aid [TDA] for the forecasting of fog. The idea is based on the Adapted Middle Wallop Technique which featured in an article in the September edition of the Weather magazine. I’ve written an application that downloads forecast data from the Met Office DataPoint web service and observational SYNOP data from OGIMET. This means that you can point the TDA at any of around 100 locations around the UK and check if fog is likely in the coming night. I’ve run out of steam with its development in the last couple of weeks, so it’s still very much a beta, all I need is to come up with a burst of enthusiasm to finish it off (fig 2).

Figure 2

As you can see, I still have work to do on the actual calculation of the forecast visibility. In this example for tonight at Yeovilton, I have it that will go into fog by 20 UTC with a temperature of 7°C, which looks way too early to me, so this is still very much work in progress. Let me know what you think of the idea and how useful it might be, I personally feel it’s just got to be better than doing it in a spreadsheet.


I have just come across a very interesting website which features a number of TDA online tools, one of them being for forecasting fog that uses the Adapted Middle Wallop technique. Isn’t it amazing what you can find find when you look!

Latest outlook from the Met Office

The Met Office have said in their latest 3-month outlook, that the period between November and the end of January 2018 is likely to be mild and wet on the whole (fig 1 & 2). As usual, they’ve covered just about every other possible contingency with phrases such as:

  • When the MJO is in this phase it favours more blocked weather types…
  •  …an increased chance of high pressure and a shift to cooler conditions in early November. (really cooler in November?)
  • …although some show more likelihood of high-pressure patterns over the North Atlantic implying a greater likelihood of northerly or north-westerly winds.
  • …there are also indications that blocking high-pressure patterns, possibly linked to the current MJO event, will
    influence the UK during parts of the month (November). 
  • …the risk of colder-than-normal conditions remain a significant possibility
  • …favouring weather patterns associated with colder-than-normal-weather

Figure 1 – Courtesy of the Met Office

The best of luck to any contingency planner out there trying to make any sense of that.

But don’t say you haven’t been warned!

Figure 2 – Courtesy of the Met Office

BBC Weather and how they visualise forecast temperature extremes

Figure 1 – Courtesy of the BBC

I might be wrong, but in the good old days when the BBC weather presenters used magnetic weather symbols (fig 1), they displayed the extreme temperatures for that day or the coming night, rather than for a fixed time as they do now, which although arguably more accurate, can be misleading at times, as it was last night.

Figure 2 – Courtesy of the BBC

When Ben Rich presented the BBC forecast in the evening (fig 2), he did mention that there may well be a touch of frost in sheltered southern areas, but the graphics he used belied this because the model was displaying much higher spot temperatures at 0500 BST, so graphically at least we got a misleading picture.

Every Autumn I get on my high horse about this one, I know it will seem small and petty to most people, but I specialise in that kind of thing. I do realise that forecasting and visualising extreme temperatures for an island nation like ours is never going to be easy, but why make it harder for yourself by refusing to use a chart of extreme temperatures rather than for just a single fixed time?

There was a good ground frost in places overnight, especially in the south (fig 3).

Figure 3

As you can see there were three places that recorded a slight air frost overnight (fig 4), the coldest of which was Hurn near Bournemouth with a minimum temperature of -0.5°C.

Figure 4