I wonder if it is an averaging thing and the max was measured over a certain time period where it oscillated between 29.3 and 29.4°C?

I think I read somewhere that they use a rounding method that rounds up half the time and down half the time. So if .35 rounds to .3 then .45 will round to .5, and so on...

That won't be it - AFAIK rounding in weather obs is standard, i.e. a .5 will round up every time, .35 would round up to .4 and .45 to .5. (The only "special" thing that'd make maths teachers squirm is -0, which is shorthand for -0.1 to -0.4 and which is used in some weather model output charts).

The answer IMO is that obs are indeed recorded to two decimal places at some stations, presumably those with the latest kit, but you don't normally see them reported with such accuracy.

https://weatherobs.com/ 

Shoeburyness (near Southend) is a good example - click it, then station dashboard and you'll see all the obs are reported to two decimal places, including dewpoints:

2026-07-05 14:40:00 26.71 14.62 

2026-07-05 14:30:00 25.74 13.99 

2026-07-05 14:20:00 25.55 13.93 

2026-07-05 14:10:00 25.85 14.39 

2026-07-05 14:00:00 25.06 14.12 

Met goes 33 32 33 34 33 33 from tomorrow. iPhone has 30 28 30 31 31 31. 

In the June spell the iPhone was much closer to reality while the Met kept going for the phantom 39s and 40s, so we shall see if that’s the case again. In the past the Met has tended to be quite conservative during hot spells but this year they seem to be overdoing the temps. 

Re the max decimal places, on Weatherobs the 1800 readings always seem to have the max to 2dp, in Kelvin. Sometimes the .x5 values (In Celsius equivalent, i.e. .x0 in Kelvin) get rounded up, sometimes down, and I have never figured out the logic. 

How many in a row can we get from today?

GFS 12Z has 30C or higher every day to the 20th, so that would be a run of 16 days including today and take us to 31 in total, just three short of 1995. And with six weeks of summer still to go!

Also based on the GFS there are quite a few possible date records coming up:

7th: 33.3C (1893), GFS has 33C

8th: 34.1C (1941), GFS has 34C

9th: 32.7C (2010), GFS has 36C

10th: 33.9C (1921), GFS has 37C

11th: 34.7C (2025), GFS has 38C

12th: 35.0C (1923), GFS has 36C

14th: 32.3C (1983), GFS has 32C

I hope not - 15 days straight (at least) of the temperatures hitting 32c or above somewhere in the south of the UK would have huge impacts - much more so than May or June. In a heatwave longevity has a big part to play on the impact.

And that's not forgetting that the 12z also serves up a good number of 36c plus days.

Possibly, but scientifically you wouldn't quote it greater than your measured precision. So should have been rounded. Of course strictly speaking they should quote the tolerances (+/- ) at a given confidence level, and they possibly do in house but don't want to confuse the public. After all it is just the weather 😊

It's as if you didn't read my earlier post! 😅They're measured to 2dp (at least in some stations), but you don't normally get to see that precision.

I used once for Nicosia July average result when it first time just closed to 40C when the result was 39.5 then checked to 2dp was actually resulted at 39.45C so only that time I used 2dp.  Let hope to see sunny days this time today onward.

still there this morning on extended ECM. But GFS wants to break it down quicker

The Met Office SAWS stations are able to record temperatures to at least 2 decimal places, but the main reason we are seeing these reported more widely is down to the adoption of WMO BUFR code in place of the old SYNOP code which only room for a single decimal place.

This is the 1400 UTC observation from Shoeburyness yesterday that Darren posted above, BUFR format first then SYNOP format.

BUFR

001001:3 001002:693 001015:SHOEBURYNESS LANDWIC 002001:0 004001:2026 004002:7 004003:5 004004:14 004005:0 005001:51.55475 006001:0.82700 007030:2.0 007031:2.0 010004:102320 010051:102340 010061:-80 010063:8 010062- 007004- 010009- 007032:1.25 012101:298.21 012103:287.27 013003:51 007032- 020001:50530 007032:0.00 013023- 007032:0.00 020010:25 008002- 020011:2 020013:1220 020012- 020012- 020012- 008002:21 020011:2 020012- 020013:1220 008002:7 020054- 008002:8 020054- 008002:9 020054- 008002- 005021- 007021- 020012- 005021- 007021- 020062- 013013- 012113- 020003:508 004024:-1 020004- 020005- 004024:-24 014031- 004024:-1 014031:29 007032:0.00 004024:-12 013011- 004024:-1 013011:0.0 007032:1.25 004024:-12 004024:0 012111- 004024:-12 004024:0 012112- 007032:0.00 002002:12 008021:2 004025:-10 011001:290 011002:4.0 008021- 004025:-10 011043:280 011041:7.2 004025:-60 011043:290 011041:8.4 007032- 004024:-24 002004- 013033- 004024:-1 014002- 014004- 014016- 014028- 014029- 014030- 004024:-24 014002- 014004- 014016- 014028- 014029- 014030- 004024- 004024- 012049-2026-07-05 13:40:00 001125:0 001126:826 001127:0 001128:217 001001:3 001002:693

SYNOP

202607051400 0-826-0-217 03693 AAXX 05144 03693 25684 22908 10251 20141 30232 40234 58008 82/// 333 55304 60005 82/40 91014 90710 91116==

The temperature from the BUFR code is 298.21K (-273.15 = 25.06C) and dewpoint 287.27K (-273.15 = 14.12C) and from the SYNOP code temperature is 25.1C and dewpoint 14.1C.

It seems that nearly all if not all of the UK stations now report fully in BUFR format, so it looks like the extra decimal place is here to stay!

Once the car was lighter, they drove it to the car space! 

Most of the GEFS 06z break things down after day 10, though we often see high pressure holding on better than forecast, so a long way to go.  But by then, most of us would want some cooler wetter weather, school holidays or not

Thank you for that detailed info - it explains it very well! 👍

That is the newer of those two formats and is the reason why I changed my various python scripts to use the raw BUFR data from ogimet.com instead of the older SYNOP format which the older versions of my scripts used, even though you can still download that data in the older raw SYNOP format from ogimet.com (which means that the older versions of my scripts would technically speaking, still work today).

As shown above as well, the temperature readings from the BUFR format are in Kelvin whereas the temperatures in the SYNOP format were already in °C (the SYNOP codes actually showed that multiplied by 10) but in my python scripts, it was easy enough to convert those temperatures into °C.

With that change in the format, also came a change in the accuracy with these temperature now showing up to the nearest 0.01K (which is just 0.01°C) instead of the nearest 0.1°C.

Airport sites (such as Edinburgh Airport, Glasgow Airport or Heathrow Airport) on the other hand, use METAR data (even though Edinburgh Airport is also a WMO site with a WMO code of 03160) which even to this very day, still only shows the temperature to the nearest full 1°C.

It feels very breezy here less than 20 miles from Heathrow and I am surprised it's that warm tbh. Local PWS reading 31 but with a dewpoint below 10C and 20mph gusts it doesn't feel oppressive at all.