UKC

Just looked at the forecast for today from the Met Office and the BBC.

I know the BBC no longer gets its forecasts from the Met Ofice so the two aren't going to be identical but, I'd expect them to be in the same area.

Met Office - every hour has a 10% chance of rain until 5pm

BBC - never less than a 22% chance of rain over the same period,

and until midday the difference is greater. BBC forecast never less than 55% and as much as 83%..

I wonder if each organisation looks at the other's forecasts and reviews what happened on the day.

Dave

I observe the same, and find the bbc forecast to be, on average, more on the pessimistic side (ie it is usually better then they forecast). With no evidence whatsoever, I reckon they err on the side of caution to stop the viewers/readers having a moan when the weather is worse than forecast. 

I've had an awful lot of days when the Met Office has said 5-10% rain and it has rained quite persistently. I've given up a bit on the Met Office for rain prediction. I still find them pretty good for the other stuff. 

Have said the same about the BBC for a while. Definitely think the weather is often better than they forecast. 

Well today the Met Office was more correct - not a drop of rain all day.

Dave

There was a "More or Less" podcast on Radio 4 a few years ago when someone from the Met Office came on to clarify those rainfall percentages - a response to a listener's query along the lines of "if five consecutive hours each have a 20% chance of rainfall, does that mean it's 100% certain it will rain that day?".  The spokesperson prevaricated and never really explained the basis of the percentages and seemed evasive about if they were actual probabilities.  I got the distinct impression they are not statistically derived from modelling - more just provided for "vibes" - to give the viewer a qualitative feel for the day ahead (big number = take an umbrella... and don't say we didn't warn you!).

Throw in metchecks useful rainfall amount and you get a better indicator.

The% used by BBC and met office isn't clear if it's % of the area likely to get some rain, the percentage of the hour that will get some rain, or the % chance of getting wet at any given location.

Generally found BBC to be useless now. I use rain radar, met office and metcheck.

The BBC provider (MeteoGroup now part of DTN) is a mahoosive US weather consultancy that does similar things for the USA that the Met Office does for the UK, eg aviation and other commercial services. They have a lot of infrastructure in the States and are probably quite good there for tornadoes and the like, but it's not clear what if anything they have in Europe other than some very persuasive salesmen.

+ 1 for Philip's suggestions. The Met Office rain radar is now pretty cool as you can scroll the slider at the bottom and go from past Observation to future Forecast, and see how they are doing.

Yes, but I also use an app called Rain Alert to give me a heads-up on area and intensity within a 20 km radius of my location. Great for advanced warning on whether I'm seeing it out in what I'm wearing or going full waterproofs. 

My understanding of the percentages is, the model makes a prediction of the set of conditions for a particular day (or hour). Things like temp, humidity, pressure,  lapse rate, cloud cover etc. Historic data is used to arrive at the percentages. So if they have data for say 100 days with the same conditions as they are predicting for next Tuesday and it rained on ten of them, a 10% chance of rain is given. 

Yes, but what is not clear is whether it is the probability of it raining at that location at a particular time or the probability of it raining at that location at some point in a time interval. The two will often be very different.

Nah. The percentages are based on re-runs of the model. All weather models have some uncertainties built in, so you can run them, say, 100 times, and get different outputs on each run. If it rains in one hour at a location 10 times out of 100, then you get a 10% chance of rainfall on the forecast. Not that they actually do 100 runs, but just to give you an idea.

It's the time interval. Numerical models like weather models are run in 'time steps' - so everything is calculated, then used as the input for the next time step (often 1 hour). So the time you see presented is essentially an average of the hour surrounding it. Some models are run with higher temporal resolution, but this is computationally expensive and a little unnecessary vis-a-vis detail for a forecast more than c. 6 hours into the future.

OT ... but try the YR app on your phone.  YR = Norwegian Broadcasting Corporation.  Rather that a %, it has predicted rain in mm per hr.  

This combined with the rainradar on MetOffice app gives a pretty accurate reading for the next 24hrs.

Apple:  https://apps.apple.com/gb/app/yr-no/id490989206

Google:  https://play.google.com/store/apps/details?id=no.nrk.yr&hl=en_GB&gl...

These days there seems to be an expectation that weather forecasters can tell us what the weather will do hour by hour in a fairly precise location.

I suspect that this just isn't possible on days when ot can be raining the the front garden and dry in the back garden.

You don't get buffeting from the BBC.  

NRK is the Norwegian state broadcaster.    Anyone remember The Weather Forecast in Norwegian?

Who host yr (which actually translates as 'drizzle') alongside the Norwegian Met Institute.

edit spelling

Ha!!!

I had no idea that's what it meant!  Learn something new every day.

Good app ... 

I've always wanted to know what the percentages mean.

• % chance that is will rain somewhere in the grid for any duration
• % chance that it will rain for the majority of the grid for any duration
• % of the interval that will likely experience rain somewhere in the grid
• % of the interval that will likely experiences rain for the majority of the grid.

I'm sure I can come up with a dozen more interpretations of what the % could mean, and they will all yield vastly different numbers.

Without this understand the % is a little meaningless.

But I suspect it's this meaning that gives the very different numbers between the MET and the BBC.

Whatsmore, even knowing the % doesn't help you understand what to do.

Take a crag, say Stanage, and a forecast of 80% for 9am to 5 pm. You could have very localised, minimal volume of precipitation all day across the area covered. From your location all you actually get is 2 or 3 light drizzles for ~5 mins. You mostly climb in overcast dry.

Alternatively it could drizzle constantly, the air damp, the rock slippery.

They both get the same look on BBC. The information presented is useless to make a judgement. Add to that it's often wrong too.

But it's okay, next year the BBC are getting a new rock on a string to improve forecasting.

The first one.

https://web.archive.org/web/20181016041811/https://www.metoffice.gov.uk/new...

Its why the radar images are increasingly useful for guessing the overall impact or some of the other apps which I guess use the radar output to provide rain graphs.

Generally  given up on the bbc so unreliable(it is a usa company) met office much better

Even if the probabilities of rain per hour were independent (which I'm fairly sure they aren't), the probability of rain at some point= 1- prob(no rain) =1-0.8^5 = 0.672. To obtain the correct probabilities you would need to know the dependence of rain between each of the 5 intervals. I'm not sure, but I expect this can be inferred empirically somehow, or from the ensemble model. 

In Glossop it’s simple: 10% chance of rain means steady drizzle! 

Yes, I've noticed BBC forecast is quite frequently wrong in our neck of the woods, and often better than forecast.

showers are very unpredictable..

Curiosity got the better of me and I can access this article regarding the dependence of hourly forecasts - https://www.insidescience.org/news/how-do-daily-weather-forecasts-relate-ho... .

Neilley added that in the summer, precipitation probabilities for an hour tend to be much lower than the probabilities for the day -- meaning they are generally more independent. In the winter, he said, hour-to-hour weather events are much more connected to each other, and therefore hourly forecasts will be closely related to daily projections.

Clearly there needs to be a posse from UKC on a visit to the Met Office in Exeter. I've been on one of these. You're allowed to look through the window at the supercomputers, and there's a big glass window on the forecasting teams area where you can bang on the glass and shout nonono Michael Fish is my hero. They have great hardware.

Most importantly, you need to ask the percentage question of the lovely young thing that will show you around. And keep pushing until you get a reasonable answer.

The best bit of the visit was being allowed into the library, where I smirked at the maths books and looked up the actual forecast synoptic map for D-Day. Very cool.