UKH

There seem to be quite a few weather forecast sites and Apps, but am I correct in assuming that the primary source of weather data collection and interpretation is the Met Office? I appreciate that some outlets are specialised eg Shipping Forecast, Mountain Weather Forecasts, Scottish Avalanche Information Service, and are aimed at specific groups, but for general national and local area forecasts I find the Met Office App is very good and  generally predicts accurately to a satisfyingly high standard.

What sources do others use getting their weather forecasts?

You can of course use your own skills and experience to build on local and national forecasts when planning a day out, but it is the primary sources that I am talking about.

My tactic is to triangulate the Met Office, MWIS & BBC.

When the stars align and they all agree you know it's bomber

There are many weather forecast models ran by different organisations around Europe. The BBC  use a European company called MeteoGroup. The inputs to the models are similar, and so the observations in the UK (e.g. radar, weather stations) are maintained and processed by the Met Office - but they are shared with other metrological services such as MeteoGroup. 

But the forecasts you and I see in the app aren't just raw model output. There's interpretation and changes made by the forecasters at the Met Office. They are familiar with how the models behave and the weather specific to the UK, so they can make educated changes to improve the output - this is why I'd use the Met Office over another service like BBC (for weather in the UK).

Site specific forecasts aren't always that useful, as it's very difficult to say if a shower will occur at that specific location, that's why you often end up with a percentage as likelihood of rain to fall. So my general rules are:

1) If there's a front passing over the location, it will most definitely rain - probably not much chance of lucking out here.

2) If it's a showery day, you could get lucky, just keep your eye on the rainfall radar (available in the app). This will give you an idea of how intense and frequent convective showers are forming. 

3) Finally, if you have your sights set on a certain location, try to find a private weather station nearby - these are really useful for getting an idea of how much rain has failed in the past 1/2 days and thus how wet it might be.

Hope that helps,

D.

​​

A sound tactic!  The Met Office run their own models, MeteoGroup (and hence the BBC) mainly rely on the ECMWF model, while MWIS (along with the majority of online auto-generated forecasts) uses the US-based GFS model, because the GFS output is freely available. 

There are good technical reasons to suppose that the Met Office's short-term forecasts (up to 48 hours) ought to be the most accurate, since they run a UK-specific local area model at extra-high resolution. Beyond 48 hours the situation is less clear - for example, the ECMWF model has some advantages because there is a longer gap between the data collection time and starting the model run, which can mean improved initial conditions.    

Lots of this was covered in https://www.ukhillwalking.com/forums/rock_talk/accurate_weather_forecast_for_t...

I use Windy on desktop to easily compare different models. You do have to know the strengths of each model as (for example) GFS ignores topology which means it may be inaccurate around coastlines and mountains.

The met office app is quite frankly useless, they just change it on an hourly basis based on what the weather is actually doing, it's more of a window onto current conditions than a forecast, you can look at it at 8.00AM and it be showing clear all day, then it rains at 11.00AM, recheck the app and hey presto it shown rain.

I also have a theory on their percentage precipitation forecast, if it shows <5% chance then it will rain, if it shows >90% chance then it won't rain

Best to just get on with things and take whatever comes !

But that's good!

All forecasts are sometimes wrong, and the longer the range of the forecast the more likely it is to be wrong.  If I'm looking at a very short range forecast wanting to know what's coming in the next couple of hours I want them to have updated it on an hourly basis.

That's a bit of a misunderstanding of how the model on the app works. The first few hours use what is called nowcasting where it is mostly weighted towards observations before fading into model data. It also reruns the first 48 hours of the model every hour, incorporating the latest observations into the model, so usually the latest forecast should be the most accurate, not just for the first few hours but for the next 48 hours.

The Met Office app was spot-on for me last Sunday for the hills above Braemar: wet at the start of our walk but quickly cleared, sun broke through after lunch then a heavy downpour mid/late afternoon - by which time we were in the car on our way home, hurrah.

I find the MWIS forecasts a bit vague/hand-wavy sometimes, but they're a useful "second opinion".  The Met Office web site (though not their app AFAICS) does have specialist mountain weather forecasts covering similar sorts of areas to MWIS: https://www.metoffice.gov.uk/weather/specialist-forecasts/mountain.  It includes a "summit specific forecast" map that you can, supposedly, use to get a forecast for a specific hill, but I'm not sure how useful or accurate that really is (I haven't used it myself so I've no comparison between forecast and what actually happened).

My experience is that the BBC forecasts are noticeably less reliable than the Met Office.  I sometimes compare the Countryfile "forecast for the week ahead" with the Met Office's forecast for the next seven days, and where there are distinct differences it almost always turns out to be the Met Office forecast that was closer to what eventually happened.

The other one that used to get mentioned on here quite regularly was yr.no.  I do still have a look at that from time to time, though they changed the app a while back which IMO made its usability worse.  Anyone know which model they use - or do they have one of their own?

"All forecasts are sometimes wrong." I apologise in advance and I've tried to avoid commenting on this statement but being a pedant I can't help myself. You do realise that this makes no sense?

To answer the main question, for mountain specific forecasts I use met office and mwis which in my anecdotal experience are generally pretty accurate between them. Mwis is notoriously pessimistic and they have admitted as much. The key point with any forecast is recognising that the exact point forecast is likely to be wrong and the important measure is the measure of uncertainty around that forecast. A good example is met check which I use for forecasts around where I live and is generally very good. This has a probability measure for the chance of precipitation (they presently call it "RainRisk" and it's changed names a few times, presumably as they try to make it more palatable and understandable to users), which I find useful. Understanding what that actually means is helpful as well. If you accept forecasts for what they are then the short-term weather forecasting these days is remarkably good.

On another point of forecasts that change almost in real time, this could be considered a good idea. It depends exactly how they're doing it but updating forecasts based on the most recently available data is generally a very good idea if it's done well. Why wouldn't you update your forecast with the most recent data to make it as accurate as possible if you can?

I'm afraid I don't, no.  If you can't help yourself perhaps you should be bolder, stop apologising and embrace your inner pedant.  By all means feel free to tell me why it makes no sense.

I like to apologise in advance. I like to play nicely and not let things escalate into name calling and abuse. Apparently this happens sometimes on ukc and I don't like it so I'm afraid I shall continue to apologise in advance of a potentially contentious statement. Sorry about that 

How can all forecasts sometimes be wrong? A forecast is a single prediction of something that might happen. That single prediction is either right or wrong. Once the event has occurred the forecast is no longer applicable. Some forecasts will be right and some will be wrong, but all forecasts cannot sometimes be wrong. In the context of weather forecasting a better statement, in my opinion, might be "it is highly likely that all forecasting is wrong, but some forecasts might be reasonably accurate". I dare say this won't settle the matter (for which I apologise) but there you go.

Are you sure you're on the right forum?  Ha. 

I see what you mean now and you're not wrong, it's a fair cop.  Your better statement is, I think, what I was trying to say but I phrased it imprecisely.

I definitely pick the threads I reply too and how I reply very carefully!

Nice to have had a pleasant conversation with you. Have an enjoyable rest of day.

yr.no forecasts are based on the ECMWF model - they also run a hi-res local model for Scandinavia, but it doesn't cover the UK. 

Output from global and high-resolution UK models with human input and interpretation - Met Office

Human interpretation of output from global circulation models - MWIS

Output from global circulation models with no human input or interpretation  - Accuweather, BBC, Windy, XCweather, yr.no

As mentioned upthread BBC weather downgraded from the Met Office to a cheaper forecast provider. I think there are a few reasons why people have stuck with BBC weather regardless: out of habit, because they don't know BBC don't use Met Office anymore (except for weather warnings), and because the highly detailed output on the BBC weather website/app gives the impression of high accuracy.

If you have the time and interest to get to know how to interpret the output from global circulation models (which you can get from e.g. Netweather, Meteociel, Wetterzentrale) then that can be quite useful for roughly planning things a week or two out - especially when there is little spread in the ensemble predictions for a given model, or between the predictions from different GCMs. This will never give you highly detailed forecasts, but sometimes all you need to know is whether it is likely to be settled/unsettled and the likely geographical origin of the airmass over the UK.

Thanks.  So basically little different to the BBC (as is now), as far as the UK is concerned?

I remember in the past a number of people used to talk up yr.no as if it was significantly more sophisticated and accurate than other forecasts.  My own experience was that it was no better or worse than the run-of-the-mill forecasts (which is not surprising) and usually less reliable than the 'national' forecaster e.g. Météo-France - yr.no always seemed to err quite a long way on the pessimistic side for some reason.

Good post.

A minor point of correction though:

Forecasters do have input into what you see on the app but only in the text forecast. All the symbols and figures you see do come from raw model output.
 

Ah ok, that would make sense. Although I believe the raw model output goes through some post processing before deciding on the weather symbols and app data.

Yes, that is correct. As I mentioned above, the first few hours are heavily weighted towards observations before merging into model data, but even then there is post processing applied to the raw model including filtering to prevent results way outside climatological means. Most of the time this helps prevent the model running away with itself and coming up with extreme values, but it does mean that, for example, temperature extremes on hot sunny days or cold clear nights in sheltered locations can be underforecast by the model.