• 2 days ago
This is an in-depth Met Office UK Weather forecast for the next week and beyond. It’s been a dry, dull and mild November so far, but things are about to change. Bringing you this deep dive is Met Office meteorologist Alex Deakin.

Category

🗞
News
Transcript
00:00Welcome along to the Met Office Deep Dive, thank you for being there.
00:04I hope you're not put off by how long this Deep Dive is.
00:07Don't get too scared, it's not going to be all about the weather for the next few days.
00:12Although there's quite a significant chunk of that.
00:14But this week's Deep Dive is a special.
00:16We've kind of done it in two bits really.
00:18Later on I've got a very special guest, Ken Mill, coming in to talk specifically about Ensemble.
00:24So that's kind of the second half of the Deep Dive.
00:26The first half will be the usual stuff, tucking into some juicy meteorology.
00:31And there's plenty of that over the next few days.
00:34It is going to get colder at the weekend.
00:37There's a lot of talk of snow.
00:39We'll break down the chances of seeing anything significant as well.
00:44Thank you for being there, thank you for watching.
00:47We love reading your comments.
00:48So if you've got any comments, any questions, then do pop them in the chat.
00:53And do give us a like.
00:54And tell other people that we're here.
00:57Let other people know that the Met Office do this every Tuesday.
01:01And here's a new thing that we've just started doing as well.
01:04Did you catch us on Friday in the live?
01:08The Weather Studio Live was back last Friday.
01:11Aidan and I chewing the fat.
01:13And it's going to be back this Friday as well.
01:15So that's another thing to add to your schedule.
01:18If you subscribe, you'll get notified when we go live.
01:22And we're going to do that every Friday from now on.
01:25One of the things we were talking about.
01:27And that was how gloomy and how dull it's been so far through November.
01:31And I've got some interesting stats to show you for today.
01:34This is the sunshine map.
01:36Or rather the lack of sunshine map for the first 11 days of November.
01:42Now this is showing sunshine compared to average for the whole month.
01:45So you wouldn't expect it to be anything other than a little dull.
01:49But this is exceptionally dull.
01:51With pretty much all parts, from southern Scotland southwards,
01:54less than 30% of their average sunshine.
01:57You would normally expect around 37%.
02:00So you would normally expect it to be in this mid-shade of grey.
02:04And parts of central Scotland are in that.
02:06And parts of northern Scotland are above that.
02:08Because northern Scotland last week,
02:10while much of the rest of the UK were suffering with the anti-cyclonic gloom,
02:14they were poking out into the sunshine.
02:16So it has been a sunnier than average month across northern Scotland.
02:19But much of the UK has been distinctly grey.
02:23Distinctly lacking in sunshine.
02:25But it has also been quite mild by day.
02:29But particularly by night.
02:30That blanket of cloud last week really kept those temperatures up at night.
02:35Now this is the rolling temperature as we go through the month.
02:39Dates going along the bottom and the extremes.
02:42So the warmest is that orange line at the top.
02:45The coldest ever recorded is this blue line at the bottom.
02:47And this thicker blue line, that is this year.
02:50And you can see that until yesterday,
02:52we were actually above the warmest on record.
02:55These were the minimum temperatures across the UK as an average.
02:58And just with some colder conditions spreading south
03:02and some clearer skies arriving,
03:04we have seen those temperatures dip down below that record.
03:07But we're still in a pretty exceptional area at the moment.
03:12For 11 days in, we are significantly above this average line.
03:16So it has been a very mild month.
03:18Particularly so by night.
03:20And it has been a very dry month as well.
03:24That high pressure that we've seen dominating,
03:26keeping a lot of cloud but not generating much in the way of rainfall.
03:29In fact, the driest start to November on record.
03:33Beating the previous record as we stand.
03:3511 days in, which I think was 1964.
03:39So it has been very dry when you compare it to the average.
03:43That's this middle line here.
03:45And again, you've got the extremes.
03:46The wettest November there.
03:47And the driest on record, that pale blue line.
03:50We're just below that at the moment.
03:51Now, things are going to change.
03:52The weather patterns are shifting.
03:54But just to show how incredibly dry, how mild it's been at night,
03:59and how very dull November has been so far.
04:02And this is, what, a third of the way through the month.
04:04Slightly over.
04:05So it has been a pretty remarkable month so far.
04:08But we deal in forecasts here.
04:10So how are things going to change?
04:12Well, they are going to change, yeah, quite a bit,
04:15quite significantly in the next few days.
04:18Here's the satellite image for the past few days.
04:20You can just envisage that chunky area of high pressure sitting here.
04:23Weather fronts have been trying to push in.
04:26You can see the bands of cloud have been trying to make progress
04:29in across the U.K.
04:30And not really succeeding.
04:32But we are going to start to see a change.
04:34Not necessarily from the west but more from the north.
04:37Let's take a look at what the jet stream is doing.
04:40It's going way up to the north, bending back round and down here.
04:45We've seen more unsettled weather across Iberia recently.
04:48And as the jet stream drives in here,
04:50there's a low pressure high in the sky up here.
04:53And it's squeezing against this block,
04:55this high pressure that's dominating, keeping things largely dry.
04:58But where we've got them squeezing together,
05:00we've actually got quite a few shower clouds moving in
05:03across the southeast of England.
05:05And we've seen some wet weather drifting in here
05:07over the past 24 hours or so.
05:10And you can see them here, the showers,
05:12just zipping in across parts of East Anglia
05:14and the southeast through the day today.
05:16But high pressure in control.
05:18And as we go through the next few days,
05:20well, we're not going to see a great deal of change.
05:22The high continues to dominate.
05:24A weak weather front topples in around it.
05:26We lose that jet coming down across the southeast.
05:29There's not as many showers across the southeast tomorrow,
05:31looking largely dry but still quite sunny.
05:34Under the axis of the high, so close to the centre of the high,
05:37that's where we're most likely to see some fog.
05:39So like last night, again, parts of Northern Ireland
05:41could see some fog, southern Scotland, northern England as well.
05:45As we go through the next few days,
05:47well, not a great deal of change.
05:48The high pressure continues to dominate.
05:50Let's put it on the bigger picture.
05:54Take the jet stream off because that's not really doing a great deal.
05:57High pressure, high pressure, high pressure
05:59until we start to see the high actually waddling back this way.
06:04And this is going to mix things up.
06:07It's a cold front, as you can see,
06:09and that's really going to change things up as we go towards the weekend
06:12and introduce colder air because it is a cold front.
06:16Let's put the temperature profile on.
06:18You can see the colder air lurking up here.
06:21We'll put the jet back on because it's this dip in the jet stream
06:25that's driving this area of low pressure,
06:27generating it, pushing that further north.
06:29And then you've got quite a contrast in temperatures here.
06:31That's the cold front.
06:33And this jet stream will push that south
06:35and all that cold air is going to flood south as well.
06:38There is a bit of uncertainty as we get to the second half of the weekend.
06:41But until then, let's take the jet off again
06:44and just walk through that period as we go through Friday,
06:48Wednesday, Thursday, Friday, high pressure,
06:50still most places dry, variable amounts of cloud.
06:54And then zooming in, here comes that cold front.
06:58Look at that cold air drifting its way steadily southwards across the country,
07:02introducing much cooler conditions by the start of Saturday
07:05across Scotland and Northern Ireland.
07:07But we'll all be in that colder air.
07:09We'll be post-cold front by the time we get to Saturday evening.
07:13That rain will be clearing through.
07:15There'll be some rain on it, but not a huge amount.
07:17But, you know, any rain is significant because it has been so, so very dry.
07:21But there's the colder air coming into play.
07:23It's not going to go any further into Sunday at this stage
07:25because, yeah, things get quite complicated for Sunday.
07:28But let's focus in on that colder air coming in
07:30because it's really going to change things up a little bit
07:33if we take a look at the temperatures over the next few days.
07:36Not that one. It's that one, isn't it?
07:39Still not quite got used to the new touch screen as we look at it.
07:43But actually, temperatures by day for the rest of this week,
07:46pretty much bang average for the time of year, 9s, 10s, 11 degrees Celsius.
07:51Obviously, if you've got fog, then those temperatures are going to be suppressed.
07:55And that's what's happened today and could happen again as we go through Wednesday.
07:58Stubborn fog patches will really suppress those temperatures.
08:01But whether it's sunny or whether it's cloudy,
08:04kind of 9, 10, 11 degrees Celsius is pretty much what we're looking at.
08:08By the time we get to Saturday and Sunday, notice those temperatures are coming down.
08:12So we're looking at single figures on Sunday pretty much across the board.
08:16Although, again, some uncertainty about exactly how quickly that cold air stays in.
08:20But you can see the trend there to those temperatures dipping away.
08:24And if we put the minimum temperature on, you don't get as much of a range here
08:28because as we see the colder air coming in, there will be still quite a lot of cloud.
08:32And we've got the clearest skies at the moment. That's where we're seeing a touch of frost.
08:36So it's not as obvious that temperature dip in the nighttime values.
08:40It's much more obvious in the daytime temperatures as we see that colder air coming in from the north.
08:45So the first thing in the morning, you may not actually notice too much of a difference.
08:48The past couple of mornings have been colder because we've had the clearest skies
08:52compared to last week when we just had the blanket of cloud keeping those temperatures up.
08:57So interesting to see that the temperatures by night don't really change a great deal.
09:02As we go into next week, we may see something a little bit lower
09:06and those temperatures could be falling down.
09:08But generally speaking, the next few days, as I say, we'll keep high pressure in control.
09:13A lot of dry weather. There's that weather front moving in.
09:17And if we take a look at that. In more detail, it's zoom in a bit across Scotland.
09:24There is that weather front and it will bring some significant rain after a very dry spell.
09:29I know some places actually still struggling because October was so wet.
09:33It was a very wet start to autumn. So even though it has been very dry start November,
09:38there's still a lot of groundwater in some parts of the country.
09:41This rain is going to come in and you can see that narrow line on it.
09:45There is a really narrow line to a sudden burst of heavy rain.
09:48We call it could be a squall line where the winds also suddenly change direction.
09:53You get a rapid switch in wind direction.
09:56The rain won't last all that long, but it could be really intense for a while
10:00and the winds will suddenly start to gust up.
10:02So this is on Friday, Friday afternoon to Friday evening.
10:05That sinks its way south as a pretty narrow band there across Scotland.
10:11So the rain is not going to last too long.
10:14But nevertheless, it will bring some heavy bursts of rain for a short period of time.
10:18The winds really suddenly pick up on that.
10:20You've got that complete change in wind direction from a southwesterly to almost a northwesterly there
10:26in a matter of minutes, really.
10:28So swinging the wind direction to some gusty conditions as that band of rain sinks its way southwards.
10:34And then obviously it continues to push its way southwards.
10:37And then, lo and behold, is that the first flurry of snow I see on the chart there?
10:44Well, not really, because we had some a few weeks ago.
10:46We've seen some snow on the Scottish mountains already.
10:49But, you know, the first kind of significant on the chart there, seeing a little bit of snow.
10:54But it is mostly going to be up over the hills.
10:57Something that Aidan showed, interestingly, on the Friday live last week was the freezing level chart
11:06and how it was particularly high across the UK.
11:10And that is still the case at the moment.
11:12So the freezing level shown here in dark green here is above 1400 metres.
11:18And that is basically the freezing level is how high up in the sky you have to go before you get to zero.
11:24And that is significant because that kind of determines the type of precipitation that you see.
11:30If it's below freezing, then it's going to be snow, broadly speaking.
11:34If it's above freezing, then it's going to be rainfall.
11:36So you've got to go right across the UK.
11:39You've got to go up through the sky at least 1.4 kilometres before temperatures are even close to freezing.
11:46And Aidan actually showed a map.
11:48It wasn't just across the UK that the freezing level was particularly high.
11:50It was right across Europe.
11:52But obviously, with that colder air coming in, this is going to change.
11:55So let's look at that change because it's quite a significant change.
11:58It's quite a flip as we go through Friday.
12:03As that cold front comes in, just look at that colder air coming southwards.
12:07So you can see that there, by the time we get to Friday night, start of Saturday, we're now into the blues.
12:14So the freezing level has dropped from above 1400 metres to less than 1000 metres, 800.
12:20In fact, some elements of 400 metres there just off the Northern Isles.
12:24So that freezing level is coming down.
12:26That's indicative of that colder air that you now only have to go up 400, 500, 600 metres before you get to zero.
12:32And that colder air, that colder air mass, sinking right the way south.
12:38Look at that. Isn't that a thing of beauty?
12:40Just wee! Oh, I could do that all day.
12:43Look at that. Cold air flooding its way south.
12:46And so that by the time you get to the early hours of Sunday morning,
12:50much of Scotland is covered in that where the freezing level is only 400 metres.
12:55And large parts of Scotland, obviously, the mountains, most mountains are above that kind of height.
13:01Now, even when the freezing level is 400 metres, it doesn't mean that you get snow at 400 metres and anything below 400 metres is rain.
13:08The snow can still fall through temperatures that are a little bit above freezing.
13:13So when the freezing level is about 200, 300 metres, you can still get snow down to down to low levels.
13:20And we're kind of talking about 400 metres. That's the lowest.
13:23Not a lot of greys on these charts. More like to get that in proper winter.
13:27Obviously, we're still only in in the autumn months.
13:29So it's not exceptionally cold, but it is much colder air than, of course, we've been used to.
13:35But that snow level, that zero degree Celsius level dropping to about 400 metres,
13:40which means there is the possibility that some of the snow flurries could get down to lower levels.
13:45Not likely to stick, not likely to cause much in the way of problems.
13:48But certainly over the hills with that kind of air, we are going to see some snow showers coming in.
13:53And that's what we're likely to see as we head towards get the right chart up.
13:58Is that the right one? As we go towards Saturday.
14:03There we go. You can actually see some snow there building up.
14:07So this is the three hour snowfall totals across Scotland is only really across Scotland.
14:13It is really only across higher ground.
14:15You could see some flakes at lower levels, but it's not expected to cause too many issues.
14:19Obviously, you know, it wouldn't take much of a swing that air to be a little bit colder.
14:23But when we see those showers coming in, they could get down to slightly lower levels.
14:27But at this stage, it looks like it's just going to be snow on the hills in Scotland.
14:30Nothing unusual about that in November.
14:33But, you know, noteworthy because it's the first proper kind of cold push that we've seen.
14:38And because it has been so mild, it really would be quite noticeable that we'll see that drop in temperature.
14:43So that's what the snow is looking like at the moment until Saturday.
14:47After that, though, it does get a little complicated.
14:52The uncertainty grows. We are pretty happy with the forecast for the rest of this week up to and including Saturday.
14:59But after that, there are more question marks than usual about what happens on Sunday.
15:06And to show you that, I am going to talk about a comparison between the Met Office model and the European model.
15:18This graph, this map is basically showing the differences between the Met Office model and the European model.
15:26You basically take one away from the other. You take the data point away from the other.
15:30So basically, if they were agreeing, if they're both saying, say, 7, you take 7 away from 7, you get 0.
15:35And that's what the white colours are. You wouldn't get a big difference.
15:38But if one model is saying 10 and the other one is saying 1, then that difference is 9.
15:42Or equally, if one is saying 1 and the other one is saying minus 10, then that difference is shown up here in the different colours.
15:49So the blobs of colours, the blobs of the oranges and the yellows, or the blobs of the blues and the purples,
15:55are showing where the models don't agree with each other.
15:59Now, this is the Met Office model taking away the European model at basically high up in the sky where the jet stream is, 250 hectopascals.
16:09So it's showing the differences in the pressure pattern up where the jet stream is.
16:13And it's showing, this is for T plus 72, so Friday, that across the UK, they're pretty much in agreement.
16:20There's no big differences here. But your eyes are drawn to these blobs, these bright coloured blobs.
16:26This is where the difference is coming off the eastern seaboard of the United States.
16:29If we go back to the model and what's going on out here, you can see around here the jet is a bit of a mess.
16:39That's for Thursday, for today, so in Wednesday. It's really complicated.
16:44The jet's kind of splitting in two. There's two arms to it.
16:48And actually, by the time we get to Thursday and Friday, it's how the jet kind of forms this trough, this dip down there,
16:56and this low pressure system as a result at the surface.
16:59But what the jet is doing high in the sky, it's a bit of a messy one from the Met Office computer model point of view.
17:06And there's just not a lot of agreement between the Met Office model and the European model.
17:11And that's highlighted here with these differences, those two bright colours.
17:15And there's quite a bit of difference also across the Greenland ice sheet,
17:18and that always adds to the complication for the longer range forecast,
17:24because that is quite tricky to model this part of the world.
17:27It's a big blob of ice, effectively, and so it can confuse the computer models.
17:32And when you've got the air coming in from that direction, that is always an added uncertainty to the forecast.
17:38If we fast forward to what's going on on Sunday now, it's the same thing.
17:43So we're looking at the Met Office model taking away the European model at 250 hectopascals,
17:48so 250 millibars high in the sky.
17:51Now you can see that we've got big differences just to the west of the UK,
17:56and actually right across the UK there are big differences as well.
17:59So whereas we're pretty happy with the confidence, we're pretty confident in the forecast for the next few days,
18:04by the time we get to this time, this is Sunday, there are big differences in and around the UK,
18:09and that makes a big difference to the pressure pattern potentially across the country as well.
18:13So whereas we've had a nice big chunky high pressure, not really shifting very far for quite a while now,
18:18by the time we get to Sunday, what is going to happen?
18:22And we can look at what the models are projecting for the jet stream.
18:28Let's take a look at what have we got here.
18:31This is the European model version of the jet stream.
18:35So this isn't one taking away the other.
18:37This is just what the European model thinks the jet stream is going to do,
18:40driving down to the south and actually generating a little area of low pressure.
18:44So we've seen the northerly wind set up for Saturday,
18:46but the European model actually wants to then bring an area of low pressure in across the UK
18:51because the jet is taking this kind of shape.
18:54What the Met Office model is doing, it's got a much flatter jet driving to the south.
19:00I mean essentially they're kind of in similar positions, aren't they?
19:02They're driving south in this, but it hasn't got that bend in it, and it's further across the UK,
19:06and as a result, the Met Office model isn't generating an area of low pressure.
19:10It's just keeping those isobars and keeping the northerly flow going for Sunday across the UK.
19:16However, the previous run of the Met Office model from yesterday
19:21actually had something similar to what the European model had today.
19:25The jet further out in this way, in this direction, it's got more of a bend in it,
19:29and that also generated an area of low pressure across the UK.
19:32So that's the big question mark.
19:34Across the UK, what are we going to see on Sunday?
19:36Will we just keep the northerly winds and keep the chilly conditions,
19:39or will we actually generate an area of low pressure across the country?
19:44That makes a big difference to the potential weather, as I'll try and show you on this map.
19:52These are the, broadly speaking, the two different possibility, the two different possible outcomes.
19:59I'm actually showing this one here, the ICON model, the German model, where there's the UK,
20:04and you can see we're all covered in the colder air.
20:06This is for Sunday. Both of these are for Sunday.
20:08The isobars there, you can see them just pointing up to the north, got a northerly flow.
20:13This is the European model, which does generate an area of low pressure,
20:16and you can see that in there as well.
20:18Now, why that's important is because with this one, I'll do some drawing now.
20:25You were waiting for this bit, weren't you?
20:26You've got the air just coming down from the north, but it's relatively dry air.
20:32OK, yes, it's cold. It's cold enough for some snow on the Scottish hills, as we've seen.
20:36And in this kind of flow on Sunday, it would be chilly everywhere,
20:40but actually it's quite dry, the air coming down from the north,
20:44and so it would produce some snow showers across northern Scotland,
20:47particularly over the hills, mostly over the hills, a little bit at lower levels.
20:51But generally speaking, a lot of the UK would just be dry and sunny.
20:55In this scenario, notice how the low pressure is sitting across the UK,
21:01so in which case the winds are going to be going on like this.
21:04So instead of the northerly winds, we're now, yeah, we're drawing in moisture
21:09because we're coming in from the Atlantic.
21:11So we're drawing in plenty of moisture from the Atlantic, and the air is warmer,
21:14so it holds more moisture.
21:15One of the reasons why this is struggling to generate much in the wet weather
21:19is because it's colder air, and colder air can't hold as much moisture.
21:22So this is more moist air, and it is a little colder than what we've been used to,
21:27but it is still also much warmer air coming in as well.
21:30So that is just going to generate much more in the way of rain or in the way of showers.
21:34So it's a very different day.
21:35In this scenario, it would be cold and sunny across the bulk of England and Wales,
21:39and indeed much of southern Scotland, Northern Ireland.
21:41In this scenario, it's looking much wetter.
21:43The isobars are close together as well.
21:45It could be quite windy at times, but the air is that much milder that most of it,
21:50if not all of it, is going to be rain.
21:51Depending on the position of the low, and there is quite a bit of uncertainty,
21:54if the low is a little further south, then there could be a bit more in the way of snow.
21:57On the hills, for example, across maybe even as far south as northern England,
22:01depending on the position of that low.
22:03If it's all a little further south, then that could generate a bit more.
22:06The colder air would be in, and on this northern flank in here,
22:10we would see more in the way of snow falling over the higher ground.
22:14So still quite a lot to play for for Sunday.
22:18Latest thinking is that this scenario is around about 40% chance,
22:26and this one is the more likely, but not much more likely, at about a 60% chance.
22:32So something to keep an eye on over the next couple of days.
22:35It's quite unusual for that split and to have that level of uncertainty,
22:40but we are still talking four or five days away.
22:45Keep up to date with the forecast because things could get interesting.
22:48If this kind of setup lingers, you get the colder air and you get more wet weather,
22:51you get more low pressure systems coming in,
22:53that could generate a little bit more in the way of wintry precipitation.
22:57But generally speaking, this kind of setup at the moment,
23:00just most likely to bring some snow to the higher ground,
23:03and it's more wet and windy weather for other parts of the UK.
23:07But really interesting to see how things develop,
23:10and it all stems from that.
23:13The differences in the computer models high up in the sky,
23:17a long way away across the other side of the Atlantic.
23:23OK, that's pretty much it for this week's Deep Dive.
23:25So please do send us your questions, send us your queries.
23:30Keep up to date with the forecast as well, particularly as things get interesting.
23:33But that isn't completely it, because as I said at the start,
23:37this is a special bonus edition of the Met Office Deep Dive.
23:41A lot of you have been asking questions about ensembles and what they are, basically.
23:46So we've been talking a little bit more about them.
23:48There's a video which we'll be linking out to.
23:51But I also caught up with Met Office expert on ensembles a little earlier, Ken Mill.
23:58And you can watch that video now.
24:01Today we are talking ensembles, and I am joined by ensembles guru, Ken Mill.
24:08Ken, thank you for joining me today.
24:11Ensembles guru isn't your official title, is it?
24:14What would that be? What's your role here at the Met Office?
24:16Well, I've been involved in the development of ensembles as a scientist for over 25 years,
24:21when I first introduced the science really into the Met Office's forecasting system.
24:26And a lot of that time also thinking about how we use them.
24:30And so my job now as a science fellow is to help the rest of the organisation,
24:36the whole of the Met Office, to make use of these ensemble forecasts in our products and services.
24:42So you know an awful lot about ensembles.
24:44And Ken was instrumental in setting up the video that we made, the what, why and how video,
24:51which we'll put a link to in the chat,
24:54in which we basically talked around this chart.
24:57So what we're going to do today is it's called the deep dive.
25:00So we're going to go into a bit more detail and build up and basically talk about what ensembles are
25:05and how they can be useful.
25:07So we'll start with a very basic.
25:10We'll pair that graphic back and we'll gradually build up.
25:13So let's talk. Talk us through it.
25:16So this is a schematic which kind of explains how we produce a weather forecast.
25:21So we're looking from left to right across the screen in time.
25:25And what we're trying to do is build up a forecast for this point in time across here.
25:32Now, the way we forecast the weather,
25:34we start by using a whole lot of observations of what the atmosphere is doing
25:38to build up a picture in the computer of what the state the atmosphere is in now.
25:43And that's not just here, but that's globally right around the world.
25:46You can't make any kind of prediction of the future without knowing what's going on right now.
25:50Exactly.
25:51And so we get at this starting time what we call an analysis.
25:55And that is the computer's picture of the starting state of the atmosphere.
26:01Now, the traditional way of doing a forecast, what we call a deterministic forecast,
26:06is that we take our computer model and that calculates forward in time
26:12how the atmosphere is going to evolve and gives us a forecast, a point on this line.
26:18So this line might represent temperature, or amount of precipitation, rainfall, or anything.
26:24Anything really, where otherwise it's just that one single path.
26:28That's the deterministic.
26:29We'll think of it as temperature for now.
26:30Yeah, so that's the deterministic.
26:32So that gives you our deterministic forecast of the temperature.
26:34This time will be x or y.
26:37A value.
26:38Right.
26:39Now, what it doesn't tell us is how confident we are in that,
26:45how much uncertainty there might be in that.
26:47Or, if you go back to your school days, what the error bar on that is.
26:50Right, okay, yeah, yeah.
26:51So, one thing we do know, and this is all we know at this stage really, is what climatology tells us.
26:58We know that, for example, here in Exeter, it's November.
27:04The temperature's not going to be minus 25.
27:06Yeah.
27:07It's also not going to be plus 30 in November.
27:09Sure.
27:10So climatology puts some limits on it.
27:11But that's really all we've got to go on.
27:13Okay.
27:15However, we actually know quite a lot about how good this analysis is,
27:19because we know how good all the observations are.
27:21And we also know how good our analysis system is.
27:25So we've got a pretty good error bar on what that analysis is.
27:28So that's a much smaller error bar than the final error bar.
27:31Yes, that's right.
27:32Okay, so we've got that.
27:33That's exaggerated here, actually.
27:34It's really very small.
27:35But it's, we know how much.
27:37We can't know exactly.
27:38We can't know absolutely everything about the atmosphere.
27:40Exactly.
27:41We can never know exactly.
27:43So, well, you've heard about the butterfly effect.
27:46And the butterfly effect means that a very small error here can make a big error here.
27:51So what we do is we make a number of small changes to this analysis.
27:56And that's not just in temperature or whatever.
28:00That's in a whole range of different things.
28:02And then we run a whole set of forecasts, starting from each of those slightly changed,
28:07or slightly, as we call it, perturbed versions of that analysis.
28:11And then each of those grows slightly different in time, because of chaos,
28:16and gives us a whole set of forecasts at this time.
28:19So we can draw what we call a probability distribution,
28:23a graph of how likely each of these different temperatures are.
28:27So each one of these is a different path that's generated by a slight tweak in those initial conditions
28:34that can end up as a different scenario at the end.
28:39Exactly.
28:40And each of those is a complete, consistent meteorological scenario.
28:45It's a complete meteorological evolution, with storms that develop and everything.
28:49Right, okay.
28:50But they're all slightly different from each other.
28:52And that is, in essence, an ensemble.
28:53So when we talk about running the computer model many, many times, that is what we're doing here.
28:58Tweaking a little bit at the start, and getting various different futures,
29:02different possible future atmospheres that then draw up here.
29:08So now you've got a range of possible solutions.
29:10That's right.
29:11How is that useful?
29:12People just want to know if it's going to be 10 degrees next Wednesday.
29:15How is it useful if you've then got that range of possibilities?
29:18So, well first of all, it just lets you draw an error bar on your forecast, if you like.
29:22So it gives you some measure of how confident you are.
29:25We'll come back to that at the end, actually.
29:27But it also lets you draw some other things up.
29:30Well let's add a little bit more here.
29:32Yes, so we've got the probability distribution.
29:34So, a little bit of maths here.
29:37The typical way that we describe a probability distribution,
29:40we often talk about the median.
29:43That's the middle of, if you take all these samples, the one that can be in the middle.
29:49And actually, that is a pretty good estimate of the most likely temperature.
29:55Okay.
29:56Notice it's a bit different from what the single deterministic forecast came out of.
30:00But that's a useful measure, very often, of the most likely.
30:05So you've got, I mean, we often run 40, 50 different ones of these.
30:10So they're not all the ones on here.
30:12We're showing this to keep it, I think there's 10 lines on there.
30:15But there'd be normally 50, even in some cases 100 different versions.
30:19So that's how you build up that distribution of that picture.
30:23And the median is, as you say there, away from the deterministic.
30:29But then we can also take, well, if we divide that total thing into 100.
30:35So the 25th and 75th are sort of a quarter and three quarters of the way through, in terms of the number.
30:42So that gives you a sort of estimate of a central error bar, if you like.
30:47Or if you want to really be, make sure you're sort of taking account of the,
30:51even the more extreme possible scenarios, you can use, say, the 5th and 95th percentiles,
30:56which will give you then a really broad picture and a better chance of making sure you capture the extremes.
31:01So, and if, just to give one example, if you are, well, supposing you, Alex,
31:08you're talking about one of those hot days in summer, where we think the temperature is going to,
31:14one or two places are going to get really high temperature.
31:17We can't necessarily pin down exactly where it'll be.
31:21But if you look at the 95th percentiles, it'll give you a good measure of what the highest it might be at any one place.
31:28And so you can, you could have on your most likely temperatures on your map,
31:33the 50th percentiles, the most likely temperatures.
31:37But you could then say, but somewhere is likely to get this, the 95th.
31:42But we don't know exactly where it's going to be.
31:45And that's useful for lots of different scenarios, again, as we'll talk about in a bit.
31:48But it is really useful. We often talk about, you know, people just want to know if it's going to rain or not.
31:53But you can still generate that from these kind of forecasts.
31:55But it's also good to know those uncertainties so that people can make judgment calls,
31:59depending on what they need to know the weather for.
32:01And rain is actually a really good example because you see many days where, particularly if it's a showery day,
32:07one of these days, you talk about sunshine and showers.
32:10And actually, we know sometimes it's very variable and it's very difficult to pin down exactly where those showers are going to happen,
32:16particularly the big thunderstorms and things.
32:19But, so it's useful to have a, it's most likely you're going to get five millimeters of rain.
32:25But you could get 25, and that's going to be important.
32:28Yeah, yeah, absolutely. OK, so that's the broader picture.
32:32And then what are we looking at here?
32:34So I mentioned earlier that each of these is a completely valid scenario, evolution of the atmosphere.
32:42And sometimes it's actually useful not just to look at that distribution of what's the range of possible temperatures or whatever at this location.
32:50You might actually want a scenario, a more extreme scenario.
32:54So it's useful sometimes to look at the individual scenarios which are lowest and highest,
32:59because then you can tell a complete weather story associated with that member.
33:04OK, so you get the absolute extremes within those possible scenarios.
33:08So that's a different way of looking at the same thing.
33:10Well, the other thing, and people...
33:12More maths now.
33:13Yeah, more maths.
33:14If you're interested in whether we're going to get over a particular threshold,
33:20say, for example, talking about heat waves again, that 40 degree threshold,
33:24the time a couple of years ago when we got 40 degrees,
33:26you could look at some threshold, like 40 degrees temperature on here,
33:30and this will tell you what is the probability, the area under this graph of being above that threshold
33:37tells you the probability of exceeding that threshold.
33:39So you can actually put a number on it.
33:41So that's really useful.
33:42Historically, in the past, we've not been able to necessarily do that yet.
33:45So we can't say we're definitely going to go over 40 degrees.
33:47No, but there is a...
33:48But we can say that there's, say, a 20% chance, a 1 in 5 chance,
33:52that you are going to get over 40 degrees.
33:55So again, it's actually putting numbers to that general weather forecast.
33:58Historically, that might not have been...
34:00Yeah, we think there's a good chance, but actually you can now put a number on it.
34:03Yeah.
34:05But one of the really big powers of the ensemble,
34:08because we could get a lot of this stuff just by looking at the statistics of
34:12how good or bad forecasts have been in the past.
34:15Okay, yeah.
34:16And we keep those statistics, of course.
34:17Yes.
34:18We measure them all the time.
34:19But the real power of the ensemble that is here is the same thing on another day.
34:24We've got the same sort of uncertainty in the starting condition,
34:27but this time they've all stayed closer together,
34:30and we've got a much narrower distribution.
34:33So we can be more confident about all the forecasts,
34:35more confident of what the temperature is going to be.
34:38And this is one of the real powers of an ensemble,
34:41is that it's able to forecast the day-to-day variability in that confidence.
34:47So it gives you that confidence in the confidence.
34:49Confidence in the confidence, yeah, absolutely.
34:51How happy you can be with the forecast.
34:54And obviously, you know, these are much more closely bunched around the deterministic as well.
34:58And it's just right.
34:59Over the past run, you can see how it's narrowed down.
35:02You often get that name with time.
35:04The bounds will narrow down as you get closer to the time as well.
35:07So you can have that greater confidence.
35:09So it means on this particular day, you can stand up in front of your television camera
35:13and say, we're really confident this is going to happen.
35:15Another day, you might have to be head-to-back.
35:17And that's where you come in and you say things like,
35:20please stay up to date with the forecast because we're not as confident.
35:22I mean, yeah, there's a good example at the moment.
35:24So that's something that really you can take on board as a presenter
35:28and communicating that because you've got that extra bit of knowledge.
35:31But you can only get from ensembles.
35:33Exactly.
35:34Excellent.
35:35Excellent explanation.
35:36And a reminder, there's a whole other video on this, which we will put a link to.
35:40Yes.
35:41In the chat.
35:42But let's talk a little bit more about, well, maybe about uses for them.
35:47Should we talk about the uses for ensembles?
35:50And we've got a few examples here that we can talk around.
35:53This is probably the most obvious one, perhaps, or the most familiar one for people.
35:58Yes. And we've done a lot of work on this recently.
36:00Some of your viewers will be aware, if they use the Met Office app or the Met Office website,
36:05that we were recently trialing some new data on that.
36:08And that is now the standard way we present, we drive these.
36:13And this is using the ensembles much more than it was in the past.
36:17And so, for example, this is a snapshot from our app, our phone app.
36:22And, well, first of all, you've got the temperatures here.
36:26And the temperatures here, whereas before they were just taken from pretty much from a deterministic model.
36:31Right.
36:32Output from the latest run of the model.
36:34These are now taken, as I was describing a minute ago, from the median of the ensemble.
36:39So on those occasions where actually the deterministic forecast is a bit out in the wings,
36:45this is going to be more central and more likely to be correct.
36:49So that's really exploiting that.
36:52When did that change? That change happened this year?
36:54It was earlier this year, yes.
36:55Yes. OK. And then we've got the percentages of precipitation.
36:58And, yes, the percentages of precipitation likelihood at the bottom.
37:03Now, we've had those on the app for a long time.
37:06But we've completely changed.
37:08These are now calculated directly from the ensemble and using a bit of clever post-processing as well to make it even better.
37:14And actually, that's the single thing which has improved the most in quality with the new version of the app.
37:23The probability of chance of rain has improved a lot.
37:27And a good example of how we're introducing ensemble forecast,
37:30most people aren't actually necessarily noticing because it's just being embedded within things that we already have.
37:36It still looks the same.
37:37Yeah.
37:38We're still essentially showing people deterministic forecasts here.
37:41Yeah.
37:42But the quality is better.
37:45Because of the ensembles.
37:47OK. Now, this is something that I use and the chief meteorologist uses here quite a bit at the Met Office.
37:53This is really, really useful for visualizing the chance of certain types of weather happening.
38:00Talk us through what we've got here.
38:02Well, this is actually coming out of the same system I just mentioned, which is driving the new app.
38:08It's a post-processing system which takes the output from the ensembles and does some more clever stuff with it
38:16and calculates probabilities and things as accurately as we can.
38:21And actually, for some things, it does a bit of calibration as well, looking at how good things were in the past,
38:26using the observations and making some further statistical corrections.
38:30But that's probably more detail than we need.
38:32So this example here, this is looking at the probability of fog, of visibility below 100 meters.
38:40And so the colors here are not showing you what the visibility is going to be.
38:45It's showing you how likely is it that it's going to be less than 100 meters.
38:50And so you can see immediately that Ireland here, we've got very high probabilities at this particular time.
38:58I think this was this morning, wasn't it, of fog visibilities.
39:05And then up in Scotland, particularly up the west of Scotland and into Wales and things,
39:10we've got probability of fog, but not as high as it was in Ireland.
39:14So there is a chance.
39:15Yeah, definitely a chance of fog.
39:17As opposed to across the east where there's...
39:20Much, much less.
39:21Much, much less probability.
39:22So again, it's informing people, giving them more information so that they can make decisions based on this information.
39:30Exactly.
39:31So whether fog is a big player for them or not, depending on whether you want to allow an extra hour
39:36or an extra bit of time for your journey to work based on how quickly you've got to get there based on this information.
39:41Or possibly take the train instead of driving or something like that.
39:43Exactly. Yes.
39:44Yeah, even better.
39:45Even better.
39:46So that's a really good example, really useful tool for us meteorologists.
39:49But I think in the future we'll be able to display that more obviously to the public as well.
39:54Yeah.
39:55What's this showing?
39:57We've shown these before in deep dives as well.
39:59This is kind of a plume or showing temperatures higher in the sky.
40:05But it's useful to kind of get trends from this.
40:07So it's really another way of showing the same forecast for a particular thing.
40:12And in this case, it's temperature.
40:15And it's a little bit like the schematic that I showed you, actually.
40:20Because you've got a plume of the different members of the ensemble, these rather faint dashed lines in the background.
40:26But the color shading is also showing you how they're clustered.
40:30So the darker the color, the higher the probability it's going to be in that band there.
40:35So here you've got your sort of central most likely value.
40:38And then the pinker colors are showing high probability the temperature will be there.
40:42You can also see the full spread of what range the ensemble might be telling you.
40:46And again, you get different ones.
40:47So if that sort of green tube is closer to the dark pink, you have that greater confidence because there's less spread.
40:55And obviously, as you get further in time, generally the spread is larger.
40:58But I've seen these ones where there's a lot of confidence because it's rock solid that it's going to be.
41:02And then it suddenly spreads because there's a big perturbation.
41:07It's a really nice illustration just of how, because of chaos, the uncertainty grows as further away you go.
41:13The other thing that can happen with these, not really happening in this one,
41:17but occasionally you can get a collapse back again, a return in confidence.
41:22And that might happen because you've got, say, the passage of a storm or something that's a bit uncertain as to when it's going to go through.
41:29But high confidence it is going to go through.
41:31And then after it's gone, confidence comes back again.
41:33Yeah, we've talked about that again in the deep dive and the 10-day trend.
41:36It seems counterintuitive, but sometimes you can have more confidence in the forecast at day six and seven than you can for days two and three.
41:45So again, it's just that variability within the meteorology.
41:47Really interesting stuff.
41:49What's this?
41:50How's this used?
41:51We've used this in our 14-day outlook sometimes.
41:54So one of the things that we can do with an ensemble is what we call clustering.
41:59So we've got all these different scenarios.
42:02But we often want to try and summarize that.
42:05It's far too much information.
42:06Yeah.
42:07And so what you can do is use an objective mathematical way to try and identify different members of the ensemble,
42:15which are actually rather similar to each other, and cluster them together.
42:18And then produce a kind of summary of, well, if this cluster comes out right,
42:24then this is sort of the average conditions you're going to get given that cluster.
42:28And that's what this is showing.
42:29So this is from a clustering thing, cluster number one.
42:33And it's showing the ensemble mean, the average of all the members that are in that cluster.
42:39So we've got clearly here, all these members are showing a big low-pressure system up in Scandinavia,
42:48well to the north of Scotland.
42:49So you've got a westerly flow over northern Scotland.
42:57Really high probability there's low pressure down to the west of Spain.
43:03And the other thing, all the numbers on here are showing the measure of the temperatures given that scenario.
43:14And there's more information over here.
43:15So 61% probability of this scenario.
43:18So that's using how many of the ensembles are within this kind of broad pattern.
43:23Yeah.
43:2461%, obviously that's quite a high chance that this will be the weather pattern.
43:26It's not telling you individually what the weather's going to be like.
43:29Not giving you a cloud cover, but with high pressure like that, it's much more likely to be dry here
43:33than wet and windy across northern Scotland.
43:35And that's really useful for those medium to long range forecasts.
43:38Yeah.
43:39So that's another use of it.
43:40And then this is...
43:41So this is a really nice one.
43:43Now this doesn't apply in the UK very much because we don't get hurricanes and tropical cyclones,
43:49hurricanes, typhoons and things in our part of the world.
43:53But what we're doing here, this is for a typhoon approaching the Philippines.
43:59It's starting out here.
44:01And for every member of the ensemble, we are calculating what is the track over the next few days
44:07of where that typhoon might go.
44:09And you can see it, same sort of thing, starts quite narrow.
44:12And then as further ahead you go, the more they spread out.
44:15The different colours here represent days ahead.
44:17Different 24 hour periods.
44:18Yeah, that's right.
44:19Yeah.
44:20So you've got then a range of the places that the hurricane might go and therefore affect.
44:31And again, this measure of how confident can we be and where it's going to go.
44:37If it's very narrow, then you're more confident.
44:39If you watch these on a regular basis, they're very interesting.
44:41Yeah, they are.
44:42They really do.
44:43Sometimes they all bunch very close together.
44:44And then other times they can go, you might have one that's shooting off here.
44:48And that's all just depending on what's happening in the upper atmosphere, which is what steers these things.
44:53Yeah.
44:54But again, really, really useful for planning and for people to be at least aware of the risk.
44:58What is the risk where you are of seeing being hit by a tornado or a cyclone?
45:04You can actually put numbers on that.
45:05And that's really useful for people prepping.
45:07OK, it's only a small chance that we're going to get hit.
45:09We'll maybe make a few calls, but we're not going to go into full lockdown mode yet.
45:14Or there's a high chance, a really high chance that this storm is coming our way.
45:17In which case, you know, you need to start evacuating people and making preparations for it.
45:20Exactly.
45:21Hurricane evacuations need to be understarted two or three days ahead because it takes a long time to evacuate a city like Miami or something like that.
45:28Yeah.
45:29So you need to take early preparations.
45:32So there's often still some uncertainty when you're that far ahead, but it gives you a good idea of...
45:37Of confidence.
45:40Making those decisions with a good understanding of what the likelihoods are.
45:43It's about making informed decisions.
45:44Exactly.
45:45It's about making those informed decisions, having as much information as possible.
45:48Now, this guy's good.
45:49I've seen him before.
45:50He's really good.
45:51What are we talking about here?
45:53So this is actually taking the same sort of thing and applying it here.
45:56We don't get hurricanes, but we do get winter storms.
46:01And sometimes we have similar sort of uncertainty in the track that those storms are going to take.
46:06We actually have another tool which allows us to track those storms in each of the members of the ensemble and produce a similar sort of picture to what we just saw.
46:14They're a bit more complicated because we have lots and lots of the storms around.
46:17And that's what you were showing here is two alternative tracks for a low pressure system.
46:23And then talking about the impact that was going to have in terms of where the snow was likely to be in this particular case.
46:30So there's a 30% chance of this happening or 70% chance of this happening and talking through those potential risks.
46:35So, again, it's about informing the public so that they are fully aware.
46:38Again, not necessarily that useful for everyone, but it's really useful for people who are planning and people who need to know about these events higher in time.
46:47Do you need to call more people in on shift to be prepared for this potential weather event?
46:52Do you need to start cancelling trains?
46:54And it's all based on those probabilities.
46:56So, again, really, really useful use of ensembles.
47:01What have we got here?
47:04So this looks horribly complicated.
47:06In a sense, it's showing the huge amount of information that we have in an ensemble.
47:11And quite baffling, actually, for an operational forecaster or whoever to look at all this information and try and extract it.
47:18Which is partly why we summarise things in clusters or as probabilities and things.
47:23But people who have looked at the other videos might have seen this before.
47:27It was the ensemble forecast for the Beast from the East storm that we had a few years ago.
47:34And the colours here are showing a rate of precipitation.
47:40In this case, it was all snowfall.
47:43And the Beast from the East itself was this strong easterly flow coming across the North Sea
47:50and peppering all of the east coast of Scotland and northern England with snow showers,
47:56which accumulated loads of snow over a period of time.
47:59And you look at every single member of the ensemble has got that picture.
48:02So we can be really confident about that strong easterly wind and those snow showers all down the east coast.
48:08But the other thing that happened, and this was a three day ahead forecast.
48:12The other thing that happened then was that we got a new low pressure system that spun up and came up into the southwest of England.
48:19But this three days ahead, we've got a few ensemble members.
48:24Well, that one's taken across the southeast rather than the southwest.
48:27But we've got a few ensemble members which have brought snow showers into the southwest.
48:32And quite a lot of others which didn't.
48:37So three days ahead, we were looking at the possibility of a really big dumping of snow in the southwest of England.
48:43But we couldn't be sure because this is what was happening.
48:46So we were, well, I think, have you got the probability map?
48:50No, not for this one, not for this one.
48:52But it's really useful to show that some of them have it dry,
48:56but a lot of them are showing that symbol for something coming up to the south.
49:00And that's the key, isn't it?
49:02Absolutely it is.
49:03And it allows you as a presenter to talk about the risk of this heavy snow in the southwest.
49:09But we can't be certain yet whether it's going to happen or not.
49:12And of course, if you just used the deterministic, if you just use one model,
49:16if that happened to be it could be completely have any snow and it could be pretty good or it could be completely wrong.
49:21Exactly. That's the thing that deterministic doesn't give you.
49:24It's black or white.
49:26And that's not really useful in this kind of situation where you're just not sure.
49:29But to have some inclination that it might happen is really useful to start prepping,
49:34to tell people to keep up to date with the forecast.
49:36And one of the ways that this was used here was that we were able to issue,
49:39I think it was initially a yellow warning from the southwest and then upgrade that to an amber warning.
49:45And then as we got closer to the event, then they all started converging,
49:48gave high confidence that that's no event.
49:50And so it was I think it was it was raised to a red warning for the southwest.
49:54But it allows us to give that early alert through a yellow warning.
49:59And then we can as we get closer to the event, we can either say, oh, well, that's actually that's not going to happen.
50:04So occasionally you'll see a yellow one. It'll just be taken down.
50:07Yeah. But other times it gets ramped up to an amber and a red.
50:11And but people have had some early warning to start being prepared.
50:16It comes with that confidence. And that's, again, anything you get with with ensembles.
50:20One question we're going to get asked a lot is how do you mix up those initial conditions?
50:25We talked a little bit about tweaking the pressure here, changing the temperatures there.
50:28But it's it's it's quite a lot more complicated than that, isn't it?
50:31I think this graph should help to show that.
50:33Yes. So what we're showing here is, again, going forward in time and we're generating the new starting conditions for this time here,
50:43nine o'clock in the evening. And we don't start from scratch.
50:47We start with we've got a previous forecast. And in fact, we also use that in creating that analysis.
50:53The starting condition for the most start over every time.
50:56We don't start over every time, but we combine the previous forecast with the observations to create the new starting conditions.
51:02But we do the same from the ensemble. Now, here's the old ensemble.
51:06And it's evolved and the spreads grow in quite a bit. And we know that the uncertainty at this point is much less than that.
51:13It's constrained by the observations. So we shrink that spread.
51:19So each of the perturbations we take as the difference between the sort of the average and the and the individual member.
51:25So it's a continuous cycle.
51:28We shrink those differences back to the typical size of the analysis errors.
51:32And then we add them back to the analysis to get us a new set of a smaller spread of new conditions.
51:39And then it goes on. And that's that's the chaos working.
51:45OK. And what have we got here?
51:48So that's a very vague picture how we change the initial conditions.
51:53Yeah. But actually, we also know that the model is not perfect in the way that it represents how the atmosphere evolves.
52:02It actually does very, very well on the large scale flow and what we call the dynamics of the atmosphere.
52:08How the winds and the pressure systems develop and all that sort of thing.
52:12But it works on a grid of points. Yeah.
52:16And those points, well in our global model now, are 10 kilometers apart.
52:21There's actually quite a lot happening inside those boxes.
52:23There's quite a lot of weather that goes on in 10 kilometers.
52:25So this is just a picture of a grid box within the model.
52:28And for example, this might be showers, rain showers happening within that grid box.
52:34Or it's also, there's also turbulent motions, gusts of wind and all that sort of thing.
52:40Small scale things that the model can't represent.
52:43Because of the grid size is too big and they're smaller than.
52:46So we have little models, if you like, within the model, which represent statistically what's going on inside here.
52:54But it is, it's a statistical average of what's going on inside these grid boxes.
52:58And so what we do is that we actually perturb those as well while the model's running.
53:05And produce, because this produces outputs which feed back up into the larger scale flow.
53:10It actually triggers chaos.
53:12Little differences in how, what's going on in this box, make small differences back onto the large scale flow.
53:19And then chaos means those can grow in the future.
53:21My brain's hurting already. It's a good thing we've got a supercomputer to do all this.
53:26That is why we need one, absolutely.
53:28It is, and actually ensembles now are the biggest consumer of the power of the supercomputer.
53:33Another question we're going to get asked a lot, Ken, is if you use probabilities, you can never be wrong.
53:39So that's, that's, that's what always people say, isn't it?
53:42Well, if you say it's 70, then you can never be wrong.
53:44But actually, we, we know that ensembles are more accurate than deterministic focus.
53:49How do we, what's the process involved with that?
53:51You're quite right. It's one of the questions I get asked, or accusations I get more than anything.
53:55You're just trying to hedge your bets to make sure you can't be wrong.
53:58And there's a certain truth in that.
54:00As long as we don't say 100% or 0%, then yes, we can't be wrong.
54:04But we can.
54:06Because if we say something's got a 30% chance of it happening,
54:13and we take lots and lots of times that we said 30%,
54:16and we look at how often within all those times it happens,
54:19it should happen 30% of those times.
54:21So you can measure that.
54:23You can do it also for 40, 50, 60, 70% as well.
54:26And that's essentially what we do.
54:28Now this is an illustration.
54:30We've got here, I'm not going to go into the details of how this particular verification works.
54:37But it's, it's like what I just described, but it's looking at all those different probability thresholds.
54:43And so we can use that over many, many forecasts to measure how good these probability forecasts are.
54:51It's a lot of data, haven't we? We've got a lot of data to verify.
54:53Now what we're doing here is something a little bit clever,
54:55is that we're doing that both for the deterministic model and for the ensembles.
55:02The deterministic model's in blue, and the ensemble in red.
55:07I won't go into detail of how we do it for a deterministic model.
55:11We can generate probabilities sensibly out of the deterministic model as well.
55:16The red is tracking above the blue.
55:18Yes, we've made the effort here to treat both fairly and equally.
55:25And higher is better on this graph.
55:28The very jaggedy lines up and down are the monthly scores.
55:31These are averaging through that over time.
55:33And you can see that very, very consistently, the ensemble in red is performing better than the deterministic model in blue.
55:40This one's for precipitation.
55:42Over the right-hand side, we've got temperature and wind speeds, and it's doing the same thing.
55:46And this is a very consistent picture across lots of different ways we measure how good the forecasts are.
55:51As a result of this, this is really the big driver behind us deciding the big change that we're going to make in our modelling over the next year.
56:00Do you want to move on to that other slide?
56:02Yes.
56:05So this is the main modelling systems that the Met Office runs at the moment.
56:11We have a global model, which we run a single realisation several times a day at 10-kilometre resolution.
56:20That's the size of the grid boxes.
56:22And an ensemble run at 20 kilometres with 36 members in it.
56:27Similarly for the UK, we do a single run at a 1.5-kilometre box size.
56:32So it's higher resolution, more accurate in theory, because it covers the UK.
56:35Just over this area here.
56:36Obviously, it takes more processing power, so you can't do that across everywhere.
56:39So we focus on the UK.
56:40And that's often the pictures that you show on the television.
56:42Yeah, yeah.
56:43From that model.
56:44And we run an ensemble at slightly lower resolution at 2.2 kilometres.
56:48That's what we do now.
56:49That's what we do now.
56:50But because of the verification I showed in the previous slide, we are actually getting away from that.
56:57We're stopping running a single higher resolution deterministic model.
57:01And we are just going to ensembles.
57:03For the UK, we're raising that resolution up to the 1.5 kilometres, the same as the deterministic is at the moment.
57:09And for the global, we're doing the same thing.
57:11We're running the global ensemble at 10 kilometres.
57:13What's the timescale on this?
57:14This is coming in on our new supercomputer in 2026, hopefully.
57:18If everything goes to plan, it's coming in in 2026.
57:21And that should provide more accurate forecasts, more detailed ensembles, more frequent,
57:26higher numbers of members, so it gives you better data.
57:29Well, initially, it will be the same number of members.
57:31But in future, the sort of improvements we'll be making in the future will not be running a higher resolution deterministic,
57:37but doing things like increasing the number of members in the ensemble
57:41and when we can afford to increasing the resolution of the ensembles themselves.
57:45But really big change.
57:47We've been running those deterministic forecasts for 40, 50 years.
57:52And suddenly now, we're only going to be running ensemble forecasts.
57:54That's a really big change.
57:55Quite a big shift.
57:56Again, a lot of the forecasts that people see, you won't necessarily spot the difference.
58:01No, that's right.
58:02They're not going to look really any different or feel any different
58:05because of the way that we can make them seem similar.
58:08But they will be more accurate, and they will give you more data, more information, and help people make better decisions.
58:14Exactly.
58:15That's the plan. Good stuff.
58:16And over time, we are planning to start bringing in a few more other products, additional products as well,
58:22which are making more explicit use of the ensembles.
58:24Excellent.
58:25So we'll have you back to have more discussions about ensembles in the future.
58:29But I hope that was interesting.
58:31I was fascinated.
58:32I learned an awful lot, Ken.
58:33We had a lot of questions about ensembles because we talk about them a lot in the Deep Dive, the 10-Day Trend.
58:37So for you to explain them here in such detail is really, really helpful.
58:40Thank you very much.
58:41And yeah, if you've got any questions about ensembles, then do put them in the chat as well,
58:47and we can always get Ken back and try and answer them.
58:49But for now, Ken, thank you very much.
58:51Thank you, Alex. It's been a pleasure.
58:52Cheers.
58:53Huge thanks to Ken for coming in and filming that earlier.
58:56And huge thanks to you for sticking with this extra special, extra long Met Office Deep Dive.
59:02Thank you very much for being there.
59:03Do let us know if you like that kind of thing, if that's the kind of level you want us to go to.
59:07If you're interested in any particular aspects of meteorology, then again, do drop us a comment and we'll try and cover that.
59:15There is another video that explains why we use ensembles, what they are and how we can use them.
59:22And you can catch that by clicking on here.
59:25And don't forget, Weather Studio Live will be back on Friday.
59:29And on Wednesday, Aidan will have the Met Office 10-Day Trend.
59:32Thanks for watching.

Recommended