Instant Genius - How accurately can we predict the weather? – Andrew Blum

Episode Date: August 7, 2019

Hurricane Sandy hit the east coast of the United States in October 2012, causing $65bn of damage. Remarkably, weather forecasters managed to predict its impact on the US eight days in advance, when it was barely even a storm. How did forecasts get to be so good? It’s a story that begins with the invention of the telegraph and ends with supercomputers. We talk to Andrew Blum, author of The Weather Machine (£16.99, Bodley Head), about the history of weather forecasting, why we shouldn’t trust the icons on our weather apps, and whether we’ll ever have an accurate minute-by-minute forecast. He speaks to BBC Science Focus online assistant Sara Rigby. Listen to more episodes of the Science Focus Podcast which we think you will find interesting: What's going on with the weather? – Dann Mitchell Could leaving nature to its own devices be the key to meeting the UK’s climate goals? – Mark Lynas Can we really predict when doomsday will happen? – William Poundstone What if the Earth’s magnetic field died? – Jim Al-Khalili Why is the magnetic north pole moving? – Ciaran Beggan Are we facing an insect apocalypse? – Brad Lister Follow Science Focus on Twitter, Facebook, Instagram and Flipboard Image: Actor and environmental activist Leonardo DiCaprio stares at a visual showing Hurricane Sandy using data from Goddard Earth Observing System Model © NASA/Goddard/Rebecca Roth Hosted on Acast. See acast.com/privacy for more information. Learn more about your ad choices. Visit podcastchoices.com/adchoices

You said this place was steps from the water. We just haven't found the steps yet. How much did we save? Enough. Enough to get lost. Or you could book a stay with Hilton. Welcome to your ocean front room. Just steps from the water.

The Hilton sale is on now. Book on Hilton.com or the Hilton app and save up to 20% to get the stay you expected. When you want savings, not surprises. It matters where you stay. Hilton, for the stay. When you need to build up your team to handle the growing chaos at work, use Indeed sponsored jobs. It gives your job post the boost it needs to be seen and helps reach people

with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this shell will get a $75 sponsor job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs. Pool days call for cookouts and lots of laundry. This Memorial Day at Lowe's, save $80 on a Charbroil Performance Series 4-burner gas grill. Now just $199. Plus, get up to 45% off select major appliances to keep dishes, clothes, and food fresh.

Having fun in the sun is easy with us in your corner. Our best lineup is here at Lowe's. VALTHUFFELSA. While supplies last, selection varies by location. See Associate at Lowe's.com for details. This podcast is sponsored by name, audio and focal. Streaming has made music more accessible than ever, but true listening is about more than ease.

It's about quality. British audio experts name audio, alongside French acoustic specialist focal, combine handcrafted tradition with cutting-edge innovation and high-end materials, delivering digital precision with analogue warmth. So you can experience exceptional sound at home. Music just as the artist intended.

Visit name audio.com to live. more. The statistic that the meteorologists love to talk about is a day a decade, which is to say that their models have improved by a day of accuracy with each passing decade. So a five-day forecast today is as good as a four-day forecast was 10 years ago, it's as good as a three-day forecast was 20 years ago. And that's a trend that just keeps continuing. You're listening to the Science Focus podcast from the BBC Science Focus magazine team. With the UK's best-selling science and technology monthly, available in print and in several digital formats throughout the world. Find out more at ScienceFocus.com or look out for us in your app store.

Oh, and welcome to the Science Focus podcast. I'm Alexander McNamara, online editor at BBC Science Focus magazine. Hurricane Sandy hit the east coast of the United States in October 2012, causing $65 billion of damage. Remarkably, weather forecasters managed to predict its impact on the US eight days in advance, when it was barely even a storm. How did forecasts get so good? It's a story that begins with the invention of the telegram. and ends with supercomputers. We talked to Andrew Bloom, author of The Weather Machine, about the history of weather forecasting, why we shouldn't trust the icons on our weather apps,

and weather will ever have an accurate minute-by-minute forecast. Here's Andrew, talking to BBC Science Focus online assistant, Sarah Rigby. So what made you want to write a book about meteorology? Well, I've always been interested in these sort of complex, technical, sort of big infrastructure systems. And weather was definitely one of those. I was very curious about where the weather forecast came from. But more than that, I was really surprised that there had been this kind of paradigm shift in how the forecast worked, and I hadn't really heard much about it.

It seemed that it was no longer about the skills of meteorologists, the kind of human intuition, but rather it was about the sort of incremental improvement in the technology of the weather models, these sort of supercomputers that are at the hard weather forecasting today. But they were also kind of a black box. I wasn't really sure sort of how to get inside and understand these sort of incredibly complex. systems and sort of began to poke around, especially after Hurricane Sandy in the United States. It was really a moment where the models worked astonishingly well. And I think everyone started to say, wow, what was this thing that we had created? Yeah. So in your book, you say that they predicted Hurricane Sandy something like eight days before it actually happened. Is that right? Yeah, it was amazing. I mean, the storm came to New York on a Monday evening. And the first inkling of it was the previous Sunday

afternoon when the weather model sort of first spit out this possibility that this pretty substantial hurricane would sort of take a left turn towards New York City. And at that far out, it wasn't a short thing. But with each passing day, it became clear that this was actually going to happen. And it turned out that, yes, that eight-day forecast was correct. It was really sort of boggled the mind that this was possible. Right. So you said that it was before it had even really started to take shape, right? So how could they predict it so far in advance? Well, of course, it's not something that any human can sort of look at a map and figure out. But the way that the models work sort of fundamentally is to run an ongoing simulation of the

atmosphere that's based in the laws of physics. That goes back to equations that were sort of first drawn up more than 100 years ago, starting with a Norwegian meteorologist named Wilhelm Bjerkness. And of course, at the time, there was no way to calculate those equations. And they wouldn't even sort of tell you what the weather would be. But in this kind of amazing iterative process, you know, year by year, month by month of scientists sort of trying out different things and testing it each day on what the weather actually turns out to be, we've really kind of made incredible progress. The statistic that the meteorologists love to talk about is a day a decade, which is to say that their models have improved by a day of accuracy with each passing decade. So a five-day

forecast today is as good as a four-day forecast was 10 years ago, is as good as a three-day forecast was 20 years ago. And that's a trend that just keeps continuing. Wow. So the invention of these models must have been a really big moment in meteorology. And another one of those that you talk about in your book is the invention of the telegraph. Can you explain a bit about how that had such a big influence on weather forecasting, please? Yeah. I mean, it's fundamentally, it's about being able to to hear about the weather someplace else before faster than the weather actually arrives. So I just, I love this notion that there, you know, it requires a kind of, you know, sort of geographically leap of the imagination to know what's happening, you know, at a distance.

And that's the kind of sort of root of all telecommunications. But one of the first uses of the telegraph was to, was to, you know, to tap out ahead of what, you know, what storms were coming. And it could only be a sort of rudimentary forecast. You know, it couldn't, it couldn't, the meteorologists of the time had theory. They sort of couldn't really predict how the weather might evolve, but they finally had this ability to send news faster than a horse or a train could carry it. So what would you say with a really important moment in the history of meteorology, the ones that

really sort of change the direction of how we forecast the weather? Well, certainly it does begin with the telegraph. It does begin with this idea that you could by sort of communicating over distance, assemble a synoptic map, a sort of coherent map of the weather at a given moment in many places in time. And of course, knowing the weather in many places at one time is the first step towards knowing the weather in many places at different times, specifically into the future. That's the most helpful time to know the weather. And I think the next step would be the sketching of the primitive equations, as they came to be known by Wilhelm Bjerkness, a Norwegian meteorologist. And that was, you know, they were based on sort of the, the advancements of

classical physics and thermodynamics in the 19th century. But he really sort of said, not so much the equations were, were they, they, they turned out to work. But the sort of bigger insight was just the sort of scientific hypothesis of it, that you could, if you could, could describe the, the sky with math, then you could sort of prove each day with what, you know, if it was actually doing what you thought it was going to be doing. And I think, sort of that notion of turning the meteorology into a science rather than purely empirical, but actually theoretical, was a major step. And after that, what it really took to enact those, it was another 50, 60 years. It took the global view that weather satellites allowed, you know,

so that you have a global map with the weather, you really need global data. And for really, you know, consistent global data, you really need that sort of view from space. And then, of course, it took computers. It took the ability to calculate those equations. Many people, know the famous story of Lewis Frye Richardson, who tried to essentially use Björchnus's equations to calculate the weather in the over World War I and into the 1920s. He was famously wrong, and it famously took him, you know, took him weeks to calculate the equations and figured he could do it faster if he could fill a stadium with tens of thousands of human calculators. And of course, we don't have that today, but we do have the supercomputing power. And really the

bleeding edge of supercomputing power is what's applied to the most active weather models. So what is it that supercomputers do differently in terms of forecasting the weather than traditional human meteorologists would do? Well, I think the number of variables required to run an ongoing simulation of the atmosphere is really the crux of it. And I think there's a sort of a misconception that the weather models are statistical, that they take past weather and from past weather, extrapolate future weather. And that's not entirely the right way of thinking about. It's not as if the observations go in and, you know, it's compared against any sort of any, any past state of the atmosphere and the most likely future state is spit out. Instead, it's kind of an ongoing concern. It's an ongoing simulation of the atmosphere that's compared against the latest

observations. I like to kind of imagine a sort of two globes kind of side by side, clicking ahead in parallel. You have the kind of, you have the atmosphere inside the computer, then you have the real atmosphere. And with each step, the sort of the most recent observations are compared to the most recent forecast of the model, and it's corrected slightly, almost like a, almost like a duet. So there's the real atmosphere and there's the simulated atmosphere. And, you know, with each run of these supercomputer models, they, they, they bring, they're brought closer together. And then, of course, the model then runs days into the future, giving us the weather forecast that we know. but I just love this notion that in a way the atmosphere inside the model is sort of more precise than the real atmosphere because we know it better.

You know, we can we can sort of pick a single point and say, yes, these are the conditions at that point, where of course in the real atmosphere we can only do that for the places where we have actual observations. Can you please explain to us what the different types of forecasts are, for example, the differences between local and global forecasts? Yeah. I think that the, well, it's interesting. I mean, one distinction to draw is the difference between forecasts and weather models. I think that the forecast is a sort of what would you extrapolate from the model. The model is just that.

It's a model of the atmosphere, which could come out of either a global model, which sort of models the entire atmosphere of the earth, or could come out of a regional model, which is often a higher resolution, you know, likely to resolve things like thunderstorms better, or and also it's likely to be refreshed more often because it takes less computing power to calculate a regional model than a global model. So you sort of you end up with a kind of hierarchy of models where you have the global models, especially the sort of king of models, as meteorologists like to refer to it, is the one that's operated by the European Center for Medium Range weather forecasts, which is the sort of pan-European operation that at the moment is located in Reading in England. But for that, where it excels is with days or even a week into the future.

Because at that scale, it really does depend on a sort of a holistic, a complete view of the atmosphere. But to know whether or not, for example, there are going to be thunderstorms in New York this afternoon, it becomes more important to have a higher resolution or regional model. And so it's really, it's about having a sort of a range of tools available. And the key point, of course, no forecast is worthwhile unless we can act upon it. And so for meteorologists whose job it is to sort of communicate to us what the weather will be and ways in which we might adjust our behavior, whether to get out of danger or to do something simple, like plan a picnic or bring an umbrella, I think that sort of ability to kind of switch between these different models,

between regional models that are more focused on the shorter range time periods and between global models that really excel at longer range time periods is sort of crucial for making this system as effective as I think it's become in our lives. I see. So would you say that sort of the rise of the internet and smartphones had an impact on weather forecasting? Yeah, it's a, it's certainly had an impact in the way that we receive it. I write in my book about, about weather underground, which is one of the sort of called a skin

of the weather company. which is sort of the service that provides forecasts, not only for their own apps and for the weather channel, but also for Google and Facebook and I think the numbers, several billion forecasts that they deliver each day. And what they discovered is that with the rise of smartphones in the last 10 years, it became, you know, people wanted their forecast more often, and they wanted it to be more geographically precise,

because no longer were they sort of listening to the forecast on the radio in the morning, but they were often checking it where they were, you know, over the course of the day. And so that really changed the way that they began to think about serving up those forecasts. So rather than having a sort of, you know, here's what the weather is going to be in the New York City area for the day. They built a system that they call their forecasting engine that sort of dips back into the data each time you refresh your app and gives you the latest, the latest most likely conditions by drawing on the kind of the combined outputs of a whole range of weather models. operated by the different sort of national government weather services. And I think that that hierarchy is really important to make, you know,

between the sort of the ongoing models, the ongoing simulations, the atmosphere run by the weather services, and the different weather meteorologists, whether at a place like the Met Office or the BBC or at a company like the weather underground or the weather company that really draw upon those models to then deliver to us the forecast. So the weather app that comes with my phone is quite useful. It predicts thunderstorms far more frequently than actually happen here. But I know that, you know, for example, if I wanted a really good weather forecast,

I might go to the BBC or the Met Office or something like that. So why is it that some weather forecasts can be so much better than others? Well, I think a lot of it is the way in which those little icons are sort of tuned. You know, I think that there, you know, I think there's definitely a bias for the app makers to predict rain when it's not going to rain, then rather than lets you get stuck in the rain. So when you think about how these things are working, the models are saying there's a certain likelihood of a storm forming.

But at what point in the kind of iconographic delivery of that information, do we change the icon from a sun or a cloud to a rain cloud or to a thunderstorm? And I think the bias is going to be towards tipping that over sooner rather than, then later for a consumer app. So it's funny, one of the things that I like about the weather underground app, which uses the exact same data as, for example, the weather channels app or even Apple's weather app or the weather on Facebook is that it kind of, it assumes that you're a bit of a weather geek. So it's less likely to sort of tip over into that next icon. But for me, it's a kind of perfect example that often it's not, it's not the, it's not the model that's wrong. It's the,

it's the sort of communication that's the challenge. And I think that that's indicative of how hard it's been for all of us, and particularly for people who have big decisions like closing schools when there's snow or evacuating people when there's a typhoon coming, that sort of evolution to trusting the quality of the forecast that we're getting. There's still an inclination to sort of second-guess it. And I think that we're starting to see a lot of examples where, you know, where major decisions are being made that have sort of life-saving,

real life-saving consequences, because not only is the weather forecast good, but it's really good enough that we've learned to trust it. So I think that is definitely the trend. I mean, certainly it's not always perfect, but I think if we kind of learn to recognize the moments that the forecast is saying maybe, you know, it might rain. You know, it's not saying, you know,

sometimes it says it's definitely going to rain. Sometimes it says it's definitely not going to rain, but sometimes it says maybe. And that's actually a pretty good forecast when you think about it. So, yes, you say that it's important to be able to interpret the weather data correctly as well as just having all the data. So would you say that as good as computers can be, we're still going to need the human element to meteorology. We're still going to need the interpretation. Yeah, absolutely.

And that's a big, a sort of big trend that's happening slowly but decidedly in meteorology right now. That meteorologists are realizing that they don't need to spend their time sort of, you know, looking at the weather maps and just, you know, sort of, you know, sort of, you know, deciding whether or not it's going to be, you know, 73 degrees or 76 degrees tomorrow, that the computer is better at that. But in terms of predicting impacts, in terms of predicting what the weather is going to do to all of us, you know, what kind of traffic it's going to cause, what kind of, you know, ice might be on the road, you know, what kind of snow might build up or flooding there might be on the coast. In order to do that, that's still where humans excel. So it shifts the job.

It's the sort of term for it is impact-based forecasting. And I think there's a bit of a letting go that has to happen on the part of meteorologists to acknowledge that their role has shifted, that it's not as much about the analysis that they might have been trained in, particularly 20 or 30 years ago, that, in fact, the computer is much better at that now. but what we still really need them for is the communication of those outputs, the sort of understanding of what the impacts might be, particularly when it creates different kinds of dangers or inconveniences. So these forecasting models have been getting better and better over the last decades, as you say, they've been getting one day better every year, every decade. So what is it that gets improved about the models?

How can you go about making forecast more accurate? Yeah, it's complicated. And I've kind of divided into sort of three different categories of improvement. There's one, the first category would be improvement in the observations. You know, how you need to know what the state of the atmosphere is to know what the state of the atmosphere will be. So new satellites, more weather observations on land, different kinds of sort of new generation of observations, things like the barometer on your cell phone taking recordings

and sending the back book through the app or an amazing system called GPSRO that uses GPS signals to measure the conditions of the atmosphere. All of those things are sort of one of the three legs of the stool. The second leg is sort of even broader, and it's the kind of science of it. It's, you know, what are the way, you know, what are the algorithms that make the models work and how well do they describe the evolution of the atmosphere? You know, how well do the algorithms say, you know, when certain conditions are going to make rain inside the model? You know, how well do they predict the temperature? And you have to picture this incredible, you know,

three-dimensional, or really four-dimensional grid of the conditions of the atmosphere inside the model that relies on the sort of the hundreds and millions of equations to see how it will evolve. And that kind of brings us to the third leg of the stool, which, of course, is supercomputers. And I think there's a, you know, inclination to say, well, if only we had, you know, better,

faster computers, we would have better weather models. But it really is these sort of these three things working in concert. the better observations, you know, better science, better algorithms that describe the atmosphere, and then better supercomputers to crunch it all together. And, yeah, you know, the metaphor that I sort of came to somewhat resignedly in the time that I spent at the European Center for Medium Range Weather Forecasts, watching them try to improve the model, was the sort of cockpit of a 747.

You know, they're constantly sort of spinning dials and adjusting levers and starting new programs to try and figure out, you know, what could make it better. but because they can sort of run the model in different ways, because they can use some of their computing power to experiment, to try different things. And because the weather tomorrow always comes and they always have observations for the next day, it really is like this daily science experiments

that they sort of tweak and tweak and iterate. And I think it is that ability to test these different improvements against the weather that actually comes that's been so crucial to improving the, the models year by year. It really, you know, there's a very sort of concrete way of saying, did this work or did this not work? So meteorology is, it's a bit different to most other sciences, isn't it, in that they've got this such vast wealth of data and they don't need to sort of go out and get it. It's always coming into them. Yeah, it's wild. I mean, especially as a

prediction problem. I mean, you know, if you're in the business of, you know, predicting elections, for example, you don't, you don't get to do it that often. You don't, you don't have a whole lot of data to use. But as one scientist in the book put it to me, you know, who I write about, a guy named Jeff Anderson in Boulder, Colorado, one of the sort of world experts on this. You know, it's really, you know, meteorology is special because, you know, you can, you can, you can sort of make an hypothesis about how you think it's going to behave and then you can be proven right or wrong tomorrow. And then not only that, but you can go back and sort of, you know, punch it up, you know, punch it in with the past 20 years of weather data. And in fact,

that's how the modeling systems are set up. They have the ability to sort of make forecasts from the sort of the data record. Again, inside the simulation. I think we're recording this on the anniversary of D-Day, and I think of there was a, they went back and they used the observations that they could collect from the lead up to D-Day, and they put them into the existing weather model, the sort of current version of the weather model. And so they using the then we're able to get a pretty good forecast.

So it's wild to think about that ability to sort of scrub forward and backward in time, both for research purposes, but then also in this way that shows up in our pocket. That's the sort of most accessible information as there is. Right. Yeah. So will we ever realistically be able to get a perfect minute by minute forecast? or is that just out of the bounds of human possibility? Well, the European Center has set the goal not only of adding a day a decade, but of increasing the rate of that improvement. And that's not a perfect forecast.

That's a kind of sort of incremental measure of skill. You know, it's saying that we want to be better than we would be just by saying predicting the weather that there was a year before. So, you know, again, it's sort of, The way that I prefer to think about it is less the sort of perfect minute-by-minute forecast and more a forecast that's really good enough to act on. I've been joking with my daughter plays softball and we've been joking with the other parents about, you know, well, is it going to rain on Sunday morning or is it not going to rain on Sunday morning? And, you know, often, you know, week by week, you know, by Tuesday or Wednesday, the forecast has been right. You know, yes, it's going to rain.

No, it's not going to rain. And that, you know, it's that becomes good enough. You know, we don't quite want to make decisions, you know, four or five days in advance yet. But a lot of times if we did make decisions four or five days in advance, we would have been right. Those would have been good decisions. So that's pretty remarkable progress. And I think that there really is this kind of delay in trusting it. And again, part of the sort of, you know, the ability to trust it requires knowing when it's not sure.

And that's a kind of uncomfortable position to be in. And I think that's a particularly uncomfortable position for the people who make weather apps and for meteorologists to really say, you know, this one is really, I'm really sure about it. And this forecast, I'm not really sure about it at all. So you should be a bit more careful. Right. Okay. So do you think it's more valuable to have a forecast further in advance, an accurate forecast further in advance rather than a perfect forecast now? When is now, I guess, would be the, you know, with the, you mean, you know, is it better to?

for the next day. Again, it kind of depends what you're using it for. I've been thinking a lot about this again, as I sort of cheer on the work of the weather modelers and the success that they've had, you know, when is it really actionable? You know, when, and often one of the great frustrations is, you know, even if the forecast is perfect, that doesn't mean you're going to like the weather. You know, if you're planning an afternoon at the pool, you know, even if the forecast is perfect, you know, that doesn't mean you're really, that doesn't solve your problems. I can kind of can kind of create more problems than it solves. But it's pretty strange to be able to predict the future.

And there are exciting moments, you know, when, you know, in New York this time of year, we often have thunderstorms coming through in the afternoon. And so when my app says it's, you know, thunderstorms likely at 5.30 p.m., and I'm sure enough, you know, thanks to this sort of broader, you know, this radar system that allows us to see over the horizon, you know, you start to make decisions about that. I will say, though, that over the years I spent working on this book, the worst days where I was, you know, staring at my screen, you know, struggling with these words and then going out and getting caught in the rain because I forgot my umbrella.

There's sort of nothing, nothing worse than writing a book, writing a book about weather forecasting and not paying attention to the weather forecast. So what would you say are the biggest hurdles or problems that we need to overcome in terms of improving the weather forecast? or getting it even better? I think the one thing we haven't talked about at all that I think is really important is the fact that the global observing system is really the result of global cooperation.

It's not owned by any one nation, but it really depends on the sort of diplomacy of the weather. And the European weather satellites, being carefully lined up with the American weather satellites to sort of cover the Earth at different periods of time, or even just the observations that are taken at surface level from all over the Earth and are sort of pooled together very deliberately in a very designed system operated and dictated by the World Meteorological Organization, the UN Organization.

So when we talk about sort of what it takes to improve forecasts, I think one of the risks is that we begin to back away from that global cooperation, that the new technology, the new observation technology, new kinds of satellites, new kinds of observations from mobile phones. The risk is that they might come from private companies, and we might sort of have a break with the 150-year-old tradition of those observations being collected by governments for the express purpose of being shared globally, shared with other governments, with other weather services. So I think certainly more observations would do it, but one of the challenges is, collecting new observations and new types of observations and making sure they're not the exclusive property of Google, making sure that whoever is collecting these observations is incentivized to share them. So that's really interesting, actually. It's not something that is sort of often brought up as an example of global cooperation.

Yes, I mean, I suppose not. I think one meteorologist, a woman from Australia, a former head scientist at the Australian Bureau of Meteorology named Sue Barrel, she said to me that recently that perhaps the reason we sort of aren't as concerned about this ongoing global cooperation is because there haven't been any wars or battles to make it happen. You know, it really is a sort of 150-year-old tradition of global data exchange. and the meteorology community has done such a good job of keeping that ongoing, that there hasn't sort of been a moment of crisis where we've sort of had a chance to appreciate what we have. I have to say, unfortunately, I don't think we're there yet, but the current inclination in the United States is towards a greater emphasis in allowing private companies to do what they want, to collect more data and do what they want in the way that they sell it.

And so I think there is the chance that this issue will come to the fore of the next few years, that particularly the World Weather Watch, as it's called, the kind of global program that sort of emerged out of the post-war era was really sort of initially called for by President Kennedy and a speech of the United Nations. But there really is the risk that the foundations of that begin to be eroded because the technological development is no longer the realm of governments, but is really the realm of these sort of, you know, large international corporations. So do you think that allowing these private corporations to have the sort of monopoly over weather data and not necessarily having any duty to the public to share that? Do you think that puts people at risk? I think it's interesting because the already, the systems that exist already are each year by year, you know, do a good job of putting fewer people at risk. We saw this amazing example from May, from the cyclone in India, that an equivalent cyclone

20 years ago killed tens of thousands of people. With this cyclone, it was well enough predicted that authorities could make the decision to evacuate people, and there were very few deaths, which was remarkable. So I think the concern is less that it puts people at risk and more that it makes a sort two-tiered system of who has access to the best forecasts. And I think those are kind of the same things, and that obviously tracks sort of broader trends in the world today. But I think, you know, particularly as we enter an era of more extreme weather, and I should say

that it's extreme weather, that the weather models are particularly good at predicting, because again, unlike human meteorologists, they're not biased by the weather in the past. They're able to sort of bite, because they're based on the laws of physics, sometimes they spit out crazy things. Like it's, you know, it's going to rain 25 inches in Houston as it did last year in Texas. And the human meteorologists say that's, that's impossible. That's never happened before. And the computer says, well, this is going to happen. And in that case, you know, yes, it did happen. So, you know, as we enter the air of more extreme weather, I think, you know, it's obviously important that there's a sort of the global good of the system that's been

developed is evenly distributed. And I think that is the case at the moment. But, you know, it's not a foregone conclusion that we can keep that. I've been thinking a lot about the, there's the famous Benjamin Franklin line where when the declaration of, when the Constitution of United States has written, somebody comes up and says, what kind of government do we have a democracy or a republic? And he says, a democracy, if we can keep it. And it's the sort of if we can keep it idea that I think is really, you know, is really crucial at this moment in the technological and the diplomatic history of weather forecasting.

Okay, thank you. And just one last question. So what would you recommend as the best forecast for individuals to use to check, say, will I need to take an umbrella today? That's a tricky one. I mean, of course, it kind of depends where you are. It depends. And I, you know, In the United States, I think that the weather underground, because it takes its system for collecting the model data and then adjusting it to your particular location, I think is one of the most sophisticated. But I think the Met Office does the same thing. And so I think the most important thing is to look really closely, sort of look past the icons. You know, often, you know, I think there still is a bit of a delay in how much meteorologists are letting us trust how fine-grained the outputs that they're getting are. So when it says, you know, it's, you know, a 30% chance of rain at 3 p.m. and a 50% chance of rain at 4 p.m. and a 70% chance of rain at 5 p.m., you know, you can, you know, that's a pretty good clue of what's going to happen.

It's different from a 0% chance and 100% chance, of course. But I think we often are, the sort of the forecasts we're presented with are kind of smoothed out in a way. That was Andrew Bloom talking about how we predict the weather. His book, The Weather Machine, is out now. Come rain or shine, why not pick up a copy of BBC Science Focus magazine? It's packed full of features, news and interviews. And in this month's cover feature, we discover how the dinosaurs could hold the key to staving off our own mass extinction. If you just can't wait to get a hold of a copy,

then check out our many, many previous science-focused podcast episodes, all of which are equally as interesting as this one. Let us know what you think with a review or a comment. It goes a long way to help get the podcast out there. Thank you for listening to the Science Focus podcast from the BBC Science Focus magazine team. With the UK's best-selling science and technology monthly, available in print and in several digital formats throughout the world. Find out more at sciencefocus.com or look out for us in your app store.

This podcast is sponsored by Name, Audio and Focal. The texture and emotional depth of music can be lost through digital sources or poor signal. Name Audio believes you can have digital precision with analogue warmth. Alongside French acoustic specialist focal, Name creates high-end audio systems, combining innovation with craftsmanship, so you can listen to music, just as the artist intended. Discover more at name audio.com. Ambition comes in all shapes and sizes.

At First Citizens Bank, we roll with your goals because we're built for what you're building. Fit for your ambition for Citizens Bank. Hey, honey, it's Mom. Did you know if we switched to Verizon we can get four phones for $0 plus four lines for $25 a line? Call me back. Me again.

That's just $100 a month for four lines on Unlimited Welcome. Plus four phones, no trade in needed. Call me. It's Mom. America's Best. network, Verizon. That's the one we're talking about. I'll send you text. America's best network based on Root Metrics, best overall mobile network performance, U.S.

second half, 2025. Four new lines and a limit and welcome and auto pay. See Verizon.com for details.