Amherst Reads Transcript

Sarah Fan: My name is Sarah Fan, Amherst class of 1998 and I’m here to introduce Andrew Blum, class of 1999 to discuss his excellent new book, The Weather Machine: A Journey Inside the Forecast, published by Echo. In this book Andrew puts the awe back into something we have come to take for granted: the weather forecast. We learn how much had to be done, the complexity of the work required, and the effort involved to have as advanced a meteorological system as we do now, to keep on improving it, and why it matters, and what its impact was and is. And yet we notice it only when it doesn’t meet our expectations. Andrew, as a writer and reporter, you are very good at exposing the interests underlying the everyday, seemingly immaterial but essential things we do take for granted and a key part of your work is visiting the places that make those things possible, whether it’s a weather station on a Norwegian island, the European Meteorological Satellite Agency in Germany, or the Jet Propulsion Laboratory in California. Could we start off with the most basic question: why weather? How did you come to the subject?

Andrew Blum: Yeah, thanks for doing this, Sarah. I kind of backed into it. My interest has always been places, and I kind of came up writing about technology after writing about architecture and cities, and so when you write about places, the weather is always there but it’s always a kind of clause that is used to introduce the thing you’re actually writing about. But is also incredibly ineffable and hard to describe and I was always incredibly unsatisfied with the way people were describing the weather. And so there was definitely a bit of a challenge that like, if I were serious about writing about places that, at some point I would have to write about the weather. And that started out as a very broad notion, and it became much narrower when I realized, with some moment of insight, that I didn’t actually want to write about the weather, I wanted to write about the human mirror of the weather, the infrastructure that we use to observe the weather and to predict the weather. And [?] stretched across the world and is technically complex but we also kind of touch it every single day. And where that crystallized in particular was with the forecast for Hurricane Sandy in 2012, where all at once it became clear that we had developed this phenomenal new skill, this ability to predict a storm, in the case of Sandy, 8 days before its arrival in New York City in October 2012. And that seemed new. I mean, it seemed like news. It seemed like a good story of how that had happened.  And so it was a way in to this white whale of a topic for me of writing about the weather and places, but in a way that felt very familiar for thinking about technology and infrastructure and how they come together.

SF: And also, in the way that your previous book did, too; it seemed invisible. You were just talking about how the weather is the preface or the ground for all these experiences that you had in places, but not discussed in its particulars. So, I want to jump to the visit you actually do make to that Norwegian island, Utsira. It ends with you writing, “The essence of a weather station is to stand firmly in place to measure the atmosphere rushing by. That contrast between the static and dynamic is a big part of what intrigued me about the weather and the ongoing human project to watch and predict it.” Can you say more about that relationship between what is fixed and what is changing and how that drove your interest?

AB: Yeah it took me an embarrassingly long time to come up with some framework for the way in which we look at the weather prediction and then the day that we were looking at the prediction for eventually arrives. Now I know that seems like a completely banal statement of the way that time works, but time is confusing, and when you’re thinking about the weather, you’re kind of constantly crossing time and place. The weather far away becomes the weather close by, further forward in time and while it does evolve, the first notion is that the weather over there is coming here. And on Utsira I had gone looking for some more profound insight into what it means to be at a weather station. And I didn’t know how to engage with that by a single visit, you know? Not watching the weather over months or years or decades or centuries, in the case of Utsira. And it clicks, truly, when I tripped on a bolt on the ground and I realized that I was looking for this genius loci of Utsira as a weather station and I found it in an extra bolt that was in the rock which kind of defined this place. And I don’t know entirely what that means, but I knew that it meant that it was a lot of what I was after in trying to understand the power of a particular place in the context of this technical system. And I could see very clearly myself as a speck on the map, which I think is a key part of a geographic imagination which drives a lot of how I look at the world.

SF: It was a physical manifestation of it being rooted in one dot on that map.

AB: And it was nothing like sort of falling to the ground, to embarrass myself—well I was alone—but to embarrass myself. There’s a bit of humor in some of these quests, especially with the weather where we do observe the weather anywhere. But in fact, Utsira was a very satisfying place because it was so specifically about the weather that the other factors at places that also observe the weather, like LaGuardia Airport, fall away. You get the pure weather there.

SF: Right, because not much else is happening on this island besides this observation.

AB: Birdwatching and people coming for lunch, basically, yeah

SF: Yeah

AB: And ghosts. There were a lot of ghosts there and people kept on telling me about them. Sheep and ghosts.

SF: Sheep, ghosts, clouds, weather. We learn a lot about how the weather model is built, which is what gives us the forecast, in your book. And you write that while you had first thought of these increasingly sophisticated weather models that create the forecasts is these machines that sort of process real time data that is being gathered at places like Utsira, to output the predictions. Instead it’s more like the model and the Earth are running side by side and the better the model gets, the closer they align. It’s almost like the model is a pocket universe version of what’s going on on Earth. But it’s also possible that the model may have more real information than the real world does. You know, you used the example of a weather event that the model can project but that we can’t confirm independently and the example you used is snowfall on a remote mountaintop in Nepal. And you write “the snow is really there but the only record of it is simulated.” Can you explain how the models are working in this instance and more broadly?

AB: Yeah. The breakthrough for me in understanding how the forecast worked and how it works so well—which it does—was recognizing that it’s not like we’re taking the weather of the present and jamming it into a computer and the weather of the future comes out. It’s more like the weather models are an ongoing concern. They’re kind of an ongoing simulation of the atmosphere of the Earth that has the ability to be sped up faster than time to give us the weather of the future in a forecast. But then with each passing hour—or twice a day depending on the model—the predicted atmosphere, this future atmosphere that’s in the model is compared to the most recent observations in the real atmosphere. And then corrected slightly. So my metaphor of choice with this is that they’re dancing together. One is leading, one is following, and they’re correcting each other. And what was striking to me is that because the model works as a grid inside the computer, each piece of information, by necessity, is filled in at each altitude, each temperature, humidity. The model is a complete grid. It has to know exactly what’s going on with its three dimensional grid segments, or four dimensional grid segments as it moves forward in time. [?] layers of checkerboards across the earth. Because of that, you don’t have the same resolution in observations. So there are places where there are no observations, but of course they’re depicted at least to a point inside the models. And so we know that there’s the popular animation online of the winds of the earth. And it is the winds of the earth but not the winds of the earth as they’re observed but as they’re modeled. And of course the model is corrected by observations, but the model is more comprehensive than our observations. I find that a very mysterious—you know, a tree falls in the forest, is there anyone there to hear it? But when it comes to weather forecasts, we don’t need to hear because the model is good enough to know what’s actually happening in every single place.

SF: Right, there are no gaps in the model in the way that there are—

AB: there can’t be gaps in the model yeah. That’s just such a strange collision of this cohesive technological system with the vagaries of political geography and where you can put a censor and where you can’t and all that.

SF: Right, it’s an inherently equal system, obviously. That’s an obvious thing to point out about the climate and the weather but as you said, weather here becomes weather there.

AB: And the globalness of it never got old. The forecast works because of this notion of this revolution in the way that we look at the earth from above and in the way we can imagine that. And the weather modelers depend on it. You know, that’s the sort of key idea, this full, complete, global atmospheric model. But we’re always in place and we’re always kind of engaging with it based on when the weather is coming to us. But not entirely because we’re also used to that. Used to the maps of hurricanes and storms from the satellite’s eye view

SF: Right and as you pointed out when you began with hurricane Sandy, you see it 8 days out and you see it changing.

AB: And sometimes what you’re seeing with hurricane Sandy is an amazing example. 8 days out we could see the animations of the storm arriving in New York, but the storm didn’t exist yet. And I love that. I think that’s the crux of what this capability that we’ve built is about.

SF: That leads me to ask something that I think in some ways is the heart of your book. Because at a certain point is becomes almost existential. Like, the forecast is so good now that the question is not so much how much better is it going to get, even though it’s constantly being refined in the way that you describe, than what we do with the information that it produces. What is the forecast for? And you write, “If the weather forecast is nearly perfect, what can you do with it?” And it seems like we’re at, or we’ll soon be at, what matters is not so much what we know ahead of time but how we make decisions based on that knowledge. And those decisions tend to be based on probabilities. And one of the things I was curious about is that—I was thinking about the Daniel Kahneman book Thinking Fast and Slow because human beings are not very good at thinking of probabilities and in making decisions based on those probabilities. And I was wondering if that is a factor in why we both take the forecast for granted and why we misunderstand it. Like, why we assume it doesn’t work because we got caught in the rain without an umbrella.

AB: Yeah, I mean, the probabilities indicate degrees of uncertainty. And we deal with that as we will, depending on how much decisions matter to us—is it life threatening or inconvenient? Can we actually do something useful based on the decision or does the show have to go on despite the rain? We are definitely unwilling to accept that the forecast is “it might rain”. That’s annoying, that seems like a bad forecast. But it can be useful. It is a day today where it might rain in New York. I think it has rained a little bit.

SF: I do have an umbrella

AB: You do have an umbrella, yes. I did not bring an umbrella because I thought I was going to be home before it rained. There are few things more frustrating than writing a book about the weather and then going out and getting caught in the rain. But with meteorologists, there’s a sort of tension in their work at the moment because—the human meteorologists as distinct from the model makers, the atmospheric scientists who sort of build and improve the weather models—as the models get better, the forecasters ostensibly have less to do because the machines are doing more. But what’s also become clear is that as the forecasts become better, our decisions become harder. A decent snowstorm could be forecast two days ahead, if that. There wasn’t a whole lot you could do ahead of time, you know? Your preparation was condensed. But what we saw this incredibly, this last hurricane season, when you have a real six-day forecast for a major storm, you really need to have a sense of when you should act and what the levels of certainty are in order to act on that. Just as a side—the forecast for Dorian was remarkable because it appeared to be coming towards Miami but everybody knew it would turn. So you have the entire city of Miami just having a complete belief in the weather models that the storm will not actually keep coming towards them but will turn up the coast. And of course it did. And that was a really remarkable negative example where in a previous era the city might have been evacuated, there might have been major decisions made. But in this era, that decision didn’t have to be made. It was clear that the storm would go in a different direction

SF: And in the previous era, the time constraint would have been different, I assume. They would have known with fewer days in advance what was the landing of the hurricane on their shores and then they would have to make these very large decisions. And now they have more time out and know that it’s not actually coming to Miami.

AB: And actually there was an interesting article by researchers that was published this week that there was some concern that too much advance notice has a negative impact on outcomes. And that seems hard to believe. I’m a little suspicious of that. But it’s indicative of how much work there is to do to help people make decisions based on a forecast that’s different in kind, not just in degree.

SF: You do talk in your book about how, as the wall of meteorologists changes that they have a greater burden to do this kind of communication. And it’s also a world historical time, you might argue, where it’s more and more important to have people understand how these forces, that we just generally call climate or weather. And that seems interesting to me to think about the requirements of what you know from these models and how you extrapolate from them. How do you think that that shift is working now?

AB: With climate specifically?

SF: yes

AB: I mean, I think the first thing to point out is the difference between the weather models and the climate models. And the main reason why weather models have become so good is because we can verify them every single day. And not only that, but you can run the weather of the past through them again. You can make a change to the model and see if all the past weather is better predicted than just this week’s weather. And with climate models we only have one cycle of that. We have a kind of first generation of climate models that are now not only verifying to actually have been accurate from their predictions 20 and 30 years ago but of course they underestimated the impacts that we’re seeing with climate change. But that doesn’t change the decision making point of it. And—it’s interesting—I have minimal patience for the notion that we’re doing a tenth as much as we could, based on the minimal probability of what might happen. And it’s like, I decided not to bring an umbrella today—and you can actually hear the rain on the window right now—I decided not to bring an umbrella today knowing the full risks of that. But if there’s a 20% chance of an existential threat to humanity or even just to coastal cities, then certainly that should be enough to act. So that disconnect between the scientific validity and the broad accuracy and probability of the climate forecast and the political inability to do anything about it, doesn’t seem like a scientific problem, unfortunately. It doesn’t seem like it’s a certainty in the models.

SF: Right. [?] other problems

AB: It sees so many other problems. Although that being said, I was just reading that there is a new effort to make a new generation of climate models because it is helpful to have a higher resolution sense of what will happen. There are more specific moments—not just global action—but more specific moments where knowing what the sea level rise or what the temperature change might be could certainly be helpful for planning

SF: When we were talking about your book earlier this summer, you wrote me “I think the banality of the everyday forecast heightens the terror of the climate forecast” and I’ve been thinking about that a lot since it’s related to what we’re talking about now. I also was wondering, earlier in the book you talk about one of the ways you measure how good the weather machine is getting is when meteorology, the production of the best climatology, which is weather as it is, as it’s observed. But it also seems like even the weather model differs from the climate model in the way that the weather model has been refined more and more and more. It seems that meteorology will best climatology primarily because the climate is changing. It used to be that the way you predicted future weather was based on past weather and I’m wondering if that is also changing.

AB: We are very well-served in an era of weather extremes by the new weather models because they are based on the laws of physics. They are not based on the past weather and there are these wild examples—I mean, Hurricane Harvey in Houston a few years ago was the scariest of them where the weather models spit out rainfall totals that made sense to no one but came true. That was actually how much had rained. And so there was no past climatology that would have said “ok 30 inches of rain, sure, that matches what we know”. Instead past human experience was inadequate but the weather models were fully up to the task based on it being more deeply seated in the laws of physics not just in past weather statistics.

SF: Right and it seems to tie into people talking about hundred year floods and things like that.

AB: Yeah hundred year floods and Houston just had again

SF: Right, it doesn’t seem as useful a mark of what is happening now, that it is what you’re talking about, that the models have become sophisticated enough to take these inputs and create these outputs although they seem to be out of scale with our expectations.

AB: And as an aside with that: I said in the beginning that writing about weather and descriptions of weather always felt inadequate to me and I realized over time that a lot of it was just these divergent registers where you’re either telling storm stories and it’s about catastrophe in that way and it has some sort of drama, or it’s the ultimate banality. Chitchat in the barber’s chair. And weather is both. Small talk and catastrophe. And it’s very hard to reconcile those two things but that is the way that weather works. And as a mode of address to write about the weather, if you’re not writing about storms, if you’re writing about the catastrophic and the banal, it requires finding a voice that encompasses both.

SF: Which I think you did. But do you also feel like you can extrapolate one from the other? Yu go to many places in this book, and you meet many people and meteorologists and you describe—I’m always struck by this scene in Germany in the cafeteria with the two kinds of coffee—But everyone’s talking. And I assume they’re talking on one level that lay people would not be. But everyone’s experience, I think, similar to the way that you tripped on the bolt. Everyone’s experience is grounded in something that is very every day or close up. But when you were writing the book do you feel like you were trying to be able to get people to draw those lines more clearly between that moment and the larger [?]

AB: I found the hardest line to draw was between the math and the wind. And among American meteorologists there’s a culture of transcendental weather moment. People spend enough time with American meteorologists to recognize that they talk about their ecstatic moments with tornados and hurricanes and things like that and there’s still this tradition that you become a meteorologist because you were somehow touched by the weather in an intense way. And that’s a big part of the way that they tell stories about what they do. The weather modelers are very different. They come at this as a math problem. And so it took me a while to not be frustrated by that and to realize that if I was going to write about the weather modelers and the slow technical improvement of the forecast that they’ve provided, I was going to be indoors in the cafeteria, not out in the weather. I don’t go storm chasing in the book. It seems so different from the reality of the way in which the system had improved that had impressed me so much in the beginning.

SF: One of the things that impressed me a lot about the book and the history about how we got where we are today is the sheer amount of data and mathematical horsepower that was required to get anywhere close to what we have. It was going to be impossible to have anything like this and now we have the computing power to be able to create these models. So I think that the math of it is very unseen but it seems obviously very crucial

AB: Yeah and as with so much today the more sophisticated the math gets, the more accessible the forecast becomes. You know like “So what’s the weather” and that kind of thing. And I think as with other technological systems like Facebook or banking or whatever, we sort of live with that opacity to our peril. It’s tricky not to know how things work when things get more and more complicated. It’s tricky to know how they work and its problematic to not know how they work

SF: And yet we feel like we have more access than ever to weather. We look at it on our phone we’re constantly checking it, we have nerdy websites that you can visit. But all of the power behind all of this information is opaque to us, I would assume. So, speaking of hurricane Dorian, one of the things that you do early in the book that I found sort of helpful is that you sort of explain this pileup of acronyms that is responsible for forecasting the weather at least in the united states, but I think for most people that would be fairly illuminating but I guess more people now know that the national weather service is part of the department of commerce thanks to president trump. Do you want to talk more about how politics and the weather interact?

AB: yeah, I didn’t realize, well I couldn’t imagine the way that politics would evolve when I started this in 2014 basically, and the last month with sharpiegate was remarkable as well. I had interviewed Neil Jacobs, who’s the active head of NOA, which is part of the department of commerce, within which sits the national weather service and then separately from the national weather service, the national satellite service. And what was amazing is that at the extent to which he wanted to privatize the weather forecast as we know it. He describes a desire for a break in the 150-year tradition of the exchange of weather data between countries. And he said to me that their concern with these new types of data, these private satellites and private observations, would be not how they could improve the weather forecast but how these private companies that provide these types of observations could be profitable. And that was a really sort of stark Trumpian moment of taking seriously what he was saying rather than assuming he had just left out the other parts. And that was pretty startling. And all which is to say that when he first threw the weather service under the bus by issuing a statement agreeing with Trump’s forecast of Dorian that it would go towards Alabama, it surprised me even more that he would become an even more unlikely hero of the resistance when he redeemed himself by standing up with his forecasters to say that they had done a great job. All of which is to say that these structures have been political but they’ve been political in the sense that they are things the governments do. They are taken as a sort of public good and weather services around the world are run by governments and the modern history of the weather services that’s been the case. And so it’s been accepted that providing the weather observations for citizens is a kind of basic function of government, and how quickly that fundamental idea could be upset. Especially complicating things further is the money at stake at the moment in an era of more extreme weather and in an era of more technological capability the census that there’s better money to be made on a forecast. So while before it was only worth it for governments to spend this money, there now seems to be the sense that tech companies could have a private forecast that supersedes the public forecast where you then end up with a forecast for the haves and a forecast for the have-nots. And that’s pretty radical—I mean it’s a 150-year tradition of certain kinds of weather exchange and at the moment we’re faced with how that might fall apart in the next 5 or ten years.

SF: Right, the beauty of the system that you described is global and it’s cooperative and it has to be a global system and it won’t work any other way. So, to think about any of these challenges that it might be facing is pretty harrowing, especially if there’s a greater risk of extreme weather that will have an impact on the have-nots. When you met with various people who are in the middle of this work, was there anxiety about this?

AB: Most of my reporting was before trump, so the realization of it wasn’t quite clear. Now they’re certainly aware of it. And I did another round of reporting in the spring in an article for Time magazine about exactly this, where many of these globally minded meteorologists were very clear about their concerns of the existential threat to this system that they were very proud of. And then again its very much born of this post war United Nations notion of global cooperation led by American ideals after World War II which are now threatened and in decline and I think that yes, that was definitely there. It’s save a little bit by the technological progress that’s been made. So yes, the forecast keeps getting better and it will take a lot to grind it to a halt fully but at the same time it does require a global cooperation but has now been shown to be more delicate that we imagined.

SF: Especially if you factor in what you were just discussing, the threat of privatization. If you have government corporation but you have billionaire funded initiatives that are private and the information is withheld, that also seems damaging to our expectations about the forecast. We already talked about the challenges of communicating what the forecast is and how we should handle it and if the forecast itself becomes a form of private property, that’s yet another problem to face. And yet your book end optimistically, I would say.

AB: It ends in a way that is admiring of the global system as it exists. And, I mean, what’s striking to me is that diplomatic system is completely entwined with the way that we imagine the earth in the last 50 years with this blue marble view from space. I love that correlation between that vision of the earth and the actual scientific observation from satellites, and the weather models that do simulate the entire atmosphere. It’s hard to see it as anything other than what brings us together. I’ll leave the optimism there.

SF: Yeah I like the concluding scenes of weather diplomacy. It is kind of the United Nations of weather in a great way, in what you would hope. Well, thank you Andrew.

AB: Thank you Sarah, it’s great fun to do this and it’s great to be included in Amherst Reads. It’s something that I look forward to the whole time I was writing this book, so I’m glad it has come into fruition.

SF: Well congratulations on your book and on being selected for Amherst Reads.