Transcripts For CSPAN2 Andrew Blum The Weather Machine 20240714

And wilder with bigger storms, hotter and more freezing temperatures and greater extremes at seemingly all times of the year. And second, that forecasters are Getting Better at predicting whats about to hit us. They still dont get it right 100 of the time, of course, but their act rahs is city has definitely accuracy has definitely improved and so has the timelessness of their reports. In fact, these days were aware not only of one weather model, but of multiple ones simulating the future space of the atmosphere. Theres a north american model, the european model, the global one and a number of others, and theyre forecasting not todays weather or tomorrows, but weather over the next ten days and beyond. Andrew blum got interested in this stuff. He wasnt satisfied with the past explanations that attributed the advances in weather science simply to faster supercomputer ands better satellites. Andrews a journalist with a knack for making complex systems accessible as he did in his previous book which was a primer on the physical infrastructure of the internet. And he saw in Weather Forecasts a similar kind of story. His new book, the weather machine, is an engaging look at what goes on behind the scenes. The international, collaborative effort that it takes to give us the daily weather report. Andrew recounts how forecasting has evolved from scattered stations to the u. N. s World Meteorological organization, and he introduces readers to the scientists, satellites and supercomputers that make up the worlds current forecasting system. Andrew will be in conversation here this evening with one forecasters, jason. Hes the weather editor at the Washington Post and chief meteorologist of the capital weather game. Jason himself has been fascinated by the weather since his days growing up in the d. C. Region. As a high school student, he interned with bob ryan who at least some of you, no doubt, will recall was for years the chief meteorologist for the local nbc affiliate. Jason went on to study Atmospheric Science in college and graduate school, then worked as a Climate Change analyst at the epa. And 15 years ago established capital weather. Com, the first professional weather blog on the internet. Ing the Washington Post absorbed the blog in 2008, and jason and his gang are now essential reading for anyone whos interested in the weather in the d. C. Area. So, ladies and gentlemen, please join me in welcoming both andrew and jason. [applause] hello. Thank you all for coming. Brad, thank you for the great introduction. I was here not here, on connecticut avenue, and really wanted to come back with a new book. I was a little nervous on the train this morning when washington was washing away [laughter] as jason definitely had a busy day. I was also particularly pleased that jason agreed to be my interlocutor tonight. We met when i first started this project several years ago. I went to meeting of the American Meteorological Society in atlanta, and sort of knew that i wanted to write about this topic and was trying to figure out what the shape might be. And having all of americas or a lot of the world meteorologists in one place was definitely a great starting point to figure out how it fit together. What i didnt realize was that there was a weather journalism community, and the second that i was there was sort of formally embraced and invited out to dinner with about, how many . About 25 people. Yeah. Sort of the American Press corps of which jason [laughter] i think a fair count. But it was a big help for me starting out, so its great to come full circle to have jason here tonight to do that. So were going to talk for about a half an hour, then open up for questions, and ill pass it off to you. Sounds great. Well, its a pleasure to be here tonight, and a funny story is that the first time i ever took an uber was with andrew [laughter] from the American Meteorological Society annual meeting in atlanta to airport back in 2014. That was the i wasnt going to say the year, but, okay. [laughter] yeah. So five years ago. Anyways, i, as preparation for this event, i did have to read the book, but i have to say it was a joy. I frequently found myself if smiling, like i was sometimes youre watching a great p concert, youre watching a great singer and, you know, a smile just comes across your face because youre moved by the, how powerful the performance is. And the writing with is really elegant, lots of excellent narratives and anecdotes. So it definitely earns my high recommendation. Anyway, so lets get started with the q a. Lets begin with the motivation for the book and what made you sit down and decide all of a sudden that you wanted to write a book about the history of Weather Forecasting. Yeah. There are a few different threads, which we have time to go into tonight. The longer answer. For my my background is mostly writing about architecture, writing about places. And you cant write about maces without thinking about places without thinking about the weather. But weather, it turns out, is very hard to write about especially if youre not a meteorologist or atmospheric scientist. The words kind of slip away, and theres this constant challenge between the banal, you know, did you forget your umbrella are, you know, am i blessed proper dressed properly for the day and the catastrophic. And i think its that combination to be a meteorologist and to write about the weather and to be dealing with the weather professionally in any daily basis, being able to switch registers from, you know, from the cheery to catastrophic, i think, is one of the Biggest Challenges and sort of had stuck me for a while. Kind of what unstuck me, knowing i always kind of wanted to write something about the weather was Hurricane Sandy in 2012 which was the moment where i think for a lot of us the weather models revealed themselves. You know, there was suddenly a sort of public conversation about the American Weather model can and the european weather model. Not only that, there was this astonishing forecast, an eightday forecast. Sunday afternoon was the first inkling for sandy, which arrived in new york city the following monday night. And that exceeded all my understanding of what meteorologists were capable of in terms of their human ability, in terms of their own sort of pattern recognition and more manual assimilation of data. Clearly, they had built systems that were astonishing and had far exceeded my understanding of how they worked, and yet to see the sort of, this question of, okay, weather models and how do they work was very hard to answer. It was an incredible black box, and that type of, that type of complex infrastructure story was really what got me going to sort of recognize that if i could figure out how these models worked, if i could understand them in a way to tell a story about them, that felt meaningful. That felt like definitely there was a story to be told there. I might have underestimated their complexity. Does anyone here work in weather models . No . No professional weather modelists . Okay, phew. [laughter] i think i might have underestimated their complexity because theres a very small pool of people in the world who really work on them and improve them. And they, the sort of barrier to entry to recognize what their parts and pieces were was significant. But it even, it sort of checked the box that was important to me of, frankly, its banal the city, this system that bay banality, the idea of the internet, this incredibly complex global system that we carry around in our pocket and yet is really a a black box was definitely a story i wanted to tell. And, of course, you cant tell it without going back in time to some of the history and how it developed. That was really the way that i approached that. Yeah. So lets talk about the history a little bit and some of those pioneering scientists in the 1800s who sort of laid the foundation for modern day numerical weather prediction. There were scientists in england and in norway who you talk about in your book. Tell us a bit about some of those early contributors and how important the work they did was to the advancements weve seen in weather prediction today. Its funny, theres kind of two histories of weather. Theres the history of the sort of Storm Chasers and the watching the sky and, you know, the sort of this kind of heroic stories of wind speeds and, you know, weather maps and all that. And then theres this other history of mathematicians. And its when i started to try to think about how to tell this other history, it really felt quite distinct from a lot of the other history. It very much begins with an original meteorologist who was kind of amazing because he occupied both histories. He came up with the idea at the turn of the 19th century, turn of the 20th century, i should say 1904 was his main paper to calculate the weather, to use the equations of physics and thermodynamics to actually describe what the atmosphere would do next. With the main insight that each days calculation could be itself a kind of hypothesis. And if you could calculate the weather, you could prove the next day whether your calculations were right or wrong and then rethose calculations on a daily basis, which is this process of improvement that has gotten us to this incredible place were at today. I feel like i should sort of lay out that incredible place. Bradley sort of framed it, but to say that the meteorologist talks about a day a decade, which is to say that the models, the forecasts have improved by a day each decade or, basically, over the last 40 years. A fiveday forecast as good as a fourday forecast ten years ago. Not that that rate is diminishing, its increasing. The talk at the moment is of a two week forecast by 2025 for extreme events. So this improvement really goes back to idea of saying, okay, if you can calculate the weather, then each days forecast is its own sort of science experiment, and you can try it again the next day. But then he couldnt calculate the weather, of course. There was a famous mathematician who tried to calculate the weather, realized he would need 64,000 computers, which is 64,000 people who he thought could be arranged in a a stadium, and they could each get, you know, their square of corresponding atmosphere, and that square they would do the calculations and sort of pass them up to front, and maybe that would work to actually calculate the weather fast enough for a useful prediction. Because, of course, you cant just calculate the weather, it has to be done before the weather comes, otherwise its not useful. So i think that inability to calculate, the lack of observations, the lack of a computer to calculate them was, its actually meant that the ideas of 1904 had to wait 60 years until computers and then satellites began to come in. And it had to wait nearly another 50 years, just to really the last 30 years, for the weather models to be useful beyond the human scale. Not just guidance, but really to exceed the capability of human meteorologists to point that now, i mean, i you wouldnt find yes, theres a sort of human check on it, but its hard to find meteorologists who will still say they dont work. Although i found one. [laughter] its remarkable how good the Weather Forecasting models are. They do trillions of calculations per second. Its amazing. One of the critical developments in the American Weather prediction, we fast forward to 1950s and 60s, and you alluded to this already, was weather9 satellites. Yeah. Talk a little bit about the history of Weather Satellites and how theyve played such an integral role in the development of forecasts and some of the sort of major discoveries and breakthroughs with that technology. I think that the key, the idea that i fell in love with really early was that for longer range forecasts, for, you know, two and three and four i think beyond a couple days you need a global view. It cant just be youre not just looking at the weather in north america, youre really talking about the weather globally, to look at the entire global fire. Atmosphere. You need a global instrument. Its it also amazes me that the satellites, i mean, the sort of month of apollo, the week of apollo, but they come out of the same conjoined sort of civilian scientific and military efforts that you have no satellites without the dollars spent for the missell race missile race, you have no Weather Satellites without the dollar spent for surveillance satellites. And the sort of two the scientific ideas and the military ideas are kind of hand in hand. Until, quite notably, until kennedy. Theres the moon shot speech, the spring 1961 speech, you know, that we put a man on the moon before the decade is out, that was bullet point number one. Bullet point number two was Nuclear Rockets for deep space travel. Point number three was communication satellites, and point number four was Weather Satellites. And its just amazing to me that you have the sort of basic impetus for a global view in the same speech as the moon shot speech and the sort of most famous infrastructure speech. And then you have it baylesed, rooted in the based, rooted in the idea of international cooperation. Kennedy liked the weather because it was a point of cooperation. And sure enough, as soon as that speech and a later speech in the fall at the u. N. , the sort of Nuclear Annihilation speech, the sort of answer that kennedy proposed for togetherness, for an they were for peace was an alternative for peace was cooperation on weather observation and control. The control part got left behind, although not to some people. But the, but that idea that from the very beginning it was about diplomatic collaboration between meteorologists from every country in order to make global data out of Global Infrastructure to eventually support global models kind of gives it its roots of cooperation that begins the 19th century. But its current incarnation was very much inspired by sort of kennedys idea of a global vision. And thats pretty, that continues today. Somewhat quietly, and i think we dont, we often dont look beyond the sort of American Weather service, but everything depends on this global pool of data. Absolutely. And so, i mean, clearly the data is so critical to todays weather models, whether youre talking about the weather balloon data, youre talking about groundbased sensors and, coffining, you just talked and, of course, you just talked about Weather Satellites. Weather satellites are super expensive, cost governments billions of dollars. But without them, our forecasts would not be where they are. So, obviously, the Weather Satellite data, the observations from groundbased sensors, weather balloons, they feed these models. And you went to two Different Centers to learn more about weather models. You went to National Center for Atmospheric Research in boulder, and you went to European Center for mediumrange and i did go to National Weather service, but i didnt put it in the book. [laughter] okay. There you go. Talk to us about your experiences visiting these centers and the appreciation you gained for these models. Touched on this a little bit, but lets hear more in depth about what makes these models work, what amount of intellect is required to run these models, who are the people who are doing this work . And how have they been able to be so successful in improving Weather Forecasts over time . Yeah. The i mean, i think this competition between the american model and the european model is sort of shorthand for it, but it was amazing in visiting both places, i mean, as a journalist the kind of cleaner, i mean, looking for cleaner stories, for more legible stories, and its amazing to me always that cleaner stories come from cleaner places. You know, the places that sort of have a coherent organization and that have a very focused Mission Statement are much easier to describe and to write about and are often, i find, more successful because of that. So with weather models, the fact that we sort of take the starting point that, okay, theres this weather model war and the european model is the best, to see to go visit the source of that, the European Center for mediumrange Weather Forecasts which is a collaboration funded by, i think its 32 yeah, 32 European Countries that are contributing, sending both scientists and money with the singular goal of running the best Global Weather model. And they do it by combining the Research People and the weather, the model operations people in a single building and in this amazing cafeteria calf tiers have become a bit of a chi they in tech cliche in tech culture. I couldnt believe that anyone got any work done, because the cafeteria was always full, and they had the most beautiful coffee machine. [laughter] but they, that constant, day by day effort to consider what the model was doing, what it was spitting out and how it could be improved as a system was really tangible. Where its built in, you know, not just in daily sort of internal wiki of comments on this is weird, why is it doing this, then codified with weekly meetings and then quarterly meetings, one of which i observed and wrote about. The sort of, okay, what is the model doing and how can it be doing it better. But i think the key point is when you talk about the weather mold, its not about, you know, a sort of exiewmp computational meat grinder, but its really about having an ongoing similar ration of the atmosphere that every six hours, twelve hours is compared to the real atmosphere and then corrected slightly to better match that. And thats sort of, you know, that sort of duet of the similar simulated atmosphere and the real earth kind clicking together able to run faster in time to give us what we call forecasts. It makes you realize that to take the forecast better means to make the simulation better. Not just in sort of some certainly not the statistical improvement based on past weather. Its more about, actually, using these equations to simulate the atmosphere in a way that is close enough to reality that it actually can then when it is run forward faster than time, it spits out the weather of the future. And to see how eager the people at the European Center were to make sure they were getting it right which isnt just to say they were getting the forecast right, but to say that the simulation they were building most closely resembled the real atmosphere for all the right reasons, you know . That the actual physics matched. Not that it was sort of just getting lucky repeatedly. Thats not Machine Learning, its not a kind of its not a predictive problem in the way we think of, you know, elections or, you know, other sort of baseball or other kind of predictive problems. Its unique, uniquely successful too. This is something we all use every single day and is mostly right. Does that answer your culture question . Yeah, for sure. I was going to ask you, did you find the scientists there, did you find that they were competitive, that they wanted to beat the other modeling all they talked about. Okay. When you say why are you doing this, they say we want to be the best. And that theres a lot of, when you, theres an ongoing discussion about the sort of modeling competition. And i think we were just talking before, there is a sense that it keeps everyone, it lights a fire and keeps everybody improving their model faster, but there is the sense that its, that they need, you know, they do it the u. S. Models are sort of convoluted by having to serve different masters in their different models. We run 12 models and they only run 1, and i think while that is true, the singular focus on this basically 10day simulation of the atmosphere makes their, sort of their collaboration and their competitiveness very productive. Right. And much more of a streamlined process there. Okay. So, you know, obviously weve talked a lot about how great Weather Forecasts are, but theres a saying your Weather Forecast can be 100 accurate, but if people make the wrong decision based on an accurate decision, the forecast is useless. That was a big when i began to understand that yeah. What makes a good forecast. Yeah. So its how does the forecast inform decision making, and does the forecaster provide the end user the information they need so that, you know, if theres a wedding reception coming up, that theyre going to, for example, have a tent or not have a tent to keep their guests dry. And you talk about that sort of Decision Support aspect in your book. Do you think Weather Forecasting centers, both in europe and the united states, are starting to gain a better appreciation, the importance of the social science . Connecting the physical science, that is the forecast, with the decision . I think, actually, its the place where the American Weather not just National Weather service, but sort of broader enterprises the weather folks call it is really excelling. The new link between that sort of forecast and the technical sense and the decisions you make based on it. And i think that, i mean, i write about a guy tame named tim palmer at oxford, the European Center. It tells a story about how sure are you that its going to rain and what decisions can you base on it. If youre have a party, a picnic and the queen is coming, you know, 20 chance is enough to order a tent. But if its just your friends, an 80 chance might not be enough to order a tent. Its that sense of where, you know, where how, whats the use of a better forecast if it doesnt help our decisions. And i think thats reasonable with the daytoday forecast just because it was clear it was going to rain monday morning in d. C. , four days ago. It wasnt clear how it was going to rain. We were talking before, you know, at what point this morning it seemed like things were getting intense, and even more significantly, what decisions would people have made differently had this days forecast been clearer. And i think that idea that you can strive technically for a perfect forecast, but if it doesnt help you make decisions, i think its kind of where i mean, meteorologists love it because it acknowledges that theyre right. [laughter] the balls in some ways out of your court. But its only out of your court if you recognize that technical achievement of the forecast itself. Right. But you still get angry emails. Yeah. , no for sure. And i think, i think Weather Forecasts have come a long way, and were really good at capturing what i, what we call synoptic scale weather systems. These are large scale systems, fronts, low pressure systems. But when it comes down to forecasting individual thunderstorms, which is what we had this morning, models are still just developing in that area. We have what we call conductionallowing models which try to get thunderstorms, small scale weather futures, correct. And that you can only do sometimes within a couple hours. I mean, even at five a. M. , six a. M. This morning it wasnt clear that that complex of thunderstorms was really going to light up and then track straight through the d. C. Area and unleash historic flooding event. Theres still a lot of progress we need to make, i mean, even as our forecasts have gotten a day better each decade. You spoke about it, its really localized, finescale forecasts where theres a lot of progress i think we can still make, which i think is a good segway to my next question. The importance of observations. In all models, they require observations and weve talked about all the different types of observations which are needed for a robust American Weather prediction system. But there are some concerns right now in the larger sort of Public Private interNational Weather community about where well be collecting data in the future. There are a lot more there are new data sources coming online, there are private companies which are launching Weather Satellites which want to sell their data rather than making them free. You have, you have Remote Sensing systems on automobiles, on different modes of transportation. You have cell phones in your pocket which are collecting data. Who owns that data, how is that data then sold to the government. So talk a little bit about this issue, and its coming up in the World Meteorological association as sort of a thorny issue as to, you know, historically whether data has been an International Partnership among governments. But now youre introducing the private sector which has a profit motive into this enterprise. How do you deal with that . And talk a little bit about how your book address that is. Yeah, yeah. It was something that i realized the folks who were involved in the daytoday of forecasting in the in particular, just because the system, the International System works so well, it operated very quietly. Actually, that was what somebody, a woman who was former chief scientist of the Australian Weather Service sort of said, you know, we havent been a squeaky wheel ever. You know, our collaboration of exchanging weather observations from all over the world, pooling them together and sending them back out over the world has worked so well that we kind of float under the radar, and peeps perhaps dont people perhaps dont appreciate how essential that is. And i think thats the case, the fact that it isnt recognized, i think, becomes a cause for concern when you have these new kinds ofwet observations. Of weather observations. And particularly observations that are no longer, after 150 years of meteorology being done mostly by governments, you have a sort of shift towards private companies collecting observations. And the tradition always has been that the government meetology offices and the question then is if they sell it to one government, can that government then share it with every other government which is a challenging Business Model if youre running a Satellite Company because, obviously, you want to sell your data to as many people as possible. The flip side though is if you cant sell that data, you cant share that day the that with other data with other governments, then you have a risk of a fracturing of the entire system, and you end up with fewer observations or, at best, less improvement because you have all these new observation streams that are no longer part of this global pool of data and are no longer incorporated into the global models as part of a Global Infrastructure offering global forecasts. And thats, thats a challenging place to be. Its a real shift marley with the big particularly with the bigger bucks involved, with cheaper sat clients, with more money at risk or more money to be made with weather extremes. Before it was sort of unheard of that a private company would spend 20 million on its own weather supercomputer model. Now those dollar values start to seem more reasonable because of this combination of greater climate risk and more capability. So the better forecast itself, i think, creates a business opportunity, and that, you know, there have been private forecasters for a while. The weather channel, accuweather, lots of Smaller Companies have been around for, i guess, 20 or 30 years, call it. But the shift would be if its no longer about a sort of global pool of data that Companies Like accuweather taking that for better forecasts, but instead a fragmentation of new kinds of data thats not been shared. You, i mean, its uncharted territory, and i think its something that a lot of the weather diplomats, the folks who are the representatives at the World Meteorological organization are thibt. And its something that the Current Administration has been very eager to push for to make sure, as neil jacobs put it to many me in an interview recently, to make sure that these new companies are able to be profitable. That is the emphasis. Its not a unique certain us im sorry, its not a unique, its not a unique desire, but its a unique emphasis. Its a new, the sort of balance shifted to saying we want to make sure that these new companies can blossom, and its not clear if, what the stakes are for that for the global data exchange, this 150year tradition of data for the public good. So its coming, the sort of combination of factors. I think the macro factors, i think, are really clear. And its very under the radar which i think is a bad weather observation [laughter] right. No, i mean, its interesting because there could be a day where ibm, for example, has better Weather Forecasts than the National Weather service, and they could might be today. Yeah. I mean, theyre developing their own model. Theyve got watson, and theyve got Machine Learning and scientists there who can create their own, who can do their own supercomputing. And so, and then they can sell that information. So Weather Forecasting, Weather Forecasts which have historically been a public good now all of a sudden if you want the best information, rather than automatically getting it for free, you maybe have to pay for it. And theres the potential that the publiclyprovided Weather Forecasts could become inferior unless the Weather Service develops partnerships with these private companies in a way which is and they can broker agreements. So yeah, i mean, and if you did, i mean, the reality is that if it worked for the, i mean, predominantly when you think about the satellite data, its really the European Consortium that runs the european Weather Satellites and the american satellites that provide the lions share of data. A lot of the numerical data that comes from the polaror bitting satellites, when you think how big that contribution is oh, yeah. It sort of lifts all boats for global meet youll. And if, i can see the sort of factors coming together to threaten it, and its strange to think about because its a 150yearold tradition. Yeah. Because if a private Sector Organization had the best forecasts and they knew a lifethreatening storm was going to hit before the National Weather service, would they be ethically obligated to provide that forecast to everyone free of charge . The Weather Service, who has a mission of protecting life and property know about it and giving it to them so they could then disseminate it. So it does raise questions like that. It was so striking to me in talking to people from around the world who are most concerned about these changes how deeply ingrained their sense of meteorology as a sense of public good is, and to see the sort of shift from public good to commodity, i think, is really startling. Right. So should we take some add audience questions . Yeah. Right. So if you have a question, just raise your hand, and ill pass the mic to you. Thank you. Fascinating talk. Is it similar in some regards to why we should care about having longer and more accurate forecasts . And i was wondering if you could do that with regard to disaster preparation. Hurricane is going to come two weeks in advance as opposed to one week in advance . I think it, i think you see, i think you see these moments where the longer its not just that a storm wasnt forecast four days before or three days before in time to make a decision based on it, but that because the forecast was so consistent for so long, the confidence in making decisions from that forecast is greater. I mean, the amazing recent disaster is from cyclone fannie in india in early may, a very similar storm to 20 years ago where tens of thousands of people were killed. Not time around not just the longerrange forecast, because i think the forecast then was several days out, but the confidence to be able to evacuate a Million People was really clear. But i think for, you know, the idea you have a high impact event with two weeks warning, you know, if that forecast is reliable and you have a weeks warning, you know, the counterexample this spring for fannie was in mozambique, righting . Yes. That was a storm that had catastrophic damage, and while it was predicted, it wasnt predicted with enough time or confidence for the local government to act on it. Yeah. I also think the dissemination system and the way of getting the word to the public and the Warning Systems in mozambique are not as sophisticated as they are in india because india has a lot of experience with tropical cyclones having dealt with them repeatedly over the years, and theyve learned from experience. Theyve had, you know, theyve seen absolute devastation, hundreds of thousands of people die, and so theyve put in place Warning Systems and evacuation plans to get people away from the coast. Its interesting though, i was want to make i just want to make one point. For hurricanes, obviously, having a lot of lead time, its the incredibly valued, and our threat forecasts have gotten so much better compared to where they were in the 1970s and 1980s. We can really pinpoint the track of a hurricane within 2448 hours with a good deal of specificity now especially compared to where we were. But whats interesting, with tornadoes theres been research which has shown with tornadoes we give people too much lead time, they actually can make worse decisions. So maybe the sweet spot for tornadoes might be 1525 minutes, so people know, okay, ive got to do something right now, because if you give them two hour, they can make bad decisions. Maybe they decide they want to wait until the last minute and drive out of the way. Theres some interesting social Science Research going into, again, its this bridging the gap between the forecast and peoples decisions. Yeah. I mean, and that the example this spring with the tornadoes in ohio and in missouri, kansas where it was, y major tornado in populated areas, and the forecast really worked. It was sort of a real victory, i think, for the National Weather Service Going back to joplin tornado in 2011, i think . Yeah. , you know, 160, i think, were killed, 200 people were killed. And, again, these are these moments where its not just about the duration of the forecast, but the fact that with duration comes confidence. Back there. Thank you. So you kind of touched on this a bit about serving different masters, but why is it that the european model is different than the american model . Maybe what is better physics is physics, no . How does one become better or what is different . Yeah, yeah. No, good aantic. Its an interesting way of i think that the basic, i mean, the crux of the answer is in the limitations of the model, which is to say that the resolution, the model is not a molecule to molecule simulation of the atmosphere, of course, so there are lots of approximations made. And i think a lot of the differences in those approximations for, you know, with a given set of weather happening, but i think it really, you know, the way that what i was surprised by is the importance of this lighting up of observed weather and simulated weather, this process that the modelers call data simulation and how challenging it is to take the data that youve collected and actually find the right time and sort of time and model scale and place and model scale to inis cert that data and make it line up. And and a lot of what, i think a lot of the advantage of the european model comes from what they, their data simulation scheme. They call the woman whos now the director, it was sort of her graduate research that started their particular way of lining up and correcting the weather inside the model with the observed weather from outside. And i think that sort of ability to constantly check both the way the model is working and to be better at slightly, at putting the model back on course with each step forward in time, i think, is one thing that they point to. Physics is physics, but the ability to actually observe the atmosphere is imprecise. Yeah, i would just adjust one point on the european model versus the american model. The differences between the models arent that great. I mean, for years the european model has been the most accurate in the world. Theres no denying that, and i tell people if i only had time to look at one model, only had access to one mold, i would pick the european model because, on balance, its the best. But its not the best by a lot, and its not the best in every situation. And so one of the jobs of a forecaster and one of the things they have to become skillful at is understanding the biases of the model and when to use what model in what situation. There are times when the american model has better forecasts than the european model. There have been hurricanes where the american models have better track forecasts. Sometimes for winter storms the american model has a better forecast for snow in the mid atlantic. So you just have to, again, you have as a forecaster, you want to look at the universe of models collectively. Theres a canadian, theres a u. K. Model, there are models in japan other interNational Centers. So you look at them all. We call this an ensemble approach. You understand, you try to look at where do they have things in common. What are they all forecasting alike, where are the differences, and then you make a forecast and you try and communicate the uncertainty based on where the models disagree. One of the ways i think about how powerful theyve become is that the models are higher resolution than reality, which ises to say that we have sort of we can say what the, you know, what the given conditions are in far more places from the model than we actually observe them. So we, you know, the models are where all the data we just have the actual observedwet is just a weather is just a very small portion of the amount of detail that the models are presenting. Its a strange thing that the model is more, you know, has more higher resolution than reality, reality as observed. Thanks. Thanks for writing the book and running a great blog. Im curious if theres any bottlenecks in the pipeline to make improvements. From my academic field, theres tons of researchers, not a lot of funding. Can you see a lot of improvement coming from just money, or do you need more graduate students doing research . So where can most improvement be made . Well, one of the things thats striking to me is this shift thats happened in meteorology, that its really become about Software Design. So i think, you know, the task of improving the models is a somewhat different skill set than the traditional meteorological training. And so i think when you is so sort of all the action is in the improvement in the system. But im not, its not clear to me, and id be curious your thoughts on this, if american meteorology programs have sort of gotten a sense of that. And, of course, as with so many other things, because the same, because the skill set of the Software Design and data science is the kind of broader term, because theres so much pull on that for the private sector, for silicon valley, i think that also is a bit of a drag as well. Does that match . Yeah, i thinks right. Only, i was talking obviously, i was talking to neil jacobs, the acting noaa administrator, and he emphasized the need for a strong class of Software Engineers and computer scientists as they continue to try to improve the american modeling system. Cheerily, you do need physical scientists. I think the every part of the modeling system which relies observations and data simulation, to keep improving that. Youve got to keep adding to the system and making sure is the observations are high quality and so forth. And the model physics, which is what the physical scientists work on, you need them involved in that, testing the model, finding where the models not working correctly, Publishing Research as to, on how to better represent certain processes in the atmosphere. Moving that research into operations which is, actually, sort of this whole research to operations question, the European Center has been taking its lunch. Yeah. Its far superior to american approach, and thats one of the reasons people who have studied this issue have concluded the european model is superior, because they have better collaboration, Better Research to operations processes in place, and theyve spent a lot more time and effort on the research. So, but i think neil jacobs, whos running noaa right now, he recognizes that, and hes trying to beef up the research to operations within the Weather Service, within their molding center so that Modeling Center so that they can continue to make progress there. Yeah. I would be a little more critical of that. I think that while there is talk of improving with the new center, it is a step in the right direction, kind of the institutional drag is e enormous. The entire structure of the american modeling system, you know, is, it has gotten sort of convoluted over years. And when you look at the clarity not only of the European Center, but even of the u. K. Met office, the Weather Service, you kind of recognize what we dont have at the moment from a bureaucratic standpoint. So thats not the kind of thing that funding for a new supercomputer solves. And while i think its the sort of first step for this new project if called epic is a step, but its a very small step. I can speak more freely about that. I wasnt so impressed. Epic, what one of the motivas with this no modeling initiative in the u. S. Is to move the american modeling system into the cloud, actually, and allow researchers at universities to run the model in parallel and test things out and then send improvements and identify errors and send those back to Software Engineers within this new Modeling Center so that they can, over time in cycles, improve the model. So thats what theyre trying to do. Time will tell when its successful. But theyre working on it. Its actually one of the key priorities at noaa under the trump administration, is to try to get the american model at least similar to european model. So thats the goal. Whether they can accomplish it in a year and a half or five years and a half, well see. Yeah, im not [laughter] hi. Question about, like, kind of just dissemination of information, like forecasts in general. Getting, you know, the blip on the phone this morning regarding the flash flooding and basically do you see it moving just towards that model completely . Are things i googled the 361212 and found an article you wrote years ago about how they were redoing that. Even local forecasts on just your channel 7 news or whatever, is that kind of falling by the wayside . Are we moving towards something completely different, or does that stuff hold on . Well, yeah. Andrew has a whole chapter in his book called the app, and you talk about the technologies that consumers use for Weather Forecasts now. Yeah, i think the way information is received is changing, when youre talking about weather, sports, news. And i think increasingly people overwhelmingly are receivingwet information on their mobile Weather Information on their mobile devices, whether a Smartphone App or on the mobile web, android, iphone, whatever. So, yeah, i think, absolutely. I i think, you know, thats one of the reasons weve been successful. You know, our weather team, because weve been Digital First since we were, since we started in terms of providing an interactive platform for providing Weather Information where people cannot only receive Weather Information, but also send stuff back at us, photos, ask questions, have access to experts. And then with the app, yeah, its turnkey, but you Program Certain cities youre interested in getting the forecast for, and in a second its right there for you when you wake up in the morning and when the weather is highly impactful, you get the arelaters. I think alerts. I think one of the things that weather Smartphone Apps dont do, which i think theres a need for and will continue to be a need for, is sort of the human interpretation. You have an icon with a cloud, that doesnt tell you that much especially when the stakes in the forecast are high, right . You want to understand what the range of possibilities are, you want to understand how this is going to impact me. And so if you want that more sort of detailed level Weather Information, you do need people to in whatever platform you prefer whether its video, text, graphics you need that information presented to you that way. And some of the automated ways of obtaining that informationen dont do that for you. So i still think theres a need for human in the decision process. Yeah. I wish we had a capital weather gang in new york. Its truly the moments where theres a lag between what the forecast and particularly the models are technically able to present, you know, stop raining at this time, and the way in which thats kind of the highs and lows of that are sort of smooshed out in the apps, you know, very deliberate kind of hedging, i think thats, i think thats quite clear. Yeah. The other thing thats clear, but in a frustrating way, is that sometimes the forecast is very confident, sometimes its not yeah. And sometimes the forecast is it might rain. That could be a really good forecast, but its not something we want to hear. And i think that lag between what the system is technically spitting out and how much were willing to trust is very much in process. You make up for that lag by presenting what the range of possibilities are. Thanks for that, yeah. Andrew, leapt me ask you something let me ask you something. Please. How do you look at a forecast differently now yeah. Since youve done the book than the way you had before . Its interesting, quite differently. But im not trained as a meteorologist, and i recognize what the limits are of the way i look at it. I think one of the things ive learned is by knowing the rhythm of the models, i can, i know i can sort of see what the trends are even in the apps. So Weather Underground, for example, has the kind of most granular detail in what its presenting, and you can if you know that its essentially being updated every 12 hours, you begin to recognize what, you know, what changes are happening in the forecast. I just, as a game, i did an experiment of looking at the eightday forecast for the publication day of my book, each day presenting it, and it was amazing to see how little the forecast as manifested in the Weather Underground app, how little it changed over the eight days. It was a bit of a cheat because it happened to be that it was 50 of rain basically for eight days, but sure rough, the day came, and it rained and it didnt rain. It was a bit of both. And so, for me, its both trusting the its trusting the percentages more, you know, recognizing that 20 percent is is 20 is 20 and 60 is 60 and not just looking at when the emoji flips over from sun to cloud and also recognizing that the day is long, and the models have a lot more resolution in time. So, i mean, today was a great example, you know . If the emoji for the day in d. C. Might have shown showers, but it was pretty clear the last several days that by, you know, 2 00 it was going to be brighter. So sort of trusting it more is, actually, really whats changed most for me. Weve got time for one more question if theres anybody with one. I just want to make a point for a second here which is, i think one sort of overriding theme which andrews book makes Crystal Clear and will really give you a better appreciation for Weather Forecasting is the public face of Weather Forecasting, whether its someone on tv, someone on the radio or a blog you read, you know, were just interpreting information. And were sort of, were just translators. The work which was done to get weather prediction to where it is today was performed by some of the most brilliant mathematicians and physicists in the world. And its incredibly complex, and these people, you know, it took decades of just incredible hard work and, again, required some of the brightest minds. Again to public, i dont think they get and i think andrews book makes this really clear you know, how sophisticated, how complicated and how rigorous meteorological science is. It requires some of the brightest minds in the world to get the weather models we have these days. You know, the people, the faces you see on tv, theyre not the people who built the foundation for this. So it gives you a really great appreciation for the science of meteorology and how weve gotten to where we are today. And i would just say, add to that, that its amazed me that the goal is, the goal is to serve people. The goal is to help people stay out of harms way and the straightforward righteousness with which the meteorologists approach that was startling. I was nervous when i was saying weve got our app, we dont need meteorologists, which is not the case at all. It was amazing to see, of course, the m is made by people and requires human interpretation particularly as to whether it becomes as weather becomes more dramatic. Well, while the weather is a complicated subject, andrew, as you did with your book on the internet, youve a made it very accessible, very readable, and i would encourage everybody to get up and buy a copy of andrews book. Or two. Or two. [laughter] thank you both, andrew and jason. Thank you. Thank you, jason. [applause] copies of the weather machine are available at the checkout desk, and andrew will be at that desk signing copies. Thank you again for coming. [inaudible conversations] here are some of the current best selling nonfiction books according to the l. A. Times. Topping the list is tara rah westovers account of growing up in the idaho mountains and her introduction to formal education at age 17 in her book educated. Its been on bestseller lists for over a year. Next in the pioneers, historian David Mccullough recounts the journeys of early settlers of the northwest territory. Then its former first Lady Michelle obamas memoir becoming, the best selling book of last year. Following that is new york staff writer Susan Orleans account of the Los Angeles Public library fire of 1986 in the library book. And wrapping up our look at some of the best selling nonfiction books according to the l. A. Times is mark mansons advice on leading a happier life. Most of these authors have appeared on booktv, and you can watch them online at booktv. Org