MODELING OUR CLIMATE FUTURE — WITH ANDREAS PREIN
WHAT NEXT FOR CLIMATE & EXTREME WEATHER PREDICTIONS?
How do we model the climate? Can we make predictions at a local level? What role does AI play? What do our models predict about the future of extreme weather events?
I sat down with world-leading atmospheric scientist Prof. Andreas Prein to pull back the curtain on how weather and climate models really work. Andreas explored the evolution of weather and climate modelling, the mechanics of prediction, where AI shines and struggles, the complex interconnections with other earth systems, and the important considerations when undertaking 'climate change attribution' for extreme weather events.
Inspired by first-hand experiences of extreme weather in his youth, Andreas dedicated his career to improving the understanding atmospheric systems. Nowadays he leads the High-Resolution Weather and Climate Modeling group at the Institute for Atmospheric and Climate Science (IAC), part of the Federal Institute of Technology Zurich (ETH Zurich).
It’s a fascinating and wide-ranging conversation:- we've all wondered what’s behind the predictions about the biggest force of change shaping our future, and this one has the answers!
Click below to listen.
As always, scroll down further for personal takeaways and a full transcript, and don’t forget you can listen to more of these wonderful insights from leading scientists, experts and future-makers by subscribing to “FutureBites With Dr Bruce McCabe” on Spotify, Apple, or on your favorite podcast platform.
TAKEAWAYS & PATHWAYS TO A BETTER FUTURE
Climate models are built on weather models
I liked the way Andreas built a step-by-step understanding of how we arrived at today’s climate models, from the earliest beginnings with human computers, then the first computational climate simulations in the 1950s (which gave estimates of global warming due to greenhouse gas emissions very close to what we have now!), through to ever more sophisticated models.
As Andreas put it, weather and climate modeling “are fundamentally linked. It's just different timescales that we're looking at.”
Climate models also connect with other models – for emissions predictions, carbon cycle, water cycle, dynamic vegetation modeling, ocean current modeling and so on – each of which is also becoming ever more sophisticated.
Many models, one direction
Approximately 60 models contribute to the Intergovernmental Panel for Climate Change (IPCC) reports. Most are derived from the same UCLA roots. The spread of custodianship and independent multiplication of effort adds signficant weight to the conclusions.
While the models are becoming ever more complex, with more possibilities for uncertainty, they continue to predict with high accuracy – another indicator of their integrity.
Bottom line: today’s models are already outstanding, and we can put high trust in these models.
A.I. WILL Not Replace Computational Modeling
AI represents the biggest revolution to weather forecasting since the introduction of computer modeling in the 1950s.
AI models are extremely accurate for weather forecastig.
AI models are also much cheaper.
AI models are, however, fundamentally different in that they are trained on historic weather data and contain no real physics.
This makes them poor forecasters of climate (historic data gets less and less relevant as the climate paramaters change) and poor forecasters of extreme weather events (rare by definition).
Computational models, built from the ground up on physical-based models, will likely remain the gold standard for accurate climate projections and for predicting extreme weather events.
Physical computational models are also important for attribution (see below). As Andreas puts it: “you can look at the equations, you can do more targeted experiments where you basically try to control some of the changes, and then you get a better understanding of what's driving what.”
AI and physical computational models are thus destined to be complimentary tools, each to be used where it is strongest. The future is hybrid models, already here, which interface AI and physical algorithms to better predict weather and climate.
Attribution is about counterfactuals
There are two types of attribution (1) attributing what effect greenhouse gas emissions / other changes have had on the climate and (2) extreme event attribution.
(1) For attributing what effect greenhouse gas emissions / other changes have had on the climate:
The method is to roll back time and run a full climate model with and without the change, then make a comparison. The models for this kind of attribution are simpler and more robust.
(2) Extreme event attribution is done differently:
Every weather event is affected by climate change because we have already altered the base state, so the question becomes, how different is event X because of climate change? E.g. how much more intense is this cyclone because we’ve warmed the planet by 1.3 degrees?
The common method is to run the weather forecasting model that forecast the cyclone, then cool the atmosphere down 1.3 degrees and run it again, then look at the difference. Then you can make an attribution. E.g. “The warming is likely responsible for 50% higher rainfall volume.”
Modeling attribution for frequency of extreme events is easier when it relates to straight thermodynamics because it's simpler physics. (e.g. climate attribution for frequency of heat waves is easier to model).
But for some things, like frequency attribution for cyclones, the interplay between such factors as thermodynamics and moisture content and intensity makes the modeling more complex and less reliable. It’s extremely difficult to make scientifically-valid attribution statements for hailstorms, and almost impossible for tornados, except to make the very general statement that they are collectively headed in one direction (worse).
The Future Is High-Resolution
High-resolution physical modeling enables predictions of local-scale extreme events like flooding, heavy thunderstorms, and so on.
The ultimate future of climate and weather modeling is many very high-resolution models, each targeted to key earth systems such as atmosphere, oceans, carbon cycle, food cycle in the ocean, etc, and all connected up together to “get more and more like an Earth system model.”
Andreas’ personal goal is to provide local-scale climate change information for anywhere in the world. He is already doing this on a targeted basis. Examples include:
Working with the US Nuclear Regulatory Commission people, using a high resolution model to provide data for resilience planning relating to flooding of nuclear power plants;
Working with the US Bureau of Reclamation, which operates dams across the western US, to model future snow in the Rockies (the key factor determining future run off and therefore river flows and dam levels);
Helping insurance companies better understand local risks;
Helping city planners with infrastructure planning.
Climate Tipping Points
We dedicated part of our conversation to discussing forecasts and tipping points.
Andreas and his colleagues are more and more looking at 3-degrees global warming as a realistic future. I’ve covered the consequences of this many times, so I won’t go into them again, suffice to say when I put it to Andreas that we are in for at least another doubling of the frequency and intensity of storm events he did not disagree. He is particularly worried about how we can upgrade so much infrastructure -- drainage systems, dams, power plants, and so on – in time to withstand these events.
On climate tipping points, Andreas emphasized that, while the earth can be considered largely self-regulating on geological timescales, the climate tipping points we are flirting with / monitoring in IPCC reports etc, are so dangerous because they have the potential to produce ‘points of no return’ with catastrophic consequences across human timescales. He further noted that the uncertainties are especially high in ecosystem models, which means we are flirting with potentially even bigger risks when it comes to non-returnability of marine life, food chains, etc.
On the changing ocean currents, such as the slowing Atlantic Meridional Overturning Circulation current (AMOC), Andreas is not an ocean modeler but points out that we know the currents have changed before and we know when they reach a tipping point the change can come abruptly. The current view is that the effect of a dramatically slower AMOC would be abrupt cooling from central Europe northwards, with the heat of course re-distributed elsewhere. We do not yet know what effect this would have on crops, food supplies, etc.
See the future you want, and work on it
Andreas’ parting message was on the “crisis of perspective” that so many young people are experiencing today — believing everything in the future will be worse and there is nothing they can do about it. He urges us to challenge these feelings and challenge ourselves instead to play an active role in the solution, however small that may be, because “we all have something to give and some role to play [we should] look into the future, see the bright future you want to have, and work on it.”
I couldn’t agree more. Thank you, Prof. Andreas Prein. A privilege.
INTERVIEW TRANSCRIPT
Please note, my transcripts are AI-generated and lightly edited for clarity and will contain minor errors. The true record of the interview is always the audio version.
Bruce McCabe: Welcome to FutureBites, where we explore pathways to a better future. I'm Bruce McCabe, your global futurist. And today we're talking about climate modeling and modeling extreme weather events, here with my special guest, Professor Andreas Prein. Welcome to the podcast.
Andreas Prein: Thank you very much for having me, Bruce.
BM: It's absolutely my privilege to talk to you today. I'm just going to give a very brief bio, and I'll include some links to some of your work as well in the show notes of the podcast. But you lead the high-resolution weather and climate modeling group here at the Institute for Atmospheric and Climate Science at the Federal Institute of Technology for Zurich. Is that, right?
AP: That's exactly right, yes.
BM: You've only taken on one of the most complex and difficult areas, I would think, in modeling in the world. It is complex, right?
AP: It's extremely complex. Modeling the natural system is very difficult, yeah.
BM: Let's just start with you. How did you get to be doing this sort of work? What drove you here? What was the inspiration? Or maybe who was the inspiration? Was there something that led you down this path?
AP: You know, I grew up in Austria on a dairy farm, and I was always interested in weather, and I can just really vigorously remember looking at the weather forecast in summer, for example, when we cut the grass and we tried to get the hay in, and you don't want the rain to come, because then you have to turn the grass again and again and again, and the hay gets rotten. So my father, I can remember that. This was always a gamble. This was really sticking with me. Also, your livelihood is depending on that. You have cows, you have to feed them in the winter. So this really influenced me. And then later on, I went to the Austrian military, and during my service, there was a big flood event in Austria, 2001. And we were caught in this area, and this was really the first time I saw devastation at this scale. And just seeing that and just also helping people there had a really lasting impact. So this really sparked my interest in extreme events. I really wanted to understand what's happening in the atmosphere when this is happening, and of course, climate change was already a topic in the early 2000s. So this drove me into physics and climate research at the end.
BM: That's really interesting. Did you have a sense from your family, the farming sort of background, that was there already a family knowledge that weather events were increasing, a sense of that, or not yet?
AP: We did never talk about that, no. This was really 1990s. I can't remember other environmental topics, like acid rain. These kind of things, they were so in Australia, you saw these pictures all the time in the newspaper. But climate change was not really that big, and I think this really changed, I think, when I went to high school. I had physics classes, and the teacher there basically brought this up, and I thought, this is fascinating. Then it just basically continued on this physics path afterwards.
BM: And important! Not only at a local level, but to me, it's the greatest force of change of all. I mean, we can talk about artificial intelligence, and we can talk about editing life with CRISPR, and there's lots of things that I do in my work, but there's nothing bigger as a force of change than a changing climate that's planet-wide, is there? It changes everything.
AP: Oh, yeah. Just thinking about the energy, like a single thunderstorm has much more energy than any atomic bomb. It's just the energy that's in the system that we are changing now. It's also mind-blowing that some small creatures like humans can impact the entire planet. And how fast this goes, actually. The rate of change is the one that... This is the most problematic thing, how fast the temperature is rising and things are changing.
BM: How quickly we can adapt. Because I often get that... I'm constantly in the US talking to people and encountering people who've been educated that climate change isn't real. And one of the cherry-picked pieces of information they use is, oh, but geologically, the planet's warmed and cooled before. And you sit there and go, “but never this fast!” Never, ever this fast. We have to adapt to this. So, yeah, nuts. So tell me about the modeling side. You've got into... To me, this is mathematically, physically, statistically one of the most... It has to be one of the most challenging things to model. The climate. So tell me about that journey, and maybe we can lead into just some of the methods and the inputs that you take into models today, or that in general we're using for better climate modeling today. Because to me, it's endless. Where do we start? Satellites? What?
AP: No, no, I think much earlier than that. The weather forecasting was always important. Actually, it started really getting serious with sailing. And of course, people on sailing ships back in the 17th century, they wanted to know how the wind blows and if there are storms coming and so on. This was in the 19th century, the first meteorological organisations started to emerge in France, in Austria, actually, in the UK. And it was for a long time really difficult to forecast the weather. And this changed once we really had breakthroughs in physics where we understood the equations of how air is moving in the atmosphere. This comes from fluid dynamics. So we have the Navier-Stokes equations. These are the equations that we use. And these are used everywhere. If you do anything with fluids or gas, these are the equations that you look at.
BM: And if you're ever looking at any weather forecast, this is behind that forecast.
AP: Yeah, so until 2023, true. But then AI came on the scene. We can talk about this later.
BM: Yeah, we'll get to that. Interesting. So there's a transition.
AP: There's a transition. But early on, there were people really formalising and they tried to, like there was Lewis Fry Richardson was the first one who imagined we can forecast the weather with these equations. And he basically put them on scorecards. And his vision, this was in the 1920, was to build a really big forecast factory with something like 60,000 computers. And back then computers were humans. And they calculate and they try to beat the weather. You have to do a faster calculation than real time because you want to forecast, not hindcast. So you have to beat the weather or the real time. And he had this sorted out, this whole thing. It never really took off, of course.
BM: I love the imagery though. 60,000 people!
AP: Graphics, like artists painted this forecast factory. And Lewis Fry Richardson wrote a book about it. So it's really interesting to read about it. But this really was the foundation of numerical weather forecasting as we saw it later. And then actually the next step was during World War II, the first computers came on the scene. John von Neumann actually, he was a student here at ETH.
BM: Really?
AP: Yeah. But he joined the Manhattan Project later on. And he worked on explosive simulations on ENIAC, one of the first computers that the Americans had.
BM: ENIAC, yes.
AP: Yeah, yeah, yeah. And he was successful. And he developed lots of these methodologies that you use to solve these equations on computers. And he immediately saw that there's a potential not only for explosions but also for weather forecasting on computers. So he actually hired a person, Jule Charney. And Jule Charney was then the first one who really wrote a, or one of the first ones who wrote a weather forecasting model and did a forecast on a computer that was okay-ish.
BM: Roughly when would that have been?
AP: Oh, this was I think '50s.
BM: 1950s. First computer weather forecast.
AP: But the point here is really that computational advancements and weather forecasting, climate modeling, were really closely linked. And still are. We still run our climate models on the biggest computers that we have.
BM: Got it.
AP: And Jule Charney, he's an interesting character because he wrote a very famous report, the Charney Report, to the US government. This was, I think, also in the '50s. You can look it up afterwards. And he warned the government back then about climate change. Because he wrote this model and he simulated climate and he basically tried to understand, if you double CO2 in the atmosphere, how would the temperature change? And he got quite close to the estimate that we still have.
BM: Really?
AP: Yeah, yeah. It's fascinating. It didn't change a lot over the years. Our models, however, really changed a lot. They got way more sophisticated. He's just started out with a simple atmospheric model. Nowadays, if we simulate climate, of course you have to simulate not only the atmosphere, but also the ocean, ice …
BM: Yes, the water cycle.
AP: … This water cycle, ecosphere, all of that interplay because there are feedbacks. These are very important. So nowadays, these are what we call Earth system models. And again, still, that's why also my professorship is named Weather and Climate Modeling. I really try to keep those two fields together because they're fundamentally linked. It's just different timescales that we're looking at. The weather forecast, you think about the next couple of weeks. In climate, you think about the next couple of decades or millennia. But yeah, still, it's still the biggest computers that we need to solve.
BM: So are we just adding inputs? I mean, I just think when we look at the fluid dynamics of how the gases are going to behave in the atmosphere, how much heat they're going to trap on a molecular basis, then at a group basis, how much energy causes how much wind, moving things around, then we're adding inputs in the water cycle and both the carrying capacity of the atmosphere for water has got to be in that model. And that changes dramatically with temperature. I think one degree, 7% more water.
AP: Exactly, yeah. The Clausius-Clapeyron relationship. Another one at ETH.
BM: Is that right? That came from here? It's such a remarkable statistic, just that one. One degree of temperature increase and atmosphere can hold 7% more water, which says straight away why we're getting more extreme weather events. It's one of the things, right? More energy and more water.
AP: Yeah, many extremes are linked to that.
BM: Yeah, and that came from here? I love it.
AP: The initial work and the theory.
BM: But you can see where I'm going. Is that what we've done? We've started just adding more complexity or expanding the model? Is that what happened over time? We've just added more inputs and then we account for re-radiation, reflectivity. It keeps going.
AP: Yeah, so we definitely made the models more complex and more complete as we also gained understanding of the climate system. At the beginning, there was very little understanding actually. And nowadays, we have much better understanding. Still not far from complete. But for example, how much energy? Most of the energy is stored in the ocean, not in the atmosphere. So you really have to capture that. Like ocean, the turnover, the circulation of the ocean is on timescales of many thousand years. It's very different from the atmosphere where you have a turnover of a couple of weeks, one, two weeks, the atmosphere changed completely. So now you have to couple all these things together. There's of course, as you said, there's also chemistry, but there's also the whole carbon cycle. So how carbon gets emitted but also really absorbed again. This is now modeled. Nitrogen cycle. So all these things, they come in and the systems become more and more complex. Nowadays, for example, we work on dynamic vegetation modeling where you really have vegetation that's responding to the warming and to the change in environment.
BM: And you've got to model and try and predict what the impacts to vegetation will be to then predict what the impacts to the climate of the vegetation will have …
AP: Yeah, because it's all linked. It's all linked. The Amazon rainforest is a good example there. This is deforestation driven to a large extent but also climate change driven. And the fear there is really that we are close to a tipping point where the rainforest will die off because if you cut down all the trees, the rainforest to a large extent creates its own rainfall because the trees are evapotranspirating. So they basically suck water out of the ground and put it in the atmosphere and this rains down. So this is really an ecosystem. If you cut down the trees, this evaporation stops and then you get in a very different regime and the rainforest will likely not grow back. So you have to simulate that, of course, if you're serious about climate change. But this is extremely challenging.
BM: It's extremely challenging. I don't know how you even begin to do that. I guess there's … if I picture you and your colleagues around the world, it's a huge community of people that are constantly contributing to the same models? Would that be fair? No, there's many models?
AP: Many models, yeah. So developing a model is extremely costly. So we talk about model families. So often what is happening, like one institute, the earliest models that we're still using were developed at UCLA, University of Los Angeles. This was one of the most earliest models. The UK developed one. Also, my previous employer, the National Center for Atmospheric Research, developed one. And then what happened over time, additional institutes developed their own models. But often these codes were copied.
AP: So somebody, you know, like this is science …
BM: We're moving around.
AP: …We basically just take the model with us and then we use it in another place. And then these models got further developed in another place, but they're still from the same family. And you can still see that, actually. But nowadays, in these IPCC reports that we have, Intergovernmental Panel for Climate Change reports, we have approximately 60 models that contribute, these global models.
BM: 60 models contribute to the IPCC?
AP: Yeah, yeah. Not all of them are on par. There's definitely models that are further developed than others, but it's a really big ensemble. And it's really important to have that because...
BM: It's independent verification, isn't it?
AP: It's independent verification. It's not clear how we, for example, just the Navier-Stokes equations, we talked about those. How do you discretize those? How do you get those equations onto a computer? There are multiple approaches how to do that. So now people are taking different approaches, and then you test what does one approach tell you about climate change and how does another approach simulate climate change? But this is almost with all components of the system. The cloud microphysics, radiation. There are different approaches to simulate that. And then you get an uncertainty out of that. It's basically, we call this model formulation uncertainty. So it's basically how well, or what's the uncertainty of how you formulate natural physics equations onto the computer.
BM: And just to clarify, was it 6-0 [sixty] or 1-6 [sixteen]?
AP: It's 6-0.
BM: 60 models, wow. So how have we been doing on accuracy? When we look at this huge history of modeling, you mentioned some of the early models are actually pretty much on target. Clearly, I imagine we're getting more accurate as we go along with complexity. And my naive understanding of this, when I read the different outputs from models, is that in general we've been too conservative with our climate change model in that some of the feedback loops have been bigger than... If we take a trend, the feedback loops have been worse than we expected. Would that be fair?
AP: It depends on what we're looking at, to be honest.
BM: Well, warming. Let's start with just warming, predicting the level.
AP: Yeah, the models definitely got better over time. So there's clear evidence for that. But they also got more complex. So you have to factor this in. If you start with a fairly simple model, and all of these models are somehow tuned. Every modeling center, what they want to see is a very good representation of the 20th century warming or the historic warming.
BM: They want to go back over the data and make sure it fits.
AP: Of course, yeah. So you tune your models to some extent to fit the historic data. Because if you're not able to simulate the past, how can you trust into the future? And tuning is a delicate thing. But the thing is, if you add more complexity to those models, you add more degrees of freedom. Which basically means your uncertainty could increase. Because now you better understand the global system. For instance, the ecosphere can do something. The feedback atmosphere and the ecosphere's feedback could kick in and you get into a very different regime. This can happen in a coupled model. So actually, it's a good thing that we're still on track. I would argue that if you have a more complex model and you can still project the future with a similar accuracy, that's a good thing. Because you understand the system way better and you didn't increase the error bounds that we have.
BM: When you talked there, it reminds me, I dug into, this is a while ago, I dug into chaos theory to try and understand a bit more. I finished a popular book from James Gleick. His book on chaos is just easily accessible for the average reader. Really, the thesis of chaos theory is that some systems don't return to equilibrium. You tip them out of equilibrium and they don't come back. By the end of the book I was utterly terrified when thinking about this particular subject because that's the possibility, right? We could tip a climate into an instability that never, even if we stop all carbon emissions tomorrow and all greenhouse gas emissions tomorrow, it could be tipped already into a place that it doesn't come back. That's sort of the fear there, isn't it?
AP: Yeah, so I think it depends on the timescale that we're talking about. Like if you look at paleoclimate records, we were in very different climate states over the last millennia and like millions of years. And the Earth system is self-regulating to a large extent. So the Earth was several times already a snowball. It completely froze over and then it melted again and then you get more in a vegetated state. It's a fourth and back. These are geological timescales. On human timescales, that's totally true. If you tip some key elements, there's a point of no return on scales that are relevant for humans.
BM: Yes, got it. We could tip it into something that's quite chaotic and with sharp increase or dramatic change, that's even much more dramatic than we're seeing now and it would be catastrophic within our times.
AP: Yeah, this is one of the biggest concerns. These tipping points and when they occur. This is also one of the main reasons why during the Paris Agreement, the international governments agreed to try to keep the warming to 1.5 degrees Celsius and well below 2. This is exactly what they say. This is not where we're heading at the moment, but the risk of these catastrophic changes are getting way higher once you get across 2 degrees. It's just a lot of … like it doesn't mean that they immediately happen but just the chances like it's a very risky game that we're playing once we get beyond.
BM: Just getting back to the biosphere and biodiversity, right now we're going through the biggest bleaching event for coral reefs around the planet ever, and I've been spending time with ocean scientists looking at that and they are very, very distraught because there's real potential to have catastrophic knock-on effects to the food chains and food production out of the ocean. It could be like a famine, an ocean famine situation, if you like. That plus acidification affects more broadly with fish. So that would be an example to me of a tip-over point which we're playing with fire around which could be very, very disastrous, I guess?
AP: Certainly, yeah. The ecosystem that we don't, in these models, we don't model the ecosystem very well. These are very complex interactions. This is, I think, an area where uncertainties are very high. Again, if you have very high uncertainties, a very high risk if you tip over. You want to stay away from the area where you tip over. Again, it's just a very risky game that you play. We only have one planet. Once we cross these tipping points, the consequences can be very dire.
BM: Now, there's lots of things I want to get into. Just briefly, we talked about this transition to AI. Can we just talk about that in modeling? Then I want to come back to some of the tipping points and some of the forecasts a little bit, some of the projections as they look today -- we'll probably scare ourselves a little bit there. More recently, we've used machine learning in these models and started to apply it. Has it become quite dominant quite quickly? Is that the subtext I'm hearing?
AP: Yeah. I think this is the biggest revolution in specifically weather forecasting since this 1940-'50 periods where we got computers and we were able to solve these equations numerically on the computers. The approach that we use nowadays is basically still the same as in the early '50s. Then AI-based models came on board. This was around 2023. Google was the first one with this GraphCast model. They are fundamentally different. There's no real physics in those models. It's basically trained on historic weather data and then use this training to predict the future. The mind-blowing thing is this works better than the physical models for weather forecasting to some extent. Often not for extreme events. Again, you train with AI in general. You need massive amounts of data to train. Then you want to have a big sample size of the things that you want to simulate, predict. Extreme events are per definition rare events. You don't have a lot of training there. Again, if you go into this extreme situations, these models can often fail. That's one of the big issues nowadays.
AP: I know our Swiss weather forecasting office, MeteoSwiss, all weather forecasting offices that I know, they're really trying to not only catch up but lead this AI development of AI models. Because there's a really big potential, but I think there's still a trust issue. We still run physical models and we will do this into the future because if you have a big extreme event coming and you have to evacuate, you really want to trust your model. With the AI models, my sense is we're not there yet. We cannot really fully trust them. It's specifically an issue if you think about climate change because we're pushing the weather and climate system into unknown territory. Historic data is less useful in 10 years because we're in a different regime. As we increase temperature and the climate is changing, this historic data gets less and less relevant.
BM: Absolutely. There's a couple of discussions recently that maybe give some clues as to where we go with that. I don't know if that fits with what you're thinking, but one is, you know, because every AI is trained once and then it can be tuned a little bit, but that's it, you know, and then you have to retrain it on a new data set if you want it to be better or bigger data set. But I recently had a conversation with Professor Richard Sutton. He won the Turing Award this year in AI. And he's basically introducing plasticity to AI models so they can keep learning.Yeah, so, and weather was one of the things we talked about, because clearly the weather's changing, the climate is changing, so a weather forecasting AI trained on last year's data will be useless, or less and less accurate over time, it has to be, in the current form. But his idea was you keep feeding in new data every day, and it's just, you know, like we learn, that sort of same bioplasticity that we have, he thinks we can get that going in another few years in AI, so that should help, I would imagine.
AP: It depends, again, if you think about weather forecasting, yes, because the forecasting is on the horizon of a couple of days, weeks, maybe months.
BM: Exactly, yeah.
AP: If you think about climate projections, where you want to simulate something at the end of the century at plus 40 degrees Celsius, again, you really have to learn how the weather looks like in this situation, and people, there are models that can do that, but they learn from this physical-based models. It seems like you anchor your AI-based solution on the solution that you get from this numerical models that we are using already, which is ultimately also a limitation. Like the big benefit of AI models, as far as I can see it now, is they're way cheaper. Like we run on the biggest supercomputers. You can run these AI models on your laptop, and they give you extremely accurate forecasts. Interesting. But again, the training is a big effort, and then how well they generalize. This is basically how well they can predict states that they haven't seen in the training. This is really a major concern. One thing I also wanted to mention is a big, for me, I'm looking at high-resolution modeling, and some of the motivations that we have there is we really want to look at local-scale extremes like flooding, heavy thunderstorms, hail, orographic influences like topography and flows.
AP: The nice thing with physical models, if you increase the grid spacing, so if you go to higher and higher resolution, new phenomena emerge. For example, we have a simulation of a tornado, and nobody told the model that there should be a tornado under this thunderstorm. It basically simulated it because it's in the solution of the physics. It's a physical-based model. If you allow it to go to high enough resolution, it will basically depict a tornado.
AP: An AI model would never do that at least not the ones that we have nowadays. You have to teach them, this is what a tropical cyclone looks like, this is what an extra-tropical cyclone looks like, this is what a thunderstorm looks like, and then it knows. But it's not really able to discover these things.
BM: It's not able to deduce it.
AP: A physical-based model would have.
BM: Maybe the future is a hybrid, I guess, of very, very powerful models and algorithms, supplemented by little bits of machine learning at the edges or local conditions or...?
AP: Like this is already there. I think that the future will be very broad. I don't see that the physical based models are going away especially for research and there will be purely AI based models at the same time. And then there will be this interface, like this hybrid models. And this is already happening. So we see this already. But I think I in local conditions. Again, I've got this question often. When I present at conferences: “are physical-based models over and AI will take over?” For me, this is really the wrong question. These are really complementary tools.
AP: We have a new tool in our toolbox now, and we should use both tools to our best ability to better predict the weather or what the climate will be in 100 or 200 years, or even 10 years.
BM: Yeah, that makes sense. You've really helped me understand why there are limits to what AI does. I mean, it learns by example, so there's an inherent limiting factor there -- if you're not getting into the systems and the systemic and the physics of the interconnections, then you can't use it to predict the outlying events and the things that matter to us.
AP: I know I'm not an expert in this area, but there are debates on if these models learn the physics or not. In some way, it seems like they did ...
BM: Yeah, it feels like it…
AP: … It feels like it, but, you know, of course, it wouldn't be able to write down the Navier-Stokes equations at the end. But there's some, again, this field is developing extremely fast, so it's really hard to know what will happen in the next two or three or ten years.
BM: I agree. I've come across all sorts of people who are doing new forms of AI or working on old forms of AI, trying to make them work still … symbolic learning, other pathways. I would not rule out AI modules that are mathematical and purely physical and so forth, and combined with ‘learn by example.’ Yeah, I would not rule that out for the future. I agree. I agree. It's an interesting one.
So let's get into... There's two things I want to talk to you about. I don't know what we handle first. One is how we get into the attribution for extreme events. And these are the tornadoes, the hurricanes, the cyclones, flooding, things that really matter, people care about today. We've got great data showing they're increasing dramatically because of climate change. But attribution is sort of a part of that. And I'd also like to get into some of the, I guess, what the models are telling us broadly about where we're going in the next 10 or 20 years. What makes sense to tackle first?
AP: We can start with the attribution, certainly. Okay, I think that there are two types of attribution. The first one, and this is also in these IPCC reports, you try to understand what greenhouse gas emissions and other sorts of changed land surface changes did to our historic climate. And of course, the problem is that we cannot roll back time and just repeat history without interfering with the natural system.
AP: So you do this in the model. So you basically don't change CO2 emissions. You don't change the land surface until you rerun everything and then you compare. This is basically … attribution is always you have to have a counterfactual. You have to have, like basically you believe in cause and effect. So there is an effect of greenhouse gas emissions on the atmosphere. And then you remove this effect and then you produce a counterfactual, that this never happened. And then you compare the two and then you can basically attribute like how much did greenhouse gas impact global temperatures. And this is done since a long time already. And like good study that there's in scientific communities, no debate. The greenhouse... Human greenhouse gas emissions increase temperature and change the climate. And then the second one is more this extreme event attribution where you, for example, Valencia, you have this massive event. And we just published a paper on that. And then you try to understand how did climate change impact. And there are two things, the intensity of the event, like how much more did it rain? And then the frequency, because it could also be how much more often or how more likely is this event nowadays. And it depends on which type of extreme you look at.
AP: It's more or less challenging to do this attribution. Everything that's related to thermodynamics, to the temperature change is easier because it's simpler physics. For example, heat waves. We have quite a good understanding. Heat waves have intensified in frequency, but also in intensity over the last 50, 100 years. And all the attribution studies would tell you the same. Once you get to more smaller scale processes like thunderstorms, where there's really an interplay, like we talked about the 7% increase per degree Kelvin, like you have more moisture. This is again anchoring this in thermodynamics because it's related to temperature change. But then there's also like the thunderstorm or this like this in Valencia, for example, this was a very cold extra tropical cyclone. How more often do those cyclones develop nowadays? That's a very difficult question. It's more in the dynamics of the system. How often do we get these cyclones? And if we have these cyclones and they have more moisture, how do the thunderstorms react? Because then you have this interplay between the moisture gets sucked into the cyclone, but then you have thunderstorms that lift the moisture and release a lot of water, but also the moisture, the condensation in the storms, they feed the intensity of the storm. So there's very complex interactions.
BM: Sure is.
AP: So that's why like heat waves, cold waves, we have pretty good understanding and can attribute those very well. Extreme rainfall, like tropical cyclones are easier and extra tropical cyclones than small scale thunderstorms. You talked about tornadoes. That's almost impossible.
BM: Too small.
AP: Too small. We cannot simulate them in the models, at least not in the weather models. And there are our understanding or even hail, like I work on hail a lot, like larger hail. We made some progress, but even there it's extreme. Like if you have a big hail storm hitting Sydney, you had hail here in Zurich. It's very, it's extremely difficult to make this attribution statement.
BM: So the Valencia example is a good one because I just, I think I left Valencia two days before that hit. And I think a couple hundred people died or something. It was unbelievable. That was such a huge event.
AP: But just to tip on that, like often it's not the forecast. I don't know if you saw, were you aware that
BM: There was no forecast?
AP: Actually, the forecast was not bad.
BM: Okay.
AP: And the same, like we had a big event in Germany, for example, 2021, many people are out of flooding. Many people died. The forecast was actually quite good. It's the warning often afterwards. Like where you say like, now you know that something is happening. How do you get people out of this region? When do you start to warn them? When do you start to evacuate? In Valencia, it was way too late. Like often, like the people were locked in their houses. The water was already halfway up the door. And then they got the warning to evacuate. And then of course it's too late.
BM: Right. And we're going to all have to get a lot better at that around the planet because there's more of it coming.
AP: So this is often the problem. Like there are enough historic examples where like some big event happened and you can only hope that you learn out of those. This is quite proactive. Like I would really hope that we got into a more proactive approach rather than reactive. So something big happens, but at least then you have a chance to learn and change. And often this warning chains are much better thought through afterwards. And when the next event happened, like we have examples for that. There are way less fatalities and things are handled better.
BM: Interesting. So when we talk about attribution for, so an extreme, a huge cyclone or tropical cyclone hits somewhere and the question hits the news media, you know, how much … You know, it's a classic thing. It happens every time. And then someone says, well, it's highly likely to be climate change enhanced. You know, what's the sensible way of talking about it? Statistically, I guess “these storms are X percent more likely to be bigger or in general, they're this much bigger now than they were 10 years ago.” Is that the most sensible way of talking about it?
AP: It depends a little bit on the event. It's really from event to event. It can be different. With tropical cyclones, extreme rainfall is often easier to attribute because it's linked to this Clausius-Clapeyron relationship.
BM: So in what sort of language would be the right type of language to use when you're talking about attribution of rainfall like that?
AP: First, when I talk about attribution, I always stress that every weather event that we have nowadays is already affected by climate change because we altered the base state. We are not in the same climate as we were in 1850 before we started the Industrial Revolution. So that's a fact, and there's enough studies on that. And then the question is, how different are these storms now? I would never do an attribution statement without backing it up with science. But there are different approaches. And for example, with tropical cyclones, what we use nowadays quite often is, it's basically to use a weather forecasting model. So you use the same model that produced the forecast, but then you basically warm up the atmosphere in the model or you cool it down. So if you want to know how much more intense the cyclone is compared to 1850, now we have something like 1.3, 1.2 degrees warmer on average global. So you cool the atmosphere down, you run the forecast again, you look at the precipitation, you do this a couple of times to just know how certain it is. And then you can make an attribution. You can say like now 50% higher rainfall intensities or more volume.
So you get this out of the model. You do the same thing with the future. You can do like 3 degrees warming and how would this look like? And people are doing this. And I think that's a good approach. It doesn't really help you with the frequency part. So this is not what like how likely or more likely was the cyclone to have like to form. It's more if it's there, like how much more rainfall comes out of it. So again, you can also see this, like if there are big events, there are always multiple attribution statements are published and they sometimes vary widely in how like how they attribute how much 50% more. It's 3 times more likely. But often with these events, at least like extreme rainfall heat, it's all in the same direction. The climate change intensified these events.
BM: You know, can I share a couple of data points? They shocked me when I had the conversation and they were financial data points. So they were just a different angle on it. And it resonates really well with my US audiences because they get the politicization of climate change, but they can often relate more directly to insurance and pricing and costs. And this came from Swiss Re, the insurer.
AP: Right over there!
BM: Yeah, they're just over here. And it blew me away. But they said over 40 years, so it's 1983 to 2022. So about 40 years, the increase of frequency for catastrophic weather events for them—now that’s ‘level two’ everything or catastrophic insurance events, really, but it's mainly weather because earthquakes were excluded from this. So it's cyclones, storms, floods. The increase in frequency is 457%. Yeah. For what they class as catastrophic. So that gives us some financial measure anyway, a different type of look at it. But then the increase in intensity, which they measure financially, so there's some caveats for the same period, 1500%.
AP: Yeah yeah.
BM: Now, a good half of that is because more people are moving to cities on floodplains and there's other reasons. You know, the claims are getting bigger. But they're shocking! To me, it brings it home. I think makes it real to a lot of people in a different way.
AP: And it's also, of course, then for politics. Like I lived in the US for a long time. And of course, you have these emergency funds and like you have to release emergency help more and more often.
BM: Yeah.
AP: And it's for all kinds of extreme events. Like it's tropical cyclones. Like I worked on wildfires, for example, with catastrophic … Like you had them as well in Australia. In California specifically. And like some of these places, they cannot get insurance anymore. Like even in Boulder, Colorado, where I live, like insurance almost doubled after we had a big fire in 2021 or 2022. So these places become uninsurable in some way. And then, of course, if you cannot insure them anymore from a private side, the state often has to bail in. And it's becoming really expensive.
BM: It's almost a financial feedback loop, another reinforcing effect which is hurting. Yeah. Really interesting.
Can we now look at maybe just some of the forecasts just looking forward? I'm just going to do a quick time check and make sure I'm not taking too much of your time …
AP: No, no, you're fine.
BM: As we look at what your models are suggesting, 10 years from now, I don't know what's the most useful, 2050? Can you give us a sense of whether there's any... I mean, my conclusions so far, so you can correct me please, but I feel like there's no chance of remaining under 2 degrees, for example, planetary warming. There is a chance of remaining under 3 degrees, but it's slipping away. That's sort of my view. I'd love to hear your views of what the models are saying, and then also what they're saying about frequency of catastrophic weather events as well.
AP: Yeah, so these emission scenarios are really developed outside of our models. This is basically like you play through...
BM: That's another model?
AP: It's another model, yeah. And then basically you develop these scenarios, these plausible futures, and then you pluck those scenarios into these physical models and then you run them. So this is basically what you do. And I agree with you. Staying under two is probably not going to happen. So what we are preparing now at the moment for is what we call an overshoot. So since the Paris Agreement will likely not be met, so what will likely happen is that we go beyond two degrees for a certain amount of time, but the hope is that we can bring the temperature back down as soon as possible. So to basically go into this danger zone for a short period of time, but then you cool down the planet again with technology. Again, this will be extremely challenging. It would be way easier not to overshoot.
BM: Way easier not to do it in the first place. And cheaper!
AP: Much cheaper. And then also, how realistic these overshoot scenarios are. This is based on technology that we don't have. Again, it's a risky experiment that we're running. In my research and many of my colleagues' research, we are now looking more and more at something like a 3-degree warming globally as a realistic future pathway. And once you look at that, changes are dramatic. You can see this already. We are at 1.2, 1.3 global warming now, and you can already feel it. For me, it was interesting. The last 10 years were really a change. Now you really live through climate change. You see these extreme events that we get over and over again. They are just out of what we have experienced in the past.
BM: I can tell you, I travel all over the world, and every country I go to, you can see it. Every single one.
AP: Yeah, everybody is affected. And it's really climate change is happening now. And now imagine with 1.3, and now you add 1.7. So the consequences will be quite dire in every kind of aspect, like extreme heat, certainly, but also flooding, the downstream effects on droughts and food production and so on. And of course, this has rippling effects on society and migration and all of that that we have very little understanding. So, yeah, we are in the middle of it, and it looks like we will continue for quite a while.
BM: And it seems like, for example, the numbers just then on frequency and intensity of storm events, it seems like we're in for at least another doubling of that, you know, just as a minimum.
AP: Yeah yeah. The problem is really also, like, for example, the infrastructure that we have, drainage systems, but also like dams, power plants, they are built to withstand historic events. And often, like, they are basically really planned with observed data that we collected over the last 100 years. And as we talked already with AI, like, this data gets less and less useful. And the more you warm, the closer you, like, the higher the extremes get and the closer you get to a chance that this infrastructure fails. And it's aging as well. And then, like, we have a lot of discussions with city planners, for example, what should we prepare for? Of course, like, coastal cities like Sydney are especially vulnerable because you also have sea level rise. So now you have to invest a lot of money to protect the infrastructure that you have and to make it resilient. Because if you build a new flood wall, for example, like, this will be there for the next 100 years.
BM: Yes.
AP: Yeah. But what will the weather in 100 years look like? And that's highly uncertain.
BM: Yeah. You know, Sydney is a good example because right now, lots of controversy very close to my home base where I live. When I'm at home, there's seawalls being built where there used to be beaches. They have to be built to preserve the properties. There's huge controversy because the erosion is moving to different areas now because of the seawalls. It's just the start. And the other one, recently, we spent some time in the Maldives looking at sea level rise there. And it's easy to say, oh, you know, it's only half a meter or whatever. Then you look at the capital, Male, in the Maldives, and there's an entire city built about one meter above sea level, you know. And it's a city. You can't elevate it. You can start building walls, but then you've got groundwater issues. It's unbelievable. You know, it's catastrophic for them.
AP: Yeah, especially for more global south countries. Like, often in the Western Hemisphere, we have the financial resources to protect ourselves, and we are basically responsible for the change, mostly. And then the people that suffer are really in these small, vulnerable areas that don't have the resilience that we have.
BM: Absolutely. Bangladesh doesn't have the same resources.
AP: Exactly. I work with people in the Philippines, and same thing, like the cyclones that pass over the Philippines all the time. It's crazy. They can, yeah, it's really extremely difficult to recover for them as well. Like if they get hit, the financial system is already quite thinly spread, and then you have to recover, and this can go into a downward spiral, certainly, yeah.
BM: So one of the tipping point ones I wanted to just pull aside and talk to you about was ocean currents, because it's my understanding that they're almost one of the most complex things to model, and they do change. And right now we're seeing the Atlantic conveyor potentially reversing or changing dramatically, or the AMOC, I haven't got...
AP: AMOC, yeah.
BM: The terminology there. But the potential change for local climate is massive when the current changes. So if the current changes off Spain, the climate of Spain will change. If the current changes on the east coast of Australia, climate. Sorry. If the current changes going down the north, south, the east coast of Australia, the local climate in Sydney and Melbourne and Hobart and Brisbane all change, and potentially quite dramatically. I just wanted your take on that. Is it something we can model accurately, or does it defy modeling?
AP: So again, I'm not an ocean modeler, but I follow this area. But it interacts with the ocean.
BM: Certainly, yeah.
AP: And one thing we know is that this can happen. We see this in paleoclimate data, and it happened in the past.
BM: Right, okay. So we know that currents can change dramatically.
AP: It can change dramatically, and the impacts on Europe, for example, when the AMOC is collapsing are quite dramatic and quite abrupt. Just a massive cooling. If this will happen is still up to the debate. At least this is what my take on this field is. Depending on who you talk to, we're close to a collapse or we are not as close as we think. The models that we have can do that, actually, but this often happens quite late. You really have to have a lot of warming, and then at 2200, 2300, you get a collapse. So you can simulate this in models, but it's quite slow. But the question is, is this realistically simulated in the model or not? Again, there is a lot of active research going on there, but if this happens, and there was a recent study, it basically would offset climate change in Europe.
BM: Yes, it would be colder here.
AP: It would be colder, and you would get closer from the temperature side to pre-industrial situation.
BM: But someone else is going to suffer.
AP: That's the problem. The heat goes somewhere else, and then we actually don't know what this would do to storm tracks or to extreme rainfall, because there's still a lot of warm air around Europe. And sometimes, and this is also what happens in very extreme events, you get this warm air moved over Europe, advected over Europe. And when this happens, it doesn't help you that you're on average colder. It means basically, and this is basically a broadening of the distribution, so you can get extremely hot, but extremely quite cold as well.
AP: So it would be a very different climate to what we experienced pre-industrial. And again, we don't actually know what this would mean for food supplies, for extreme events, and so on.
BM: And again, it's the speed of change, which is the potential disaster. It's all about accommodating that change. Infrastructure, crops, food supply, all of that stuff has to change dramatically.
AP: The AMOC already slowed down, so we know that from observations. And also the models show that. What we also know from observations is that it slows down for a long time, and then it suddenly gets way lower. So it never shuts down completely. The shutdown is bad terminology, because it's just operating on a way lower level.
BM: I see.
AP: Yeah, but this transition goes quite fast. This is the tipping point that we are concerned about.
BM: I'm struggling to remember the term. So it's Atlantic. What's the M? Meridian?
AP: No. Meridional Overturning Circulation.
BM: Okay. And basically that's taking warmer water north. I'm trying to remember.
AP: Yeah, yeah, north.
BM: North, and then the cooler goes south.
AP: That's the reason why the UK or Scandinavia is so much warmer than the east coast of the US at the same latitudes. It's this transport of very warm air to those regions. So it's really from central Europe northwards. This would be the regions that are most impacted by collapse.
BM: So, where do we go? I know some of your work. I'd like to just capture a little bit of your future research sort of direction, because I know some of your work is about multi-level modeling ...
AP: Hierarchies.
BM: Yeah, what's all that about?
AP: Yeah, so as a scientist, I'm really interested in the fundamental processes. I'm a trained physicist, and I want to understand, if I see climate change, and for example, a tropical cyclone gets way more intense in the model, I want to understand why. That's also the power of physical models, because you can look at the equations, you can do more targeted experiments where you basically try to control some of the changes, and then you get a better understanding of what's driving what. And you often do this, if you really focus on single processes, you do this in quite simple models. Well, you can control, like you basically turn off, for example, the condensation and heating. So if you condensate moisture, you release a lot of heating, like a lot of heat, and you can just turn this off in the model and see what would happen. And if you do this in an idealized model, you can basically go into the equations and find out physically this is what happens, and this is the approach often. So we do this, and then we look at more complex models and see is this the same thing that we can find there.
And so this is basically how we gain trust.
And also my models are simpler models than what is used in the IPCC reports. My group, we use really more weather forecasting models that are very high resolution, but often not coupled to the ocean or have not very sophisticated carbon cycle and things like that. But they can really do thunderstorms very well, for example, or very local scale extremes. And the goal for my research, we're building this models now up to get more and more like an Earth system model. So at the moment, we're actually coupling the high resolution atmosphere model to a high resolution ocean model, trying to get biochemistry in to see the carbon cycle in the ocean, but also the food cycle in the ocean to simulate that at very high resolution. And for me, the primary motivation to do that is really to provide local scale climate change information globally. This would be the ideal...
BM: So if I'm a policymaker or a farmer or a resident or an industry person in Barcelona or in Colorado or whatever, I should be able to come to that cluster of models and ask specific questions about what might happen in my area. Is that what that means?
AP: Yeah, a little bit like that. A classic example, I work with the Nuclear Regulatory Commission in the US, and they are very concerned about flooding of nuclear power plants. We actually got quite close a couple of times. And of course, climate change is making extremes more extreme. So there's a real concern there. For example, we work with them using high resolution model to provide additional data that they can use in the resilience planning. This is a typical example. We work a lot with insurance companies, for example, as well, that they have a better understanding of risks, but also with city planners, for example, especially if you think about infrastructure. But of course, it's also true, like the farmer on the ground, to understand how extreme can a drought become in this region, or how much more likely are hailstorms that ruin your grape harvest, for example.
BM: But it goes forever industrially. I was just talking to people building data centers, and their key issue is not only electricity requirements now, there's huge exponential increase in electricity required for AI in data centers, but water. So they're all interested in water availability and forecasting at a local level.
AP: Yeah yeah, water. I work a lot with the Bureau of Reclamation in the US, and they are operating all these dams, Lake Mead, the Hoover Dam, or all these massive reservoirs that they have in the Western US. And this is, of course, this is a major concern. If a region runs out of water, you're in a really big trouble. Again, Western states often can cope better. But yeah, water availability is, and there, for example, in the Western US, it's all about, it's mostly about how much snow there is in the Rocky Mountains. Again, this is changing a lot as well, because the snow line is going up because it's warming, and you get less and less snow and more and more rainfall. It's running off more and more quickly. And there are really good studies that show every degree of warming contributes, I think it was 14% less runoff in the Colorado River.
BM: Wow.
AP: Yeah, so quite stark changes.
BM: And that's an example of a very localized insight.
AP: Yeah, you really have to simulate snow then in these mountains. Again, the models that I'm using can do this quite well. So it really opens up a complete new area of research where you can look at very small scale processes, like how snow accumulates during the winter season and how this changes into the future, but also rain and snow transition. So yeah.
BM: So can governments and industry come and work with you, consult with you on a project basis with the centre here?
AP: Certainly, yeah. We work a lot with industry. But of course, we also have research grants. If you have a research grant, you basically work for the government, we try to do everything that we do open. Yeah, it's public.
But it's really important, if you really work with stakeholders, to get them involved as early as possible. In a city planner, for example, you really don't want to just do your research and at the end of your study, you say, “this is the data, use it or do whatever you want.” That's really a lost opportunity. So often you get them involved very early on. You team up with them. You talk to them what their interests are. You get a mutual understanding of what you can do and what they need. And then the outcomes that you can deliver are way more impactful. And of course, it's often a trust relationship as well.
BM: So is there anything else about the centre that you wish more people would know about that we haven't covered?
AP: I think what I would hope, I don't know, what I often try to convey is the modeling capabilities that we have nowadays are really stunning. And I think we can put more trust in these models nowadays. Many people are still very sceptical. And rightly so. There's good reasons for that. But we really have, it's a new regime of modeling that we do. It can provide very localised data. For example, extreme rainfall is a good example. Often people working in the hydrologic community are very sceptical against model data. And they rather use some observations and modify the observations for the future than using something that comes out of the model …
BM: Interesting.
AP: … and I think we can do much more and much better than that nowadays. But it's again a trust issue, a communication issue. But I think there's huge potential, especially when you think about resilience and to improve our resilience to extreme events with these new types of models.
BM: So if there was one thing that you could have leaders do, you know, to make this sort of, make the future or create a better future, what would you pick?
AP: Leaders, yeah.
BM: I don't know. Just on the back of our conversation.
AP: I really like, I'm more what I can do rather than what leaders can do.
AP: Okay.
BM: What's the biggest thing?
AP: What's really concerning me is if I work a lot with students, young people. And when I was in my teens and 20s, I always had this vision of the future. The future will be better than the past and the current. This seems to erode nowadays. It seems like young generation have, they have really a crisis of perspective or... And when I talk, like, it's really shocking to me when I talk to them, like, everything in the future will be worse. The good times are over.
BM: They've given up.
AP: It's, that's, you know, it's heartbreaking because it's really like this is self-fulfilling prophecy. If enough people think that, we go there. And it's really like, often this hopelessness is really bad. So what I try to, during my lectures, are talking, working with young people, empowering them. Like, be part of the solution. You have something to give. You have something to, a role to play. Don't give up. Look into the future. See the bright future you want to have and work on it. And I strongly believe in that. Like, if enough people think like that and we work together, we can achieve that.
BM: For sure.
AP: Yeah. And it really starts at a very fundamental level.
BM: Yeah, yeah. I like that very much. And, you know, before we started recording, we were talking about how, well, you were saying, you know, we have all the information we need. We don't need any more information, really, to fix this. We have enough data. And we have the technologies, you know, to fix most of it already. So it's just a matter of acting on it.
AP: It's very true. Yeah, like my take on that is adding more information has a marginal, leads to marginal improvements nowadays. Like, we know what to do. We know where we're heading. We know it's not a good place. It's now really to take action and act on it.
BM: Professor Andreas Prein, thank you so much for spending so much time with us this morning. It's wonderful.
AP: Thank you Bruce, this was a pleasure.