180. Der Meteo-Nerd - Tobia Lezuo

Or you're saying I really have a good feel for the weather and it's enough for me to look at the general weather situation, look out the window, and then I basically know what's going on. This skill is getting a bit lost nowadays with all these detailed tools; people sometimes forget how important and also how exciting it is to actually build your own model in your head.

Podz-Glidz, the Lu-Glidz podcast.

Stories from the cosmos of paragliding. Today

with Tobias Lezour and I'm Lucian Haas. Anyone who seriously aims to do cross-country flying, of course, wants to know when the best days for it are. This becomes clearly recognizable with the help of thermals forecasts, especially those that put potential flight distances for a day—shortly PFD—directly in front of the users like a lure. Once, this was the domain of the region-based REC-Therm forecast from the DWD. Until they discontinued the model in early 2024. However, two other providers, Burnair and XC-Therm, are continuing it. And not only that, they have refined the REC-Therm regions and given them completely new layouts in some cases to better map the thermals conditions along classic paraglider flight routes.

The driving force behind this development at XC-Therm is Tobias Lezour, 28 years old, from South Tyrol and currently living in Davos. Tobias calls himself a weather nerd. But he is also an enthusiastic paraglider pilot and can translate the complex world of meteorology very well into the world of the average pilot. In this episode 180 of Podz-Glidz, Tobias tells, among other things, how he was interested in the weather even as a child. He explains how the REC-Therm model creates regional thermals forecasts and how it differs from other weather models. And we also talk about other weather forecasts. We also talk about why it is so important to develop an understanding of the processes behind weather patterns instead of blindly trusting the models.

I can only present special conversations like this in Podz-Glidz because there are supporters who financially support my work on the podcast and blog. Join in! You can find all the necessary information about this on the Lu-Glidz Gleitschirm-Blog, specifically on the "Fördern" page.

Tobias, have you checked the weather today? Not just out the window, I mean, but really into the weather maps. Like how a meteorologist would do it.

Actually, I haven't. Today was just a day where I planned to quickly prepare for this podcast episode; it was a spontaneous thing. But I only looked out the window, and here in South Tyrol, there's fresh snow.

But isn't checking weather sites part of your daily routine? It doesn't have to be every morning, but yeah, I definitely check the weather sites once a day. What does your weather routine look like? I mean, how long does it take you to go through the weather sites? What do you look out for? Well,

It really depends on the purpose, so on the intention I have. On one hand, I check the weather sites for professional reasons. There, of course, my goal is to validate the forecasts and see if everything is working. Those are our weather forecasts and thermals forecasts, of course.

I also enjoy spending a few hours analyzing the weather. Mhm. Yeah. But when I check the weather privately for paragliding, it usually goes pretty quickly. I look at a weather map for the general situation, then I check the wind maps, the precipitation forecasts, and then of course our thermals forecast. And depending on the flight, it naturally takes longer. If I'm planning a long thermals flight, I can spend a few hours on that again. Even though I'd say it's not always sensible. Then in the end, the preparation is longer than

the actual flight.

That was actually rarely the case, but it can also happen. Yeah.

What's your go-to weather site that you look at when you say, "I want to do a relatively quick weather check"? Like, yeah, I want to go flying and I want to quickly check. You said you want to look at the general weather map first. What do you look at there?

I actually look at the weather maps from the DVD. So the

Surface weather maps, just?

Surface pressure weather map. Okay. Just for a first idea of what it looks like, what the overall synoptic situation is. But then I move pretty quickly over to our weather site, Exaterm. And I actually find everything I need for flight planning there. I might also add Windy, which has additional layers for temperatures, and if it's about skiing or something similar, I find the snow amounts, the accumulated snowfall, are represented quite well there. Exactly, those are the weather tools I mainly use.

Would you also say that's the bare minimum you actually need for flight weather to assess it? So, first a look at the general weather situation, then maybe checking the wind at different altitudes again, and then looking at a thermals forecast?

I think the minimum is everything you mentioned. Without the thermals programs. The thermals forecast, if it's purely about safety, becomes relevant when you're interested in extending the flight and really pushing it. But for safety, the thermals—you can find the information regarding thermals, well, the safety-relevant information about thermals, can be found in tools like Windy and so on, in plenty.

Now, you work for XC Therm, providing the thermals forecasts that the whole paragliding scene sometimes misses. It's true, because they say, oh, there's everything, you can see that long distances could be flown and stuff like that. Isn't it enough, at least for good flying days, to look at such a thermals forecast and say, if it's really that good, then the surrounding weather must also be that good? So, in terms of variables, can I say I can go flying if I see a region where good distances are predicted?

I would say for an experienced user, yes. Because the experienced user can't... he can recognize what a truly good thermals forecast is. If a user isn't experienced, they might be misled by good thermals forecasts that might have a hint somewhere that it might not be so good after all. Let's say the thermals forecast says strong thermals, predicted with a high base, but you ignore that the small cumulus clouds can grow very high. Exactly. And the experienced user sees that immediately and knows that means overdevelopment. And that probably means there will be thunderstorms in the precipitation forecast. An inexperienced user could ignore that, and then there's a real safety risk involved.

But for the experienced user who already knows all this, who knows when the wind,

these wind signs, these small arrows at the very top are a bit stronger, then you should definitely look at the wind maps. And then the experienced user will also recognize, for example, Föhn currents or strong valley winds. The inexperienced user might say, "Yeah, there doesn't seem to be much wind," and won't look at the wind maps, where you can then see that there's an overflow of half the main cell or that the valley winds are going to become very strong.

That means the more experienced you are, the less weather checking you actually have to do.

I wouldn't sign off on that just yet, but my observations and conversations with experienced pilots actually show me that they are either very routine with their tools and know exactly what they want to look at, or they really say, "I have a good feel for the weather and it's enough for me to look at the general situation, look out the window, and then I basically know what's going on." This skill is getting a bit lost nowadays with all these detailed tools. People sometimes forget how important and also how exciting it is to actually form your own model in your head and learn what a certain pressure distribution means, or what a frontal passage specifically means for me at home, specifically for my flight or my potential flight.

That's sometimes a shame, and of course, I see us as providers in a certain role of responsibility, that we still give people the opportunity—actually, to force them—to build their own model and to engage their own heads and knowledge to make decisions.

Do you have your own model?

Yes, absolutely. I find that exciting too. It's my passion, always trying to understand a bit more than what the model is showing and actually knowing a bit better how the model is forecasting something. Otherwise, I'd just be blind. I'd just believe the models and then I wouldn't have anything left to do. I actually want to improve the models and provide feedback, on one hand with the weather models—we are of course in contact with the developers there, at the major weather services like MeteoSchweiz and the Deutscher Wetterdienst—and it's naturally exciting to give feedback, but also with our thermals model. You need your own model or models, your own ideas of how processes work.

To be able to improve the models at all.

Would you say your model is better than the ones, I mean, all those other computed models you work with? Do you beat the models in many cases?

I haven't heard anything about there being an XContest. Well, there is the weather XContest, where you basically give tips in Davos.

I live with people from the SLF avalanche research institute, and we do a snow depth prediction game there. And of course, we try to be a bit better than the models, because you could just take the model output statistics. But we try to incorporate our understanding of the processes as well. There are contests like that, but I don't know of an XContest regarding flight weather, which is of course ultra-complex. That's why I think I have to answer the question with yes, otherwise I can't say with confidence that I can make a contribution.

That sounds like a lot, that you have a real weather contest. How did that actually happen? I mean, when did you start getting into weather?

Yes, that was actually very early. I don't know why, either. It's probably because I grew up here in South Tyrol, where I am right now, and it's pretty rural, I'd say. A small village, Wengen in the Dolomites.

And you're constantly in contact with the weather in the countryside. And I can remember how my... how my grandpa told me about the wind directions. There's a small valley here and he said if the wind pulls upwards, what we would call valley wind now, and there are still little clouds hanging in it, like wisps of fog, that often happens after a thunderstorm, then it rains again, then it rains again. And the process behind that is actually that there are updrafts again and that something new can brew together again. And if the wind comes from above after a thunderstorm, then sure. Then it actually picks up afterwards. And I started writing these kinds of climatological season and weekly reports as a child, like what the weather was like, how much it rained.

My family isn't a farming family, but of course, there are many farmers around. And it was naturally about how well the grass had grown, because we also harvested for our little farm. And it was about which flowers were already blooming in the spring. And in the winter, for example, I measured snow depths. So yeah, I was a weather nerd pretty early on.

And was it decided early on for you that you wanted to work as a meteorologist later? Actually, no.

That developed very gradually over time. I originally wanted to study for teaching, always in a scientific direction. I was always very interested in physics and mathematics. Then I worked in a school for a year, but realized that it might not be the right place for me after all. And during my stay in Peru, which was for half a year, I got in touch with a glaciologist, a Peruvian glaciologist. He noticed that I had a lot of interest in the environment, in topics surrounding the environment. And he said if I wanted to study that, I should definitely go to Zurich because they have a good...

...a good... study program in environmental sciences. And when I got back, I couldn't help but apply there. And then I got into the environmental sciences program. And there's a specialization called atmosphere and climate sciences. And I had already been a pilot. And probably because of that, I had an incredible interest in the atmosphere and wanted to understand how it all works better, and I sort of drifted into it.

So your grandpa, your aviation experience, and a glaciologist in Peru led you to meteorology—or to the degree, ultimately. To put it simply.

You could say that. They were definitely signs on my path, yeah.

When you studied it, was that what you expected? Or did you realize, oh meteorology, at least during your studies, is a lot of statistics, mathematics, and other stuff? And then maybe it didn't have that much to do with what you used to do outside. Or what you do now with aviation outside, where you say, I really feel the weather and can then measure snow depths or whatever. Instead, it's a lot of theory and also things where you have to say, these are large-scale weather models and whatnot. Things that might not interest me very much locally or aren't that important at all.

I never actually felt like it was a burden, because it certainly took some time for me to accept that meteorology is actually mostly computers and very little nature, very little being outdoors. The cool thing, of course, is that I found my own way to combine both. But yeah, in my studies, there was a lot of math, physics, and programming. But I have to say, I found the joy in it too. Or this combination of the super technical and yet very practical. I mean, I do program. I'm not programming some abstract large language models or training some machine learning algorithms, but I'm programming something that I can then feel with my own skin when the wind picks up, or that I can use with my own paraglider.

That's what makes it so enjoyable.

So you can sort of validate your own programming too and say, does this actually achieve anything or not?

Exactly. So I always say... It's a workday when I go flying, because I have to go out and validate how the forecasts are holding up.

Actually, I'm more focused on my leisure time, and I try not to spend my time in the air thinking about the model all the time or wondering what it has to do with X-Term, but rather to really have some time for myself and actually make it some leisure time.

But does it still happen to you sometimes like that? For example, you say you want to go flying this afternoon or this morning. You look at your own X-Term forecast for the region where you're going to fly and it says the cloud base is supposed to be at 3,200 meters this afternoon at 2:30 p.m. or something. And then you take off and you notice, okay, I'm only hitting 2,900 so far. And you think, okay, is my model wrong or am I just flying like crap and do I still need to find the right thermals? That I can still manage to get those corresponding few hundred meters of altitude. So does it happen to you sometimes that you're really thinking mid-flight, okay, am I the error or is the model the error?

I actually don't think the thought "am I the mistake?" compared to the forecasts—that thought rarely occurs to me. Probably because I have a special sensitivity to how many uncertainties are involved in the model itself.

And I'm more like, I'm in the wrong place, in the wrong region, for example.

Now, in our case specifically, the forecasts are average forecasts for a region, and that depends very strongly on which thermals I'm currently in, whether I'm reaching this base height or whether I find these strong thermals or these weak thermals there. And that's why the thought "am I the mistake?" comes up less often, unless I make a mistake where I say, "But there, I'm acting strategically again." And I say, "I know, I made a strategic flight error." That was last autumn, I saw our bomb thermals forecast for the Engadin and wanted to fulfill this dream of soaring over the Bernina or on the Bianco Glacier. And of course, I drove from Davos to the Engadin, spent some time there, took the cable car up, and then took off.

I was actually the first or second one to take off and then I just flew past a thermal. I simply didn't fly to the right spot, there were several reasons, but in any case, it was like, yeah, of course I was too stupid, right? And that had less to do with the model or validating the model and more with the fact that, yeah, I just didn't fly into the thermals, but

...over. That means the others who started after you, they're geared up, they had this great base, and you're just standing somewhere on the ground. But do you get angry, or is it something where you say, okay, I just have to analyze and validate this now, what happened there?

I got pretty annoyed because I realized I'd made that mistake less than, let's say, ten seconds after it happened, but I couldn't turn back because it was such a small pass; I couldn't fly back into the thermals. You were already too low, you couldn't get over the pass anymore, right? Exactly, I was already too low, I couldn't get back, and that annoyed me quite a bit. And yeah, there were incredibly many pilots in the air and of course I checked the live tracking and saw, yeah, the forecast is right, but I'm on the ground. But that's part of our sport, isn't it? It happens

everyone.

Speaking of the forecast being accurate, you see that then, when you say I look at the live tracking to see how high people are actually climbing and stuff like that. Then you can say, yeah, the base altitudes we calculated seem to be reasonably accurate. Do you really do that so regularly that you maybe at the end of a day, at least on the good days probably, actually look at XContest again and pick out the good flights from the important regions and look at the altitude profiles and where they flew and how high they got and when, at what time?

Yes, absolutely, that's part of our work. We're constantly developing the model together with Olivier Liechti. He's the model developer, he's been developing the model continuously for over 40 years, and we've also been developing the model in recent years. And our role, especially in paragliding—he's a sailplane pilot himself—is to provide feedback, and we actually do that, well, mainly I do it. We look at the flights flown, XContest, sometimes even in the live tracking; we look at them, compare them with the altitudes, with the climb rates. I rarely do this with individual flights because a single flight isn't representative. But if there are several flights in a region that are part of the assignment, and we see that something is off, then we switch that on specifically, we discuss it together with Olivier, and then we can, of course, immediately make adjustments to the parameterizations, i.e., the time-dependent parameterizations.

How

do you do that on XContest? Is there an option to show you flight buddies—people who flew together or in similar regions—all at the same time and then overlay the altitude curves so you can see, okay, they all reached roughly the same thing?

Exactly. There are also these spots, hotspots, I don't know exactly what they're called, where you can actually see which launch sites were used the most.

I think it's called Daily Hotspots.

Daily Hotspots, exactly. And that's actually where I check most often, because it already gives you an initial validation of the potential flight distance, right? Did we predict the right areas where the best flying happened in Europe on that day? Exactly. And then? Then I look at, as you say, several flights at the same time. I also download a few flights there and do... well, as an IGC flight file, you can of course do further analyses with the individual thermals, the climb rates, and so on. That's actually a bit cumbersome right now with XContest. For glider pilots, there's WeGlide, and that's a cool project; everything there is basically open data, which means I can get the flight data there via programming through an API and basically have it spit out a validation for me every day.

And it helps incredibly that you don't have to do everything manually.

XT-Therm, tell me how you actually ended up there, because you've studied now, but if I remember correctly, you also already worked with XT-Therm during your studies. How did that happen?

Yeah, I was doing my Master's, and a mandatory internship is part of our Master's program. And that's when I started working with them. I thought to myself, what could I try out during an internship like this that I might not be able to try out when I look for a job in the future.

And I knew XT-Therm back then; I used it for my flights. I first got to know it in Switzerland when I went there to study. In South Tyrol, it wasn't that well-known back then, and I thought, yeah, I'll just reach out. Because what they do seems really exciting with thermals forecasts and generally weather forecasts for paragliders. I sent an email, and Dani from XT-Therm wrote back to me. An internship would be difficult because he's the only one in the company and only works on it part-time, so a mentorship, as the university requires, isn't possible. And that would involve weekly meetings and such, of course.

But he said, yeah, maybe. Maybe another type of collaboration would emerge; he constantly needs support, also in the meteorology field, since he's a software developer himself. So, that was just in the back of my mind for a while, and then in 2023, Dani asked me and said, yeah, he had something concrete now. It's about the DWD shutting down the production of the Rektormodel and making sure the production still works. And now we have the flexibility to adjust the regions, and he wants to do that, and for that, he needs someone who understands the process and can sit down with Olivia Licht, the model developer, to adjust and further develop these regions.

And yeah, then I did an internship in the summer of 2023, a remote internship, because I was here in South Tyrol. I'm organizing a small reggae festival here, so I'm often tied up here in the summer, and I worked remotely during the day on the regions that the X-SATAM users now use. And that worked well, it fit; working with Dani was super cool and exciting. I got a lot of insight right away into how X-SATAM works, what all there is to develop, and that naturally hooked me, I'd say. It really motivated me to keep going, and after that, I was able to continue working,

still on a part-time basis and the areas of responsibility have naturally shifted a bit, a bit more towards the Meteo side, which suits me. But at the same time, I mean, we're a small reality, we're a team of four,

at the same time, there's also a lot of communication. And yeah, that's really motivated me. And yeah, marketing, storytelling, and stuff like that.

So, what's your role at X-SATAM now? Are you the CMO, like Chief Meteorologist, so Chief Meteorologist?

I've never been asked for a title, I'll have to ask Dani, but I think that could fit.

We always talk about X-SATAM now, you say it too, yeah, there are these thermal regions and we had to adjust them and stuff like that. Maybe just explain what the difference is between X-C-Therm with its Rectherm forecast? In Burner, it's also stored in a similar way as a model, where it also calculates in regions. What's the difference? Why does it make sense to calculate in regions for thermals, and not like other providers who basically offer area-wide thermal forecast maps that aren't divided into regions?

The listeners who also talk about X-C-Therm with its Rectherm forecast, who were at the presentation at the DHV seminar on high-resolution weather models, heard that a model must have a certain grid resolution to resolve a process. For larger processes, for example föhn, a resolution in the kilometer range is already sufficient in large valleys. With thermals, we're talking—we all know this—about very small processes or the radius of a thermal. Let's say the small Huyas thermals are perhaps very small, but if we say,

we resolve the model very, very, very finely, then we see these thermals. But as I said, we would have to calculate in extremely high resolution, like around 30 meters for example. And that is simply not possible computationally. At least not so far. And the approach of Rectam—and the approach is already quite old, Rectam has a 40-year history—is to solve this problem by saying, okay, we can't predict every single thermal individually, but we can group regions and in these regions we can then account for the terrain that triggers the thermals,

resolve at a high resolution. I mean, in full resolution. And the Rectam model then sees within this region, sees that the valley goes very deep down, that there are shadows there, it sees that there's a lot of area somewhere that's irradiated from the south and has forest on it, and therefore produces good thermals. It sees that there are high mountains with snow on them, which of course also influence the thermals. And all these details simply cannot be in a weather model, no matter how high the resolution, even at a one-kilometer resolution. That

So, just from a conceptual standpoint, Recterm, at least for its thermals calculations, basically calculates this with a terrain first, where it says, I'll lay a much finer grid over it. Do I understand that correctly? Exactly.

How fine is this grid, at least where he maps out the terrain and says, "Here I have these south-facing slopes and there I have shadows in the valley depending on the sun's position, if it drops below 30 degrees or something." So, how fine does Recterm calculate that internally, and by extension, XT-Therm?

That's actually 30-meter resolution. Okay. And that's the huge advantage of having the option to summarize by regions. And that makes a big difference compared to weather models that have one-kilometer boxes. You can imagine that in a one-kilometer box, the vertical rise—meaning the thermals we use there—won't be that precise, because you either get the average rise from that box or you calculate a specific thermal on the profile. But you also have to define how large it is to get a rise speed. With Recterm, however, these are really realistic thermals being predicted.

And they're also calculated with finer vertical resolution, meaning it calculates with 100-meter resolution there. The output format shows 200-meter resolution. Because now we're talking about uncertainties again, or you could go incredibly into detail and show everything, but that conveys a false sense of security. That means it also calculates much more precisely vertically than what a weather model has for vertical resolution. And

What goes into a thermals calculation like that? I mean, does a Recterm sort of simulate an actual thermal bubble and say, okay, early in the morning they're still smaller? In that case, we say they get an average resolution. They have an extent of 100 meters, and in the afternoon they just get bigger, and there we calculate with thermals that have a volume expansion of 300 meters, and because of that, they'll rise better just based on the volume alone, or how does that work?

What's actually more relevant than the size there is the solar energy coming in, what happens to the ground, how the ground warms up, what that does to the air, and then of course the stratification, right? Yes. And a very important point is that Recterm also calculates its own stratification. That means it doesn't just take the profiles of these parent models. It needs output from these parent models, so from the weather models themselves. Recterm isn't standalone. It always needs the input from the weather models, but it then calculates its own stratification, which then results and also takes into account the valley winds, so the winds between the regions.

subsidence, as we call it in meteorology. So, how the air mass sinks as a compensatory flow for the thermals and from where it rises. How does it actually calculate an internal vertical profile? And it usually just matches much better than the weather models because of the resolution again.

But do you calculate a vertical profile per region, or do you say, "Wonderful, we have a resolution of 30 meters"? I calculate a few thousand meters distributed across the entire region. These are profiles that are all 30 meters apart.

Internally, all of this is handled in great detail, but what we use for the final product is one profile per region. That also makes sense because the region—well, Recterm's assumption—it works that the heating and the thermals conditions are dependent on the valley wind systems, because the valley winds actually bring in the air. Depending on where your air is coming from. Yes. Yes. Depending on where your air mass comes from, you have different conditions. We see this incredibly well here in the Dolomites. If you fly here in Alta Badia, where I learned to fly, you always see the region around Cortina, which has a southern inflow with the valley wind, with malpine pumps.

The air actually comes from the lowlands, from the Pianura Padana, and further looking, from the Mediterranean. It has a much, much lower base.

The air being delivered there is much more humid.

Much more humid air from this area, stable layering, invasions that are well-known below 1500 meters there. It is very difficult to dig yourself out when talking about the Pustertal, as everyone there has already dug themselves out below 1500 meters. So there, you can really see the obvious difference in where the air mass is coming from. And Recterm's approach is precisely to summarize these regions respectively. Yes. And then to simulate this air mass advection with valley winds. That's why it also makes sense to summarize them and not have a profile in a single region.

Now you have to explain something to me again. The valley winds. If I understand correctly, most weather models are actually still too coarse in their spatial resolution to be able to map many valleys at all. The large valleys still fit into the grid, so they are represented, but finer valleys are not. That means these valley winds aren't actually in these weather models. So how does Recterm get its valley winds?

Interesting question. With Recterm, it's actually the case that it doesn't work with the dynamics, meaning the wind fields of the parent models. The wind fields from the parent models are only used for displaying this wind in the region, which is very uncertain. The

average wind that's there, yeah.

Exactly. It enters there, but not into the calculation for the valley winds, for example. The valley winds. So conceptually, the valley winds don't just work, but here we might have to make a few concessions now when we talk about it, but because of the pressure gradient that arises between the lowland in the classical sense and the valley, because more air mass rises there. Then there's the volume effect, where the same energy is shared across a smaller volume of air, and therefore the valley atmosphere heats up more and this air rises. And that's why it sucks from below. And exactly this pressure gradient can, of course, be calculated by Rectan. Because Rectan calculates the rising of the thermals and because of that, a suction effect is created, so to speak, in every region.

And then this region pulls air mass in from the neighboring region. This happens, so to speak, with a wind vector between the regions. So, it's not about knowing where the valley wind is. Where exactly it flows through. Rather, you know this region has stronger ascent. Total. So, it's not about individual thermals now, but about the total air mass moving from the surface upwards, and because of that, it has a suction effect from the other region.

Air is basically being shoveled from one region to the other. Exactly. And Rektan can model that. But Rektan doesn't show, "oh, the valley path is like this and I'm following the exact air parcel as it slips through there." It just says that because Region A is currently developing better thermals, the air pressure there is decreasing or is slightly lower than in Region B, so it's now getting additional air supplied by Region B.

Exactly. Very well put.

Okay. Ah, then I misunderstood something again. Maybe the listeners will too.

This Rekter model. If you're calculating base heights there, you're saying that's the average base height for a region. Is the average calculated by looking at the entire region and saying, okay, we see in the front area that the thermals might be worse, but in the back, it's pulling really well on this day, and that's why we calculate the average accordingly? Or is it just an average point that you've set once in the region, where you say, that's our standard point that we calculate from?

No, the average value really takes the entire region into account. So it's not a characteristic point. Instead, it considers the whole region. It's obviously a difficult statistic in the sense that if there are thermals in one area and not in another, it would pull the average base height down significantly, for example, if there's high fog somewhere, and so on. So it's a bit tricky in that sense. But it definitely takes the entire region into account. It's not an example point where you could say, hey guys, look, there's this example point. With this information, you could so to speak extrapolate and say, I'm that far away, probably the height is lower for me. You can actually do that if you have experience with the tool.

Then you know that my home field usually only rises about 200 meters below this predicted altitude because of the local conditions. Of course, solar radiation plays a role too; if it's an east-facing slope, it'll have a higher base earlier and then a lower one again later. Because it stops getting energy and so on. Those are the local effects. And there we're actually back at the point that despite all these exciting models, local effects remain something the pilot has to take into their own hands a bit, have to stay informed, have to build their own model, and have to inform themselves about local conditions.

And I think that's also what makes flying so exciting.

Of course, it also involves risks, and you have to know the local conditions. But I actually think that with process understanding—not with models or forecasts, but with process understanding—you can already figure out a lot. So I don't really have to call every flight school or every tandem club to ask, hey, what's the valley wind like there? Because with a process understanding, I can already read quite a bit of the air and the lee sides.

Process understanding means for you, okay, I know where better thermals develop and the air flows accordingly towards them, and then I can imagine the lift zones, thinking to myself, because I just say, if it's pulling that way, then the air has to come from somewhere, and the lower mountains above it will probably be overflown by it. So that's the process understanding you have there.

Exactly. On one hand, it's about the thermals that can create lift and lee. Of course, understanding valley winds is also crucial—how does a valley wind work, which direction is it pointing, especially with special topography? That's where process understanding might get a bit trickier. Sometimes it's even counterintuitive for me, why a valley wind is going in that direction, the Maloja wind. But with process understanding and some information, you can get very far. Also, a valley wind has a certain inertia; it's not as agile as us glider pilots who can just quickly turn left when it flows into this valley. It first goes up, then shoots up a opposing slope, and then falls back down, and maybe it also makes those kinds of meanders that you sometimes observe.

And here, the models can actually help, but I always say they don't help with the forecast. I don't look at them and see, "What's the valley wind like tomorrow at 12:00? How strong is it? Which way is it pointing? Where is it creating lift? Where is it creating lee?" But I can also read a lot from the model forecasts for valley winds, like how the valley wind works and where it makes those kinds of meanders, for example.

This region you have there—you mentioned earlier when you got to XC Therm, it was also about redesigning the regions. And I think you've put a lot of work into that. You've invested a lot because now you're not just covering the old ones, but XC Therm covers most of Europe now, if I saw it correctly. You've defined regions everywhere; I even saw thermals regions mapped out in Iceland and stuff like that. I don't know if there are XC Therm users in Iceland, but at least there are regions there that can be had. What was the work like for you? How do you go about saying, I know, this has to be a meaningful, functional thermals region? Meaningfully functional in the sense that there's a somewhat sensible forecast at the end, because you say this air mass in this area should be somewhat uniform, so I can say if I take the average, then it fits too.

Yeah, that actually brings us back to this assumption that the uniform thermals conditions

in areas are uniform, which are limited by the valley winds. So if I have a basin somewhere, if I have a basin that is limited on all sides by high mountains, then I can actually assume that the air mass that will rise there, as well as the surrounding profile, is shaped by this valley wind coming in from below. How strong it is is another question. And then I draw my regional boundaries exactly along the latitude. It's a bit more complicated with the pre-alpine, alpine, and inner-alpine areas, because the same thing naturally plays a role there. So the alpine pumping plays an important role. And it doesn't stick to the rules of the valley wind, so the valleys themselves, but it likes to flow over mountain ranges and especially if the valleys are transversal, i.e., transverse to the alpine pumping, then you can't think in terms of valleys anymore.

And there, it's simply a classification into pre-alpine, alpine, and inner-alpine areas.

And of course, pilots' experiences come into play there. Because there are local peculiarities, like knowing that this mountain is washed over all the time, so you can put it in the lowland region. And from this mountain, there are actually proper Voralpen thermals. And from here, the base is suddenly 1,000 meters higher. That also comes into play, of course. And then, what we're also trying to optimize are flight routes, to make them user-friendly. So sometimes there just isn't such a clear boundary. For example, if you divide a valley lengthwise, like in Valais, we have Lower Valais and Upper Valais.

There, there might be a more distinct boundary. But if you take the Inn Valley, for example, and divide it somewhere, there isn't such a clear line. It's not that the base is suddenly higher; it actually changes gradually along the valley. And there, we also try to take the flight routes into account and see how a user can get all the important information about the region in the most ideal way with as few clicks as possible. In the lowlands, it's naturally even more of an issue to take the flight routes into account because the terrain's constraints are much weaker there. And in the lowlands, the concept is actually more about grouping together areas where the topography is similar—like a mountain range or a group of hills.

That's because in the lowlands, you don't have any given... there's an inversion, you don't have any topography, no surface area that can heat up above the inversion and lead to thermals, but everything is below it, so the thermals basically stall at the inversion. You can't really fly. But if there's a hill or mountain range somewhere, then the surface area above the inversion can heat up and can also produce very good thermals.

That means, probably for you now, if you were allowed to define and redefine these thermal regions, you'd say, okay, with my knowledge from mountain flying, I'll go along this ridge, that's a closed valley, it comes in from there, so I'll sketch it out like this and say, it's limited here—that's relatively easy to understand for the mountains. Now in the low mountains, maybe partly, but the flatter it gets, the more difficult it will probably become at that point. Does that ultimately express itself in how accurate the REC-Therm forecasts are? I mean, does REC-Therm work better in the mountains than in the lowlands?

I would say actually no. So the regional division in the lowlands, because it's flat, plays less of a role in how the regions are set up. Of course, something I forgot is the ground surface. If I have areas that are forested and next to them I have areas that are farmed with irrigation and so on, that naturally also has an impact on the surface conditions and on the thermals. But let's say we do it as well as possible with all the information we have, satellite data and so on, then I wouldn't say that's the reason why the forecast in the lowlands should be less accurate than in the mountains.

REC-Therm naturally has a history, so historical REC-Therm, developed for the mountains. And it also brings all its strengths into the mountains especially, because, to put it generally, thermals in the lowlands are somewhat easier to simulate, to forecast. And in the lowlands, as I mentioned before the advantages of this topography that we can completely resolve within the region, if we expand that to the lowlands, then REC-Therm with this method doesn't really have an advantage over another. Over a normal weather model. That applies now to the topography. The other thing I mentioned, with the vertical resolution, with the explicit calculator,

the thing with the vertical resolution and the explicit calculation of parcel lift, that also applies to the lowland area, of course, there are advantages there as well.

What I notice, I live in the lowlands, so here in Bonn, the average height is like 60, 70 meters, then there are a few low mountains, then it goes up another 100, 150 meters. And then it goes back down into the valleys and stuff. But from above, it's already pretty flat, so to speak. And what I often notice here with flying, when you look at the forecasts, including the thermals forecasts, is that sometimes even in relatively small areas—much smaller than your regions—there can be significant differences in the thermals. Where you can see, there's really the classic stuff you see on a large scale, there's the low where it rises and in the high it sinks. And you have these small subsidence inversions and areas that are more capped, where not much comes up, and other areas where it can flow through nicely.

And sometimes there are inversions or moist capping layers hanging in there that aren't in the models at all, even if you look at some temperatures on Windy. You notice it completely while flying—you think, "Today is forecast to be great, but I just can't get off the ground here."

Is there a way to improve that, from a forecasting perspective? Or, looking at it differently, how could I as a user perhaps anticipate something like that, so I can say, "Hmm, Lucian, factor that in, maybe it won't go as well today as the XZ-Therm is telling me right now," or whatever service I'm using at the moment.

Yeah, I think as much as we wish for there to be a perfect forecast that solves all these challenges with microclimates and so on, I actually believe—so the higher resolution will help, it's already helping now and will continue to move in a direction that provides better forecasts in the future, but we will never be able to do without our own experience and our own attention to looking at such things. It could be, for instance, an inversion that's just annoying because you can't climb. It could also be something safety-related. And I think we should always be aware that no matter how great the forecasts are and become, we will always have to rely on our understanding and our experience.

Now, for your specific example, I think it can be a super high resolution. In meteorology or research, they talk about LES, which stands for Large Eddy Simulations, and there they talk about a resolution of 100 meters, for example. And there, of course, very small processes become visible—slope winds, individual thermals, and so on. The point is, if the model doesn't have input at that grid resolution—meaning observations for the starting point, for the initial conditions—then running this model will only bring in so much, because in this basin, as I understand it, where you're going to fly now, there's this inversion because cold and moist air might be gathering there.

If I don't have an observation there and maybe my station isn't enough, and you really need a vertical profile, then this higher resolution might not provide the benefit we're hoping for.

You'd have to measure much more. You'd have to measure much more. Not just... the measurement is one thing. The big difficulty in numerical weather forecasting is assimilation. How do I bring all these measurements together? For me as a non-experimental atmospheric scientist, a measurement is the right thing. Maybe for you too, maybe for most paragliders, when they look at a measurement, they say, that's the truth, that's reality. But reality is much more complex than this measurement. The measurement is at one point, the station can be biased in some way, because it's now for the wind in some lee, or behind a hill for temperature in the shade, not over grass, but next to a house.

The device can have problems. I had to learn that during my studies, in the lectures I took in the field of experimental atmospheric science. What was really shouted at us is: measurements are wrong.

You always have to be careful. In assimilation, it's a huge challenge to bring all these incorrect measurements into a consistent picture. Which measurements do I exclude when they contradict each other? MeteoSwiss has a very high standard for which weather stations they assimilate—meaning, which ones they include for the model's initial conditions. They don't just take all those many temperature and wind stations that exist; they only take the ones that meet their highest quality standard. Because they have to rule out the possibility of causing even greater damage this way. If they include incorrect measurements, it causes more damage because the model actually knew better. And then they tweak the model until it's wrong.

So basically

the whole time, they're calculating with all sorts of pi-times-thumb rules, where every weather service—if you say assimilation, it's about taking in data and dumping it in somewhere according to certain rules. But in the end, a lot of it is actually just estimated stuff. I

I feel that sentence right now. So, after...

...estimated stuff based on certain rules. If you imagine it like this: you have a measuring station on the ground. It tells you super WMO standards and all that. I calculate really perfect temperature values for this point. The solar radiation is measured exactly. The wind is measured accordingly at 10 meters and so on. But sometimes it doesn't know anything about what's exactly 500 meters above this point. Yes, and maybe moist air can pull in there, or something else that can't be measured down there at all. And that might just be a relatively thin band of moist air that pulls in there for a while and causes a disturbance, or creates a kind of inversion there, or something like that which can't be mapped at all. And sometimes, when I imagine all of this, I'm surprised by how good the weather models actually are in the end with what they spit out, how well they still match reality.

I think that's exactly how many listeners feel right now when they hear all this—this skepticism toward the models. Many might wonder how the forecasts can be so good. And it's not my intention here to say the models are bad, so that everyone suddenly feels like they're actually not that bad. But it is true that the forecasts are getting better and we know more and more, and the statistical methods—the machine learning methods—that are involved continue to improve everything. But we're still talking about a chaotic system. That actually comes directly from this idea of initial conditions, so the starting conditions. If we don't know them perfectly, we can't predict the future; eventually, it spirals out of control.

You know this as the metaphor of the butterfly's wingbeat on the other side of the planet. I think the double pendulum—maybe you've heard of it or seen a video, maybe we can link it in the podcast episode—I find it very illustrative. Depending on how you set this pendulum at the start, and it can even look like it's being started exactly the same way, patterns and things happen in this pendulum,

which look completely different, and it's similar with the weather system. Very small differences in the initial conditions can have completely different consequences further down the line, at least over several days.

You mentioned earlier that for XC-Therm, you always do a validation. You check if it actually happened the way we predicted. What happens if you find out it wasn't? Is something in the model tweaked immediately? Is something turned at some point? And what interests me is, you have hundreds of regions,

across Europe, and it could be that it fit in many regions but not in another. Is the model actually tweaked in such a way that there are adjustment knobs for each region, where you can say, okay, I know now that in my region in South Tyrol, in the Dolomites, the cloud base is actually always 200 meters higher in reality than what we've been outputting. So we always automatically add those 200 meters for this region. Do you actually do that? Or are there actually region-specific adjustments? Or do you just say, well, I have to cut the region differently so that it fits better?

So there's both: adjustments on a global level for the entire model. But it's actually very rare to do something like that. That would be a completely new version. And that certainly wouldn't be validated based on a single day, but over several days with different weather conditions. Because of course you can optimize the model to fit, let's say, for high pressure in the summer, where the longest flights are made. That is actually one of the focuses. It's true that compromises are made, for example, to ensure that the forecasts fit well in the summer when the long flights are happening. And for that, you might have processes and mechanisms running in the winter or spring that are less well simulated.

Those are conscious compromises because there are limitations in the formulas. These are physical formulas in there, after all. And of course, you can make case distinctions, but somewhere it's limited. There are adjustments at the model level, and there are adjustments at the regional level. If I see a region is constantly off, then something is wrong. You can of course make adjustments there, tweak different parameters. And what I also have to add is that we rarely make adjustments based on a single day. As I said, you have to make sure the model works in different weather conditions and not just now, for example, with high pressure, because then a front comes and in post-frontal weather—meaning when the air direction is unstable and good—we want the model to work there too, because you can also have some great flights then.

And that's why, whenever we make a change, we always test it over a longer period of time—meaning we also run it against historical data to see if it would have worked in the past. And only then is it released. Of course, it takes time, but experience has shown that sudden and quick adjustments aren't actually productive, because then you just turn the screw back the other way and it doesn't fit somewhere else.

Do you also calculate ensembles with Recterm, or have you ever tried calculating ensembles where you say, I'm not just calculating one solution, but I'm calculating ten solutions with very slightly different initial conditions or something like that, and then I look at how large the range is, so that I can ultimately say, I might only output one result in the end, but I can say, okay, the probability of this result being correct is very high for today, and on another day it says, okay, I predicted a base of 3.6 here, but the probability of that being correct is maybe only 60 percent today.

Recterm was also run as an ensemble at the German Meteorological Service. There, they had ensemble PFDs—potential flight distances—distributed by probability. They had boxplots of the potential flight distance. With our current setup and the amount of data that would have to be loaded, that's actually quite challenging. We didn't do it that way. Of course, there are other uncertainties, for example, the model or the parent model used. We can also estimate the uncertainty there a bit. That's the uncertainty between the models that is then taken into account by Recterm. That naturally also creates an uncertainty.

What is the base model for Recterm then? What do you use there? Icon D2, or is it now Alpenraum Icon CH1? Or what do you have as your basis?

In the production environment, it's Icon D2 and Icon EU. Those are the models. People always talk about compatibility and model scales. And every time I, so to speak, include a different model—let's say I include a 1-kilometer model—it naturally simulates much more of the processes like thermals, valley winds, etc. And then it can be that the input data for Recterm is different than with a coarser model, and therefore adjustments have to be made in Recterm. That's why we continue to use the slightly coarser models, Icon D2 and Icon EU, and we actually have pretty good experiences with them.

But you have a weather site where the CH1—that 1-kilometer Swiss model—already accounts for the winch and things like that. Do you have plans to say, at some point, we also want to calculate the thermals with this 1-kilometer model? Or would you say that ultimately, for our regions, which are already somewhat larger, this 1-kilometer or 2-kilometer—meaning Icon CH1 or Icon D2—actually doesn't make much of a difference anymore regarding the quality of regional thermal forecasts? Or does it even matter at all?

I can't tell you that yet, because that's exactly what our current research and investigations are about—specifically looking at how the rotor behaves with the higher-resolution model.

So, internally you calculate something like that and then you see if it actually helps or not? I mean, it might come along at some point or it might not, because you say it doesn't really do anything. It's just as good as before. Or maybe even worse, because some errors might build up more easily there or something. Exactly. Cool. But what I saw, what you have new on the xD-Therm side, is that you now have wave forecasts. That's probably usually more interesting for glider pilots who want to fly Föhn waves now and want to know where the main wave is, and the second and third, where can I get in and how can I maybe get up to 5,000 meters or whatever. Is such a wave forecast in any way also interesting for paragliders?

Yes, as you say, the wave forecast is naturally intended for glider pilots. They can do incredibly exciting and long flights in the waves. We've already been able to reach quite a few glider pilots with this wave forecast and received very positive feedback that the position and timing are very accurate. The wave involves a completely different process than thermals. I would say it's very predictable. The uncertainties are much, much, much smaller because this process takes place over scales—temporal and spatial scales—that are simply much larger. A wave is rarely there for ten minutes and then gone again, as is often the case with thermals. A wave has very laminar lift over a large area, and a wave is triggered by topography or, of course, by inversions and so on, but primarily by topography, like how the air is pushed over the Alps,

the air parcel reaches an altitude where it's no longer comfortable, where it's actually too cold and wants to go back down, it goes down on the other side, still in the wind, so it gets transported, goes down on the other side again and then shoots back up because it was pushed down too far, and that's how a lee wave works. And because this mechanism can be represented quite well with high-resolution weather models, even the mountain peaks are in there and relatively local effects are included, the predictability you have with the wave is very good.

But in that case, are you calculating with Icon CH1, so with this 1-kilometer model? Exactly. Or do you calculate even more finely, because you say that with Rektern we go down to, so to speak, 30 meters, and we then calculate our wave model internally maybe with only 100-meter resolution or something like that?

Rektern can actually calculate waves too, but the problem there is that the output is only possible at a regional level, which means I can know if there are good waves, but I can't see where they are. And for glider pilots, knowing exactly where the center of this lift, this wave, is, is obviously a huge help. Then they can fly directly there, and they can also piece that together from experience, but a forecast like that is naturally a huge help. And for paragliders, we've thought about for a long time how we want to incorporate that into our system, whether we want to make it available just for glider pilots or for everyone, and how we communicate it. A vertical wind forecast, like the wave forecast, could also be misleading, and with our communication, we've now tried to make it as clear as possible what it can and cannot do.

This vertical wind forecast can show on convective days—days with thermals—where there are good lines, where the air mass rises particularly well, or where alpine pumping kicks in. For example, I observed this last summer; in the Engadin, people know this: in the morning, you can fly well on the west side of the valley, and then the north wind comes in and basically washes over everything. You can see these kinds of effects very well in this vertical wind forecast; the same applies to the Abendtal in South Tyrol, where the thermal north actually washes over it in the afternoon, and you can see those kinds of effects. And I'm saying it again here: you can use this to learn something about the process. I don't look into it to see exactly at what point and at what time it washes over, because we're talking about very small-scale processes that can only be partially resolved.

The same applies; you can see the Montana Flush in Valais on this vertical wind forecast, or you can see the Grimselschlange, how it's so to speak being pushed down from the Grimsel Pass. And for things like that, it's naturally very, very exciting for paragliders. The wave forecast—so when there are waves—is of course also a sign, a warning sign for us paragliders not to go flying if there's strong wave activity; it's Föhn. And the waves are really that feared turbulence that pushes you up and down and is also very intermittent, meaning it changes very quickly. And the wave we're displaying is a hourly average. It might look quite calm there. You could think it's rising there, sinking there, and nothing is happening here for me.

But that's of course an hourly average. And the conditions, the wind gusts and so on, we know those from the Föhn. And you shouldn't mix those up. And I was surprised that some pilots, even very good pilots, wrote to us and said they really enjoy the wave forecast. And from what I understand, there are just a lot of nerds in our community. And of course that makes me happy, because as a nerd myself, I also enjoyed seeing the vertical component of the wind visualized. And of course, it can also be a benefit simply out of curiosity and from learning how the atmosphere works.

If a non-nerd paraglider reads somewhere, from whatever source, DRV weather or something else, that it could be foehnish or that a foehn flow exists—so they have this little warning in the background but don't quite know how to assess it—could it be helpful to say, for example, I'm looking at this wave forecast and if I see that no heavy waves are predicted, then this foehn situation can't be that intense? Or would that be a false impression, because you'd say, no, you can't just make it that simple.

Absolutely wrong. The shallow Föhn, for example, was already a topic in your podcast, and it doesn't work exactly like that. There isn't strong wind shooting over the mountains and the inversions that then propagate the waves; instead, it's actually a shallow overflow of a colder air mass from one alpine side to the other over the passes. And that doesn't actually lead to waves, but more to—you can imagine it like water flowing over it. And yet, it can still cause very high wind speeds, especially at the valley floor. With shallow Föhns, you can sometimes fly over it quite well if you don't go down to the passes and into the valley bottoms where that's the case, but that has to be enjoyed with a lot of caution. And in these cases, I would expect the wave forecast to perhaps show negative vertical velocities in the pass areas.

So, you get a hint, but you also have to know that it's not just the waves that can be dangerous there.

I mean, in this wave forecast, you show—on one hand, you show wave wind forecasts for different altitudes, in absolute altitudes, which really go up to 5,000, 6,000, 7,000 meters or something like that. But you also have a forecast level in there where it says wind at 100 meters AGL, so above ground, and wind at 400 meters above ground. If I imagine now, your thing is relatively high resolution, there's the Brenner Pass, for example, or something else, then I should actually—I would have to—be able to see a clear subsidence zone in such a forecast if that 100 meters above ground, for example, sinks back down as a cold air mass behind the Brenner Pass.

Actually, in the wider valleys like the Brenner Pass, in the important valleys, in the main elevation corridors, you can see that, of course. But then we're back to the point of horizontal resolution, and that only applies to the larger valleys, whereas for the smaller valleys, you have to look at it with a filter—that same problem with the "Kalen" again. When I look at the Brenner and the Inn Valley, I see the Foehn pretty soon, but if I go into the Ötztal, I need better resolutions to see it there.

That means I can only take something like that as a warning, but not as a definitive statement of what the situation is.

Absolutely. And here we're only talking about the horizontal resolution uncertainty, and as I said, this chaotic system that is the atmosphere is still on top of that. That means even if the model represents a process perfectly, which might be almost the case with a wave, it can still be that the initial conditions weren't quite right, that the conditions change again, that the Föhn breakthrough comes earlier, and so on.

Let's change the subject a bit. What kind of pilot are you, actually? I mean, we're staying with meteorology, but now I want to talk to you a bit as a pilot. What type of pilot are you in terms of meteorology? There are those people who look at the meteo, see the word "Föhn" somewhere, and immediately say, I'm staying on the ground. Without specifying a bit more whether it's strong or shallow, or high-reaching or whatever. For them, "Föhn" means old school. No, you don't fly there. Are you one of those who says, no, I want to experience what the weather actually does. I'll fly more often then, even if I know it might not be completely clean, but I want to see how it feels in reality and how that reality fits with whatever was in the forecasts.

Actually, I'd say I've become more of a fair-weather - and hammer-day pilot by now.

So you look at the

...date, and there has to be a nice big PFD somewhere on the route where you might want to fly.

That I'd take the whole day for it and really put in the effort to drive somewhere, yeah. I'm lucky, I live in the mountains, so I live in Davos and I'm also in South Tyrol every now and then, and I'm lucky to be in an area where it's flyable very often. That means if I just feel like getting up in the air for a bit and circling some thermals, I can do that relatively without much effort. But to take a whole day and say, today I'm going on a cross-country flight, I need a few conditions to be right for me. I don't feel comfortable in the air when it's windy, and my wind tolerance, I'd say, is pretty low when I compare myself to colleagues who fly for similar lengths of time or fly similar distances.

And for me, it just works that way, that I say I'll go flying when I know the large-scale winds are weak. Actually, I'm a pilot who doesn't try to fly as far or as fast as possible. I like being in the air for a long time. That means I fly relatively defensively when I fly further. It also means that I rarely come into contact with valley wind systems because I try to stay as far away as possible and fly in the furthest mountain gardens. And that's why it doesn't matter that much. So days with strong valley winds don't bother me much, but days with strong meteorological wind do bother me more. And of course, there are still days where it's a bit on the edge—so actually low wind, but with such shallow lift or such overshooting and those kinds of effects.

And there, I actually rely on my knowledge of the terrain and my skills to make sure I don't fly into the wrong places. So of course, I've also been in the air in Valais, where there was a north tendency.

and I had to feel what it's like to shake and rattle afterwards. But basically, I'd describe myself as a defensive pilot. My goal isn't to break any records. My goal is actually to become an old pilot who can still gain a lot of experience, and I feel like I have the time for that.

and who can talk about his experiences for a long time.

That would be a nice perspective. But even though you say you want to be an old pilot who can tell great stories and share a lot of experience, have you ever had weather-related experiences where you thought, "I don't want to do that again"? That was really on the edge for me.

There were two situations. One was right after my training. I started paragliding as a 17-year-old, and looking back now, I have to say, I wouldn't recommend doing it that way again. You're obviously in a different stage of life, with a different willingness to take risks, and of course, it's a lot about trying things out, but it was a type of trial-and-error that, looking back at myself now, was very unstructured and also very uninformed. I did the training with the standard weather training I had, but it might be a funny story for the meteorologist to tell here: on my first thermals flight, I flew directly into a cumulus.

and I just let myself get carried away.

Unconsciously. You

You didn't say, as a meteorologist, I want to know what it's like inside a cloud. I'm going to fly right into one.

I had an A-glider, but the A-glider still climbs pretty well inside a cloud. I can still remember, I was super hyped about this thermal I finally managed to turn into and I went up—and it was a wide, nice thermal. Above me was a nice, but quite high cumulus that was pulling, and I kept turning happily, and at some point it starts to get a bit hazy. We all know that, right? And I was perfectly under the cumulus, of course, because I let myself be carried by the thermal and wasn't at the edge or anything. And the first thing that occurred to me wasn't "shit, I'm in danger," but the first thing that occurred to me was, "ah, the forest is moist, it's rising right now, now it's going to get even faster."

And it took a few seconds for me to realize, oh, but that's also dangerous, I shouldn't fly up there. Yeah, and then in my first thermal flight, before I had ever practiced spirals, I tried with ears, of course, but it's still sitting there at eight, nine meters per second in this cumulus, and then I used a spiral and a displacement out of the thermal, thinking to myself, can I get myself down from there, and then I flew out again. That was one. As far as I'm concerned, I've luckily never had that much risk again, I'd say. A second situation was in Zurich at Uetliberg; I went flying there and I had already been there with local pilots before.

and they told me, if no one is at the launch site, then something is wrong. I was alone at the launch site and I ignored that warning too, thinking, no, actually the wind isn't that strong, it's fine. And I took off and I'm telling you, it really... I had very little backup behind me, maybe two meters to the trees, it catapulted me vertically up, 200, 300 meters, really vertically, I was a speed bar and it just went up. That's dynamic lift there over the meadow, and luckily everything went well. Up top, of course, that wave or that soaring is weaker and I was able to fly out of it up there and land safely. But there were also a few incidents for me that naturally made me the pilot I am today, and why I say I definitely won't take certain risks anymore.

Did you have experiences like that where you'd say, "Ah yes, that's exactly why I want to study meteo or get so involved with it"? For me, it was like this: when I started flying—it was one of my first flights after getting my A-license—I was in Greifenburg, flying thermals, not into a cloud like you, but it was the first flight, the first really nice, longer thermal flight I had. It wasn't that long, maybe half an hour in the end or something. But it just kept going up, up, and up, and I was flying, and at some point, I'd had enough. At some point, when you're not that experienced yet, the adrenaline level rises faster, and then you think, "Okay, it's actually fine, it was a nice flight, now I want to feel the ground under my feet again," and then I wanted to go down, but it wouldn't go down anywhere.

It wasn't a day with any major cloud development, but it was a day where something that sometimes happens in Greifenburg occurred—a kind of convergence forms in the valley. Usually, in meteorology class, you learn that it goes up at the mountains and down in the middle of the valley, but I was flying in the middle of the valley and it wasn't going down; it was only going up. I didn't understand what was actually happening at all. Everything was safe, but it still took a long time until I finally got down, even with the ears down and everything. I couldn't do a steep spiral back then, like you did with your cloud, and when I was down... How did it suddenly happen when it gets dangerous? That might be it. In any case, when I finally got down, I said to myself, "Hey, Lucian, you felt completely at the mercy of the weather up there."

You didn't know what was happening at that moment, and that was the reason I said, no, I want to understand this better. That was actually the point why I taught myself all my weather knowledge autodidactically, which I have today. Certainly not nearly as deep as you as a trained meteorologist, but I do understand a fair amount of what's going on. Did you ever have something like that, where you'd say, hey, that was really the driving force?

I don't think it was a single event. I think I was already predisposed, as mentioned before, by my grandpa and all those traces of nature all around. But the people who know the Dolomites and the complex valley wind systems there, I think that was something. It wasn't one specific experience, but being at the landing field with the club, with the flight buddies, and the wind suddenly shifts and everyone just says, the weather does what it wants and no one can really go and say, that's the explanation for it. I still can't sometimes. It still happens now, but I think it already inspired me quite a bit to acquire deeper knowledge.

Understanding this complexity, these complex wind systems.

Now, you're someone who has delved very deeply into this, including the very small-scale weather that we as paragliders, of course, always have to deal with. To what extent are you, in principle, still an exotic in the broader meteorology scene? Because from what I gather, many meteorologists are experts in large-scale stuff with their big models and all that, but with these really small-scale processes we deal with and all the chaotic stuff that happens on the slope and such, they actually have no clue. Or very little clue.

Actually, you are an exotic, but also not, so there is a community, a scientific community, but let's just say the industry for this area is super small. We pilots are, of course, an audience; we are very interested in the weather, we want to understand it, we need good forecasts. Otherwise, most people don't care whether the valley wind flows into or out of the valley, and they don't care how high the base is, as long as the hikers can still see something at the summit. But in science, there is this community because it actually, even though people don't know it, has an impact, right? It has an impact on how well the models predict the base height.

Not because it's important for ordinary people to know where the base height is, but because it's important to know how a thunderstorm develops in the mountains, which is of course just as relevant for the safety of the general population. That's why there is science, that's why there is funding in science for these areas, and there are mountain meteorology communities, and there are also conferences for this where you can meet like-minded people, and I have to say, I don't feel so exotic there anymore. I'm an exotic from the industry. Most people there do research and develop the weather models, but in terms of interest in these processes, you feel very much at home. My master's thesis supervisor is Stefanie Westerhuis. I did my master's thesis with her at Meteo Schweiz, and she also combines these two aspects of practice—from paragliding—and the technical, scientific side. And there are many figures like that there, where the people who move in this small-scale mountain meteorology world are also such outdoor enthusiasts.

Well,

Earlier, relatively near the beginning, you mentioned that you sometimes provide feedback to the meteorological services based on the experience you gain with XC Therm—where you might say, for instance, that their model wasn't very good, or whatever the basis is. Maybe where you say our cloud base, which we reported, was completely different from what an Icon calculated, or something like that. How do they handle such feedback, and have you ever noticed that the large models are actually being tweaked in some way as a result of your feedback?

So of course we're a small fish in all of this. A weather model like that has a huge budget and also a huge user base. There are user meetings for the Icon models organized by the Deutscher Wetterdienst, which we attend, but we're more like listeners than we are here now. We're not the ones who would do active thermal reporting. And actually, it's more about personal contacts at the DWD, at MeteoSchweiz, and if I'm being completely realistic, the way they handle it is that they listen to it. Most of the time, it's stuff they already have on their radar.

and they often say, you guys are the ones looking at it much more frequently and we don't even know it. The model developers themselves then say, yeah, actually I don't know where exactly the biases are. The forecasters know that better, the ones who say it every day. And I wouldn't be able to promise right now that the Icon model will be changed because of our feedback. But we can make a small contribution to it, and I think we paragliders are attentive model users. We don't just look at point forecasts and say clouds or a cloud or raindrops, but we also like to look into more detail, and such feedback is of course very valuable for model developers at every level. And maybe I can also send out an invitation to all the listeners here to give feedback.

It's always helpful when it's positive. It's always helpful, even if it's negative. That helps us directly with the rector, and we can pass that along. Let's get to...

To wrap up, I have one more question. From all the knowledge you have, both regarding the paragliding scene and the meteo scene, as well as your general meteorological knowledge—where would you say the biggest gaps actually exist in the intersection of those? In other words, where should the paragliding scene engage more with meteorology? Where do you think the biggest knowledge gap actually lies, and if we could somehow bridge that, it would really help the scene a lot?

I don't know if I understood the question correctly. Where does the paragliding scene need to catch up to be better prepared in terms of meteorology?

Yeah, so where would you say, you probably know what paraglider pilots are told in the weather training and how simplified many things are, or a relatively simple or sometimes even outdated Foehn theory or something like that. Which areas would you say, hey, I'd most likely start there to improve something and somehow teach people weather a bit differently, so they can understand the processes better or what? Or which process would it be where you'd most likely say, that should get into people's heads?

I don't have any specific process I would mention right now. I think making decisions based on the weather is a major point. From my experience with pilot training, you learn very little about how to make a decision—how to get from the forecast to a decision? To go flying or not? Or where to go? And so on. And of course, the various tools available help there, but you still need that own model and your own understanding of the process. And that's why I would say, in the area of decision-making, that people should learn to visualize from the forecast to the decision. I imagine it now: based on the forecast, I'm sitting inside, it's dark in the evening, I imagine now,

how the sky will be, how the thermals will feel tomorrow, and so on. To bridge that gap from the forecast—from all these abstract and complicated numbers and maps—to reality, and I mean really the reality we experience, the subjective reality we experience while flying, to bridge that arc. And I think, on one hand, the tools help, but on the other hand, or are also necessary, to constantly keep learning or attend training sessions. Lucia, you're one of those who also does weather training. I think that definitely helps with this understanding of the process, having your own model in your head. That helps every pilot. We've been doing the Mountain Meteorologist project with Verena since this winter, and we also do courses there.

Verena Stoll, right? Exactly, Verena Stoll. I met her quite by chance at a natural hazards conference where my main employer was, and I met her there and we got along very well right away, and we said, let's do this together. We noticed we do a lot of similar things. On one hand, forecasts for expeditions, but also for races and projects here in the Alps. For example, we're doing a forecast for Olympic teams now, but also lectures and courses. And you can see that the demand is there. And I think that just helps to have this understanding of the process and not fall into the traps of these models and this false sense of security.

In which area would you most like to further your training? Meteorologically, where you say you still have a lot of room for improvement. I mean, for you, not for me.

So on a personal level, because we were just talking about Verena, I want to learn from her how to do what I call old-school, classic weather briefing. Using weather maps, the large-scale weather situation—to understand that even better. As you said, I retreated into the small-scale stuff very quickly, which I also enjoy a lot. But I don't want to neglect the big picture, because ultimately it always plays a role. I can only understand the small-scale stuff if I also understand the large-scale weather situation. Unless it's a high-pressure system in the summer, then I can ignore the large-scale situation relatively well. Yes. A high-pressure weather situation, I meant to say.

And on the other hand, but also the technical side, I think with the X system we are constantly evolving, and on a personal level, I'm interested in how far we can develop thermal forecasting towards even better hit rates. And of course, there are different approaches, like using flight data or machine learning. But that's all still the current state of our development and research. And we hope to be able to contribute something to the paragliding and glider scenes.

Could you imagine a system in, say, a few years—I don't know how long that is—but looking at the rise of AI and all the machine learning possibilities, that we eventually have something where, you also have live tracking, you could pull live tracking data, see how high the corresponding flights are. And where you could then, basically live, because AI models can calculate relatively quickly, say I could deliver updates on certain things live every hour throughout the day, based on what was just flown.

I don't really want to say too much about it, other than the fact that we are, of course, working on projects like that.

and that all the listeners should keep an eye out for your podcast, for example.

Let's just leave it at that for now. Then I'm personally curious to see what comes from you, from X-Therm, or whatever else comes out in the future, that makes paragliding for us paragliders—in terms of the weather—more fun on one hand, and above all safer and perhaps even more predictable in some ways on the other, for the people who really want to know which days it's actually worth going out to have the flying experience they're imagining. Very cool. Tobias, thanks a lot for sharing. Thanks, Lucian Haas, had fun.

In the show notes of the Gleitschirm-Blog Lu-Glidz, you'll find some further links to the topics discussed here. If you enjoyed this episode, tell your flying friends about it. Give it five stars on the podcast platform you use or write a corresponding comment on the Gleitschirm-Blog Lu-Glidz. Also, please become a supporter of my work. On www.Lu-Glidz.blogspot.com, you'll find all the necessary information and links under the "Support" section on how you can further provide a basis for this project with a freely chosen, large or small financial contribution. I want to say thank you to all the supporters. Take off early, land late, fly far, and keep an eye on the weather processes. See you soon, your Lucian.