 Good morning everyone and welcome to press conference four of EGU 23, which is the annual meeting of the European Geosciences Union. I'm Jillian D'Souza, EGU's media and communications officer, and I'll be your host of today's press conference, as well as I will be facilitating all of the media interactions that will be happening during this week. I will shortly introduce speakers for today, but before that I wanted to add that each of the press conferences will be recorded live and streamed for journalists who are also joining us virtually. And the press conference will have time for speakers to make the presentations one after the other, followed by a combined question and answer period at the end of the briefing. So if you're joining us virtually, I ask that you mute your mics through the press briefing and we will then have you unmuted during the Q&A round at the end. I'm now going to go ahead and introduce our esteemed speakers for today. So our press conference is titled early warning for extreme events, earthquakes, droughts, floods and livestock disease. Our participants for today are starting from my right, Marta Hahn from the Swiss seismological service ETH Zurich. Then we are joined by Pedro Lima Alenzar from the Technical University Berlin Institute for Oncology, Germany. And our virtual speaker, who's number three is Frederick Hutoff from HKV and University of 20 Netherlands. And finally, we have Paola Nassisi from the Central Euro Mediterranean. So I'm sorry I'm mispronouncing all of these words. Come via mentee, climatici, less Italy. So we are ready to begin and we will follow the order that I just announced our speakers in. We already have your slides queued in, you have the clicker if you would like to begin your presentation, we will first hear from Marta. So over to you. Thank you. So hello everyone. I don't know about you, but yesterday, when I was going back home from here I got totally wet. So in the morning today I made sure to check Vienna weather forecast. Now wouldn't it be awesome if we could do the same for earthquakes. Unfortunately, I will not now look into my crystal ball and tell you the next magnitude of that and that will happen then and there. This is currently not possible. However, I think there is still some things that we can explain and that are worth looking into. For example, as we are here in Vienna, you're probably not so worried that a big earthquake will strike us now but if we were in Chile or Japan. It will be a different situation, I guess. So on the European scale, this is exactly what we have. This spatial variation is in seismicity rates explained by the European seismic hazard model that gathered a lot of historical and physical data. In this visualization, we see the space varying like, let's say probability or likelihood that an earthquake will occur in different places. However, if a bigger earthquake did happen now, first stage responders would probably be interested in knowing is it safe to enter this building and rescue us and so on. So there is a time dependent component to this and I think it is worth modeling and looking into and this is not something that I invented. In the US and New Zealand. Here we see for Italy, and in Switzerland within the Swiss seismological service, a model has been developed and tested that issues these time dependent forecasts for Switzerland. In these images we see that after magnitude 4.7 near Basel, the probability of another earthquake above magnitude 2.5 occurring in that area during the next day is much higher than normally. So, I just want to state very clearly that the European model is not meant to overrule any forecast issued by the national and regional agencies they're still in charge of their respective areas it's just meant to provide something harmonized baseline, such as the hazard model that we saw before. That does in a time independent manner. So in order to make this model we need data on the past earthquakes. This data is gathered by many different agencies in many different areas in different ways. We need different methods to account for this when making our model, but we need to know what these differences are so what is the precision that we record earthquakes to how low magnitudes are we able to detect and so on. Fortunately, this data has been gathered within the hazard model development. So we are using their catalog and the expert solicitations for these data properties that I just mentioned. And having this data, we establish a model so we use epidemic so called epidemic type aftershock sequence models this is also not something that I invented. It's been around for a long time. And the main strengths of these models is in explaining the aftershock behavior. So we know that after big event has happened it triggers another events those event trigger their own aftershocks and so on. So in these aftershocks we know that they decrease in time and space as we move further from the main event, and but they increase in number as the magnitude of the main event increases so magnitude six will have far more aftershocks than magnitude three. So these models ethos models are agreed upon by the experts worldwide to be suitable for issuing these forecasts. There has been an initiative by a CD that's with seismological service to gather the experts opinions and experiences in good practices. There is a national earthquake forecasting. There is a survey result shown here that basically says says that there is an agreement within the expert community. These models models are okay for issuing the forecast. So here you don't have to worry about much. It's just to say that we model these dots that is our data with these lines. I will explain that the number of aftershocks decays in time and space, as we move further from the main event and increase in productivity. But here in this third plot, you can see that the line is a bit below the dots, which means that our model, which is represented by the line tends to under forecast the number of aftershocks of the large magnitude events. And that we are addressing that has been observed in literature also before. Also one important modification that we want to make to our model is to make the background seismicity consistent with the hazard model that I shown in the first slide. So one of our products that we want to have is the operational forecasting which means that it's regularly updated and available online. So, one example of how we could we could visualize this is shown here. The plot on the left displays the expected number of events or earthquakes per time and space unit. We see that it's spatially variant and it changes from the left to right. What happened in the meantime is magnitude six in Greece and there is an mistake on this slide this was in 2015. Not so recent. So, here maybe you cannot see that much because of course this event doesn't happen doesn't affect anything affect anything in Iceland. But if we zoom in to this area. We see that the expected number of events above magnitude 3.5 has increased significantly in the area around the main event during the day following that event. So in order to make these forecasts as accurate as possible. We are working on some modifications as I said improvements to the model. We also observed the sequence that happened. Unfortunately recently in February in Turkey and Syria. This has inspired let's say the further work on sequence specific updating of our model. There's a poster on that that will be displayed on Thursday. If you are interested. So, that's it. Here are the links to the abstract of this talk that's also happening tomorrow and the poster on Thursday and my email if you want to contact me. Thank you. And now we will hear from Pedro when your slides are ready to go. Hi everyone and thank you for having me that's kind of my first press conference so I'm a little bit more nervous, so please forgive me if any mistakes. This I would like to present to you today some of our latest research together with Professor if a patent in the eco hydrology and landscape assessment department of technical University Berlin. And what we're interested is in developing a framework that can define better extreme events and we're looking specifically to droughts. We're talking here about forecasting but to forecast some event we first need to know what is that we are forecasting what is actually defined as a drought, or any other kind of extreme events like flash droughts it's one of the things that I'll be talking today. So we are kind of giving a step back and looking into the landscape of research and definitions of flash droughts and and seeing what it's actually a drought. So what you're proposing is a new framework. And the idea is that it's new framework is a way to improve our communication to the society. It easily takes into account personality because when droughts happen also matters not only if they happen and how intense. We also interested in seeing what kinds of threats are related to each type of drought or extreme event. And also to be able to adapt this definitions and frameworks to new and emerging extreme events caused by climate change. So, initially, what we observed is that the usual way to identify and to define extreme events, such as droughts, is very data centered. So usually researchers, they observe an event, and from that it is built into a model and then derived impacts and maybe communication to society. There's a maybe. And this is very meaningful and relevant information because if we don't do that we have nothing, but there is maybe a better way to do. We observe that initially when looking into flash droughts which is a new topic in the drought community, where if you look at this graph here, we have each role showing a different method to identify the same type of event flash droughts and what we conclude from this graph is that they don't agree quite often they agree sometimes but sometimes they just fail. And so it begs the question, which one is correct or any of them are correct for all of them. So what we propose then to try to conciliate this data with events and impacts as inverted pyramid of priorities where we first communicate and dialogue with community with society to understand what are the impacts of extreme events for their different users. So we use this knowledge to build our models assess impacts and thresholds sorry and then assess what are actual events, and then communicate back to the society. Okay, so for those impacts for this perception of impacts, those are the kinds of events that should cause trouble or hazard. So, into a little bit of an example of how it looks like. I'm going to talk about flash drought, which is different, different from the conventional droughts that are large areas and long periods of time. This is a short lived drought that's mainly occurring in the top soil. So you have a rapid depletion of so much. So we using data from Brandenburg, Potsdam in northeast Germany, and also the crop of barley, which is one of the main produce in the region. We built also a dashboard that you can go there and assess how is the historical production of grains in the region. And we want to then identify or define flash drought, considering this region, Brandenburg and this use barley. And this, what we can have is, let's say a producer in Brandenburg expects to produce around 5.3 tons per hectare of barley in a specific in any year that's kind of the goal of the general average historically. Maybe you have one producer that is very cautious and sees that if he produces if they produce less than 10% or, I mean, they have a loss of 10% of produce only 90% of expected. This is already an impact. But maybe you have another one who maybe has a better insurance, or it's just more prone into going to risky situations and see that an impact for them is a 50% loss. So these results, I can point out, but this will then derive different events. So we have a flash drought for the risk prone producer and to the risk avoidance producer are different, and they also are different from the off the shelf methods that we have today when we are trying to identify flash droughts and I try to represent that in that graph. So overall, you have less events that are perceived only by that are perceived by the high damage acceptance, high risk acceptance, and if for the low risk acceptance, you have much more events than the high risk, and also then the off the shelf, which is kind of And what is important just to close my talk is that this off the shelf methods they often, as I said, are very data centered so they don't take into account. What is the lens use what are the local conditions of the soil vegetation, and, and also often they are used at the just enough estimated by specialists but they are not necessarily derived from any urgencies. So that's what I wanted to point out. And in summary, we have your proposal in your method that improves communication and improves our understanding of water droughts based on impacts, and the society demands. He is also a way to contact me, and thank you for your attention. Thank you Pedro. We will now hear from our virtual speaker so Frederick whenever you are ready. Yes. Hello. Thanks for having me. First of all, I'm going to try to share my screen. I believe that now worked. Is that correct. I'm waiting on looking at your screen. Not yet. Oh, one more button click. I believe. Yeah, there you go. Okay. All right. Well, thanks. Well, hello, my name is Frederick Rudolph. I'm a researcher and consultant in the area of water management and disaster risk reduction. On behalf of my colleagues at HKV and University of Twente in the Netherlands, I'm going to present some results on a global approach for hazard forecasting. Yeah, something that we've been working. So first, some background on early warning systems. It's been all over the news at the recent COP 27 in Egypt. It was emphasized that a cost effective way to reduce the impacts of natural disasters, specifically floods is to invest in early warning systems. And, well, these can then warn vulnerable populations in time and emergency measures can be carried out to harmful impacts. So in the action plan, early warning for all 3.1 billion US dollars has been pledged with about one third assigned to observation and forecasting to achieve global early warning services by 2027. Now this is only a few years away, but surely not impossible. But if you realize that technologies for such systems are readily available and are being used at many places around the world already. Now the big challenge is to make these work everywhere. And this is really where it becomes difficult. In many of the most vulnerable countries various attempts have already been made to introduce introduce highly advanced technologies, but still early warning systems have not become part of standard operational procedures and also many times have not really left a lasting impact. And sometimes the problem is that such systems are not well aligned with existing operational procedures, technical or local technical capacities available data, etc. So what we think that is needed is, or yeah what we think that is needed to make a real jump forward is to establish a basic approach that works well everywhere. Not only that, but also that functions on globally available data that is not difficult to understand, or to operate, and from which you can grow and expand. Now here's an outline of our proposed approach. We establish static hazard maps based on terrain characteristics that's that's on the left here. So where possible we supplement these with experience from recent local extreme events. So on the left you see a landslide susceptibility map that we constructed based on terrain slope, land cover soil type and elevation. And we constructed a similar static map for potential flooded areas based on elevation and drainage network and that map is not shown here. So next, in the middle figure, you see real time or forecasted precipitation values and rainfall from global meteorological models. And this rainfall is of course the driver of the hazard. In this case for landslides but it works similarly for floods. And this information is actually available globally for the next 10 days, the global forecasting system. It's not ours but it's available. And on the right, we combine the static hazard map from the left with the, let's say real time or forecasted expected rainfall from the middle and emphasize those areas where landslides or floods may occur because of the actual rainfall conditions. Now and then places where potential hazards and rainfall quantities are high. Here, a certain combined threshold can be exceeded and that leads them to emergency. So there's your, your early warning. Now this type of calculation is actually very simple and can be done quickly on a global level. Essentially what we do is we immediately translate rainfall to different hazards, and wherever threshold is exceeded, you can then issue alerts or warnings. To just stress again, the information for this is available everywhere globally and can do forecasts for the next 10 days. And I already mentioned that this can be done for landslides and floods, but similarly also droughts and forest fires could be addressed. Now of course there's some need for disclaimer, detailed processes such as overland routing and smaller flow obstructions in the terrain are not taken into account, but to get a first indication of oncoming dangers under very extreme conditions. This is surely better than what many of the most vulnerable countries have available right now. And then another key characteristic that I want to point out which works very well in this approach is that there are no artifacts in the hazard estimates from artificial boundaries. First of all, by using global data sets, political boundaries do not play any role in the hazard estimates as they shouldn't. Next, for the case of flood forecasting, it is very common that advanced flow routing models take on boundaries of river basins. Now this makes perfect sense for minor flood events, but there are many examples known were under extreme flood conditions, water from neighboring river basins become interconnected. The example shown here for the Likungu River in Mozambique, where in 2015 flood waters extended all along the coast, even towards the Zambezi River Basin. And in black here you see the Likungu River Basin boundaries. And here on the left, you see the actual satellite detected flood waters that occurred in 2015. Right here. You see our flood hazard prediction from global data. Now it shows that these agree very well. And the predicted flood extent would not have been possible with more advanced hydraulic flow models that are limited to these basin boundaries. Okay, then. Next I want to briefly show that our proposed approach can indeed also inform and help with early action. So here's an example for the Manambola watershed in Madagascar, where you can see that if you zoom in on the derived flood hazard map for individual communities. There's a quite distinct indication of vulnerabilities. In this image on the right in the blue areas you can see that certain parts of the community are actually affected by flood hazard, much more than others, and also that certain access roads. These are among the first crucial infrastructures to be affected. Now, early action of course has to take those aspects into account and focus on precisely those areas. Okay, then finally to summarize. So we propose a global early warning approach for various types of hazards based on readily available global data sets. We applied this to several test cases in Africa and Central America showing good results. The approach can easily be expanded to a global scale, because all the data is available, even allowing forecasts for the next 10 days. Next, based on the spatial extent of the hazards it's even possible to estimate affected populations or infrastructures or other assets present. So this still requires quite some work to get this done on a global scale, especially to define suitable alert levels for different places around the world. But we can confidently say that this is possible for a fraction of the pledge budget from COP 27. So my final message. So let's focus on developing such systems that work everywhere, and then gradually improve from there when better local information is available, and local capacity can carry it. Many attempts for early warning have failed in the past. And now let's get it right and make sure that early warning for all is indeed achieved within the next four years. Let's use what's available, what works already. That is robust and simple to set up and build upon. Thank you. And final comment I want to make. So obviously I'm not in Vienna at the moment, but I will be there in person on Friday. Also, thanks. Thank you Frederick. And now we will move on to our last speaker, Paula, we're just loading her slides in a couple of minutes. Thank you. I will introduce you. The main objectives of an early warning decision support system for disease outbreaks in the livestock sector. The livestock sector contributes substantially to the European economy with about 168 billion annually and the 45% of the total agricultural activity. Provide the diversity of production system with the 4 million of employees and 400 billions of European industries linked to the animal production. However, climate change with its variability and extreme events. The livestock sector in many aspects, ranging from animal well being production of milk and meat reproduction, but also diseases and their spread and food quality and availability. For example, very long period of high temperature combined with excess of humidity may affect negatively affect the animals because the perceived temperature would be very, very high. But also called extremes or extraordinary windy conditions could negatively affect food and the animal well being of course. In this context, the European funded project Sebastian aims to provide support for a more effective and sustainable management of the livestock sector in Italy. And in particular for cattle sheep and goat breeding. This is some numbers related to our project. We have a number of farms monitored in different ways and allow us to have a vast amount of data related to different different species of cattle sheep and goat. With about more than three millions of monetary animals that allow us to to to do analysis important analysis of this data with a lot of samples gathered. So the aim of the other project is to put together weather and climate data sets with territorial aspects, for example, the presence of vegetation around the farms. The slope of the soil, but also animal welfare indicators and sensor based data coming from IOT devices, sorry, installed on the animals to detect environmental temperature and humidity around the animals. The position of course, or it and the movement, but also the skin temperature. And these data are collected and summarized through advanced techniques using statistical indicators or machine learning algorithms. And to provide the user tailored information for the stakeholders for breeders, but also for researchers, policy makers, but also companies interested in this sector. And the main goal is, of course, to promptly notify breeders. When ads are those conditions occur. Now, just a brief video that nicely nicely summarized the main objective of the project. The production and products of animal origin make up 45% of the EU agricultural annual back. The livestock sector accounts for 36% of the total agricultural production. And this changes having a negative impact on livestock management. Changes are more valuable than extreme, and it is important to anticipate and mitigate the effects for employing suitable farming practices and planning to update the ever changing environmental conditions. In this context, the main goal of the Sebastian project is to deliver the support system for long and short term decision making. Sebastian is a tool and service work platform, gathering a huge amount of data about Italy, that is integrated, I'm on my sense, some time, thanks to the difficult indicators and machine learning algorithms. The aim of Sebastian is to cross them access and to use existing data indicators and tools as well as creating new ones. The other is geofacial material to pull up observational and environment data and statistics, new and reuse are the bread and butter of Sebastian bottom. Sensors can more is a well being in real time. Climate simulations and future projections over Italy are responded as a very kind geographical revolution that my data detector and remind different features of the vegetation structure. I want to make sure that you don't want the context. She goes for us, researchers, educators and features. I see deep companies and many more involved in the creation and testing of the Sebastian platform services. The main goal is a single point access to data services and informative content for you. So using livestock farming and free vegetation for environmental conditions and production units. Learning about approaching or projecting dangerous environmental circumstances. Data that are really seen throughout past years, revising updated risk maps and parasites, and he's asking us not last long time management in a challenging environment. Okay. Thank you. Thank you for that. Very insightful presentation. I will move to the next part of our press conference, which is the question and answer round. So I open the floor to questions from journalists both in the room and those who have joined us online. If you have a question and you're in the room, I will hand over the mic to you. Please introduce yourself and let us know what your question is. If you are in the room, you have, if you are joining us virtually, you have two options. If you have a question in the chat or you can use the hand raising function on zoom, and we will come to you for your question. So over to you now. Do we have any questions coming in. Okay, we have one already. Hi, I'm Nicholas I'm a journalism students in the University of Paris and thank you for your presentations. So I have a ready, like general question for all of you. There's more and more cooperation between people working on different hazards as droughts or floods etc since some of them are quite really related like drugs can really like to heavier range for example in the Mediterranean so I was wondering whether some different hazards may come to cooperate a little bit more since the models are getting more and more complex, or is it so much special specialized that you don't even have the time to cooperate or to create like common models with other researchers. That question makes sense of course I can answer. I think that. Yes, it's, it is quite difficult to operate, but we have to because these topics are strictly related to each other. I was thinking to like the livestock sector, for example, it is that depends on droughts presented by Pedro before. And I think that the European Commission is investing to create more cooperation between these topics, and also between the infrastructure that provides support to this sector so yes, I think that we are going toward a more integrated solutions to support this diversity. Yeah, I mean, maybe I can just add on my colleagues and so there is indeed a pressing necessity to more cooperation, especially in the also when you not so new but the compound event realm of extreme weather when you have either fires or followed like sequential events, and these are just a cooperation for instance I work more with droughts, and maybe you have forest fires developed and after forest fires you might have storms that lead to higher land degradation, and then you have three different disciplines that have to cooperate together. So, yeah, I think that's what we are all walking towards but yes it's challenging with the current requisites of funding and all that. If you have any more questions. Okay one more. Hi. So my name is Caitrin and I'm with the EG press office. I'm not sure if this would be a question for Pedro or for someone else but I was very interested in the section of your presentation that talked about how inaccurate impact assessments can affect the accuracy of modeling. And I was wondering if this could also apply to other kinds of hazards like flooding, flood modeling. Well, it's a very good question. Yeah, the, what happened I was actually just having this conversation this morning was a convenient session earlier, and a colleague from the US talking about how farmers in a specific pixel from the US drought monitor have problems because that pixel is kind of unique, and it's not well represented by all the other drought definitions. So they often have hindered expectations and get frustrated because the drought monitor doesn't flag the regional impacts in droughts but then they go to their fields and see the vegetation suffering. And that's why we need to look at actually at least my point of view that we should first look at the impacts. And I mean that's not uniquely my view it's something that the community is being paying more and more attention. And I think that's in the floods. I will more advanced and well played than in droughts. Actually, some of the techniques that I've been trying to implement. I inspire from flood analysis. Thank you. I hope that answers your question. Okay, great. Do we have any more questions? I have questions from any of our virtual attendees. Okay, so if we have no more questions then we are ready to conclude our press briefing for today. Thank you once again to our speakers and even to Frederick thank you for joining us virtually and like he mentioned he is going to be in Vienna as well so if we have any questions for him interviews for our virtual attendees as well as in person. Please reach out to me and we'll be happy to facilitate this entire press conference is going to be recorded and will be uploaded to our YouTube channel later today. Thank you once again and I encourage you to check out our press packs, the printer and digital to know what our next press conferences are we're done for the press briefings of today but we have some exciting ones lined up till Thursday so to tomorrow and on Thursday. Thank you so much and have a good day.