 Okay, good afternoon and good morning for some of you. We are just starting these units in Persian to pose with geo sciences. We will present in detail the cook on dinner at the end of the presentation. And for the present time we will start as much as soon as possible with the first talk. And the general idea is that we have this session because the pandemic is a major trauma for humanity. And naturally call for scientific base respond to it and to mitigate the risk and big residence. But at the same time we have also longer view about this problem, because it will be not presumably the last one. And therefore, we want to put geo sciences in a post covid perspective. Just to give an idea of the importance of the problem. And this is a copy of the dashboard of the Center for Systems Sciences and Engineering at the John Don Keynes University, which will be discussed. And there is, of course, a terrible figure of yesterday over three million of deaths across the world. The first speaker is Dr Judy Omumbo, who is a specialist in climate and Victor born diseases. She's from the African Academy of Sciences, based in Nairobi, Kenya. And she serve on several national, regional and global committees. In particular, on the joint covid 19 tasks team of the WMO, well, mental organization and W H O, what is organization. She's a co chair of this task team. And she will present today is the first report of this task team, which be on the methodical and air quality factor affecting the covid 19 pandemic. I think I'm just in time for the five minutes introduction. And please, Judy, the floor is for you. Thank you very much for that introduction. So has we all aware covid is continuing continuing to ravage the world and of course there are questions around being a viral infection, whether there are any signals that we can see in the in the in the metrological or air quality factors that may be impacting, may be impacting the epidemic. So in September, 2020, the WMO Research Board constituted an international multiple disciplinary tree that was established, essentially to provide relevant knowledge to decision making communities on the met and their quality factors that may be affecting the dynamics of this pandemic. So basically, it was to look into what we know of the impacts, possibly what we should be monitoring. And for the decision communities which are very, you know, really important in this pandemic. What could they possibly do, you know, how can how can data from from the Met Services and from the design climate community assist the response. So here's a multidisciplinary team very ably led by Ben Zatech, who's on this webinar. And we're doing quite a few things since since that time, as you can see that from many, many institutions and many, many parts of the of the globe, and also very strongly supported by a good team at the secretariat at the joint problem program at the WMO and WHO based in Geneva. So I'll go straight away into some of the learnings that have come out of it, the achievements of the of the task team, just focusing on those outputs. So right at the beginning, it was really important to engage the scientific stakeholders and that process began with an online conference. We supported an online conference in August last year, I think it was August last year, yes, and about more than 400 scientists or stakeholders, decision makers attended and participate in that conference. This was sort of just to launch the exercise and to just inform the community of what the task team was doing. And then were around that time as well, there was a submission, we asked them to submit any relevant research and the team also collected any relevant research, and we had 100 plus submissions of research around around COVID. The exercise led at the end to the publication of recommendations for good practice, and that publication is available there there's a link in the slides, or if you visit the WMO public website, you can find a link to it. Now the aim of the task was of course as I said before is to provide policy relevant science. So the task team was put together a list of key questions. And the community to be interested in answering. And we followed with a process of distillation of these questions, and out of these questions 16 questions informed the priorities, really fall for MAC factors, and in COVID. And we were able to to demonstrate a way of a good practice in engaging end users to produce the main statement of priorities. There are examples there you know things like through what are the mechanisms of what can we tell about COVID. MAQ is a meteorological and air quality factors from our knowledge of climate drivers, really looking at what we know about other other diseases that are transmitted the same way viral infections. And how much can we tell from the data and the research that's been done around of possible impacts of these factors. The next program a public next process was the publication on on good practice for for interdisciplinary research. This is really learning from the from the process and it's published in nature communications last November. The topic in here summarizes the process that we use for for for developing this framework for the good practice, looking at data doing an analysis interpret interpreting it, and really also focusing on the communication which is the very important aspect of it. So the main report has finally was finally published in February, February this year, and it provides a literature review of all the research that we had an overview of the current understanding of Mac Mac influences on COVID 19 pandemic, and also the best practices that we that we learned. It's taken quite a while validation of the report was done in September drafted towards the end of the year, went out for open review was presented a webinar for example in Beijing, the virtual webinar that was held in January, and then finally published in February. So really comprehensive process of seeking peer review for it. What did we learn. So I'll focus on some of the outputs of the learning that you had from that report. So, if you look on the first, the first column on the right hand panel my my left hand panel sorry. These are the main factors that we considered both from the history and also the understanding of epidemiology of viral infections, temperature humidity, possible effects of ultraviolet radiation, and also quaint air quality. There are multiple mechanisms that these factors may be impacting the, the virus viability, the host immunity that's the immunity of the human population, and also human behavior, which is also related, possibly related to to risk of being infected. So, what we found is, what we find is that the temperature may have I mean this is these are all possible strong influence from temp from temperature possible strong influence from humidity. So, radiation. Yes, particularly in around around for host immunity, and also considerations for crowding indoors due to heat or cold and precipitation. And then, air quality factors which are very closely related to, to human behavior. It's like, in many countries we find that in cities where the air quality may be lower they tend to be more, more infections. Now this is all a lot of it is based on anecdotal evidence but there's quite there is some science that have been found related to that. So what are the key findings. The results are conflicting. There are a lot of consented confounding factors which are related to to human immunity and also to population factors behavioral issues transition transmission seems to be more likely controlled by interventions that that governments are putting in, because we tend to see a lot of changes in transmission levels depending on what sort of interventions are in terms of lockdowns. Social distancing that governments have put into place. And it's difficult to tease out clearly what the max signatures are. So one of the conditions that the COVID transmission transmission is possibly seasonal, based on studies of similar, similar viral infections and the dynamics of those. It's again difficult to tell we've only had really one and a little bit longer seasonal cycle of the pandemic so it's not really clear to really understand that that that seasonal cycle. But we do, but this seems to be what we are seeing, you know, many countries are seeing a seasonal uptake again at the beginning of the year. So additional key findings that include the fact that the impacts that drive season seasonality. They appear to be those that directly affect viral survival. So to kill the virus, then you're able to to control how it spreads, and also human resistance to infection and seasonal changes in human behavior. Lab studies have also shown that the virus survives longer when it's cold and dry, and when they're low levels or ultraviolet radiation. And it's not clear. You know, these are based on lab studies, of course, so it's not clear whether this is also true the real situation is the situation in real life. And then there's some indication that air pollution may worsen symptoms and increase the risk of mortality from the COVID virus. There's no actual peer real peer reviewed evidence of this though. So I think one of the key summaries is that there's still a lead a need for a lot more modeling. I'm going to get some of the recommendations of this report. There's some indications that the transmission is seasonal seasonal and it suggests that monitoring monitoring seasonality will become important, and perhaps useful. Some additional research needs to be required to expand the data on this, particularly in investigating also the compounding factors. The data record will it will improve the prediction skill and improve also process based modeling attempt. Additional peer reviewed research is critical, and very importantly, clear and improved communication between research and the critical decision making stakeholders. So where are we now in the future what we'd like to do. Continue looking at the evidence it's still insufficient as we continue to we need to continue update knowledge and monitor monitor the pandemic, and also engage this decision making communities more. I think the task team is really looking at how to to put together a lot more public faced meetings. And then really focus on sharing the information. I think over the past over the first 12 months, the strong, you know, the strong scientific evidence that supports environmental environmental conditions as drivers and transmission. I think that the work of the task, the task force is just is just really beginning to look at that evidence and continue collating it to make a better reassessment, which should we are hoping to do in June. So thank you very much. Thank you very much, you will be able to summarize a long report in a short time. Therefore, we have more time for discussion now. I'm sure there will be several questions, including from the co-combiners. You can see that Benjamin is thinking a lot about this report, and you may have some question about that. I certainly do but I think that Jacques has his hand raised so perhaps Jacques would like to lead off. Well, while Jacques works on the live stream conflict, I could ask a quick question of the speakers a follow up because I know this was Judy as you well know this is something that's been discussed a lot by people who have responded to the report, which is about understanding these sensitivities in geographic context. And so there's obviously a lot of work on Europe, the United States, Japan to some extent these temperate countries, right, very data rich environments. And I think that's a little on your perspective on what some of the big challenges or perhaps opportunities are for improving our understanding of the climatic sensitivity in, for example, Sub-Saharan Africa and other other tropical regions. I think for Sub-Saharan Africa, it's important to, it's an opportunity because I think the seasons are a lot more defined. There's not that much variability. It's warm, it rains, as you know, the seasons are not as complicated as they are in temperate countries. But the big challenge is that there's not much data. There's not much data. So there's a need to sort of think about that and work with decision communities to put the data together. Particularly now that the vaccination programs are being rolled out, let's understand. I mean, we have several countries where we still haven't really seen that wave of different variants coming in. So where we can look at those, you know, those areas where the transmission is relatively simpler and it's relatively more like it was at the beginning. I can think of my own country here in Kenya where we are beginning to see variants coming in but there's still quite a large population where there's just one. There's also opportunities in Africa to look at that because also genotyping is being done across the continent and just really working with decision-making communities, particularly those who provide climate information to work with the health sector and start teasing apart what the seasonality is. Julie, I got a question from Jacques who's chat box. So his question is the following, UV8 and dry in activity in the virus, but what about the risk to generate variants under these bad conditions for the virus? Well, Jacques, that's a difficult question. I mean, I think, first of all, we don't know what sort of extent of UV radiation would actually limit the virus. And in the lab, the sort of conditions that levels of conditions, levels of UV radiation to kill the virus, I mean, we don't normally experience that in the human population. I think this is an area that just needs to be studied. I don't know whether Ben has any comment on that. Not beyond what you said, I think that's a great, great answer. If there is no question, I will ask a general question. This report is mostly based on a large review of literature, scientific literature at a given time, of course. Can you give an idea of the amount of literature which has been searched through? So 100 plus, about 100 plus. And I think the literature is still coming in. Yeah. But remember, this is the only, only the peer review literature. I think there's still quite a lot that's in the realm of grey literature that it's difficult to report on because it's not peer reviewed yet. Okay. Is there any other question? Okay, probably I would like to have it. When you make such kind of seasonal variation on the weather effect study, how about the data quality? Because the data quality is different in the beginning of the pandemic and also different region. So it could be difficult to separate this, some changes due to data quality change or due to the real change. So, thank you, Mata Toshi. Are you asking about the clinical data or the climate data? The infection data and other things. Okay. This is a perennial problem for infectious diseases. There's no way of assuring that it's good quality data. I think many people just rely on what's been provided by the Ministry of Health. So that is very country dependent. There's absolutely no way of assuring that it's this quality. I think as these scientific community, all we can do is provide guidelines and then hope that the communities pick those guidelines up and use them. It's hard to say, from my perspective as a health person, that the quality of data changed at all. I mean, it's either positive cases or negative cases. They tend to be collected from those who have sought treatment in hospitals or have gone in for a reason because they have symptoms. The data, none of the data right through the pandemic have been representative of healthy communities or the public. Are there any normalization projects within the WMO? If we can normalize the data for some way so that the many different scientists can use the same data set, that will help in understanding more. Are there any such kind of projects? Well, certainly in theory one can do that, but it would mean making countries do field studies. They have to go to field studies, select random sample populations that in different settings and have standards that are applied across the globe. Practically, it doesn't work. It hasn't worked for other diseases and it's unlikely to work for something that is affecting governments so much now that they don't have the bandwidth to really be doing field studies. Thank you. So the question asked by Julia with respect to impact for COVID in the cities, how to distinguish the impact of air quality from the impact of from a higher density of population. Was that question for me? Yes, again, again, the only way to distinguish is to have data from those different settings. If you look at data in the crowded areas, we have to have good air quality data, which I think outside really big cities in temperate countries, there are not much, there's not much data for that. And the fact that populations are so mobile, I mean, you have to have the air quality data taken from the site of infection for each individual and that is an almost impossible thing to measure because people are moving all the time. We'll be looking at air quality data maybe in homes. Impossible task. Okay, thank you very much Judy and also the people who asked question. I think we will discuss more about the problem which has been evoked question of data and so on, during the general discussion at the end of the session. So now we can move to the second speaker. Thank you. Okay, thank you again. So Theo is the Emeritus Director of the Max Planck Institute for dynamics and self organization in Göttingen is also professor of theoretical physics at this same university. He has been received different prizes, including the prestigious the Godfrey Wilhelm Leibniz one and he's member of the Academy of Sciences and Humanities of Göttingen and fellow of the American physical society. He's very well known for his research on non-linear dynamics and chaotic system. I met him 20 years ago due to that. And especially about the discovery of Levy works for people who are familiar to Levy flight it's quite different. And he has been working many different physics field like quantum chaos and nanostructure and the spread of epidemic and theoretical neuroscience. And his presentation is on top of epidemic and human mobility. So this time I'm giving the back to screen. Yo, it's a pleasure to be at this meeting meeting, meeting colleagues again and giving a talk at this human session. And thanks, Daniel for your introduction. Actually, I can add that I've just submitted a manuscript on the swing feeling jazz some kind of psychophysics experiment. But that's not the topic here. But it was a lot of fun. It's become commonplace that human mobility plays a role for the spreading of epidemics. And so, but how, what, how does it play role. What does it help them know about human mobility for the focus of epidemics. There are cases where it actually does not play a role. So, for infectious diseases that are transmitted by so called vectors, where the viruses are transmitted for instance by these others mosquitoes. So like the tiger mosquito here or the Egyptian mosquito, the yellow fever mosquito. And they transmit diseases like the yellow fever with anger fever and she can go on and others. Now, here, the human mobility doesn't play much of a role, because you can only infect yourself in areas where these beasts live. As you travel there, you will not catch the disease. But that's different for the coronavirus epidemics. So in, they have shown that the severity of epidemics can increase drastically as they become transmissible between humans. So in the case of the different coronaviruses, the, the original SARS coronavirus 2003 Merce and SARS-CoV-2. In all these cases, we know it's very likely that the viruses stem from reservoir of pets. They can be transmitted to other animals like the civets, the promenary and the pangolin, very likely. And at the moment when they could be transmitted to humans and between humans, this became a game changer and the disease can, can potentially develop in a pandemic as we have seen. Now, the virus can use our human traffic networks for its spreading. And therefore it's important to understand the statistical properties of human traffic in order to forecast the spreading of an epidemic. How the transmission and the spreading are usually imagined is depicted here. So in terms of overlapping home ranges, so each person has a home range overlapping with others. And so the virus can spread either if somebody brings it to the home range, home range of somebody else, or if this somebody else goes to another home range and picks it up there. But this picture was good for the description of historical epidemics like like that in the 14th century in Europe, where the spreading is mainly in terms of wave fronts reminiscent of reaction diffusion equations. Sorry. These wave fronts that you see here. But what are the home ranges in our modern world, of course, they are different. As shown here. It's our modern home range based on the aviation network. Color coded light color for high intensity. So it shows that you can reach almost every spot on the globe within a couple of days. So, our modern home range is more or less the entire globe, but we can refine this picture of course, not every spot can be reached. And so the virus will not spread equally fast to any spot. And so this is part of my topic today. First, I will talk about the role of the aviation network for SARS-CoV-1 and SARS-CoV-2. And then in the second part of my talk as the question, can we assess human mobility by using proxies. We mentioned my co-authors, Jack Brockman, he's a professor at Tumboldt University and the Robert Koch Institute in Berlin and Lars Hofnagel, who is at the EMBL in Heidelberg. The original SARS coronavirus disease started in the Chinese province of Guangdong. The virus arrived in Hong Kong and from there it could spread to more than 30 countries in the world. And it was understood, of course, very early that this was due to spreading on the aviation network. Now, therefore, let us look in more detail at this aviation network. Again, light colors represent intense passenger transport and dark colors, little transport. First, let's look at structural properties of the network. And well, maybe the most important one is the connectivity of the network. Because the distribution of the number of connected nodes from 1 to 200. And what you can see is there is a strong variability in this connectivity. There are a few nodes, a few airports that have many connections. And there are, sorry, there are a few that have many connections. And there are many that have only few connections. Now, this is a variability in the form of an inverse power draw. I would say this is like a scale-free distribution. But of course, it's limited on the left and on the right. So you only have a few, two degrees of, two hours of magnitude for this scale-freeness. Here are some examples. Some examples, Frankfurt and Chicago. Frankfurt has many connections, as you can see here. And Chicago, oops, sorry, Chicago has comparable capacity, a number of passengers, not many more, but has much less connections. Now, with this aviation network, you can model the spreading of epidemic, for instance, an agent-based simulation assuming such a chain of Markov chain with waiting time distributions for the infection to onset of the disease and the waiting time distribution for the time between onset to admission to hospital and so on. So if you do this as shown here, then you can forecast the spreading of the virus, which is shown here in B down here, the simulation after 90 days and compared to the WHO reports also after 90 days. And this is color coded according to the number of cases that showed up in a given country. And you see, it is relatively accurate. I mean, of course, there are fluctuations, but it is accurate in the sense that you can predict which countries will be affected most and which countries will not be affected as much in an early stage of the pandemic. Now back to SARS-CoV-2. One can predict the most probable aviation spreading routes. And this was done by Dirk starting in Wuhan, as we know, you can also use other starting points. Using the aviation network predict the most probable routes of spreading that is on which airline connections to which airports and from there to which other airports. Now, one can transform this network into a network where you have an effective distance between the different airports. The effective distance also depends on the strength of the connection. Now using this, one can, and then this is shown here, the effective distance is shown along the vertical axis. And this is also related to the arrival time of the disease at another metropolitan area. And that's shown here for different continents and different colors. So green is the US, the Americas, and Asia is violet, Europe is orange. What you can read off here already in this image is that the largest, the areas with large airports are affected first. So, Lex, Frankfurt, JFK, Charles de Gaulle, Daniel. So, and then the small airports are affected much later. And this is also something that we've observed already in in SARS-CoV-1, where we quantified this effect. We're going at the largest, the 10% largest airports and the 90% other airports. And this is shown here in infection time in days. And what you can see here is that the largest airports spread the infection much earlier than the smaller airports. So, large cities trigger infections and this can help us also in the development of vaccination strategies. Okay, but modern travel occurs on all length scales, not only on the aviation network, but also by cars, by trains, etc. So, how do we get statistics on the human mobility that also uses these means of transportation? Well, there used to be data by the American National Bureau of Transportation Statistics. And formally, they just could ask people, if they traveled from A to B, where do you come from, where do you go to? And of course, this led only to a very poor statistics that wouldn't help us much. And therefore we're looking for other means of getting more data. And so, can we assess the statistics of human travel also on smaller scales? And we were the first who pioneered the use of proxies for that purpose. And how we did it? Well, follow the money. This is here a dollar bill. The dollar bills have the serial number. The dollar bill also has this stamp track me at versegeorge.com. This is a website. Here you see other such dollar bills, where people have registered these dollar bills with their serial number that you can enter here. Also the zip code where you've got this dollar bill. And this way, this database has more than 300 million dollar bills, meanwhile. Well, we use this database to extract the trajectories of dollar bills and of people that transported them. So we have the reports of the position at some time T1 and the report of the position of the dollar bill at some later time. We had the total number of trajectories of 11 million. We didn't use all of them. But I'll come to this. We can do two things with these data. One thing is we can use it for agent based simulations. I mean we have the travel statistics between all counties of the US. And we can use this for just for simulations. And we've done this, for instance, for the swine flu predicting the spreading of the swine flu. But here I want to look at some other properties I actually want to study the mathematical properties of, of this human traveling. And we used in cases where we do not have such data available, but we still have the mathematical equations. So, we can look at, we looked at to get the trajectories we only looked at events where dollar bill was reported twice within three days. And that way we could be sure that it didn't travel twice between the two reports. And this shows the length distribution in kilometers log log. t is an inverse power law, and the exponent is minus 1.6 for this distribution of length distribution of jump length. So in another distribution, we notice that the probability of being reported at the same site once again, after a time t is also an inverse inverse power law. So this survival probability also indicates the inverse power law with an exponent minus 0.6. And based on this, I think I should hurry up a little bit. And once we, we build a stochastic model with with jump length delta x and distribution for them and waiting times delta t and their distributions with these exponents alpha and beta. And let me get to the final result here. From this model, but maybe walk model and continuous time random walk description. Let us to a simple diffusion law. But by fractional diffusion law with fractional derivative with an exponent alpha. And this is what some kind of generalized diffusion coefficient, and also a fractional derivative with respect to possession with an exponent beta. Okay, now, and this could actually be confirmed by a data collapse here for this typical scale free solutions of this diffusion laws. And this is confirmed for the whole data set of the 11 million protectors we had at that time. Now let me switch to more other proxies that one can use of course meanwhile we use mobile phones as proxies, although in Germany it's not so easy. The first one to manage to get such data was Barabasi and his group Martha Gonzalez and Dalgo. And you can get other information from cell phones as proxies for instance, on the number of positions that an individual travels. And that's shown here, number of different positions traveled goes up to 50 even but only in a very few cases. And, and this is the probability that an individual travels to one up to 100 positions. Now, finally, back to COVID-19. Of course now it makes no sense anymore to use the aviation network or the use mobility for the break the spreading now that it's prevalent in most countries and but cell phones can help us to extract information on your mobility. For instance, Duke again has determined the changing mobility in Germany based on cell phone data, compared to 2019. And what you see here over a year is that a year ago in March, the mobility drop considerably by 40%, but then get back slowly again in summer even exceeded. And now it's, it didn't get back in Germany to these money minus 40%. So this tells us somewhat about a little bit about the compliance of people. And another very actual topical information is contained in this data set, because as you might know in Germany. In our parliament in a long fight, a long political fight among many idiots. But finally, our parliament decided about a lockdown measures this national lockdown measures this week. In particular, there will be now a curfew among these lockdowns a curfew between 22 hours and 5am in the morning. And look at the mobility average over three weeks in March in Germany by daytime. So over the hours. And what you can see is that 6464% 64 million so 9092% of the mobility happens during daytime, and only 7.4% during nighttime. And this tells us that a curfew will not reduce the mobility much. If it only happens at night. But so you could wonder whether it will have much of an effect, but I, I still believe that the curfew will have indirect effects by suppressing parties or reducing the number of parties. I think my time is over I should stop here. I will skip this this is a recent unpublished work of ours on open populations andemic diseases in refugee transit camps and nursing homes and so on. I think there's no time to go into details of this. And I resume understanding human mobility is key for understanding and containing the epidemic spreading in the early phases and enormously tracking proxies banknotes and mobile phones can provide important information. Thank you for your attention. Thank you very much. We have few minutes to three minutes. I'm sorry. It's fine. Is there any quick question. There is a question from the question. Good. Asking that where this mobile phone data come from is that from the company or that just he want to ask the question. I can hardly hear you but so the mobile phone. Nowadays, different people use mobile phone data when he started this, he kept it secret. It was just too, too dangerous. He actually got criticized a lot in forums at the time. He was kind of at the boundary. And so he didn't, he didn't say in the science paper where he got the data from. Meanwhile, like in Germany. One can get data from T mobile, for instance. There are anonymized and one gets only reduced data you don't get. You can follow, for instance, when an individual leaves one cell telephone cell and enters another telephone cell. So this is what is recorded. So the information we get at least in Germany is very limited. Probably the Chinese authority have, have more information on cell phone data than we get in Germany. Okay. There is another question. Asking about instead of mobile phone. Is it possible to use a seismic data to check the mobility. The second part of your, of your question is it. The other type of the data, for example, the, the, the, the seismic that kind of the noise on the due to the traffic. Can you use such kind of data for the mobility. I'm sorry, it's very bad acoustically I couldn't get your question. Can you get what kind of data. Yeah. Okay. I'm, I don't know how you would get your mobility from seismic data. I know how you can use bird mobility to predict the seismic data. It's actually done with with sensors on the birds. But I don't know I have no idea how you could use the seismic data for human mobility. One more question about the super spreading event. How much is it affected by this mobility, the super spreading event super spread. Well, as super spreading can have two, two aspects. One is just the number of travels that a person undertakes. The second aspect is that particular person can transmit a disease much more easily than other persons. So there, they are two aspects when you talk about super spreading. The first aspect of the second aspect of course cannot be dealt with in such a moment but the first aspect is can be handled by the cell phone data for instance. I showed you this distribution now, you know, I can show it again where you see the number of different locations visited by individual people. And so that's what you can detect with cell phone data. But not say how infectious a particular person is, of course. Okay, now a new question come that how the flight ban stop the further spread or is it too late? How does what? I'm a bit afraid we are out of time. I'm sorry. I'm supposed to take this question. I'm tuning up my loud speakers here. Okay. We will come back to this question on the during the general discussion. Because we have to move to the next speaker. Yeah, with the Mongeau. Just a short presentation. So Jack is president of the French society of theoretical biology. He has been director of different CNRS laboratory CNRS is the center for national research in French. He has both medical expertise and I would say, physics expertise, which is something very special. And he has been working on many topics. He has been, for example, chief of the mission of same biology and complexity of CNRS and involved in a large French Japanese project and so on. And his presentation will be about the geoclastic demographic and socioeconomic determinants of the COVID-19 prevalence. Jack, the floor is for you. Okay. I will speak about some variables. These variables are companion of the COVID course. And you can distinguish epidemiologic variables. Very well known. You can get data, for example, from sites like WorldOmeta or Johns Hopkins, you know that. What's the daily in your cases, committed cases and daily numbers. And you can calculate from that by using different approaches. The basic production number of zeros effective revolution number, the initial slope of the auto correlation function if you are using an ARIMA time series approach. If you are using the raw data, you can calculate, for example, the initial slope of the log linear regression of the exponential regression of the data is different ways. The ratio economic variables, the given variables you can find in some sites from WorldBank or OECD site, for example, are GDP health expenditure percentage. That is the percentage of the GDP devoted to health expenditure. The consumer confidence index, the index related to the confidence of consumers in the market, in the general market, and the gene index, which is a kind of inequality index. And from that you can calculate the social health index, the ratio between the GDP health expenditure, and also index the social fracture. In the beginning, the difference between the income of the rich people and divided by the income of the 10% of the lowest, the poorest people. The demographic variables, the given variables, you know that you can get them from WorldBank and OECD, pyramid of age, birth and death rates, density. The related variables are the median age of a country, the relative death rates, that is the risk to die from the disease divided by the natural risk to die. And geoclinatic variables are essentially given variables. You know that you can get from a weather atlas or climate tents, temperature, humidity, elevation, sunshine hours, etc. And the phenomenological approach allows us to get the epidemiological variables. The Arima time series description is easy to describe. So you are attempting first to stanchionalize the signal, the variables, the variables of, for example, the daily new cases, and you can after subtracting a trend and a cycling component, you are trying to express the daily new cases at day G. And you have a combination, linear combination of the daily new cases at day G minus K plus a certain noise. And the Arima, for example, the Kuwait, I have just calculated in red the trend and the red stationery is a black signal. And with that we can do, of course, some forecast and also study in this black curve in the stationery. You can study the different coefficient relating the new cases, for example, in FAD, relating that to the new cases in days G minus K, until a day R and R is called the degree of regressivity of the Arima. Very often we are using a historical phenomenological model. It's a model by Bernoulli. It's in fact the logistic Bernoulli has invented the logistic by modeling the Viola epidemic in 1760. And from that, there is a lot of studies, but practically all the ideas are in the paper of Bernoulli. A big paper, 80 pages, my advice is for you, you have to read absolutely this seminal paper. And from that, it's relatively easy. We have done that with the Bordeaux team with Pierre Magal. Finally, to try to fit the data, the data are daily new cases. And after that, of course, you have a good approximation only by using this logistic Bernoulli approach. It's very easy to fit the cumulative data country by country. The discrete modeling is giving something different. If you have an estimation of the R0 and if you are, for example, data about the way in which the zero patient is contaminating progressively, Theo has shown that the other patients, if you are taking into account the length of the contagious period of an individual, in fact, you can play with more precise reproduction numbers during the contagiousness period. You have each day a local daily reproduction number and R0 is only the sum of this daily operation numbers along the contagious period. It's an interesting parameters. As often in many countries, we are recovering this V-shaped structure for these daily operation numbers observed for many respiratory diseases like influenza, etc. You are among contagious at the beginning of the contagiousness period. There is an improvement and you are renewable of contagiousness after some days. We can study that for all countries and you have four different groups of countries. Some be phasic like influenza, some decreasing, some inverted be phasic, some increasing, along the increasing the reproduction number along the contagious period. Geoclimatic, it's an example of variables here temperature, elevation, density and median age. It is possible by studying all this data to do some statistical analysis I will give now. The geoclimatic factor, our first, we have published that in February or March in biology last year. We have observed that the gradient, the northeast, southwest gradient of temperature in France was exactly represented by the region in France in which the occurrence of the disease was very high. And in fact, the public health policy in France has followed the occurrence of the disease and we were, for example, the third of May last year, we were in a lockdown, in a geographic level, respecting the gradient of temperature. Of course, it was not based on the temperature, it was based on the occurrence of the cases. And by analyzing France region by regions by looking at the mean temperature of these regions. In fact, if you were calculating the Pearson coefficient of correlation, in fact, it was for each of these days in March, etc., the correlation was significant. If you are putting all the heat here, the OECD countries, well, you have a gradient for if you are looking at the annual temperature, we have the same kind of gradient we have observed for the French region. What is strange, if you are comparing the first wave to the second wave, the correlation is negative for the first and positive for the second. And it's a bit difficult to comment that and we are trying in countries changing a lot as their occurrence of daily new cases like Rwanda, for example, we are trying to relate that to the dramatic change in these countries. But we have to do these analyze couple of first wave and second wave data, country by country, we have not finished this work. But apparently, between the two system waves, we have a big difference. Now, the demographic factors, you know that mortality is depending on the age, young, old people. And if you are looking at more 80 patients, the death rate is growing until 15%. It's very, very huge. And if you are considering this median age, for example, for the first wave, there is a positive regression and in which, for example, countries like Iraq having convex, a very young population. Before the demographic transition, you have the shape for the pyramid age. And Iraq, of course, has in the first wave a small, if you are considering the slope of the log linear regression of the daily new cases. The slope is small. And in countries in which you have after the demographic transition, a lot of median and age class represented in the pyramid of age by you have the, you observe, not for the first, but for the second, you observe the contrary. It's the same. It's this difference between the first with positive regression coefficient and the second with a negative regression coefficient. It's a bit strange. You have absolutely to enter in the couple of data, data obtained for the first, data obtained for the second, country by country to interpret that. But for example, here for the second wave, the regression coefficient is significant. It's minus 0.42. And for the first, it was also practically the same but positive. Of course, it's very tricky because in order to render, take into account this data, you have to complicate the model. For example, if you are playing with the macromic model, you have absolutely to divide at least into two classes, the old and the young, but here we have in order to estimate the different coefficient of this Bernoulli, Mackendall, Ross, et cetera, model, it's difficult to get the data. Very few countries are giving the data age class by age class. The socio-economic factor is the gene index. The gene index is, roughly speaking, a distribution curve. If, for example, all people have exactly the same income, the curve is like that. It's a distribution function representing the, oops, I'm sorry, this part of the square. If there is a good repetition of the income, the curve is diagonal. And in fact, in general, you have that. And by estimating the area of this difference between an harmonious repetition and that very inequity repetition, the repetition corresponding to that curve, you have a certain evaluation, quantification of the inequality. So the cis gene index is a problem with my, oh, I've changed something. Why I cannot. So if you are looking at the repetition of the gene coefficient in Europe, you have, of course, many differences between countries. This repetition indicates, for example, some countries having a good repetition of the richness on the north, about in the south of Europe, and in the middle, of course, it's in between. The consumer confidence index, I will pass quickly on that. The GDP health expenditure, it's the percentage of expenditure, the percentage of the GDP you are devoting to the health, and it's the same as there is a, it's a bit similar to the repetition to the gene repetition. And some countries here in between have the cis percentage, very important, and the major country is unfortunately France. This percentage is equal now to practically 50%. Well, the US are the champion with 18%. So if you are using, for example, cis GDP health percentage, if you are using that with the epidemiologic variables, and if you are doing just a PCA, a principal component analysis, in fact, you see that the GDP health will appear in the principal component too, and those are our dispatch into, for example, here is the second wave slope of the log linear regression curve, and we can analyze the components and with three components, you are explaining practically 60% of the variance that is considered as good. And if you are considering all countries developed and also developing countries, you observe a positive correlation between the GDP and the first wave slope that is the way in which you are accelerating in the exponential phase of the first wave. If you are looking only on development countries, you have the same kind of observation of regression. If you are looking at the second wave, it's different. You have the same phenomena of difference between the first and the second wave. The second wave is regression is giving a negative, a negative coefficient. If you are using not the initial slope of the log linear regression curve, but the maximum of R0 observed during the first wave that you have for developed countries for the first wave are positive regression, and it's still available for the second wave. If you are looking at all the countries, you can divide the countries into two groups of countries. The first group is a group containing France, Norway, Denmark, etc., in which if you are looking on the PC1, on the principal component 1, in fact, you have essentially negative coefficients in the constitution of the linear combinations of the PC1. For these countries, Kazakhstan, Moldova, Ukraine, Belarus, it contains a certain quantity of countries coming from the old USSR. It's exactly the contrary. It's probably due to the fact that the health policies are completely different in these blocks of countries. Or you can introduce the other GDP health percentage, socio-economic variables, the consumer index, the gene index, etc., that change the bit. The explanation is perhaps a bit better by looking to the three first PC components. These first components are now essentially the GNI, the income of the 10% of poor people, the income of the 10% richest, and the social index you can construct from that. The variance is essentially concentrated on these variables. Yes, the conclusion. Ah, we are practically finished. We can, from that, also, like in the first analysis group, the countries in clustering using the PC, principal component, and France, for example, is with Norway and Finland. So the perspective that is public health policy variables, although to take into account, we could do benefit risk analysis in risk groups, but we have absolutely no data. The best analysis is analysis by crossing age classes and commodity classes, no data. And of course, to do this analysis for each variant, I have done that by mixing the variant data, but we have to enter in the data. It's a teamwork with many people who have published, etc., about 16 papers. And okay, I have finished. I'm sorry if I have passed the time. Thank you very much, Jack. We have only very short time for quick question. There was one in the Q&A, which was about the modeling of variance. In fact, you mentioned a little bit that in your conclusion. Yes. The problem of the variance is very interesting for the forecast, because the variance, for example, is useful to detect the rupture point between the waves, because the modeling now is excellent for the waves by using classical approach. But in between the waves due to lockdown and other public health policies, you have period purely stationary periods with either a big R0 or a small R0. And in order to detect rupture points, time, it's time points, between the date at which a wave is finishing or at which a wave is initiating, is starting. But we are using the variance. The variance is a very, very important. The variance of the raw data is a very, very important statistical parameter too. Okay. If there is no immediate question, we will move to the next talk. But in any case, during the general discussion, we will come back to several questions related to your presentation. Okay. I'm sharing the screen now, just to present Lauren. So Lauren Gardner is associate professor at John Hopkins Whiting School of Engineering. And she's well known as the creator of the interactive web-based dashboard. And which has been used by many, many people, including me and other researchers all around the globe, and which strikes the outbreak of the novel coronavirus. And there is a record of about 3 billion page views. Okay. She briefed the Congress about that. And she was named among the time most times 100 most influential people of 2020 for a democracy that I'm feeling a void of public leadership during the pandemic. She was also including DBC 100 women list 2020 and other ranking like that. She's a specialist of mobility in spreading disease, but she was more realistic approach and very various diffusion, taking into account climate plan use and other contributing risk factors. Our presentation will be on tracking COVID-19 in real time challenges face and lesson learned. Okay, the floor is now to Lauren. Okay, as soon as to get it to it. Okay, to leave the screen. So Lauren. Got it. Thanks, Daniel. And for the introduction, the opportunity to speak here. And, like Daniel mentioned, I'm, I've been leading the mapping and data collection efforts behind our CSSC dashboard. Since we started this in January of 2020. So it's, it's hard to believe it's been about 15 months at this point. Definitely not what we expected. I think at the start of this. And so I will talk today about the evolution of this dashboard and some of the processes that we've put in place and like Daniel mentioned challenges that we've faced along the way, and some suggestions for moving forward as well that we've taken away and learned through this experience. And I'll try to get through it a little bit quickly so that we have some time for questions afterwards. All right. So just to make sure everyone's on the same page if anyone's not familiar with this dashboard and these efforts, we've been tracking cases deaths and recoveries on this dashboard. For about 15 months, the spatial resolution varies. So for the US, for instance, we collect data at the county level. And then it goes all the way up to the country level and with quite a few countries at subnational levels, which are equivalent to a state province level. All together, this is about, there's about 3500 points or locations on the dashboard. And we're collecting multiple variables which amounts to about 10,000 variables an hour that we pull on a continual rolling 24 seven timeline so it's a pretty substantial data and effort. And there's all sorts of other pieces and math layers that are provided as well, which is all interactive. And, and, importantly, all the data that we pull and push onto this dashboard is made publicly available. And, and so what I wanted to focus on is kind of why we did this and how we did it. And one of, and why we did this is actually kind of twofold but one of the things which I think aligns well with the talks that have been in this session is that my background is really not in collecting data it's using data so I model infectious disease spread and risk, and addressing a lot of the same kind of data and concerns and research questions that were posed by this, the talks before this. And so I'm really interested in understanding emerging infectious diseases novel pathogens what's going to happen, especially in the earliest stages and so I'm very acutely aware that there's a big gap. And that's why we did this before this type of data. And it's important and necessary for making evidence based decisions and guiding policy, especially again, early on. And so this new kind of these new cases of pneumonia that were rising in China back in December, January, were something that piqued our interest and with my one, one of my first year PhD students who is from China, his friends and family were being affected directly. And so we were following and tracking these new cases and collecting this data internally and decided to just start building out a dynamic data set and sharing it publicly, so that people like myself and groups like us and other users could access to it. And so we actually decided to do this one day in January, making it public and build the map alongside it to visualize the data that we would be collecting. And so we actually built the prototype for the whole dashboard that evening and shared it the next morning and that was really the start of this whole effort and it was just two of us for a while, running this thing. The initial architecture supporting it was also very simple. So, this was a time where there was no public health dedicated pages to coven data. So we were collecting data from kind of untraditional data sources. And so the original offices were posting about new cases on Twitter and Facebook links for instance, there were news and media articles all over the place so we were really kind of aggressively trying to manually track what was going on, and push this data into the dashboard and also into what was a Google Sheets at the time. And in some ways we were also getting information back to us through some proud source efforts so no one was able to update the map themselves outside of our group. But because of the attention it was getting we were getting informed through communications of new cases as they were occurring in real time and we could validate those and add them to the map. So, this we started when there was just a few hundred cases in China, and only a couple countries outside of China that had reported a case or two so it was feasible to do at the start but clearly was not a sustainable process as the outbreak grew like we all know it did. And even in just that first week, we started getting a lot of attention for what we were doing, we had you know a few hundred a few thousand hits the first day up to 10 million hits by the end of the first week. And so the dashboard was really already starting to be picked up and relied upon by international and domestic media and individuals and data users all over the place. So, as this was growing we kind of saw this exponential increase happening with the outbreak itself and also with the interest and and demand on the dashboard and so we knew we had to re strategize. And so we teamed up with a larger group, most critically with Esri who was this generators of this the mapping software that we were using. And that helped support the visualization infrastructure. And then the at Hopkins we have a lot larger team which went beyond my group at Center for Systems Science and Engineering, and included the applied physics lab at Hopkins which really helped us build out a much more robust and resilient data pipeline to start collecting this data as it was becoming available and more and more and more sources. And so we up to this architecture and the kind of next stage of it was really focused on two different things. One was expanding our source set so you know from the start we were doing these kind of ad hoc sources as things were being reported. And so as actual authoritative sources begin reporting on this data, we needed to be collecting and pulling data from those and so we began quickly seeking out and expanding the source set by identifying them and validating them, and then also automating the whole effort so obviously we couldn't continue to do this manually. And so we needed to automate this data collection process and develop scrapers and pull this data and collect it and clean it and curate it. Before we would push it out to both the dashboard, and also just to the GitHub repository in general where it was hosted and could be pulled by any interested party. So this architecture was obviously it was automated it was much more sustainable, but there was still a lot of challenges with it because there was still this manual air recovery process that we were dealing with where we weren't able to really automate the reading and the data and and validating it through the system we had to still do that manually. And so this was happening while everything was just still expanding and growing exponentially. And so we went again from hundreds of thousands to millions to billions of hits on this dashboard a day which was happening back in March. And, and this demand was again coinciding with spread from, you know, one country to four to now close to 200 different countries and like I said we're pulling from about 3500 different locations independently for the data on this dashboard. And users were also really broad. So we had individuals that were utilizing this data directly through the interactive dashboard and pulling it from the GitHub, based on access to it, and exposure through the media through all the different platforms. But the dashboard data was also being relied upon by, you know, our highest heads of state. And I exemplified by one of my favorite photos of our dashboard photo bombing Mike Pence on the HHS watch HHS watch floor back early last year. And so, while it was really great and exciting that the data was providing this public service and providing this public good it pose a lot of challenges for us so on one hand I mentioned the air recovery. So the data was being updated so quickly that new kind of delays in our reporting or corrections of even an hour we're kind of being we're being fed back to us to us through different, different platforms through through the media for instance and so these were happening faster than we could internally detect and correct them. At the same time, there was a huge communications challenge around what we were doing. We had to deal with issues from things like how do we report data for locations where there's disputed territorial boundaries. We had issues around naming conventions. And we had issues with just false accusations of us putting this data behind paywalls for when there was just delays and outages for technical difficulties. And so trying to continuously communicate issues around what we were doing and also just clarifications was almost a full time job alongside this and this was all again being done while we were just kind of almost as volunteer group trying to provide this public service. So we knew with this growth and exposure and demand that we needed to continually update this this architecture which we're still very much doing. And so again we continue to this day to expand the data sourcing. So we are going from lower to higher spatial resolution and a lot of places and seeking out some national data. So we're installing data from the most authoritative sources for any place that we're reporting for. And then we have a much more resilient pipeline to pull this data and collect it and curate it into an open source open data product. And there's a couple critical components that are in here that we've developed and one of them addresses that air recovery that I was mentioning earlier. We have this automated anomaly detection system that we built from scratch, which reads in those 10,000 or so points an hour and looks for patterns to identify which ones may be error prone and holds them back for manual recovery and validation before we push them into the open data product. And these errors can come from all sorts of different causes from upstream data entry errors at the actual sources, or because websites change the structure of their reporting and therefore we may read in a variable and assign it incorrectly. So we try to detect those before they go, they go public and we also have different data fusion logic in place, because of some of the inconsistencies that we've encountered with the actual data that's being reported. So this is something that again is still very much in development like this experience I keep saying is like building a plane while flying it in a lightning storm like it's just all being done in real time. And, and it's, it's very messy and so we're continually kind of bandating in this, this pipeline to try and make it work as best as possible. And, and so that's the system that we kind of have in place, and we started with this case and death data collection effort in my center which expanded in terms of the team but it's grown into a much larger effort by john Hopkins, which is hosted under the Resource Center. So this was built around our data for tracking coronavirus but is expanded to be a huge collaboration across the university that also looks at initiatives around testing and contact tracing and vaccines as well as looking at providing analytics and insights and so just generally what's going on with coven. And the data uses to date are are really broad I think today we've kind of this data serves a lot of public good which is really exciting. Really surprisingly to me I guess. Well, not now, but at the start I wasn't expecting it to have such a major role for the general public. And so this data is accessed and accessible by individuals all over the world. You can get it directly from the site it's integrated into things like Google Maps to help inform individuals on the risk in their kind of in their surroundings and help them make better decisions at an individual level. It's obviously integrated into mainstream media so it's, you know, lives on CNN 24 seven and it's, it's relied upon by a lot of other major media sources and, you know, MPR Wall Street Journal, etc. And then I think really, for me, excitingly it's, it's really being used to drive public health policy, and it's being relied upon by research groups all over the world. And so, for instance the CDC forecasting hub uses this data to both build and validate the models that contribute to this, the coven forecast. And so my group's also actually one of the contributors to these forecasts so we're not only generators of the data but we do also use it. And then internally in our group we're using this data for all sorts of applications as well. Trying to understand some of those, the questions that have been brought up in this session around the role of, you know, human mobility and land use and climate and socio demographics and human behavior and the risk and spread of coven. So just a couple more things, one of them. I think it's probably obvious to most people but I just want to kind of hit on this like why this effort has been so hard. And the reason it's been so hard is because standards matter a lot in these in these data collection processes, and we've been living in an environment that completely lacks standardization. There's inconsistencies and instability across the board and this, what we're trying to do. In my hand there's variations in the data structure and mechanisms on which this data is being provided. It's not always provided in machine readable formats. There's retrospective reporting happening all the time and we're seeing a lot of that in the US right now as there's audits going around on the death counts and there's being new cases and deaths that just get dropped in at different points in time for different locations. It's really hard to get a good understanding of what the actual epic curve looks like what's actually going on with this outbreak and at different stages in time. There's discrepancies in the numbers reported amongst authoritative sources which is really challenging and disheartening. So in the US in a state like Texas, the state of Texas on their dashboard will have one number for a certain time and the county might say something different about itself and these are really challenges to distinguish between and decide what we should be reporting. Again, it's a pandemic so there's time of day and frequency reporting that varies by location all around the world, and then really critically there's huge ambiguity and parameter definitions that we're collecting so, you know what's a confirmed what's a probable case how do those vary by location that were of reporting our cases and deaths reported based on, you know when they're when the tests were done when the, when the actual infection actually occurred or when the report was made available to the public. There's new technology so confirming cases using different types of diagnostic versus antibody test provides a huge challenge, and of course tracking recoveries is just a huge mess. And so we really need to be able to be kind of nimble to address and detect these kind of challenges and when changes occur and then respond and react to them. And so I think, you know, it's all of us, I think, think this already and know this but there is really a huge need for open data principles and standardization to support these kind of events. We need a standard reporting system, and you know not for COVID alone but again for emerging infectious and notifiable diseases across the board. This needs to be standardized and made available publicly in a way that's actually actionable so these are spatial and temporal scales that prove useful enough to use for planning and modeling purposes. And they need to be provided in a really timely manner and machine readable formats as well, and in a systemized fashion so that these kind of efforts to centralize and aggregate this data in the future can be done, much more efficiently and effectively. Lastly, I get asked this a lot and I have thought a lot about it is why we were so successful in doing this and again this is the fact and why there was such a demand for doing it in the first place so I don't think there's going to be a world in the future where this kind of information isn't just available to everyone when something happens. So I'm just surprising that it felt like that was even a gap in this day and age. And so I think some of the reasons we were really successful in doing this are one we acted really early so coming from my background and just knowing the value of this data, and also having the personal interest from my group. We, we acted within, you know, essentially days or weeks of when there was just some first hint of a problem. And, and our guiding principle from the start was really open data and open science so not only were we collecting this because we knew it was important but we were collecting it and making it available to anyone that wanted it in a format that was easily accessible and feasible. A second thing is we were really lucky to do this in a very supportive environment. We had funding available from the start, which let us continue these efforts as they grew exponentially in an uninterrupted fashion which would just not have been feasible if we didn't have that support. And so this kind of support really needs to exist. And there needs to be a better mechanism for science agencies and governments to financially support these efforts quickly when needed. We also, again, I was really lucky to be surrounded by people that are so capable with the technical skill sets that we needed for this work so these were engineers and computer scientists and software developers and spatial analyst and and experts that we relied upon, and these skill sets really need to be standing skill sets that are invested upon and integrated into public health agencies, as well moving forward, if we want to have high quality data available to us in a timely manner. And the last thing that I think is really interesting is, we were, you know, a private institute doing this, just on a voluntary effort, like I said, and so we really had the freedom to make executive decisions in real time on how things should be done. There was no bureaucratic internal approval process we had to go through. And, and that can definitely potentially pose some challenges and be problematic but I think this was really critical in the early days when things were changing you know hourly daily at least. We could just make decisions as it was happening and do something to provide this information and share it. So I know not everyone has the luxury of working in that kind of an environment. But I think that if there are better open data standards and infrastructure and processes in place to make this information available, then it could a lot more cleanly and efficiently be collected and shared publicly and a timely manner. It could be done by other groups as well. And so this is to effort by me. It is a huge group effort. Frank and Cheng Dong, who goes by Frank, I just want to highlight who really kind of pioneered this effort from the start. And it was led by my team in the Center for Systems Science and Engineering and our close kind of collaborators now across the applied physics lab, also the Sheridan libraries. And then again at Esri as well and so this group of people is is includes all the core people that have been involved from from the earliest days and none of them have slept near as much as they should have been the last years. And then I also just want to acknowledge in the support that we've had for this and this is from industries philanthropies also the science organizations that have been supporting our research that utilizes this data as well. So, hopefully I left some time for questions. And I'll go ahead and stop sharing my screen now. Well, thank you very much. Of course, it's very impressive what you have done with your group. And I really enjoy your conclusion which were not only limited to what you have done. But for the future, raising important question about open data and requirement to be done. Okay, I'm sure there should be some questions. Yeah, two participants raised their hands. The first one is the previous one. So the only second one. Jack Demangeau is the first maybe. No, no, please. Hello. I believe the question question is that do you have any collaboration with the WHO dashboard. Yeah. So we're we're not working together on that. I know that our partners at Esri the spatial modeling software, they are working with WHO so indirectly through that there's been a bit of communication but they're completely independent efforts. And they're really doing, they collect data differently through their kind of on the ground organizations that they have in place ours has been purely through collecting publicly available data directly from each of the different locations country states cities that we pull it from so it's done completely independent. Yeah, well thanks very much for for giving us these insights into your extremely valuable work. And I'm a little pessimistic. I understand that you, you call for standardized reporting systems and I would be extremely valuable. And when you see what's going on within Germany, we have 16 different states and health issues are are not federal, that says, will buy every state individually. And even within Germany, the report doesn't work. They, they are still using faxes to report to the robot car institute. And this, this makes me very pessimistic that we would be able to get this standardized on on a global scale and also unique politicians for that and when you see just looking at what happens in Germany. The unable to get together to take decisions for for for Germany as a nation. Every Prime Minister acts as he wants and not listening to scientists. They, they say they're listening to scientists but they also have other aspects to take into account and so on and and then they do what they want. It's pessimistic that we will be able to improve that soon. Yeah, well, given the last 15 months of my life I share a lot of the same frustration with you. And I hear you about Germany, I'm in the US, specifically in Texas right now. And people just say that we share a lot of the same concerns on that front, we have 50 states and over 3000 counties, and they do things differently. And so, you know, I kind of can say, this is a huge problem. We can't get alignment anywhere. How are we supposed to know what's going on. I think there's organizations here like the CDC that couldn't, you know, that, you know, should not be political that could enforce some standardization and they've been trying to do better lately. What but it needs to be in place earlier globally I agree it's a huge challenge things are done differently, different technologies are available are available surveillance systems are different. And all that we know is that the way we did it this time, like we just shouldn't do it this way again it does, you know, it was so it doesn't make any sense so we need to do something better and I hope we use this experience to kind of drive and move that way and baby steps at least to be able to get a better understanding to compare apples to apples because right now pulling data from different countries. I think in this group we're so interested in what's going on globally on a relative basis and trying to understand things like the role of climate that you have to have some good understanding of what's happening in one country relative to another. And if you don't have access to the same kind of reporting, then you know it's impossible to actually draw accurate conclusions about what's going on so I hear you but I think it's something that just it needs to be worked towards. Yes, have you access to some hospital data coming from COVID patient files, because in Johns Hopkins you have an excellent hospital information system built by Professor Marion Ball, have you access to this. This. Yes, so some definitely there is great connection with the hospital systems in the university that's completely independent of our data collection efforts. This map has no individual level data it's all anonymized aggregated publicly available data individual level patient data is just such a different realm in terms of using that and sharing it. So it's it's not part of this project at all. You could play with them without rendering them. Yeah, using that kind of information to understand more specifics and contextual things like demographic distributions and age and comorbidities and underlying. Absolutely that's being used. It's just again it's not going. It's not part of these larger scale kind of global data collection and mapping efforts but hospital data across the board is being collected and provided an aggregate anonymized forms. And I think HHS has the best version of that for now. We were working with COVID tracking project previously and put to put up hospitalization data. So what's filtering these data you are receiving that after a filter done by hospitals or public health services, etc. Absolutely that's a huge thing, everything that we put up. We cannot say that it accurately reflects the real world we can only say that it accurately reflects what's being reported publicly. We definitely can't go validate what Spain is saying about its COVID patients so yeah that's not yeah that's not our role right it's centralization of the best data available, but that's it. So, and I kind of feel like that's really important and then that's the first step, and we're not in a position to try and adjust and filter it to say what we think is really going on that's a modeling project and you need to know, you know, everyone needs to know what the, you know, what the same ground truth data is before they can work with it and then make their own assumptions and adjustments for whatever modeling purposes there are but yeah. You are right but for benefit risk studies, you probably need this information. Yeah definitely and I hope people are taking our data and then doing that to share it and I know in our group we do a lot of that we use the data and we post process it to do things to put into models to do decision support and predictions and forecast but again I feel like that needs to be done in a second stage so everyone has the same baseline information. Yeah. Thank you very much. If there is no other question I have one, which is a bit more optimistic than the point of view of tail is the fact that the data becoming public. It could be some feedback to the providers who certainly cannot hide big uncertainties. For example, in France, we have two main public institution to give the numbers for their deaths with a difference of 10%. It was public until now, now it's public, and I believe in this, what you have done help a lot to maybe to create some difficulties to these hidden facts, because now they are not too much hidden, they are a little bit public. Yeah, I think we caused a lot of people a lot of headaches with doing this. Absolutely and we talked directly to these public health entities all the time we coordinate with France, all the way down to like Harris County, Texas, trying to understand how they put data out that we think doesn't make sense. What's really going on, how should we be doing it how should we be reporting this and trying to align it with them, but saying you know there's these inconsistencies we're seeing and working through it and it's also a huge issues when they do big data dumps and then we need to figure out how do we back distribute this. It's, I think you're right I think it just, it just brings to light a lot of the things that we're not doing well, that hopefully some of them will start doing better. Okay. Thank you. There is one question from Sheila. Give any information about accuracy and efficacy of the test and trace programs. That's a big question. I think that, yes, it could, but it would require lots of kind of second stage data collection and modeling to do that and actually collect all those other, all those other variables the test the technologies in different locations to understand how they're doing things and what's really accurate. So the data alone doesn't answer that but it could be used in those, in those kind of to answer those kind of questions. Okay, if there is no other specific question about the dashboard and the one that policy which has been set on by Lauren on this group. Well, we can move back to the other question which we are not fully answer in the short discussion and to have a more general discussion. For example, I believe you're my you collected few questions which were raised by people about the violence. Yama. Oh, okay. Yes, the question from the, I knew that question is that the cannot fly to bounce stop the father's to spread. According to your model. Yeah, this is a question to tell you. This is for Jack or Theo. I think this is to Jack. Yeah. Now, the, the variance is take a bit into account in the principal component analysis, you know, but, but, but in general, we are reducing the information and expectations and other moments like variance Q&S, etc. We are not taking sufficiently into account. I agree with you. But by introducing tools like principal component analysis, I think we have the way to introduce the variance as a major parameter statistical parameter. And then also another question from a very not going to the tail that can can we model how the variant have penetrated to the Europe through this flight. Yes. I was just gone for some reason I was interrupted. In principle, yes, in principle, it's possible to model this, but I guess the data that one has available are not very accurate. Imagine. Actually, this reminds me of a coin sped spreading. empirical study that was done in Europe. It was originally, similarly to what we have done with reverse charge and dollar bills. There was a study on coins on European Euro coins were the different coins were issued into differently in different countries so you could see the origin of a, of a coin from which country it came from. So of course, after a while, all this is mixed up, but so in a certain transitory phase you could study this. So that that's just an analogy. Here, I think what one needs in order to model this accurately, of course is the, the, the transport between different countries. And this changes on a daily basis. And it, it is different in different, even in different states. Imagine, we have before states in Germany that have borders with friends, and every state has a different rules for for Germany from France. And so getting these data accurately will be difficult. Otherwise, it's possible to model this of course. If not, not any question I have one for, I believe, more or less all the speakers, especially those who are using stochastic modeling. Is it so easy to to deal with. Decision makers arguing with the help of stochastic models. I'm afraid they are more interested by the domestic model. So what is your experience about that. They are looking to our forecasting 90% intervals, etc. With a lot of curiosity, but I think they are not using our models because they are saying of course the cone of uncertainty is very large. And they are not really understanding what is representing forecasting uncertainty, interval, and I think they have a culture, a statistical culture, very, very small and very poor. And it's a problem of stochasticity. The statistic are stochastic, but it's probably better or easier to understand what is an average. I mean, the variance is a bit more complicated. And after that, if you want to enter really in an ARIMA in a time in series modeling. I think they have the culture for me in France. But the president is apparently learning a lot, he's saying he's learning. Mathematics and we can hope in the future he could understand all from our models. I have my own question to Roland and Judy, that is it possible to get the data of the time of the day. Could I still answer the question before, excuse me. Okay. I have a question to the Roland and Judy about if we have an answer on the previous question. They would like to answer to the previous question. Yeah, well, as you might know, in our country, we have a chancellor who was trained as a scientist, actually, actually as a theoretical physicist and theoretical chemist. And, but she was very open. And she still is. And she acted very early on in the pandemic in last year in March, without her, it would have been much worse last year. So there are open years right now at the very top of our nation. But it's the federal system that is and the fact that we have election in many states that make it so difficult right now this year. And of course only a few people are able to, to make sense out of the stochastic modeling. But our chancellor does that's for sure. Okay, thank you. And it was a comment about this question. Dr. Merkel has to teach our president in the domain. You cannot say it. Judy about that you have some feeling about how to deal with the decision makers about modeling different type of modeling. Yeah, I think it's not just for COVID but for any for any types of diseases they really anything that's the word uncertainty means that for them means that it's you don't know the answer so they won't use it. Yeah, I think statistics are difficult for you know the non statistical population in general, and I think you know governments decision makers often they expect to be sure about doing something because there's always some funds attached to it. So if there is uncertainty then they would rather than not make a decision rather than to make one and be accused of something later on. So the answer I guess would be no not really. Okay, something about the site the time of day. If you can get the data that when the infection time, because nobody the infection data is playing day of the infection fits their infection happen but if you can get the date estimate of the time, the time of the infection, then you can really see that the time we should burn for going out of whatever that's so such kind of is such kind of data available or is that possible to get such kind of data from any source. I think you know that is is this close to impossible. Nobody knows when the period and you know that you're infected when the symptoms start to show and that's possibly four days afterwards there's no telling what time day. I think I think providing daily data is actually dangerous I tell everyone that uses this that they should be smoothing it at least over a week for multiple reasons just issues of reporting of cycles over the course of the week and the way testing is done the way health, you know, test public health authorities are actually providing data. I mean, I want to trust anything coming out that said it knew what was going on on an hourly basis. I want to trust things that say what's going on on a daily or weekly basis so. But if you find that let us know that's great. You have, for example, only when zero patient, and if you have an ideal transmission, it's possible to extract from the sequence of the daily new cases number, the daily reproduction numbers corresponding to each day of the infectious virus is possible. That is called the fever as it's a general Fibonacci series. But there is a lot of noise. Yeah, they are not precise. There is no zero patient. Yeah, I totally agree that you can, in theory, do all of these things. The thing is that the data that underlies the models that you need to use is so noisy that there's no point to making daily predictions for anything. But yeah, Daniel, I just have a quick question. I am going to have to drop soon. I wasn't sure when this ends. I thought it was finished. We are a little bit behind the schedule, about 15 minutes, and we have to stop. I would suggest that now we will ask to the continent to present themselves. I will give them the floor and I will share the screen with them. So that I think the main conclusion we have is that we are only at the beginning of something very important. But okay, something was done but there is much more to be done. And that there is done there are two other opportunities to discuss that during this conference. I will mention that briefly. In any case, we have first to thank all the speakers for what they have done. Thank you very much. And I'm sharing the screen with what is the last slide. So, if Alexander can say two words about himself, that would be nice. Okay, I can see you. Daniel, thank you for speakers. It's really a very interesting presentation and good achievements. I'm Alexander Klanoff and working for Science Innovation Department of WMO here in Secretariat and also professor affiliated in the University of Copenhagen. In my field is modeling of atmospheric pollution and meteorological processes. And first of all, I just would like also to highlight one issue that was not touched. Actually, most of the presentation, especially this first WMO research board report, considered impact of geophysical factors and meteorological and air quality factors on disease with COVID. But we have also to remember what this pandemic and especially lockdown period affected on many aspects on geophysical factors. In particular for WMO, a very important issue, quality of forecast and observation system, which was also one of the other group is analyzing at WMO and to minimize impact of, for example, reduce of observations from aircraft. During this lockdown period and reduction of flights on the quality of forecast. Another group of the Global Atmosphere Watch program is analyzing effects of lockdown on atmospheric pollution. It was also quite interesting results and professor Soki who is leading this group is now preparing a big overview paper is I think more than 200 papers analyzed there and will be published soon. So it will be good maybe next time to also ask them to consider. And it's many different impact and different direction for example ozone it's opposite to particulate matter and for greenhouse gases. It's also interesting topic to discuss of course now we don't have time but just. Thank you. Thank you. I'm afraid that Paul Bourgine is not present and the same for Stefan Otinty. Paul Bourgine is the head of the UNESCO UNITWIN on complex system. And it's interesting to pursue this question with the help of UNESCO so you will have some news in short time about that. Stefan Otinty is from the University of Bologna and is the physicist dealing with seismicity. So it's too bad he could have better answer maybe to the question of seismicity and I don't remember the COVID discussion, I think. And on the contrary Benjamin that chief from Jonathan University is here. So if he wants to say a few words that would be nice. I think just briefly Daniel and I want to thank you and the other conveners and and of course all of our speakers for a very engaging session. I think you summarized it appropriately that there's been tremendous effort made and so much more work to be done so I'm sure we will be continuing this conversation in the future. Okay, thank you very much. Gabrielle Manoli from UCL London University College London. Hello everyone. I'm Gabrielle Manoli I'm lecturer at the University College London as I would like also to thank you you know the speakers and the conveners I really enjoyed this session it was amazing and I particularly enjoyed the discussion on uncertainties right now we should communicate and deal with it when we're trying to inform policy of decision makers I think it's really an open challenge and I guess it's a new very interesting discussion that we can definitely continue during the next sessions and town hall meetings in the coming days. So thank you everyone really really enjoyed today. Thank you Gabrielle. Klaus Friedrich from Hamburg University. He was around Klaus. Yes. Okay, let's see. Does it work? And we can hear you. Okay, okay I try to do the video here. Okay, thanks to all the speakers I really very much appreciate that and what comes to mind with the last talk is that there may be an example for positive outlook like the WMO has been created in connection with weather forecasting and if we at data analysis associated with that. And so worldwide we have a really good data network, which works tremendously and it could well be that WMO could do the similar thing. Plus we need an intriguing start for politicians and politicians maybe have two problems. First, they do not really believe in weather forecasting although it's very good and secondly, maybe we need a next generation. Okay, thank you. And last but not the least. Okay, I'm quite different. I'm a space scientist and then doing the data collection and the data analysis. But from both data collection and the data analysis that kind of common background or anything global data and the corona data is really such kind of global data. My feeling is that the more scientists from the many different fields, if it is handling the global data or regional data and also analyzing data that would be important to think about the corona. Even we see off time on time doesn't matter. That's kind of thing. And that is my background. Thank you. And I just want to mention that we have two opportunities to meet again. One is the town town hall meeting on Wednesday afternoon 530, and which is mainly about the engagement of the geoscience communities by geoscience it should be the term should be taken in the large sense. Anyone who is concerned a little bit with geo so almost anyone. And there will be today after an interdiscrimination ITS one on COVID-19 pandemic as urban system and geosciences and there will be 26 papers with a wide diversity of topics which will be presented there. And of course we are thinking about a special issue to keep track of what has been done today and during the next to manifestation of this year at the EU. It's not a remark we are closing the session we are just a little bit late 20 minutes or 26 minutes late. Okay. Okay. So thank you again for all.