 I would like to talk about the assessment of risk mitigation strategies for texts or ridges and following our scheme that we talked about in the decision scenario, then what the model results and the value of information with the conclusion. The work I am presenting was done in the short term scientific mission. This is one tool, one networking tool we have in this cost action framework and I was going to UCASA in Australia. They are actually the expert on security issues for structures for civil engineering structures can be found. And I was very happy to work with him together on this case study here. So we analyzed the risk mitigation measures for terrorist attacks with provides the explosive devices. And the structure we are looking at is an iconic structure, two million euro investment. This is what we assume the decision maker is a public authority responsible for the safety of the infrastructure and of course of the infrastructure users. And the decision point and time is the design phase as I think this is also an important feedback we have already collected. SHM should be the decisions are taken in the design phase. We can also for a second scenario or a second risk mitigation strategy we can also take the decision in the operation phase. Life cycle phases is the design phase and 100 years of operation we there we annualize the costs over the 100 years with a forward certain interest rate and the performance is the improvised explosive device IED loading that we are considering. And the objective is the realization of the risks and expected costs or posing the objective differently are the measures are the investments in the safety are they cost efficient do they really help to reduce the risks. So we have here a little different decision tree so our base strategy here is it's no risk mitigation we can do a protect strategy as one. So that's basically implemented in the design a measure to enhance the collapse or to protect against the collapse of the bridge. And the second strategy is control here we have surveillance information which provide an indication that the structure is safe or there's a threat. And if there was a threat then the decision is or the action is to close the structure and to allow for detection measures or certainly getting rid of the explosive device. And this is then connected to the system states safe and collapse for the bridge. So the basic model for the assessment of the collapse probability we are having here is that we work with the probability of a threat. So are there terrorists at work developing an improvised explosive device or maybe the point is that we don't know researchers don't know engineers don't know but maybe there's some information in the police departments and security services. So this is basically to a large extent this is an unknown. And then we have the probability that there is a hazard given that there's a threat. So this goes basically to the performance of the improvised explosive device. And there's quite some data on the performance of an explosive device. So it doesn't ignite, it doesn't explode and it's the damages. And this can be assessed on previous incidents and it's around 20 percent. 20 percent of the explosive devices really do a damage. And then what is the probability of collapse here was this hazard. So again in the perspective of that our decision analyzes provides the interface to the or between the decision analyzes and the decision parameters to the more sophisticated models. We analyze the protective measures in a way that they need to risk reduction and they cost and the higher the risk reduction data are here the higher they cost and this is non linear. So what we find out is that the optimal or we can identify the optimal risk reduction together with the costs. So this is these two columns here and it depends on our unknown threat level. So we basically assume the threat level of our threat probability. This is 2.10 to the power of minus 3 and then we would need a risk reduction of 46 percent and we have these annualized costs here. So we see that for low threat probabilities the optimality goes with rather low performing risk reduction measures and low cost risk reduction measures. This is this is optimal but if the threat probability is higher then we the optimal is here to implement high performing risk reduction measures and also more costs. So this was basically the analyzes of this part and now we are coming to the control strategy. So the control strategy is that we have surveillance information which can give an indication or no indication and the indications depend on the true state. So there's also false indications taken into account. We have the cost of the surveillance system investments operation and replacement every 10 years and the bridge closure for detection actions and costs are accounted for. For instance due to a threat. Again we have we want to identify an optimal strategy in terms of the probability of detection in dependency of the costs we see. Again we have assumed and we have calibrated basically our model on real numbers in a way that high detection or detection performance goes with nonlinear high costs. What we find is that here and strategies surveillance strategies this was all decision variables with a high performance are optimal and also with associated higher costs and both strategies in comparison basically what we can see here is this is the relative value of information. So we divide by the expected costs of our base scenario and we see that for a low threat probability we can identify optimal strategies but there's basically no effect on the overall costs and risks. This is different if the threat probabilities are higher than we can have very significant risk production. So the conclusions are any risk mitigation strategy should be implemented with the knowledge of the threat probabilities. So there must be an interface to some who knows the threat probabilities. The strategies protected control are cost-efficient for threat probabilities higher than two times 10 to the power of minus three despite we can follow our threat probabilities identify optimal measures but they are not significant email field implemented. So below 10 to the power of minus three protective measures and control strategies should be implemented. And this goes to more detail in a publication for the YEPAS in Berlin this year. I would like to know from you what is the experience with the security measures for bridges from your side and are they somehow enforced or how they are accounted for in the design or are there requests in the operation to account for security measures? So this is the first question and I start here. I think it's a very interesting study especially because it's about a type of theory that is maybe not really explicitly considered in design. One thing that I'm curious about is how you think about the capability of security services of assessing that threat probability because from literature it appears that we are notoriously bad when estimating small probabilities where it used to be orders of magnitude of. And in relation to the last presentation is about the catalog of different monitoring devices that catalog is also based on the failure modes that we consider in the design but there might be other failure modes such as an explosive device that might also be relevant. So I think that's also maybe a good thing to consider if we purely base our monitoring strategies on the things that we consider in the design it might give us a really good overview for those things but it kind of lacks the robustness to cover those things that we might not have looked into in design and for those issues it might be even more interesting but yeah it's a I tried to answer the question but I couldn't manage that so. Yes. So Max and I have a comment here. How do you calculate the risk reduction? You have safety measures and of course you have to optimize them and so on but according to experience most of the safety measures are measures which has not been proven in the past by statistical data so you have let's say how effective they are or how they work. How do you calculate this risk reduction especially in these cases where the technology is going fast? This risk reduction is an interface here so we describe what risk reduction is in combination with the costs as optimal and of course this should be substantiated by the assessment of the measures and here we need to come from this overall level more to the specific measures like like a boss offense for instance or whatever measure is or simply strengthening the dimensions and account for the pressure wave and have a set of scenarios where the explosive device can be placed could be placed and then the risk reduction goes with the probability of a reduction and this is what we can assess with the structural reliability methods. Thank you. In the last slide you created a clear threshold based on the thread probability and that probability is purely subjective there is no real realistic definition of a thread probability so how much does it make to create a clear threshold based on such a definition? It's a frequentistic deputation of the tread level but the challenge here is here we can we can calculate from the incidence we have observed at the number of both buildings so there is incidence a number of incidents in the past 20, 30, 50 years and then there's a number of buildings and that's the frequentistic probability of tread and that's in the order of 10 to the power of minus 6, 4 buildings so first day spaces, frequentistic spaces but the more important point here is yes civil engineering, natural hazards, we can look at the past and with our egoticity and stationarity assumptions we can predict the future but do we have egoticity and stationarity in our models which are causing these hazards so that's something like human mass behavior models and where these conditions we are working with usually in civil engineering are not fulfilled at all. So obviously you did this study while in Australia and Mark Stewart has been studying those sort of consequences of bombing and so on for a while right and assessment of risks so Australians are clearly scared about that I don't think they actually have any bombing incident terrorist attacks on Australian soil but anyway my question is really kind of related to your about the probabilities right is it something that Denmark I don't know any other European countries perhaps more UK right because well terrorists very small kind of visible but is it is it a problem for us here in Europe or are you in Europe because UK is not going to be in Europe soon so which kind of relates to what is the probability of that really happening on a theoretical kind of basis okay it's a study in making those decisions right but in practice I mean the practical examples in another context but the practical examples we are seeing is that we have in the areas where there's a lot of pedestrians we have protections like big big plants with big concrete containers or yeah well actually security measures are implemented for this example and there's also also this is papers by Max Ludd an enforcement of security measures for large governmental buildings in US but it's US or Australia right I'm asking in European context should we be scared right or is there a regulation we just don't know or I don't know about but I'm not a person to answer whether we should be scared did the only thing I can do is help me to study such studies and point to official measures which was a good answer Michael well one thing one way to go about the unknown threat probabilities is to look at the unknown probabilities or let's say discretize that you have there in your model which is the typical way to try to assess what if we have this level of threat or the other one but then to consider these to be outcomes of possible unknown systems so given we have one system the the threat level is this another system is this such system as this but we don't know and then you can introduce let's say a uniform prior over the unknown systems and you can optimize decision making looking at the future of course but also looking at just like if we are looking at climate change scenarios uh buying options for adapting to uh knowledge which will come in the future of relevance to understand better which system is actually realizing itself right and I think this is the only thing we can do maybe to ensure that we are able to adapt to the future as it evolves and we get more knowledge right and take into account the different possible systems to the best of our knowledge and if we don't really know then in principle we have a range of different threat probabilities to to to take into account in that way we can optimize decisions and begin identify decision alternatives which are robust