 It would be really useful for us to hear the feedback. Are we just adding model on top of model, one error on top of another error? I'm sure many of you may feel that way. Or are we actually providing new insight for policy makers? Is getting to those PDFs on growth for Mozambique going to be something which can be used for policy making? Is it a step forward? So following tradition, maybe we start on the left and we move to the right. There's a question right at the very back and then we'll come to the front here. Josh Busby, University of Texas. My question, I guess for any of you about the probability distributions, are they based on multiple model runs? Is that how you're able to generate those? And then the other question was to elaborate a little bit more on return frequencies with respect to either drought or floods. What that means. Thanks. Thank you for your presentation. I just tried to complicate more the model. Another layer. Because for example, country like Mozambique you're thinking you're given two for policy makers. For example, suppose the government of Mozambique you look at this, okay, what kind of decisions I have to make having this information. And one issue is like countries like Mozambique, they are kind of climate takers. They don't influence much the uncertainties and also the levels of policy that would come. And for these people, huge policy uncertainty. They don't know what's the probability it would be level one, level two, level three, level five, and even the business usual. And for them, you know, much of what they present is interesting, but they don't know what's the best way to do it. If for example, suppose they choose, okay, level one would be the most likely and they invest a lot of money in kind of very complicated, then they cost three times more and they'll be locked in actually is going to be business usual. And how you bring in this kind of uncertainties to the model that actually be, I think, more useful for policy makers. Thank you. This is Sudhakar Jaila from India. You have taken mostly the sectoral approach in assessing the impacts as such, especially the CGE modeling I'm talking about. There could be migration coming out as a result of sea level rise or there could be migration coming out of impact in agriculture, and that certainly can have a large impact on the economic performance of the country. I wonder why that interaction between sectors is ignored in assessing the overall impact on the global, I mean, the economic performance or value added to the national economy. The second thing is there could be some positives coming out of, say for example, impact in agriculture. There could be changing agriculture pattern may end up having a crop which is more economically viable, say, biodiesel plantation, for example. Any such possibility could be adding to the economy rather than adding negatively to the economy. Did you try those things in your model, please? And one more question, sorry Clemens, the person behind you will come back to you in a second. Thank you, great presentation. My name is Raba Haraski, IMF. I realize that you do scenarios without unchanged policy, but I wonder whether in your model there's any private sector reaction. And the reason why I say that is, I mentioned earlier this morning that of course one of the big developments these days is these large land acquisitions of the consolidation of land for the purpose of potentially using more modern agriculture. So I wonder if in anticipation for higher food prices, higher commodity prices, the private sector may not force the modernization of agriculture and in a way mitigate it, but I'm really separating the mitigation from the public sector rather than the anticipation that the private sector may react may indeed lead to some sort of positive spillover out of the risk generated from the climate change. Thank you very much. I think what we'll do is we'll turn back to our panel. We have six questions now and I'm going to have an attempt to try and allocate the questions, maybe to make it more efficient, but feel free to chip in on others. Adam, maybe there was a question from Joshua on HFD and multiple model runs and then also a question on, global policy uncertainty, which I think is an extremely good question. That's an uncertainty we haven't got rid of in the analysis. And then turning to Ken, a question again from Joshua about return periods and flooding and also on agricultural zones and how they may shift and to what extent we're capturing that in the analysis. And then finally Channing on migration question from the front and also on the role of the private sector in the analysis. Okay, so the first question about how I guess really essentially how these ensembles would generate, it's actually, I'm really glad that was raised because I really didn't have time to sort of really get into the details. So essentially what happens is we take the economics model and the earth system model and run them linked. They are linked by this emissions, by the emissions coming from the economic model. And the idea again is going back to those, the one slide where I showed the sort of the parameter space, if you will, of certain aspects of the climate system that we feel are uncertain. So one of them is climate sensitivity. We have a range of climate sensitivity based on work that I showed and we use a Latin hypercube sample of that parameter space. And so essentially when you do that you can make your bins really small and your resulting ensemble size can be on the order of 10,000. So if you take each one of those uncertain pieces of the puzzle, choose your bin and do a Latin hypercube which basically means you're separating these distributions under equal probabilities and run them all, you get a very large ensemble. We've done approaches similar to what Channing mentioned, the Gaussian quadrature and we found that when we run about 400 through this Latin hypercube sample we can span the range of parametric uncertainty in our model. So it's not, again this is not particularly a model, it's a model framework. It's a framework that's trying to capture the uncertainty of the structural uncertainties, the parametric uncertainties that we see in climate models. So yes it is one model, but the approach is done deliberately to try to span the range of the structural uncertainties that we see in climate models. Hopefully that addressed that question. The next one was about the, yeah so, wow, so I agree with you that fundamentally what we do with policy is very exogenous and what do I mean by that, is that we are predetermining what the policy future is and we have no way of really letting that be dynamic. In other words, we march along and we say well wait a minute, that didn't work, so let's stop what we're doing and do something else or try a different tact. I guess probably the best way to answer your question or address your question is that I would say holistically if you could take this approach and run it across a range of policies that you deem are plausible, realistic, could happen, maybe not happen, you might get a sense of what the policy uncertainty is. But the problem is you're just adding distribution upon distribution and you may end up with something that's completely white noise. There's really no, there's nothing to glean from it other than anything is possible in the future. So I think at least in terms of trying to understand what we see in the model and what we can interpret from the model, the approach of looking at the stabilization scenarios is again, resonates with the IPCC community which is we need to avoid this much warming or this much accumulation of trace gases in the environment. Those are the sorts of approaches that we've taken. We have used the model to look at things like biofuel policy. What if we embrace some penetration of biofuel into the energy sector? What does that mean for land use change and the dynamics of land use change between what we need to plant to feed ourselves and what we need to plant to generate as much bioenergy as we think we need. So our approach is certainly fluid in that sense. We can adapt to different policy scenarios but we certainly do have to pick a path before we put this model ahead. But it's a very important issue that you raise. To add on to what Adam was saying about the policy uncertainty. One of the things it does is it allows the countries themselves to look at how vulnerable they are in particular in terms of if we're just looking at reservoir design and we see that over all the ranges of them you still have a significant negative probability or frequency of events the country can make their choice of how to hedge that bet themselves in terms of design. And one of the big things it's leading to is since we don't know where we're going is this whole idea that's coming out in the world of infrastructure of flexible design. And we maybe instead of overbuilding and have a regret of overbuilding or underbuilding without room for adaptation or adding to it is spending a little bit more now to be able to add on. So an example would be we see it in many places if it's a foundation we make the foundation big enough to handle another 20 or 30 feet of meters of dam on top if we do get this extra water or it's very cheap to leave a hole for another turbine but it's very expensive to put the turbine in if it won't be used for 50 years or maybe it's never going to be used. The other thing I think it happens is it helps the one of the things we've got feedback from is people involved in the climate debate and the mitigation debate to see how their country is faring under different policies. So if they can see that this policy they're a climate taker that they're going to benefit from these policies going on or if they're not. So it allows them to see those possibilities and use it on the mitigation as well as the impacts and adaptation side. So that's something that happens there. In terms of the return periods there is a a nice thing that has happened with this work is that there is a standard for design in structures in water resources and in roads an example is that generally for roads depending on the type of road you'll have something which you'll call the 1 in 10 year return period which is probability of 10% every year that your structure would fail. And we designed to that in the OECD countries we use 1 in 50 for flooding we use 1 in 100 in certain high important high valued areas. What we're finding in developing countries is they may have these standards in handbooks on the shelves but they're not being implemented. So basically developing countries are under ensuring their infrastructure but we understand why is because of the lack of capital or what we see happening is instead of building 200 kilometers of road in a five year a five year plan they'll want to build 4,000 kilometers in that period to a lesser standard but what happens is we're finding that's the damages so as we look at these return periods we're seeing that there is not enough protection to current climate variability and the best way to adapt to the future climate changes is just to adapt to the current variability you're doing and as you look forward we're finding out in certain cases it's only a marginal cost to go a little bit higher as Channing pointed out in a study in La Sabre in Honduras to make the 1 in 20 year storm event for putting in the entire drainage system for the town of La Sabre it was 5% more than doing the 1 in 10 year storm so you're getting 20% 100% increase in protection now and the projection is by 2050 the 1 in 20 will become the 1 in 10 so we have to look at these things in sectoral areas as we do it and working with the ministries of planning and the ministries of water resources or transport to look at how we deal with these climate issues so that's kind of how we look at those and then in terms of spatial areas the models are looking at the spatial areas and we model the impacts on all crops so that if in the model there is an advantage or certain crops are losing their advantage or gaining them those are in the model for the economics to choose them with the autonomous adaptation so that information or those response curves, the change in the production functions those are in there in the models making those choices as they face them in the future I'll try to finish off quickly in terms of migration labor is moving all around within the country we don't have an endogenous they're not going to South Africa more and remitting back and that's just a choice that we've done but labor is definitely urbanizing throughout the whole period it would move out of areas that are not doing as well and into areas that are doing as well and that gets to the private sector reaction the private sector is reacting looking at what's happening to return to investments across a whole slew of sectors and so if certain sectors are doing well then investments going to go in that direction an important thing to point out is the model has productivity or whatever your impacts are getting is one determinant of investment the price is another and so it can be the case that climate change is driving down the productivity of a certain sector but it's driving up its price even more so instead of investing away from a highly affected sector you're investing into the highly affected sector and that does happen within the model Thanks very much Clemens we promised you a question and then we're going to have to go over to the right and we'll try and come back if we have time but I'm not sure if we do Clemens in the middle I'm Clemens from IFPRE Adam I really like your representation of the wheel I was really tempted to start spinning the wheel see what's going out of there now my question to Jenning first of all what does a wheelchair wheel look like for Mozambique could we represent our choices in terms of the wheel that's the first question and the second question if we put that wheel in front of a policy maker would he not say do you know how many wheels I have to spin all the time and since this is a conference on climate change and development policy I'll just give you an example if that is too abstract a policy maker may say what happens if I don't invest in education then my education level is going down if I don't invest in health then my child mortality will go up and so on there may be broader choices to make thank you Thanks Clemens over to the right person in the very back on the right and then we'll come to you Anthony Milner from the London School of Economics I just have a question about your sort of modelling strategy so it's obviously an extremely complex model and as we know models with very high dimensional parameter spaces very often are subject to overfitting and may in fact do less well than much simpler models of predicting aggregate variables so I just wondered the climate problem and the fact that you sort of dealing with the problem that we haven't seen yet whether you have any thoughts about the possibility of using past data to validate your model in some kind of out of sample test or at least components of the model if any comments on that would be very welcome thanks right the gentleman here on the left sorry I'm making we'll come back to you in a second the woman here with her hand up okay thanks for your presentation about the model first could you please introduce more about the regional response to the global climate change especially how you model this is this GCM model if it is can we access to your model and second is for the CGE model how many sectors and regions do you consider in this model and also is your CGE model published can we read some papers about your model thank you very good I'm sorry we sit just here in the middle and then we'll come to you thanks my name is from the Center for Human Rights University of Pretoria now my question is simple you know I can see somebody on that slide I don't know whether I say he or she you know is this on firewood or on a head then the question that comes to my mind is how do we make this model meaningful to local populations who are all bear the brunt of this development that we talk about you know is it just elitist model or is it a model down to heart we're talking about balfors we're talking about hydroelectric projects all these projects negatively impact you know people particularly in Africa you know it promotes it promotes land grabbing which is a phenomenon which is rampant in Ethiopia, Uganda where all these projects are being implemented now if we're if you're a faction in a model I want us to humanize this model let's put humanity on its face let's put the feeling of these local populations you know let's put their image to it let it make sense to them and I think I don't know whether it's a comment or a question it's the first time our models have been called elitist again elitist in our ivory tower it's a good US political issue here in the front just behind you my name is Nidhi, I'm from Liebnis University Germany I have two points first point for the first speaker about the model see generally it's been noticed and argued that the global models don't perform well in tropical systems and as your area of interest is the tropical region so what was your observation and you think that's CMAP 5 which is the update on CMAP 3 will have more possibilities and potential to project better in tropical conditions point number 2 is about we've heard perspectives on biophysical climate and economic analysis but often in discussions issues on livelihoods poverty, gender remains under or unaddressed we're getting a holistic approach towards development policies so inviting a thought on that I think what we'll do we've got again six questions and so we'll put them back to the panel we have Clemens' question about the wheels and how many wheels can we spin at the same time or how many wheels can we give to policy makers to capture the trade-offs we have a question on historical validation of the models and on the building strategy Channing maybe you handled the multiple wheels can if you talk about maybe historical validation although I'm sure everybody wants to chip in on that there's the regional response question and how that's factored into the IGSM so Adam maybe you can tackle that one the elitist model question which I think the question boils down to to what extent are we capturing the human dimension in this analysis in what is already a supermodel framework can we go one step further again I'll leave that for all three panelists to comment on a question on how effective is the model in capturing tropical climates and that's definitely for Adam and then finally on and again it's somewhat I guess related to the human dimension which is to what extent can we incorporate gender and poverty considerations and models that project out to 2015 so Adam do you want to have a crack to begin with or shall we start with Channing oh okay multiple wheels okay yeah I agree with Adam that he made the point once you start to do too many wheels then things can go awry we could easily build a wheel just like they've done we could take okay what GDP do you want and look to divide up the PDFs that we had and stick them on wheels and we would have two wheels just like the climate wheel and it would be the GDP wheel right so you would you get you spin and you get this wheel or you spin and you get another and the L1 level one stabilization wheel has a lot more favorable outcomes and a lot less dispersion you know that would be possible I think you know one of the things that came to mind as you were asking the question is to you know what we're coming up with with the scope of impact by 2050 you know in the very worst case we had 10% right around there and remember the baseline growth rate is 5 so in the worst case we're two years delayed right by 2052 we're going to be at where we where we would have been okay that that's the so it's something and they're losing it every year so when you present value it it's it's it's there but but there this is this is the level of impact that that we're getting now that's out to 2050 and I think this speaks to think one of the panelists this morning was saying oh there's a there's a difference between what the economists say and what the the social scientists or the the climate scientists saying and I think part of the reason is that the climate scientists talking about you know 2090 or 2100 and and we're talking about 2050 right if you get if we get seven degrees of centigrade of warming by 2090 I mean my imagination is not that fertile to model that I don't know what to do it's it's it's so far beyond anything that we've seen that you know we've just I haven't gone there so I think that that time to mention is really important to remember on on overfitting this is a I'll take that one the these are structural models so we're not we're not estimating them we take them out you know and and we're following in a structure and this allows us various policy parameters I wasn't was I assigned the validation question you go ahead and then I'll chip in so I'll stop there to to tag team here and take the baton in the biophysics we are using the standard water resource and agricultural engineering tools that are a state of the art for doing design now by engineering firms that go in and use them and so the first principles of those is we validate them with historic conditions so all of our work is validated to start with the conditions and there's a whole way to do that and then we move forward with the climate driving them there is the the risk that some models of hydrology or there are others that the change of climate and this has again is towards the end of the century can be so far out of what we've ever seen that how valid are our models that's a risk that we take but generally by 2050 we are not seeing things much different than some of the extreme values we've seen in the past so we feel pretty confident that these models are representative of what's going on and on the other scale is there is the ability to bring in to some of these models depending on the scale human aspects and one of those particularly is flooding because those who are hurt the most about flooding are the poor and so when we do things with flood mapping and showing where you're getting flooded are generally where the poor are and why they may not show up in the GDP they show up as numbers of people affected so we try to work on that as best we can and as you can see in some of this work as we're looking to economic development we're aggregating up but you can go in the other direction and these models can assist in that as well it's all on the question being asked but it's an appropriate question and that's why human health and some of these things are being brought in and have been used and some of the work that Adam's showed has been used for looking at health impacts of some of these impacts as well. Adam on the regional response in tropical validation. So I'm going to actually my response will sort of touch on both of the questions which I think were very poignant and related and really get at the heart of essentially what we're trying to do here which is I would say in a generic sense you could you know lift the hood on any model and find a whole can of worms it struck me that you sort of point to the tropics because I have a number of colleagues who feel like the tropics are our best hope for climate models that you can actually get a signal out of the noise if you will the intrinsic model noise from climate models that actually can indicate some form of impact that we can really rally around and say yes all of our models are consistent with this signal and it's coming out of the tropics and people in the tropics are not used to a lot of variability and climate to begin with and if you see this signal come out of the noise and this is something that we really need to point to so I find it very striking that you feel as though that's a weakness of the models and yet there's another big chunk of the community this is a great opportunity so I think that falls under the general categories of as I said before you know there's no perfect model and I think we all wish that we could hang our hats on that one model and say this is the one that we're going to go with but I'd like to think that our approach holistically tries to encapsulate all of these picadillos if you will better in models and as you pointed out there is another round of the C-MIP of the C-MIP exercises that's out we certainly didn't mean to be disrespectful of C-MIP 5 in our study it's just that the study happened a few years ago and the C-MIP 3 data was what was available to us but certainly we are looking to the C-MIP 5 models along these lines. I will say though that some of the initial findings of C-MIP 5 against C-MIP 3 show that for the most part the climate models show a big difference and I'll generally describe their skill we don't really see a real salient change in the way that these climate models are operating. I'm not saying that that's necessarily a bad or a good thing because these models are far more complicated than they used to be so there's a lot more information that we can take from them but in a generic sense with the climate variables that we're comfortable with and we have a lot of experience with we don't see any real difference in their skill so far with regard to the regional information that we've garnered from these runs the regional details of the model are you could say an algorithmic extension of these more sophisticated climate models. We don't actually run well in our framework we can run a climate model but to get these thousands of different realizations we have to use an algorithmic approach this Taylor expansion and all it basically says is that we find we try to find some emergent behavior on a regional sense from these climate models and we put it in in terms of this Taylor expansion form so there's not really a model running per se but we're trying to discern some emergent behavior or response from the model. In terms of its availability we're sort of a research institute right now we have the global model results available. In terms of these regional results yes the data can be provided gladly to those who are interested for other regions it's sort of a well if this is our new focus then we will generate the data for that region the issue here really and as much as I hate to admit it it's just sort of storage capacity you know we generate terabytes of data to provide for these simulations and it's hard to sort of keep that metadata available at any one given time so for those of you who are interested in the information that we've used the climate information that we've used for this study I think that's something that we can arrange. Okay well thank you very much we've reached the end of the day and we can certainly say thank you very much to the panel.