 So, we'll be talking today about a portion of the research that I focus on which is really how to understand the health, environmental and climate change consequences of different types of strategies that we may pursue in the electricity system and focusing on the United States for part of that research. So, as you probably know, energy services contribute to the largest amount to greenhouse gas emissions and we currently in the U.S. energy system have the emissions being split kind of 50-50 between transportation and electricity and in that regard the transportation emissions just surpassing those of electricity by a tiny amount right now. And so the question that I'll post to you in terms of starting thinking about different types of interventions in the grid is tell me what you'd go for. So, should assume that you have a budget constraint and that you need to think about where to cite the next set of PV panels. Are you going to install those in California or in Pennsylvania? So, I want some real answers. What do you think we should do? Both? Assume that you can do both. Your budget constraint, the next few set of panels you'll need to put them in one location or another. What do you do? California, Pennsylvania. So, a few of the folks that mentioned California, why California? It's sunny. The resources are good. Pennsylvania, why Pennsylvania? What? Too many panels already in California. Okay, what else? A huge demand on these panels. Okay, more. Yes? It's cold. It's cold? This place? Cold. Okay. So, you could have answered me with another question which was what's the goal? Why are we adding those panels in the system? And if the goal is to maximize electricity production you'll go for the places with the best resource. But if the goal is to displace health and environmental impacts and climate change as I had in the first title, maybe we want to think about what is being displaced and what is being displaced, namely at the margin as we incrementally add new technologies. Again, let's think about another technology. Electric vehicles, great for the environment, right? Should we invest in electric vehicles or hybrids in the United States? And does our answer change if we think other geographical contexts like China or India? Right? So those are the sorts of models that I want to build in the dress. Where do we get the largest environmental benefits from deploying wind? Is storage green technology? And are we really helping the environment by deploying storage? Can we think about novel ways of pursuing mitigation strategies? For example, shuffling the ways we use data centers for video streaming. What would be the potential from that? I'm not going to be able to go over all the examples but I'll pick a few for us to cover. But the idea is how can we measure how different interventions are going to affect the emissions? But do we care about the emissions themselves? No. We care about the damages incurred from those emissions. Let that be greenhouse gases or criteria pollutants. And the first point that I want to come across is that we can't talk about the greed in terms of average emissions factors because we have a big diverse greed where the carbon intensity and emissions intensity is quite different from state to state or a subregion to subregion. And we can't talk about damages from air pollution as one single number because that too differs widely across the country. So the same ton of SO2 emitted from a stack may have health damages in terms of premature mortality ranging from $1,000 to $15,000 depending on where you are. Why is that? Why does the same ton of pollution has such dramatic different effects? Again, not the rhetorical question. You guys can actually jump in. Population density. That's a big one, right? And then also the patterns of just the dispersion of the pollutants and where you have the highest concentrations. OK, so we have those effects that are different regionally. And so in order to track what the implications of our choices are, we need a couple of building blocks. And the first one is to try to understand how do power plants operate and which plants are operating at the margin in different regions and in different hours. So in order to do that, one of the strategies that you have is just use simulation tools and build dispatch models or even better representations that account also for transmission constraints in the grid. But a very simplistic model will go something like that. You have a merit order dispatch where you're dispatching your plants from the cheapest to the most intensive, the most expensive. But you want to bear in mind what the emissions intensity, and here I just have the example for CO2 of those plants. So imagine that you're just pursuing an energy efficiency plan that is decreasing demand just a little bit at the hours of peak demand when demand is 12 gigawatts in this made up system. OK, maybe the case that you have an actual gas power plant at the margin. And so those are the emissions that you're avoiding. Whereas if you have another intervention or energy efficiency plan for lighting that is actually changing load at an hours of moderate load, it may be the case that you have a coal power plant at the margin with much higher emissions intensity. So you have a way to track that. And ideally, track that as close to real time as possible and also taking into account how the operations of the system are changing due to plant retirements and new plant construction. So this line of work basically does the following steps. One, to have a representation of the grid operations for all its power plants that are fossil fuel-based and to identify which plant is operating at the margin at different hours and different locations. Second, merge that with information about the health, environmental, and climate implications of those plants by coupling it with air quality models. And finally, plug-in interventions that are changing at the margin. What is the plant that is being operated because you either have more solar PVs, more wind, vehicle electrification solar, more efficient buildings, and so on? So let's go back to our original questions. So do we add solar panels in California or in Pennsylvania? Well, a little bit more on how we went about modeling this. First of all, we had information about the damages in terms of dollars per ton of premature mortality associated with the emissions from plants by stack height and for each of the pollutants, SO2, NOx, and PM2.5. Again, this is kind of mapping how those damages will be different across the country. And for that, actually, in the current research, we're using several models at the time. And for the result that you'll see, we're using AP2, which is the model that is used also by the national academies. It is a model that accounts for the dispersion of the pollutants, the chemistry in the atmosphere, the changes in concentrations, and then couples that with those response functions of people exposed to those pollutants, accounting both for health and also environmental effects, though the health effects are the predominant ones. And finally, you need to have some judgment assumptions somehow, because ideally, you want to put everything in dollar terms to be able to compare strategies once across another. So we monetize those impacts by using the value of a statistical life. And for the damages incurred from greenhouse gas emissions, that, again, could be an entire lecture and discussion about what's the value that you place on the social cost of carbon. At the time, we use the guidance of an actually fairly low price of $20 per ton of CO2, which was the recommendation from the US agencies. For all those plans, we have their geographical location and hourly emissions. This, again, is something that is a little bit easier to do in the United States versus other locations. The United States, through the EPA, publishes hourly measured emissions from the stack for all power plants that are larger than 25 megawatts, both for SO2, NOx, and CO2. It does not include direct primary PM2.5 emissions. So for that, we did need to do some assumptions of, basically, we assumed that the emissions of PM2.5 were proportional to the plant generation output. Much more research can be done in that regard. OK. And so instead of building a simulation model based on dispatch and transmission constraints, what we've done was to actually take advantage of that we have all this granular historical data on how the power plants have actually been operating. And so the first step was to have a metric of the hourly damages associated with the plant by simply multiplying the damages by the hourly emissions. And then we did a regression-based approach. So we parsed the data into 20 generation means. Let me zoom in. This is an example for Texas, for ERCOT, and just for SO2. And what we're trying to capture is what's the change in damages that you incur, as you have slight changes in generation? So we observe that in hours of low demand, you actually have a factor of $20 per megawatt hour for SO2 damages. So this means that in those hours of low demand, we're capturing some of the coal power plants that were operating at the margin. Whereas in hours of high demand, it is virtually zero. Because we don't have coal at the margin. We have natural gas. And so there are no SO2 emissions. So we do that for the different hours and different regions. In a nutshell, if you want to see two-figure, again, back to our made-up dispatch model, we're trying to capture what is the change in the emissions intensity that we see in a certain time when we have a small variation in total demand or actually total generation, in our case. So this little change in emissions based on the change in generation. We do that for each of the agreed subregions. And we select the agreed subregions, because those are historically regions that are fairly independent. So there is not a lot of exchange. Taking it into account imports and exports and their emissions intensity is quite challenging. So we kind of left it out. And so by doing that for each region and each pollutant, ultimately, we're able to have a profile of what are the damages in dollars per megawatt hour as a function of the total generation that you have in a region. Again, this is for Ur-Cut. So this first step enables us to map the damages for the electric system in the United States as it is currently operated. Now the second part of the puzzle is how do things change as you incrementally increase renewables or storage or you have vehicle electrification? So in the case of the solar wind, we had hourly simulated output for 35,000 locations. And so we match the output in those locations with the electricity generation that gets displaced or avoided to be generated by the fossil power plants in those same hours. So now some of the questions that you could ask is, OK, but are you really representing the real system by doing it that way? And it's a pretty good approximation because we're using this historical data on how plants actually reacted and how they operated at the margin when they saw those fluctuations in electricity generation. So at least for a fair amount of renewables, it is not a bad proxy. And so we match in each of those hours of output what are the damages being specifically avoided by each of those pollutants. And then we aggregate it all up to have a sense of the damages avoided per megawatt of solar PV capacity or wind capacity or per megawatt hour of electricity generated during an entire year. So let's jump in a few results. In the case of solar PV, this is not surprising. This first slide is probably not surprising to you. Of course, we get the most output in the regions where we have the best solar resource. But the interesting thing is that those don't align with the best locations. If your goal as a decision maker is to avoid the effects of climate change or improve the health and environmental consequences associated with air pollution. So if your goal is to maximize electricity production, then the best resources will be Arizona, New Mexico and Southern California. Indeed, the solar PV panel in Arizona will provide something like 45% more electricity than one in Maine. Now, if you want to avoid CO2 emissions, the story is quite different. And you'll locate your solar PV in Kansas, Nebraska or the Dakotas, right? The solar resources aren't that great, but at the margin, you'll be avoiding the generation from coal power plants. And finally, yet again, you select to place your solar PV systems in a different location if you want to reduce the health and environmental damages associated with air pollution. Indeed, a solar PV panel in Ohio, which you wouldn't think as the first place to deploy solar PV, would provide 17 times the health benefits of a solar PV in Arizona. So let's look at the story now for wind energy. So in this case, you'd be better off by citing your wind power in the great planes and through West Texas, you have really good capacity factors in those locations. And they align reasonably well with the locations where you'd have been able to avoid the largest amounts of CO2. The Midwest has good wind resources and at the same time still coal power plants operating at the margin in some hours. But yet again, if you want to maximize the reduction in damages from premature mortality, you would locate indeed your wind farms in different locations. A few examples with quantitative results, a wind turbine in West Virginia would display seven times the health effects of a wind turbine in Oklahoma and 27 times those of California. Okay, we can pause a little bit and think about we have all those sets of incentives that are being provided to energy technologies are they paying off? And so one of the examples will be to think about the annual benefits that we get from wind farms that are installed already. And so we did this analysis of now looking at existing wind farms, looking at their hourly production, matching it with our model and competing the overall health, environmental and climate change benefits that would have over the course of a year thanks to having those wind farms. And we come up with the rough estimate of 2.6 billion. And now we have PTC, the production tax credit subsidy that was provided to wind developers for wind generation and a rough figure for those would be $1.6 billion. So this is good news. We're getting more in terms of the health and environmental and climate change benefits than what we're spending in the incentive. Now let's zoom in a little bit and look at two different regions, Pennsylvania and California. We're picking a little bit on those two. Well, the health and environmental damages that would be avoided in Pennsylvania would be almost $90 per megawatt hour. And they are just receiving the 22 cents per megawatt hour. Whereas in California, because their grid is already so clean, you're actually avoiding a fairly low amount of damages. You're still getting the 22 cents per kilowatt hour in terms of subsidy. Very briefly, a lot of people think about storage in the grid as a green technology. That would really depend on how we use storage and how it is operated. So there are a couple of policy pushes for using storage, namely over here in California with the goal of having 1.3 gigawatts of storage by 2020. Now, let's think about one of the uses of storage which would be for energy arbitrage. What's this? Well, the storage operator is gonna charge the device when prices are low, and it's gonna sell the stored electricity when the prices are high to make money. What is generating electricity when the prices are low at the margin? Well, in large portions of the country, it would still be cold. What is generating electricity at the margin that you're displacing with storage? Well, it may be natural gas or even in some locations, renewables. So overall, if you use average emissions factors, you're assuming that storage is a green technology. If you account for the timing of charging and discharging and what's at the margin, that's not quite right. And the second effect that is a little bit smaller but is that you'll have energy losses by operating your storage device that also consume electricity and so you need to account for that penalty too. So we've developed a linear programming optimization model for energy arbitrage. You run that across the country by using price data and the marginal emissions factors that you've seen previously. And the first thing is that, yeah, you can get a little bit of money from using storage for energy arbitrage, but the key message is that you increase emissions of CO2, NOx and SO2 virtually everywhere across the country. It's not a green technology anymore, at least not for the energy arbitrage applications. So I'll end with a couple of notes. The first one is that we decided to make all of these data and results publicly available for any modeler that is interested in understanding the effects of other energy technologies. A few folks are already using this for real decisions of procurement of renewables and where they want to do that for their climate action plans, which is great. And a couple of final, more big picture notes which are that we really need a major, enormous transition to sustainable energy systems, not only in this country and globally. And in order to do so, really looking at this in silos, just at climate policy, just at air pollution or just at waste really makes no sense and you can actually head into directions that are misleading and actually leading to unintended consequences. We observed that location, temporal patterns and behavior are all jointly going to determine the health, environmental and climate change effects of these interventions. And today, like never before, we have enormous amount of data that we enable us to make wiser decisions. This was not the case 20 years ago, but we can do that now and at a fairly low cost. And finally, that openness in the stack of research is, in my view, critically important so we're making all of this available to all of you. Thank you so much. Thank you very much. Yes. The unclean storage of energy, if you only charge it during the day between 9 a.m. and 3 p.m. when the solar is at the highest and it's definitely like a clean thing going in and then you just charge it 4 to like 7 p.m. Then it might not add and it's reduced CO2 emissions? Correct. So yes, things may as well change and they are changing. Like some of the results that I showed you that pertain to 2013 or 2015 in the figures may be already outdated given how rapidly the degree is changing. And so one of the goals is really to keep track. One of our projects is actually trying to track how much does this need to change for storage to be at least carbon neutral or emissions neutral. So you're quite right. It will change. It will depend on the charging. It is the fact that for the conditions that we saw until very recently, and if you're using it for energy arbitrage because the purpose was to be revenue maximizing, you would have problems and the emissions would increase. Yeah. And there was someone here. My question was purchased from not to, this one is that, is the fact that battery storage can in fact increase emissions only true in an energy market where there are still cold power plants? For example, in another country where there are no cold expectations, are still battery not good for the environment? Yeah. So the results are definitely specific to the conditions that we saw at the time. So to the US market for energy arbitrage and for the emissions profiles that we had at the time. If we think about France, for example, and charging the storage device with nukes, basically you don't have any emissions when you're charging it and you'll be displacing perhaps natural gas at peak hours. So it will be specific. The results that you'll find for different services and for different countries or region may not hold the same way that we've seen here. They will not, yeah. Hi, my name is Namdi from Nigeria. Curious as to how you monetize the damages like premature mortality and with the dollar value, you just set a random reference point and compare different areas or is there a way you do that? That's a great question. So we used the recommendations from the, I think at the time it was the EPA of $6 million and I won't remember the base year that that was a triple little too. So the $6 million would be more or something like $10 million right now. So adjusted to today's dollars. But this is such, it is a judgment call and when you're deciding how to evaluate those things. Now, since we used one single number across, we then rerun the model also using different values for statistical life, the higher bound that we've seen in the literature, the lower bound we've seen in the literature using different dose response functions too. So all those sensitivities and ultimately for the air pollution, you could present everything in terms of numbers of additional mortality and that's premature deaths rather than putting a dollar value, but if you want to start conversion, that discussion with the one of the damages from CO2, you need something. It may not be dollar units, but some sort of weighting factor that brings everything into common units, right? So we went to that, but that's a big part of the discussion. Should we go with that? My strategy has been to really present all the research in those papers using also lower bounds and higher bounds, but then defer to the actual policy makers on what they should use to guide their ultimate decisions. Well, thank you all and once again, welcome.