Climate Projections as a Way to Illustrate Future Possibilities

Video Transcript
Download Video
Right-click and save to download

Detailed Description

This video is a recording of the webinar “Climate Projections as a Way to Illustrate Future Possibilities" that is part of the USGS National Climate Adaptation Science Center (NCASC), in partnership with the National Conservation Training Center (NCTC), 2019 webinar series. Previous recordings can be found at


Date Taken:

Length: 00:52:45

Location Taken: US

Video Credits

Adrienne Wootten, Presenter
Laura Thompson, Introduction
Elda Varela Minder, Organizer



 John Ossanna: [0:05] Welcome. My name is John Ossanna. I'm here at the National Conservation Training Center in Shepherdstown, West Virginia. Welcome to this edition of our web series in partnership with the National Climate Adaptation Center for the US Geographical Survey. 

[0:22] Today's webinar is titled "Climate Projections as a Way to Illustrate Future Possibilities." We're excited to have Adrienne Wootten from Oklahoma University with us today. To introduce our presenters, we have Laura Thompson who is a research ecologist at the National Climate Adaptation Science Center. Laura, take it away. 

Laura Thompson: [0:42] Thank you, John. It's my pleasure to introduce for today's NCASC seminar, Dr. Adrienne Wootten, who's a postdoctoral research associate at the South Central Climate Adaptation Science Center. 

[0:59] For nearly 10 years, Adrienne was a student researcher at the North Carolina State climate office and worked closely with the Southeast Climate Adaptation Science Center on projects involving both scientific research and stakeholder engagement. 

[1:15] She earned her PhD in atmospheric science in 2016 from North Carolina State University, and joined the South Central Climate Adaptation Science Center in 2017, as a postdoctoral research associate. 

[1:29] At the South Central Climate Adaptation Science Center, or CASC as we call them, Dr. Adrienne Wootten's research focused on the accuracy and uncertainty associated with climate modeling and downscaling, with added emphasis on the use of climate projections and impact assessments in decision making. 

[1:50] She also works with the Tribal Liaisons and other South Central CASC staff, in outreach and extension efforts to help ensure that research meets the needs of stakeholders across the South Central US. 

[2:07] Adrienne is also the editor in chief of one of the founding members of the Early Career Climate Forum, which works to promote communication and collaboration among graduate students, and early career professionals, and climate science, and related disciplines across the United States. 

[2:24] Adrienne, we're very happy to have you. I will hand it off to you. Thank you for being here. 

Adrienne Wootten: [2:29] It's my pleasure to be with you all. I thank you for the invitation to present to you this afternoon. The climate projections are powerful tools in climate adaptation planning. They can tell us many stories of how the future climate could unfold, and illustrating the possibilities that we could have to adapt to. 


[2:48] What I'm going to do today is to give a little bit of understanding first on the climate projections and stories they tell. Why we get so many different stories and possibilities that come from the climate projections? Then illustrate a bit from my research on what this means for projections of precipitation in the South Central United States. 

[3:07] Then talk a little bit more at the end here about what that would mean in terms of impact assessments, decision making, and some of the things that users should take into consideration. 

[3:19] On the first hand there, of course, when we're talking about climate change, we're talking about changes in the measures to climate over decades or longer. In measures of climate we could be talking about the average annual temperature, the frequency of extreme temperatures, or extreme precipitation. 

[3:36] All of this is well and good, but why do we worry about understanding a changing climate? Why does understanding a changing climate matter? A changing climate changes the frequency of droughts. Then we have to consider what does that mean for ecosystems and agriculture across the United States and the livelihoods of the people therein? 

[3:57] It changes the average high temperature which could mean potential detriments to human health for multiple communities across the United States. It changes the frequency of heavy rain events. Which again presents another challenge for protecting life and property, as well as protecting infrastructure maintained by the federal government. 

[4:18] Perhaps on a more subtle level, it also changes the lengths of the growing season or changes the timing of first leaf and first bloom. Particularly for USGS and interior interests presents challenges with phenology and ecology, with regard to the management of our natural lands and different endangered species. 

[4:39] All of this, of course, leads to the wants and desire that has been increasing in the last 10 years, in particular, have high projections available for use and assess the impact of a changing climate across the United States and indeed increasing globally. 

[4:56] This is projected change, the annual average high temperature here in Oklahoma City. I should say this is one projected change. One future story that could unfold. This goes all the way out to 2100 and shows the change in temperature in degrees Fahrenheit. Now, this would be great if this were the only climate projection we had. It'd be something really easy to plan for. 

"[5:18] It's going to increase by about six degrees Fahrenheit at the end of the century." That's great, one item to work with, but would it be the right thing to do? In reality, in working with the climate projections, what we really have is a picture that looks a lot more like this for Oklahoma City. 

[5:35] This comes from my collection of climate projections produced in-house here at the South Central CASC, of which that one black line in the middle is a single member of that set of projections. We have many future storylines for how the climate could unfold. 


[5:51] The interesting thing here is if we plan for just that one in the black, then does that potentially leave you vulnerable to something worse? An increase of up to 15 degrees Fahrenheit by the end of the century, according to one of these particular storylines, is a possibility. 

[6:06] That leads to concern of just planning for one story is not necessarily the best way to do it, but looking at the nice stories that the climate projections can tell us, and being able to plan for multiple scenarios that could be good or could be really bad. 

[6:22] Of course, the question that I get next is, "Why are there so many different future storylines? Why do I have to deal with this particular mess that we have to work with?" There's a few different reasons for that. The first of those reasons is human action, the hope, or trying to represent what human action will be, at large, globally. 

[6:45] Society at large, we have multiple options for future action with our own development as a society, as a planet, from continuing with the higher emissions to changes to less-intensive energy sources and sustainable development. These societal choices are reflected in the climate projections through what are called Representative Concentration Pathways. 

[7:09] The RCP here, 8.5, reflects the situation where there's continuing higher emissions of greenhouse gases. On the other hand, the RCP, 2.6 here, represents a situation where there is a decrease in emissions, sustainable development, or a rapid increase in technology for carbon capture, for instance. RCP 4.5 here is the middle ground. 

[7:34] The information from the RCPs is provided to the climate models, which then describe for us how the climate would respond to those human actions. From that whole mass of storylines we had before, we have at least three now to work with based upon just that information with the RCPs. 

[7:51] The interesting challenge then is there are actually many climate models that exist. This particular map has one circle on it representing each of the different modeling centers that has produced a climate model for the Intergovernmental Panel on Climate Change's assessment reports. 

[8:08] In fact, this is a little dated now because there's more centers that are now contributing for the Sixth Assessment Report, which is upcoming. To date, all total, there's more than 40 climate models that exist. Why do we have so many climate models when we actually have a really great understanding of the fundamental of the climate? Why do we have so many of these? 

[8:30] To understand that, it's important to go back to, "What is a climate model to begin with?" The climate models are built now, I think it's almost 150 to 200 years of scientific research and fundamental physics that describe the atmosphere, the ocean, chemistry, physics, thermodynamics. 

[8:46] The equations with the fundamental physics include things like conservation, mass, momentum, and energy, and then also things down to describing surface temperature and surface precipitation. 


[9:00] All of those equations are converted to a form that computers can solve and then solved over a grid over the entire planet to then give us projections of temperature, precipitation, and much more, going out across the globe to the end of the century and, increasingly, in the assessment reports for IPCC, beyond 2100. 

[9:21] When I'm referring to a climate model, often what the formal name is, we call it the Global Climate Models. These are mathematical representations of these many physical processes in our world. They work by dividing the planet into an immense grid of boxes, not just in X and Y, the latitude and longitude, but also on the vertical, up into the atmosphere and down into the ocean. 

[9:43] For each of these grid points, the model calculates all of those equations representing all of those physical processes, going out through time. It's a massive computational effort, and often uses an intense amount of super-computing for all these physical interactions and processes. 

[10:00] The interesting thing here, of course, is that with all of those physical processes involved, we can get back to, "Why do we have so many different climate models to begin with?" We have a great understanding of the fundamental physical processes behind the climate, like the greenhouse effect and the interaction with those greenhouse gasses and the atmosphere to begin with. There's also many more processes represented in the global climate models than just that. 

[10:27] In fact, of the matter, everything that you see in this graphic is represented in one form or another in the global climate models, everything from the formation of clouds to the air interaction between the air and ocean, to wind, and volcanic activity. 

[10:41] It's in all of these different smaller processes that are so important that we find the reason why we have different climate models. While we have great fundamental understanding of the changes to the climate and why the climate is changing, there are some processes that we don't completely understand yet. 

[10:59] As a result, there's competing theories for how each of these individual processes work, which will lead us to different equations representing those physical processes. Each of those equations can be used to create a new model. 

[11:13] That might sound like a little bit of gobbledygook, but let me give you an example of what it means using clouds. I'm going to say upfront these are not real papers. Please do not cite them. They are not real papers at all. This is just an example of what I mean by what can happen with a physical process and competing theories. 

[11:33] Let's say Smith in 2003, after doing a lot of observational and modeling work looking at cloud formation, determines that there is this particular physical relationship between cloud formation and temperature represented with this particular constant. That gets published into one paper, and that could be incorporated in one climate model. 

[11:53] On the other hand, you may have Jones who comes in a few years later. Then based upon their work, trying to add to that theory behind cloud formation or with their own theory on cloud 


formation, Jones comes in and says, "No, I think the actual relationship is something more like this, with a different constant involved." 

[12:12] Both of these are valid theories for cloud formation. Until one theory is disproven or another, they could be both used in the representation of cloud formation in different climate models. This is a really simple play example, but cloud formation's actually a pretty important process because it does affect temperature here at the surface and also the values of precipitation. 

[12:36] There's more than just two different theories on cloud formation and different physical processes therein, particularly with respect to convection and thunderstorms. There's also many, many, many other processes in this graphic where this is the case in some of these small physical processes that are very important to the climate. 

[12:56] If we go back then to our data for Oklahoma City, suddenly now, I've added a few more red lines to these. Each of these are from a couple of different global climates models. Which one do you work with? There's a couple of different stories in here that you could go with. 

[13:12] The German model says one thing, a little bit warmer for this year in Oklahoma City, whereas the National Centers for Atmospheric Research model here in the United States, they're suggesting perhaps a little bit cooler future than the other ones at the very end of the century. 

[13:27] That's another reason why there's so many different stories and projections. If you think about it, we have three RCPs here, at least, and there's more than 40 climate models to work with. There's potentially 120 stories there to get started. 

[13:44] Increasingly now, there's also the question of whether or not one needs local information. On the left-hand side is temperature outputs at a typical resolution for a global climate model. What I mean by resolution, for those unfamiliar, is the spacing of the grid in the model. 

[14:01] These are about 200 kilometers apart. These are 200-kilometer-wide boxes. What you can tell, if you look at this image, is that it's very pixelated over the United States. It doesn't give us a whole lot of detail. Global climate models, by their nature, run at the global scale and release information globally. 

[14:18] Because of the computational expense involved, they can't be very fine everywhere. They can't give us a beautiful detailed picture of what's going on locally. That would require a tremendous amount more computing power to be able to do that. 

[14:32] The thing is that we know there are physical processes that aren't going to go away, that happen at a local scale, but they're not going to go away when the climate changes. For example, we know that the temperature changes dramatically with elevation. 

[14:45] Here in the Rockies, there should actually be some more dramatic changes in temperature than what you see here. You also can't see the Appalachians at all in this particular section. 

[14:58] Another particular physical process is that there's a relationship between temperature and the proximity to the coastline, to the ocean water supply, to the ocean, but that's not visible here in the model either, in here, over here, or anywhere along the Atlantic Coast here. 


[15:18] What we want for more applications' purposes is a picture that is a bit more like this. You can now see the dramatic changes in temperature with elevation. You can see how Florida shows up much more clearly with temperature over land compared to the ocean surrounding it. You can finally see the Appalachians in this particular picture and where they are based on temperature. 

[15:39] To get from what the global climate model looks like, to a picture that is more useful perhaps for adaptation purposes and studying local effects of climate change, we use a process called downscaling. 

[15:52] Downscaling translates the global climate model's change to the local scale, and re-incorporates all those physical processes that we know are going to be there or, I should say, the effect of all of those physical processes that we know will be present in the future. 

[16:06] Indeed, this is a downscaled climate model output at roughly 25 kilometers between all of those little grid boxes. I should say 25 kilometers is one resolution, but there's some publicly available data now that provides results at 6 kilometers, even finer than this. 

[16:24] The interesting thing though is, just like there's competing theories for different smaller physical processes in the global models, there's also competing theories for the best way to approach doing downscaling and that translation to local scales. 

[16:37] As a result, there's not one downscaling technique that is preferred or the absolute best in all situations. As a result, if we go back to Oklahoma City one more time, we can see here now that suddenly we have many more lines on here. I've added three or four downscaling techniques on top of the global climate models that I put in here. 

[16:58] Now we have RCP with global models, with downscaling techniques involved. There's much more information to work with and many more stories. If we put it all together with the other two RCPs, we get back to that picture that I showed you at the beginning and why we have so many more storylines in the projections than just one to work with. 

[17:19] This is great because it does give you a range of possible futures to plan for. It comes back to what kind of risk you're willing to tolerate as a decision-maker, what you want to explore as the worst-case impact if you're doing the impact assessment, or maybe the best case of an impact. 

[17:38] Now that I've provided you a bit of a graph of some of the reasons why so many projections and that so many future stories can be told, I want to get into some of the more subtle details that can change the storyline, particularly for precipitation. 

[17:52] To describe that though, I'm going to have to give you a little bit more detail behind what downscaling is and show some results from my research in the South Central United States. 

[18:01] There's two main types of downscaling, one of which is called statistical downscaling. Statistical downscaling works by training a statistical relationship here, between observations in a historical period, and a global climate model simulation during the same historical period. 


[18:19] Once that relationship is trained, then it can be applied to the global climate model to then give us an estimate of what those observations would look like. 

[18:30] If we're going to wait and see what the depth of skill of our downscaling technique is but with respect to the future, we could apply that same statistical relationship to the global climate model projection and that gives us then an estimate of what the future observations would look like. 

[18:47] The interesting thing now in the world that we are doing downscaling in, the statistical form of downscaling, I should say, is that we're not using weather station data for the most part to do this. Not directly anyway. We are using what's better-called gridded observations. These are interpolations of weather station data. 

[19:09] We do this in part because we want to overcome some of the limitations of weather station data. Weather stations are not spatially complete. On one hand, they're dots all over the United States. 

[19:21] Perhaps we want some information on how the climate is going to change in a particular park or in a part of South Texas, or somewhere where there's not a weather station data, we can do that with statistical downscaling. 

[19:34] One could use the gridded observations as they have interpolated the weather station data across the United States. Just like the statistical downscaling techniques themselves, there's not one approach to doing the creation of the gridded observations. 

[19:48] As a result, there are several different sets of grid observations available publicly. Several of them have been used to create publicly available projections, including those used in the Fourth National Climate Assessment. 

[20:03] The other thing, of course, as I already hinted at is that the downscaling techniques are not standard. There is not one standard choice because we have competing theories for what works the best in our particular efforts to do the downscaling and translate global change to what it means in local scales. 

[20:22] The important thing in all this then is that the global climate model is one input, the observations are another input [inaudible] , and the downscaling technique is the third choice to make in this particular collection of information I'm going to talk about. 

[20:36] All three of those have an impact on the raw precipitation data but can also have an impact on the change signal for precipitation. In other words, some of these little subtle details can change the story of the future climate, particularly with respect to precipitation, quite dramatically. 

[20:53] Let me show you some data to illustrate one of my points. I'm sure everybody's thinking aren’t all observation's a bit of the truth. Can’t we just use one and have it tell us the same exact truth all the time. 

[21:08] On the top here, we have three global climate models and their solution of the historical climatology for annual total rainfall across the South Central of the United States. 


[21:21] This is in millimeters, and we know they're not perfect, absolutely, but we cannot do without. They catch pretty well the rain precipitation across the South Central United States from wet on the east side, to a much drier area in New Mexico. 

[21:37] You can see this particularly in comparison to this bottom left panel, which is one set of gridded observations that is referred to as the Daymet. Then we have another set of gridded observations in here. 

[21:51] They look very similar on a first glance, but can you tell, perhaps, that there are some subtle things that may change the story just a touch in particular places, particularly in New Mexico or [inaudible] . 

[22:02] Or we could go further. This one comes from Oregon State University, the PRISM data set. Can you notice here something a little subtle with the changes in precipitation? It doesn't look like much, but maybe it does matter if we're using these things in downscaling and changing the story of future precipitation just a little bit. 

[22:28] All of this is a lot of itty-bitty little details. Most of the time, it's people like me who do a lot of downscaling, evaluation of downscaling that really care about this. But why do I care about these things? 

[22:39] I think about this with respect to impact assessments in particular because changing these kinds of subtle details can affect the timing, frequency, and the amount of precipitation that is coming out of each of these kinds of projections. 

[22:54] Changing the story just a little bit and then providing to an impact assessment. It may mean quite a difference in what the projected impact is of a changing climates. It also then affects our derived variables to get to extreme precipitation and drought. 

[23:10] We often produce projections of daily precipitation and then derive from that the changes to the precipitation extremes or changes in drought. 

[23:20] The other reason I look at this is to make sure that we have all the information that we could possibly need and want to provide the proper use of the projections in multiple other applications. 

[23:32] Now in the next few slides, I'm going to get into some of my research results, which are part of a paper that is under review right now, and show you some of these things to illustrate my point. 

[23:45] This is a set of boxplot. On the left-hand side here are the three gridded observations and their representation of the annual total precipitation across the South Central United States. 

[23:59] The three boxplots in the middle here are for those three global climate models that I showed you before, and their representation of the annual total rainfall during the same historical period. 


[24:10] The final three boxplots here are the future rainfall from those three global climate models out at the end of the century. This is 2017 to 2099 here at the end. I should say right up front, these come from simulations done with the RTD 8.5. 

[24:28] What I want you to take away from this, in particular, is that, "Sure, there's some differences there in the annual total rainfall but it's not all that much." You might get a story here that's pretty similar. There's some subtle differences between the global climate models and some subtle changes with respect to precipitation, or at least it looks so. 

[24:49] I'm going to show a few maps a little later that might emphasize a little bit that the changes -- the way it looks. In the aggregate, there's little difference between the gridded observations for annual rainfall. 

[25:01] This is not the case for variables related to precipitation as I can demonstrate right here. This is the same set of boxplots, but now this is for the annual number of rain days, so the number of wet days, for the South Central United States. 

[25:18] Do you note something interesting here? The set of the Livneh observations here has many more wet days in it compared to the Daymet and the PRISM gridded observations. In the analysis, we've found that in some cases, this is up to twice as many rain days compared to the other two gridded observations. 

[25:39] The global models here have more wet days and the observations and also suggest a decrease in number of wet days as you go forward. 

[25:46] The network of the global models is pretty well known in literature and I'm not going to dwell on it too much today, but if you're interested, you're welcome to ask a question about it later. That really works to correct a lot of those particular features. 

[25:59] This is something that's been lesser-known. Why are the observations here so different? All of these are things that we would consider "the truth," when we're doing downscaling or evaluating our modeling work against observations that could be used for that also. 

[26:18] Why is the Livneh here so different? Let’s go back pretty far to when Livneh observations are created, before we even do interpolations of weather station data. 

[26:30] The gridded observations made use of both automated and volunteer station data. The automated data is, as the name would suggest it, technologically-driven. They automatically report electronic sensors and the whole works there, report their 24-hour totals of precipitation at midnight. 

[26:50] The volunteer stations, on the other hand, are manned by people and people don't necessarily want to be up at midnight to report a 24-hour total of rainfall. They'll be happy to get up at seven o'clock in the morning and do that, but that means their 24-hour total is from 7:00 AM to 7:00 PM. Not midnight to midnight. 

[27:07] That creates a temporal mismatch between volunteer stations and automated stations. Both are used to create gridded observation data sets. That is one particular mismatch that can be accounted for when gridded observations are created. 


[27:23] Some creators of gridded observations do count for this, but others don't. Would take a different approach to the one that the Livneh observations do. The method that we use to create the Livneh dataset splits the rainfall total over two days. This increases the number of rain day but also decreases the extremes. 

[27:43] Now, let me walk you through a little bit of an example because this might make absolutely no sense at first blush. 

[27:48] Let's say we have an observer that on 7:00 AM on day two reports a total here of one inch. What the adjustment here would do is that, knowing that there are seven hours in the second day in that overlap there, [inaudible] 24th of that one inch of rain, and can keep it on day two, but would rather keep the other part of that rainfall, the remaining large amount of rain, on first day. 

[28:16] Before that adjustment is done, the observer record will look a little bit more like this. Where on day one, there was no rain, day two, there was an inch of rain reported by the volunteer. 

[28:28] After the adjustment and before the interpolation of the weather station data, you would see something more like this. Where now suddenly, there's almost three-quarters of an inch of rain on day one, but a third of an inch on day two. 

[28:41] In the aggregate, over three days, it looks exactly the same. But in that three days now, there's two days of rain instead of one to work with. This affects and increases the number of days and it also decreases the extremes. 

[28:55] This is all well and good. Now it's time to see, does this have any effect on the downscaled data? Does it change the story at all when we get to the end with the projections? 

[29:07] This is the same boxplot I was showing you before. This is the annual number of wet days for the historical period. But I'm going to start revealing a few things here for you. 

[29:15] These are the three observation data sets again, right now. These are the number of wet days in the three global climate models that we were using. 

[29:25] These are our nine sets of downscaled climate projections, nine possible stories to work with, and what they now show for the number of rain days during that same historical period. 

[29:37] Can you tell now which set of observations was used in each one, how the story is just a little bit different? These are climate projections produced here in-house, in part because we were interested in exploring this research question of, what does it mean when you change the gridded observations when you're doing downscaling, providing them to users? 

[29:57] As I mentioned previously, there's many different downscaling techniques that exist, and many different sets of downscaled climate projections available publicly. My wonder for this as I was preparing this talk, and a couple of others was, "Can we see the same feature in some of the publicly available projections?" The answer here is that you can. 


[30:21] These are actually data from twenty global climate models, but two different other downscaling techniques that were used to create a set of publicly available projections, and not just for the South Central US but also available for the rest of the country. 

[30:36] I'm going to highlight this one in particular. This is three global models from the LOCA statistical downscaled data set. That was the data set that was used in the Fourth National Climate Assessment. 

[30:49] This is why I bring this up, to have it as something interesting. The story there, with LOCA, is perhaps a little bit different if LOCA had been created using a different set of observations. 

[31:04] Both the observations and GCMs are used in downscaling. They're both very important inputs. To that end, it's also important for us to realize that the story can change a little bit. The downscale projections will inherit any of the unique features from the observation data that's used. 

[31:21] We've seen this, now, in the raw data, but it's a good time to go check, what does this mean in terms of future change? If the same issue is there in past and future, does it really change the potential, how much rain will change in the future, how much it will decrease or increase, how much the number of days with rain will decrease or increase? 

[31:42] This is a little bit complex and busy, here. Let me walk you through this. What you're looking at in the box plots on the right, is the projected change, now -- not the raw values -- in the annual total precipitation as we end the century, 2017 to 2099, compared to our historical period of 1981 to 2005. This is, again, for our CPA .5 base simulations. 

[32:06] The three box plots, here, on the left-hand side, are for the global models by themselves, no downscaling done to them just yet. The 18 remaining, here, are from the projections done in-house, here at the South Central Climate Adaptation Science Center, together with our partners at the NOAA Geophysical Fluid Dynamics Lab. 

[32:24] They're grouped, here, by the observations that were used. You can see the first six are from JMED. They are sub-grouped by the different downscaling techniques. For the sake of time, I'm glossing over a little bit what the exact downscaling techniques and things are. You are more than welcome to ask me in the comments session. 

[32:44] The main thing I want you to take from this particular set of box plots is that it doesn't seem to matter, for this particular variable, what downscaling techniques you use and what observations are used. What seems to matter more in the story of the projected change in total rainfall, is what global model is used, here. 

[33:05] This is further emphasized if you look a little bit more at some of the individual simulations, like this particular map from one of the global climate models, compared to this particular downscaled version, one story, of the change in total rainfall across the region, compared to a different downscaling technique across the same region. They are the same. 


[33:28] The story didn't change there as much as it changed across the global climate model. It was the case for just the raw data. This is not true for other variables, and particularly the number of wet days. Can you tell how much the story is very different, going across those set of 18 box plots, now? 

[33:51] Similarly, there's a range for quite a decrease in the number of days with rain, to, in some places, a dramatic increase in the number of days with rain. I should have said that first, this is the same layout as the previous slide, but you can tell now, there's much more spread in here, in the story told from the projections. 

[34:12] This is also reflected when you go back into the map, particularly to changing stories. Changing downscaling techniques changes the story quite dramatically. This is for that one particular global model, showing a pretty strong decrease in the number of days with rain in the future. 

[34:28] This is that same global model downscaled with one particular downscaling technique in here, which shows about the same, to a slight increase in the number of days with rain across the domain. 

[34:42] If one switches to the other downscaling technique that we have, the story becomes a bit different again. We go back to having a decrease in the number of days with rain across our domain in the future. 

[34:53] This is a pretty interesting issue when we're talking about how do you plan for these kinds of things, if this is an important variable for you. There are other variables where you can see some similar issues. I have some few things on extremes up the road that will also be quite interesting for you all. 

[35:09] The thing here that also makes this graphic very interesting is, what we talked about with the gridded observations doesn't go away when you move into a projected change. For example, if you look at this particular box plot, here in the green, this if for one global climate model, downscaled with one technique, but using the JMED observations in there to provide the training data. 

[35:35] Then, if we look over here, to using something with a little bit of training data, here in the orange, you can see how there's such a difference in just switching what observation data was used. That goes back in part to that feature, where Livneh has quite as many rain days in it, with smaller amounts of precipitation, in general, for a given day. 

[35:59] It makes it very interesting. If you change the gridded observations used, you change the future story. At the same time, there's something very subtle and simple. 

[36:10] In the real world, we recognize now that some variables that we're really interested in for adaptation and planning, might be very sensitive in terms of the downscaling technique and gridded observations that are used, the annual number of wet days being a really good example. But it's not the only one. 


[36:25] Others, though, are not as sensitive. Annual total rainfall, it doesn't matter entirely what set of observations you use, what downscaling technique is used. It's all about which RCT are you looking at and which global climate model are you looking at? 

[36:40] You might have noticed in this particular plot, it's still the annual number of wet days. I just added a few things in there that you can't see right now. The projections I've shown you thus far are from our in-house projections, here at the South Central CASC. 

[36:55] I want to add back, for you, those projections that are publicly available. This one in the red is, again, links that are projections available from the LOCA which was used in the National Climate Assessment. You can see, there's a spread of stories that is still continuing there, with the different sets of downscaled data from the public versions. 

[37:16] [inaudible] when we start giving maps to users and talking about what does it mean, the projected changes. For instance, if we go back to annual total rainfall again, we have in here, all of those 18 that I showed you, all averaged together. 

[37:34] These two different maps, here on the right, are averaged together by the different downscaling techniques that we've used. Not all that much of a difference, again, between downscaling techniques, to make a big point of emphasis here. That doesn't mean there's no spread in the stories. 

[37:50] I apologize, there's so many little mini-maps that's making it hard to read. The point that I want you take from these is that, you can see that there's some variation there, across the individual 18 members -- anything from quite a bit of increase in rainfall over here in the eastern part of the domain. In this one, there's huge, dramatic decrease across much of the domain. 

[38:09] The picture gets a little bit more different when we talk about the annual number of rain days, where we can see the total mean gets a bit muted. That's in connection with the difference between these two different downscaling techniques, with one method showing a slightly less of a decrease in the number of days with rain in the future, on average, compared to the other method. 

[38:36] This goes back, again, to looking at some of the individual members and picking up the stories. The scale in this is a little bit different. This tops out at a decrease, here, of more than a month, fewer, in rain days, in some of these simulations. 

[38:50] Ask a question, for those of you who might be watching from this particular region -- what would you do if you had a month fewer of rain to work with, a month or more where it was dry? 

[39:02] What would you do if you were in Louisiana, when you guys had just had quite a intense rain event, with Barry passing not too long ago? What would you do if you suddenly had many more days with rain to work with? Both of those are possibilities, in here. Both of those are stories that could unfold. 

[39:21] I didn't talk about it much, with the box plots. I figured I'd leave you with something a little interesting and different. These are the annual one-day maximum rainfall projections. Think 


of your heaviest rain events here, how much heavier it could be, or, as this one downscaling method would show you, how much less it could possibly be, across the region. 

[39:44] That's an interesting story. What would you do if you had your heaviest rain increase a lot? Would you be happy if, suddenly, the heaviest rain events decreased quite a lot in the future? How would you plan for those two very different options? What would you want to plan for? What would be the worst? 

[39:59] You can see that's reflected again in the individual stories. I'm not going to belabor the point with this, but what would you do in the situation? The scale in these little ones tops out at an increase of 60 millimeters or more, in some of these. 

[40:14] Take your heaviest rain event and add another two-and-a-half to three inches to that particular rainfall event. What would you do if it was much heavier, potentially leading to more flooding events? 

[40:26] My point in all of this is that, one projection is only one story. It's not necessarily the wisest thing, depending upon what you need as a user, to look at only one story. The projections provide us many stories that we can use for planning. Depending upon what you need, you need to look at a lot more, perhaps. 

[40:44] Annual totals, it's about the same. You probably just need to look at a couple of different global models and a couple of different RCPs. On the other hand, occurrence in precipitation extremes can very quickly widen. That's important. 

[40:57] If you're worried about a drought situation occurrence as an important aspect, if you're worried about flooding, extremes would be of interest, there. What could all of this mean, perhaps, for impact assessments? 

[41:10] We know the choice of downscaling technique, now, and the observation data, can affect the raw daily precipitation, can also affect the change signals for precipitation variables. It's interesting for me in particular, when I'm working with impact modelers nowadays, that a lot of them use daily precipitation as an input. 

[41:30] Ecosystem, hydrology models in particular, crop models do make use of raw daily precipitation and temperature projections as their inputs. 

[41:39] I am wondering, at the moment, what some of the impacts would be, and how some of the stories would change, if we gave many more stories that were connected to different observation data and different downscaling techniques to those modeling efforts. 

[41:51] I'm hoping to find that out soon, with another project that I'm working on a little bit with the New Mexico Water Science Center, taking some of the same projections that I've shown you and running it through the PRMS hydrology model. 

[42:06] Looking forward to seeing what they come up with to changes in water supply, and run-off and stream flow on the Rio Grande -- the [inaudible] Rio Grande, I should say. I'd love to see more, if people are interested in exploring that. 


[42:19] I also like to include a word of some caution. Climate data can't ever tell you everything. The projections here can tell you a lot of the future storylines of how the climate can unfold from many different things, at this point, but they can't tell you anything about how risk-tolerant you are. 

[42:36] The projections can't tell you anything about how vulnerable your systems, communities, species, anything that you may be interested in, are to the changing climate. Thinking back to all those pictures around the beginning, I can't tell you how much your community would be vulnerable to an increase in extreme heat. 

[42:52] I can't tell you how much your community would be vulnerable to an increase in heavy rain events, or an increase in the frequency and intensity of droughts. I can't tell you, nor can the climate projections tell you, what would it mean to change the length of the growing season, or first leaf, or first bloom to species that you're interested in. 

[43:12] You all know these communities and systems the best, those of you who are in management or doing impact assessments, you know these all the best yourselves. It's always good, based on the stories that the climate projections provide, to ask those what-if questions. 

[43:29] If you suddenly had 60 more millimeters of intense rainfall in your heaviest rain events, with more flooding that could happen, what would you do? What would you plan for? How would you approach handling that situation? 

[43:41] Would it be really detrimental, or really good? You may know that. If you don't know that, that's where climate scientists can also be of help, as something to figure out -- is your community really vulnerable to that, or not -- community, or system, or whatever you're managing, really vulnerable? 

[43:58] Great for you if the extremes of precip decreased. That's another option. How would you approach those kind of questions? With that thought-provoking question on how you might plan for some of these many different stories, we can shift into taking questions. 

[44:18] If you want more information on the climate projections that I showed, that are in here, in-house, I would ask, please email our email. Particularly, if you're in our region, we would be happy to work with you, and getting you access to those projections, if you're interested in using them. 

John: [44:34] We have a few questions right now. For those of you who are online, feel free to throw them in the Q&A box. We'll get through them all. The first one comes from Rowland. She asked, "Challenge of dealing with the imperative to consider multiple future storylines," she is so ably articulating, "can be addressed in practice by using scenario planning." 

[44:58] Looks like a little bit of a mistake. Thank you. " [inaudible] planning explicitly versus one to consider multiple options." Good point. 

[45:05] David has, "Can you talk about any thinking on the best approach today, to the main challenge or highlight -- possibly averaging, or better, selection of reanalysis data sets to find the best one -- and what would that mean?" 


Adrienne: [45:26] I see the question now. Let's see, "Not counting RCPs." The way that we have approached this, typically, in the past, has been to...We understand, 3 RCPs times 120 global models, times 30-plus downscaling techniques that exist, times however many gridded observations -- that becomes completely unmanageable from the user perspective, or unfeasible to work with. 

[45:56] What we've often recommended, and continues to work well, and meshes a lot with real scenario planning, is to, working with the climate scientists, look at the possible spread of the storylines, and select a few out of those, which would work well for your climbing efforts. 

[46:15] 3 to 5. 3 is the middle, we never suggest it. 5 to 10, or more, depending upon what you're capable of working with, doing that in conjunction with a climate scientist, or one of the boundary organizations. 

[46:31] I do like averaging, in the sense of, you get a good glimpse or snapshot at what the average change would be. I don't recommend using just that, because you could be ignoring the extremes and the worst case scenarios for the system that you're trying to manage, or the impact that you're trying to assess. 

John: [46:49] Our next question comes from Ryan. "Is it fair, from your analysis, to conclude that we should maybe avoid LOCA and other downscaled projections that are trained on Livneh, if we're looking at precipitation metrics other than annual mean?" 

Adrienne: [47:10] [laughs] Hi, Ryan. [sighs] I have been hesitant to recommend LOCA, knowing this right now. I don't think it's necessarily a bad thing, because the projected change, there, is not outside the realm of possibility. From that perspective, I don't completely discount it. 

[47:35] I have not also seen, yet, what it does in the context of a hydrology model, something "trained with Livneh" does with that. That's what I'm concerned about. [inaudible] . I caution it, but I don't think it's necessary that we avoid it altogether. I just place a caution flag there, instead of a penalty flag. 

John: [47:59] James says, "Edward’s aquafer Texas River gained a five-year [inaudible] changes to temperature interpretation may affect our aquafer recharge over the next several decades. Our plan has used downscaled GCM projections." The [inaudible] looks like it got cut off. Sorry about that, James. 

Adrienne: [48:19] It looks like you are in our domain region. You'd be welcome to email us if you want to work with us on that whole process. 

John: [48:26] There it is. I see the second half of it. He's asking, "Can you recommend contacts?" That answers our question. Marissa has, "Thanks very much for a great presentation. Your research looks very interesting. Can you tell us, what type of activism have you been engaging in to get your science heard by non-scientists?" 

Adrienne: [48:43] I haven't been putting myself out there as much, with the exception of doing presentations and being asked, then, to do more presentations, not just to scholars, but to non-scholars and non-scientists around the region. 


[48:57] I have not done a huge amount of activism. That said, the one way we did advertise -- it was mentioned in the introduction that I am the editor-in-chief in the Early Career Climate Forum. That's one way we've been advertising, it's been through that forum, otherwise known as the ECCF. 

[49:14] If you have any early career students who are interested in connecting with other students working in climate, or post-docs working in climate across the country, please feel free to email me about that. I can put them in connection, there, with that. I have not directly done any activism. 

John: [49:32] Thank you. Angela has a question. "What kinds of graphs do you find practitioners are most receptive to? For example, are people most receptive to box plots, scatter plots, maps, or something more creative?" 

Adrienne: [49:46] That's an excellent question. It does depend a little bit on the level of knowledge that some users already have. In the general public, a lot of people like maps. We tend to use a lot of maps with that. That has done really well. 

[50:02] We had this conversation, not too long ago, helping a couple of our tribal nations in the region prepare some climate adaptation plans of their own. They did very well with those maps, also with some box plots that we worked with, with them, showing them that. 

[50:19] What we found with all the different stories there, if you're willing to work with people and help explain it to them, they tend to get it. People do understand that there's more stories there than one. It's just a matter of working with them, to help them understand how to use it. 

[50:37] Maps have easily been some of the best to work with, alongside box plots. We haven't done a whole lot of scatter plots. Maps have been the favorite, seen a lot of people being really receptive to that. 

John: [50:50] Great. Oshka -- I hope I said that correctly -- "Have you looked at any drought variables, particularly low frequency, persistent dry events, in the context of uncertainty from downscaling [inaudible] ?" 

Adrienne: [51:04] Have I looked at dry variables? Indirectly, yes. In the paper, we have also looked at the length of the longest dry spell, and the length of the longest wet spell. There is an effect there, also from the observations and the downscaling techniques. 

[51:24] I don't have that in this presentation. I'm happy to share more of that with you if you'd like to email me. But I think... 

[51:30] [crosstalk] 

Adrienne: [51:31] We haven't looked directly at the drought indices like SPEI or STI yet though that's something we've been planning, considering of doing. 

John: [51:40] OK. Thank you. We'll take all the questions that we have. Thank you, Adrienne for your presentation and thank you everybody that participated today. This was a great webinar. I'd once again like to thank Elda and USGS for the contained score of this webinar series. 


Adrienne: [51:59] John, can I add more thing before we go? 

John: [52:02] Absolutely. 

Adrienne: [52:04] What I wanted to mention, one thing to be clear on, is that not all the downscaling techniques and the observations are equal, in short. Depending upon the circumstances and what you want to use it for, some things may be better than others. 

[52:19] That's why it is really important to work with other CASCs and with the different boundary organizations. We'd welcome having that interaction with you, which is why I also put our email up there. If you're in our region, please feel free to contact us through that, also. 

John: [52:35] Great. Thank you very much. Everyone have a nice day.