 who is Kerri Callaghan from Lamont Columbia University and she will talk about coupled groundwater and dynamic lake modeling using the water table model. Kerri, go ahead. Okay, there we go. Thank you everyone. Today I'm going to share with you the water table model. I developed this model during my PhD and I'm currently working on a postdoc where I'm applying the model to simulate past and present water table. So I'll start off with a few quick acknowledgments. First of all, Andy Wickard was my PhD advisor and he of course played a really big role in this work and was instrumental into getting all of this done. Then Richard Barnes has been a close collaborator throughout this project and he really kind of revolutionized the way in which I was thinking about the projects and thinking about coding in general. And a lot of the work that I'll be showing you today is stuff that he's had a very close hand in as well. Ying Fan Reinfelder is someone that we've worked with and Ying has done a lot of work on global water table and she's just been a font of knowledge and information throughout the entire process. Crystal Ng is also a hydrogeologist and she's also really been helpful, especially as we figured out how to get evaporation working in our model. And then Jackie Osterman is my current postdoc supervisor. Many of you probably saw her talk earlier in the week. She specializes in sea level and the work that she's doing is going to become more intertwined with what I'm working on at the moment. Kerry, your video's off just FYI. Yeah, for some reason I can't figure out how to get it turned on. My Zoom has kind of minimized itself and this is never happened to me before. So I don't know what to do about that. Okay, sorry, I can't, normally I can say request to start video but it's not showing. There we go, let me just do that and then I'll reshare my screen. There you go, okay, sorry about that. Yeah, that was kind of a weird thing that happened. Yeah. Okay, let me reshare that screen. There you go. Okay. Okay, so here's a quick teaser of the type of results that we get out of the water table model. As it says in the name, we're trying to stimulate the depth to water table and our focus is on large spatial scales such as this entire continent. So we'll come back and examine this results a little more closely in a few minutes. However, what I want you to take away from it at this point is simply that we have both groundwater and surface water. So the blue that you see kind of in the shape of lakes and so on is surface water that is simulated by this model. So before we get back to that, I'm gonna tell you a little bit about why we would care about past water tables and the water table in general. So first of all, when we look at a global water distribution, of course, most of the world's water is in the ocean. But when you look at the portion that's not in the ocean and particularly at freshwater, we see that about two thirds of freshwater is stored in glaciers and ice caps. And then about one third is stored in groundwater. So that's a really large proportion of the world's accessible freshwater. And when we look back into the past back to a time period like the last station maximum 21,000 years ago, there's a significant body of work examining changes in sea level and changing ice sheets over this time period. And there's some phenomenal work being done, but there's much less of a focus on changing groundwater storage. And as you can see from this figure, the amount of water stored in the ground is also significant and it's really important. So to put some numbers onto that, groundwater stored in the upper two kilometers of continental crust is about 62 meters of sea level equivalent. And when it comes to lakes, there are over a hundred million lakes globally. And these lakes of course represent locations where the water table is above the surface. So basically when I'm using the water table model, I'm trying to think about this full water balance. So we have water that's stored in ice, water that's stored in the ocean, water that's stored in the ground and in lakes. And so you can kind of see all of these components here in this figure. So some of the problems that we're trying to think about with the water table model is first of all, when we think about the last station maximum, can we close this water budget at that time? So modeled estimates have last station maximum sea level being depressed by more than can be explained through ice volume from ice models. So could it be possible that some of that water was stored in the ground or in lakes? Another question is just, how has that terrestrial water storage changed throughout the entire deletion period, both in terms of total volume and spatial distribution of this water? The water table also indicates locations of wetlands, such as the one that we have over here on the right hand side of the figure. And changing wetland locations is really important for carbon cycling and that's something that our collaborator, Jean-Fran Reinfelder at Rutgers University is looking into. Way on the right, we also see how some groundwater is entering the ocean. So something that we can ask ourselves is, could groundwater that's entering the ocean play a role in multi-water pulses during the deglaciation? You can see a couple of really large lakes on this figure. And so one thing that we're also looking into is the effect of these lakes on glacial isostatic adjustment. The actual volume of the lake itself would have also had an impact on that. And then finally, we can compare long-term water table records at local sites to the results from our model and see how well it does. And in this way, we can validate not only our model, but perhaps we can provide some validation for existing climate models which provide the inputs into the water table model. Okay, so here's kind of a schematic of the specific things that we're focusing on with the water table model. So this is sort of a simpler version of the figure that we were just looking at where we have this green line that indicates the topography and then the blue line that indicates the water table. And this blue water table is basically what we're getting out. And we have two main components within the water table model. So first of all, we have the groundwater component and I'm indicating that here with QGW. So Q means discharge of groundwater. And this is a function of the porosity, hydraulic conductivity, slope and winter temperature and of course the amount of water that's available at that location. And then wherever that water table gets above the land surface, we have overland downslope flow and we simulate these flat lake surfaces. Then we have of course precipitation entering the system and the evaporation removing water from the terrestrial system. So to give a little bit more details on that groundwater component, in the groundwater component, we move groundwater using the 2D horizontal groundwater equation. And here I have a small little example where these three columns represent three cells within a digital elevation model that are next to each other. And we're focusing on this central cell which is why it has a star above it. So we calculate the groundwater discharge Q from that cell and this is forced by climate topography and sea level and we will move water for user selected time state. So we do this little calculation for each cell and once we've calculated discharge from all cells we'll update the water table. So it might look something like this where the original water table was the blue lines and then after moving water for one time state we have an updated water table at these red lines. And now you'll notice in the right hand cell that the water table is above the land surface. So now we need to think about what do we do in this type of case? So here's another kind of example topography where we have water above the land surface that is creating these large lakes. Now any one of you who has had to perform flow routing or compute a flow accumulation before which I imagine is several of you in this audience may be familiar with a depression removal step where an assumption is made that any depressions in the digital elevation model are data errors and they are removed through processes such as a flood phone like I'm showing here while the process such as carving. But what if that surface water is actually really important? It may be going to form part of a major lake like Glacial Lake Agassi that had major impacts on things like global climate and also aesthetic adjustment or it could even just be part of a smaller lake that will still impact local to regional water table and groundwater storage. These lakes are very much present in our world. They are important. And over the digital time period that I keep referring to these lakes really changed dramatically. So we needed to come up with a new algorithm to actually think about these depressions and the lakes that will form within them. So we do so in two steps. First we use a depression hierarchy to break down the topology of depressions within the landscape into a simple binary tree structure. And second, we use an algorithm called Falsible Merge to actually move the water into the depressions. So for the depression hierarchy, feel free to scan that QR code if you want more information on this piece of the work. So first of all, we define a depression as being an area in the landscape that contains a low point called a pit cell. And here on the left in panel A I'm showing one example of one such pit cell that doesn't have any downslope cells around it. And then the depression also has an outlet cell which is the point at which that depression would overflow either into another depression or into the ocean if it were to be filled with water. A depression can then also be contained within a larger meter depression. And we'll see an example of that in a moment. So we start off by labeling each one of these depressions based on where those pit cells and outlet cells are. So we'll label each one of those one, two, three, four and five. And then we start labeling those meter depressions. So here you can see how the larger depression six contains one and two within it. And then if you look over at the right, you can see how we start building that binary tree where depression six points towards its two children, one and two. And we continue doing that for the rest of the landscape. So we add depression seven that contains four and five within it. And we add depression eight that contains three and six within it. Then when you look down at panel B, which is a map view of the system, you'll notice this dotted line and the dotted line indicates the direction of flow of hypothetical water. So you'll notice that depression eight would overflow into depression seven. So we add an arrow to indicate that. And then depression seven would overflow into the ocean, which we've labeled zero. And so there we're done. We've labeled all of the depressions in this topography. And the ocean zero can have any number of systems of depression that would overflow into it, which I've indicated with this group of lines. The depression hierarchy also records depression volume, area, number of cells, location and elevation of pit and update cells and other useful information that could be used for any number of applications. So once we've constructed this depression hierarchy, we use flexible merge with the hierarchy to rapidly root water across the landscape. And I've put up a new QR code for flexible merge. So flexible merge, here's a conceptual look at how it works. If you imagine that water is entering this topography from the left-hand side, the leftmost depression will begin to fall with water. Once it is completely filled, it will spill over into its neighboring depression. And once that depression is completely filled, the two will merge into the immediate depression as water continues to fall up. So this process will continue until either all of the water has been used up and then you might get a situation such as we have here in the third panel with some depressions being filled with water, some having a little water, some not having any water, or the process will continue until all of the depressions are completely filled and additional water continues to spill into the ocean. So of course there's a lot more to flexible merge, but in the interest of time, I'm leaving it at this basic concept, but it's a really cool and versatile algorithm. So I'd be more than happy to discuss this more with any of you later on. And so now we'll look at how we put these pieces together in the water table model. So the water table model basically just takes the groundwater and the surface water components that I've described and it couples them just by running one after the other. So we start off with some kind of initial condition that might look something like this where we have the water table basically at the land surface. Then we run that groundwater components. And so we have groundwater flow and you can see how the water table has moved down beneath hilltops and we've had some exfiltration along hill slopes and in the valleys. We then run full full merge to move that surface water down into the bottom of the valleys and we get those flat lake surfaces. We add precipitation minus evapotranspiration so you can see how that water table has moved back up in some areas and in some of the lakes the lake level has gone down due to evaporation. And then we just cycle through this many times move groundwater, move surface water at P minus ET and eventually we get some kind of final water table surface where we've got those lakes and we've got that groundwater. So now I'm gonna loop back to this result that I showed you in the beginning and we'll look a little more closely at this. So this is where I ran the water table model for North America for present day pre-industrial climate and topography condition. And so you can notice how we've got some recognizable lakes. So for example, the great lakes with their kind of recognizable shapes we can see that they're there. In the drier west, we have a deeper water table represented by the red colors and then in the wetter east and north we have this shallower water table represented by the yellow and green colors. Now, when we compare this to actual observations here I've got observed water table in red and simulated water table in blue. And these observations come from a data set that was gathered by Ying Fan-Rin Falder and her co-authors and they gathered over 500,000 observations of groundwater depth across the continent. So as you can see the simulated water table depths tend to be shallower, especially in that very shallowest bar. This is a trend that has already been observed in other modeled water tables. And it's likely at least partly because of a selection bias in observations as a result of well location because generally speaking we're not going and putting a lot of wells out in wetlands. And so those very shallow water tables out in the sample. Additionally, these observations also include human impact such as pumping which lowers the water table and the model does not because it is a pre-industrial simulation. When we look at those surface water areas here I'm comparing my simulated lakes with lake bathymetries from a data set by Corson Nova that was created in 2012. For many of the smaller lakes the data set includes only a mean depth and not a full bathymetry which makes the comparison difficult and leads to some of the mismatches that you see. However, when you take this limitation into account I think that this comparison looks very reasonable. So finally, I'm coming back to talking about lake water loading in glacial as a static adjustment. So I did mention this also since the beginning and here I'm showing some examples at different times during the deglaciation of how those large proglacial lakes may have looked. And so the way in which I'm combining this with the water table model you can see here. So I've started by creating a topography with glacial as a static adjustment based on I6G ice model. I'm then running the water table model and that's the step that I'm currently on with my postdoc. I'm running water table model on these topographies throughout the entire deglaciation. I will then create new topographies with a glacial as a static adjustment based on a combination of I6G and lake level loading from the water table model. And I will repeat this process a few times and what we should get out of this is an improved estimate of past topography as well as our first estimates of changing water table throughout the deglaciation. Okay, so a quick take home message. The water table model uses fossil merge and the groundwater module to simulate water table level and it is inclusive of lakes. By coupling the water table model and sea level models we can obtain the first global simulation of changing water table lakes and the impact of terrestrial water on terrestrial as a static adjustment since the last terrestrial maximum and we will be able to answer the question what impact does changing water table have on sea level since the last terrestrial maximum? Okay, and that's all I have. So thank you and I'll be happy to take any questions. Thank you very much, Gary. Great talk. So let's see if there are any questions please raise your hand or pass them in the chat and we have time for a question. Maybe when people are typing I have a question for you, Gary. So I was wondering, and this maybe has to do with the time scales you're assimilating on. I'm just not 100% sure if I captured that well. Does the water table model include energy equations where when you're changing seasons and for example, you're in the winter season summer slowly starts. You got the sun is higher at the horizon and you provide more energy to the water surface and therefore those lakes that you can create with your model, they would be once frozen but then melt. Is that, can you simulate on that time step and if so, have you incorporated those kind of energy balances or? Right, so generally speaking the inputs that we've been using for the water table model have been annual averages. So we're not really capturing that seasonality. However, we have kind of thought about the potential impact of sea ice as something that we, or not sea ice, excuse me, lake ice as something that we may want to include in the future depending on kind of how well the model performs with simulating lakes. So it's something that we've got in the back of our minds but it's not really within the model at this stage. Okay, thank you. And I see another question in the chat here from Nicole. Thank you for your talk. I'm sure I missed it, but how do you know what is a lake and what isn't? Is that part of your algorithm? Okay, I'm not sure if you mean by that like what is a lake and what is actually a spurious depression. That's a great question. And it is something that's challenging because that's kind of a spatial scales that we're looking at. Yes, you will sometimes get something that is not a real life lake. There's not necessarily a good way to differentiate between these. When we look at a modern day topography we can get rid of some of those things that we know are not lakes by kind of going in and manually adjusting that topography or by using river networks that we know exist but that's of course harder to do for the past. So our current plan is to not remove any of those depressions. Assume that every depression has the potential to be a lake and we'll kind of see how these results look and see if there are improvements that we can make later on in post-production.