 Our student keynote this morning is by Mariela Perion who's a PhD student here at the University of Colorado and her talk is predicting the influence of floodplain vegetation on the geomorphic effects of large floods. I'm getting an idea. Okay, that's fine. Do you know how to run that? So it's forward, backward, and then laser, forward, and then forward, and then forward. Actually, do you have AVI movies? Was it doing it? Okay. Oh, keynote. Yeah, that's fine. We can try it. Cool. Awesome. Thanks. Cool. So a significant source of uncertainty when applying our models to natural systems, at least in the case of sediment transparent rivers, comes from our choice of implementation of vegetation. All plants or individual patches of plants can be very complex in their geometry, so simplifying their behavior to a nice, clean set of parameters or equations can get very complicated. And the choice of simplification could affect the final result of our model. So we need to understand how those choices affect the results of our models. And one way to do that is to compare them to natural experiments where we know what happened. So choosing that simplification will depend, of course, on the scale that we're looking at. So if we want to model a whole reach of river, that image is about four kilometers across, vegetation could maybe potentially be simplified as one big carpet of green something. But if we look in more detail, we can see that actually there is some variability in that big green mat of vegetation. And if we look even further, we can see that, while it's not totally uniform and there's different types of vegetation, and if we look even further, we can see that, well, it's actually just a big jumbled mess of cylinders that are not the same diameter and don't have the same spacing. So we need to understand how much we can simplify and ignore the processes at the lower scales to model at the larger scales while still preserving the final results. And as I said, one way to do that is to compare it to natural systems. Our goal is to develop a model that can predict the effects of a flash flood on a sand-bedded river with a vegetated flood plain at the scale of whole reaches. And for that, we have chosen a natural experiment on the Rio Parque in New Mexico. So I'm first going to show you some data we have from that site and then quickly give you the structure of our numerical model, run through some numerical experiments that indicate what level of complexity we might need and then show you some simulations of what a flood looks like in our model on the Rio Parque. Okay, so the Rio Parque is an incisorio in central New Mexico with a fine sand bed and that really nice uniform geometry with banks that are aligned with vegetation. We are interested, sorry, on a 24 kilometer reach on the lower Rio Parque, which is... Come on, there we go. The banks of the Arroyo used to be lined with San Baruela, which is that light-colored vegetation that looks like that. And behind that and on banks along the flood plain, there's a lot of Tamarisk, which is a really dense invasive woody plant that looks like that and you can't actually walk through. Because Tamarisk is invasive in 2003, the local government came in and sprayed herbicides on 12 kilometers of the river right there and killed everything. Both were low on Tamarisk. And then in 2006, there was a big flood. The area that was sprayed saw extreme erosion with a channel locally and the back walls of the Arroyo in place is migrating up to 20 meters. All of that sediment then moved downstream and deposited on the downstream reaches where vegetation was still preserved. And you can see there that's a Tamarisk and all of that is new sediment from that one event. From now on, whenever we look at the Rio Parque, we're going to look at specifically that little area right there. So what makes this a wonderful case study, apart from the fact that the only variable that changed is the density of vegetation, is that we have two airborne LiDAR datasets for that area just downstream of the spray reach, that bracket, that 2006 flood. And from that, we can get a measure of exactly what happened during that one event. So we did a difference in study and that's a small area. And what you can see is that it's in blue, aggradation, and in red erosion. So we have a very detailed image of what this flood did to that landscape. And comparing that to what we know about the vegetation, which mostly comes from aerial photographs, we can get some idea of how vegetation affected those patterns of sedimentation. And I'm not going to go into much detail, but what we found is that in general, aggradation occurred where there was vegetation. But there wasn't a very strong correlation between the thickness of that aggradation and the density of vegetation. If there was about more than about 25% coverage of vegetation, sediment accumulated. What we did find, however, was that there's significant correlation between the variability in that sediment thickness. So the mean is the same, but it is a lot more variable when it's in tamarisk than when it's in willow. So there is some process that's inherited to the morphology of those two plants that is affecting those patterns of sedimentation. So when we develop a model that should predict the effects of this event, we need to, in order of increasing complexity, well, hit the high water marks, so our hydrodynamics should be good. Reproduce, embolk the changes that we saw in the Rio Puerco, both in patterns and in volumes. And then, hopefully, start matching those patch scale patterns of aggradation. So if there's vegetation somewhere and we see a certain amount of aggradation, our model should be able to match that. But what we really want is to be able to match that local variability that we see within the tamarisk, but not the willow. If we are able to do that, then our model can actually reproduce what's happening in reality. So we're building a morphodynamic model to simulate these processes. And it is built on top of the existing hydrodynamic model in NUGA, which simulates a depth average flow and was developed to simulate tsunamis. So it handles those sharp waves of a flash flood really well. On top of that, we built a sediment transport model that treats erosion and sedimentation separately. And this is important because it means that we don't have to impose a transport capacity. In the Rio Quarco, it is known that it can carry very high sediment loads. So this is important. And there's also now a vegetation component which treats areas that are vegetated as fields of cylindrical stems that impose some drag on the flow. The complexity that we're putting into this model comes from a collection of sources and things of momentum that we're considering in your model. And that is from things like pet topography and drag on bat forms and from concentration gradients and exchanges with a bed, but also from turbulence on the wakes of cylinders. So this is all happening at a scale that's finer than our grade, what we are including it in the final results. These are some outputs from the model of flow coming down a ramp that is vegetated by some density of vegetation. And the bed is not evolving and we are looking at two measures of turbulence. So here on the x-axis, there's increasing vegetation density, so the stems are getting closer and closer together. And on the y, it's a normalized measure of turbulence. In black, it's a turbulent kinetic energy and on red, it's diffusivity. And as stems get closer and closer, the turbulent kinetic energy keeps increasing. But there is a point where the diffusivity actually starts dropping. And this happens because there is some point where the scale of the eddies go from being the depth of the flow to the size of the stems. And this is important for the Rio Puerco because both types of vegetation fall on opposite sides of this line. So what we don't, we don't actually know if this is what's driving that difference in the variability of sediment thickness. This is a process that we might need to consider if we're going to simulate the behavior of the sediment and the flow. This is also some results from flow coming down a ramp. Although in this case, instead of having uniform vegetation, there is a patch of vegetation of increasing density as you go towards that way. And the bed is allowed to evolve and the water is coming as a flow front. So this is a video and we'll see what happens. Maybe not. So it's a video of the depth, maybe. Nope, apparently not. Let me try this again. Can't see, there we go. Okay, so this is a video of the depth of flow as it comes down this ramp. And it's not terribly exciting, but what you can see is that when the vegetation is denser, their water is piling up more on the upstream side than when there is no vegetation. On chetted relief behind the depth, you can see that there has been some erosion behind the vegetation. So what's happening in this case is that the vegetation is increasing the energy slope and that's driving flow to move around the patch and increasing erosion. And I'm going to show you also what the concentration looks like. So again, this is a sharp flow front coming down a slope. So the water has a really high concentration right at the front. And here vegetation have just drawn the outline so you can see the concentration. So before we even start the movie, you can see that the concentration is actually refracting right at the edge of the patch. And when it gets very, very dense, it's effectively stopping sediment from going through and just piling it up right at the front. So when I play that, what you can see is that the concentrations of the upstream end of the very highest densities are really, really high while sediment of the lowest densities are just flowing through. So in this case, effectively, we are starving the area just downstream of sediment. And this all comes from differences in the stem spacing that are not that high. It's basically half a meter, which is a variability that we might see in the field and average to a single number. So I wanted to end by showing you a simulation for the flood going down the Rio Parco looks like with this model. Well, I haven't actually proven to you that this is exactly the right level of complexity that we need to simulate these processes. I think at least we have a hint of the need for these sources and things of momentum and the need for special variable vegetation. Here on the left is the topography is going to run over. The relief between the bottom of the channel and the royal, the valley surface is about 10 meters, and there is the vegetation pattern. So we're imposing on the flow and you just have two categories of vegetation which have some generalized values. So again, I have to get out of here, find the video. So this is a video of the flow depth. It's about five minutes of flow. And the first thing that should come out is that, well, it's just running down the channel where there's no vegetation. And I'll play it again, going much more slowly on the flood plains. And it's not coming as a flood flow, but it's following some path as it tries to skirt the highest densities of vegetation as it makes its way downstream. And here on the final image, you can see that there are areas right next to the channel where the vegetation is actually very dense, where there is no flow. Because it's just taking the longer path around to avoid the size densities. There you can see how the topography has changed during this run. Most of the erosion concentrates along the higher royal walls, just like we see in the differencing. And the position occurs primarily in the channel and on the flood plain in areas of very dense vegetation. This matches really well with the differencing. Although with only five minutes of run, it's not. It's not a very strong inclusion. So what we have found is that the subgrid processes such as turbulence behind stems might have some significant effect on the reach scale patterns of change we see in the flood. That small details in the morphology of vegetation could have a significant effect on the results as well. Things like a change in the spacing of stems. We need to test our models against natural experiments because that's the only way we can know if our simplifications work. And to end on a motivating slide for that search for natural experiments, this is the Rio Perco right at the boundary between the spray reach that eroded and the area of the stream. And there I marked the height of the 2006 flood. Well, it turns out that the storms that hit Colorado last September also hit New Mexico. The Rio Perco had possibly its largest flood on record. We don't know because the gauge got washed away. That's Irina. And her arms are at the high water mark for the 2013 flood, which I'm going to mark right there. So the 2006 event is a great natural experiment for flow with emergent vegetation, vegetation that's coming out of the flow. But the 2013 flood, and you can barely see it, that's a high water mark right there. Could be a great case study for the flow with emergent vegetation, which has much more complex dynamics, and we really need to test. Thank you.