 that on? Can you hear me? Okay. Well, first let me thank the organizers for inviting me. Well, I'll say on a technical note, I had some external animations I'm not going to be able to play, but that's probably a good thing because I had too many slides anyway. So I have other animations that are embedded. No, no, no. I might just use the thanks. Yes, I can't see it on here, so I'll put the screen. So, yeah, I'm going to talk about modeling or service flow hazards with D-Clock, which is software we've developed. So to give a little background of where this software comes from, the word I'm going to describe in this talk is really a synthesis of two lines of research that have been developed over decades, the first being development of numerical software. It's folks here and I think most of them are here today. And then the other line of research that this is a synthesis of is the development of physical and mathematical models for landslides and debris flows. And that's work that's been done primarily with Dick Iverson at USGS and also the reflow flume crews, some people here at work. So just briefly, the software to give the phylogeny, I guess. So Clock Pack is an open source software package that was developed by Randy Levec and others, which is for modeling general hyperbolic systems or wave propagation problems. We then created GeoClaw, which is a subset or extension of POPAC, which is devoted to tsunamis and wave propagation problems. Originally, we called it Tsunami Claw, but then we realized to add features that were useful for other free-service flow problems. And then D-Clock is the software that's specifically aimed at granular fluid flows or landslides. Now, historically, in the 1980s, there was really spurred the interest in being able to model debris flows for hazard assessment. These events here really increased the interest in it. But in developing the model, I'm going to use the terms debris flows and landslides fairly generally and interchangeably. People in the geology community sometimes have a very specific nomenclature for different types of landslides and so forth, but I'll use the terms sort of generally. So we're talking about saturated granular fluid mixtures. So I'll briefly give a model summary. Obviously, I don't have time to go into the gory details, but it's described in these papers here. But this is a depth average flow model. It's two-phase, so we're modeling solid fluid mixtures, and there's a poor pressure evolution which affects the resistance or the mobility of the flow. So our variables that we solve for are the depth, velocities, the solid or fluid volume fractions, and then the poor fluid pressure. And the feedback between the solid volume fraction and the poor fluid pressure is really one of the unique features of this model. So our motivation was to be able to simulate landslides from the initiation process all the way to the deposition. So simple early models often use fluid models with some sort of elaborate single fixed rheology. But however, those have a difficult time capturing this transition from a stable mass on the side of a hill slope to a highly mobile debris flow. If you think of a sediment mass on a slope, there's a force balance. When that force balance is perturbed, the driving force has overcome the resistive force that then fails. So that's in the realm of slope stability, but it really doesn't tell you anything about the fate of that flow. It could slump and stabilize, or it could evolve into a high-speed, highly mobile flow. We want to be able to model that fate just given this initial force perturbation. So the basis of being able to do that is this co-evolution of the poor fluid pressure and the solid volume fraction. So if the poor fluid pressure increases, that reduces the effective stress of the resistance of the flow. And then that leads to a mobile flow. If the poor fluid pressure decreases, that stabilizes the flow due to higher resistance. So to model this, we use the concept of dilatancy. And that's what allows us to couple this evolution of solid volume fraction and poor fluid pressure. And it's based on if we have a fairly loose granular mixture and it begins to shear, it contracts, drives up the poor fluid pressure, drives down the resistance, becomes highly mobile. Conversely, if you have fairly dense material, it shears, it dilates, drives down the poor fluid pressure and increases the resistance that I can stabilize. So in developing this model, it was important to, or one of the features or facilities that allowed the development of this physical model is the USGS reflow flow, which was built in the 90s. This 90 meter channel that's heavily instrumented. And it's in a measure of things like the evolving depth, poor fluid pressure, shear stresses and so forth, allowing us to develop that physical model. I can't play the movie, but we have different types of experiments we do here. We load the hopper up with sediments, saturate it, open the gate doors, measure what happens. And then we can then, we can then simulate that to validate the model, for a validation tool. But more importantly, it, like I said, it allows us to develop the physical model. A different type of experiment that's done here are what we call these natural release experiments. And in this case, there aren't gate doors, hoppers loaded with sediment that's dry. And then it's slowly saturated. And then eventually the poor fluid pressure increases to the point where the driving forces overcome the resistive forces and it fails naturally. So here's a simulation of this, one of these experiments. So you'll see here, as I played this animation, it goes really quickly. You'll see a blue line increasing that represents the poor fluid pressure. It'll reach a certain point. It'll hit that threshold where there's failure and begins to flow and the flow will then quickly liquidify highly mobile. This would be more compelling if I could play the video of the experiment. That's okay. I don't have time anyway. Now, we do essentially the exact same experiments, but compact the soil more. So these are essentially repeated experiments like the highly mobile one, but with dense soil. I won't go into these quantities here, but the initial porosity is slightly higher here. It's denser material. And in this case, I won't play an animation because you can't really see anything happen. But we rise, we manually raise the poor fluid pressure like before. At this point, which I've called time equals zero, it's actually reached that threshold and this material is actually failed and started to slump. But as you can see 60 seconds later, nothing really has happened. So these are then comparisons of these loose soil experiments and the dense soil experiments showing the evolution of the pressure, comparing the experiment and simulation, see if it becomes mobile as it moves down slope. In the dense soil case, you look at the time scale here. This is 100 seconds or so. The flows only move less than a meter or so and the poor pressure is slowly decreasing. Okay. So moving on to some applications, we most recently have been looking at flows which involve the interaction of these debris flow granular fluid materials with bodies of water. Because the decoy equations are two phase, so we have evolving volume fractions. That means if we initialize with zero solids, the equations actually reduce to the shallow water equations, what we use for tsunami modeling, and you can have volume fractions intermediate to a debris flow and pure water and use single simulations to seamless simulations to model those events. So we've been looking at these sometimes called cascading flow hazards where you have these interactions. So some examples are you might have tsunami inundation which entrains debris as it inundates shorelines that essentially becomes like a debris flow in the front. Landslide generated tsunamis are another example. Landslide impacts pure water generates tsunami. Some other examples are say the formation of a natural dam. A landslide might dam a drainage, might have a lake built behind it, then you might have a dam breach and failure, flash flood that can entrain debris downstream and become a debris flow. So I think I'm going to skip this application. This is the first application we looked at in simulating these interacting flows but it's all external animation. So I think I'll move on to the next example. So we've been looking at this problem of a glacial lake outburst flood in Oregon. There's a marine dam lake, Carver Lake, on South Sisters Volcano which could potentially be an outburst flood and it could result in flooding in the town of South Sisters down here at the bottom of the slide which is about 20 kilometers away. We this came to our attention because there was a study in the in the 1980s of you're sort of zooming in on this Carver Lake which has a marine dam here and the South Sisters here. In the 1980s the USGS and some other groups did some studies and brought this to the attention of the community and it was a somewhat of a kind of an alarmist approach in some ways that the community became very aware of it and heightened awareness to the point where they were thinking that this could be an imminent disaster that this marine dam could fail at any time. So backing up a little bit to show the domain here. So this box here is the same area as this slide here. You can sort of make out Carver Lake here at the peak of Sisters and then about 20 kilometers downstream this box here is this here where you can see the community of Sisters. So it's a fairly large domain and it doesn't really you know first glance it doesn't really seem like this small lake here poses much of a threat to the community and so the the community has been aware of this previous study for a while and they want to revisit it. This is the original study from the 80s which shows the potential flooding in Sisters. This is the Whitechess Creek drainage which goes through town. It's hard to make out but this this black line here is what they predicted might be the flood extent from a Carver Lake Albers flood and it it goes right through the middle of town so it's actually a fairly significant hazard based on this this study. Now looking at the the lake here some I'm not a geotechnical engineer but various groups have gone up there and it doesn't really seem like this is the marine dam here and it doesn't really seem like it has much potential to spontaneously fail based on rising lake levels. So we've we think really the only thing that could cause a lake Albers flood here is a large landslide into the lake. If we cause overtopping waves over the dam then that could cause the dam breaching process to move forward and so what we've decided to do is is just to model that basically. That's it that's really the most likely cause of a dam breach. We might as well model that that worst-case scenario. So what we did is we we just simulated a we came up with a volume of landslide here up above the lake which would be large enough to drain the lake essentially. So we're not predicting or suggesting that this is a likely landslide. We didn't identify a slope and say hey this looks like it's gonna fail. We just said well if you want to look at the worst-case scenario then that's what we'll try to model. So so we we initialize the model with a solid fluid mixture for the landslide and the lake with pure water and basically let it go. This is a 5 million cubic meter landslide which which is fairly large for south sisters but not unprecedented for cascade volcanoes. Be quite small on larger volcano like Mount Rainier. So you can see the slide come down. It's the lake generates these waves and then as these waves over top the dam we allow entrainment of that solid material in the dam breaching process to occur. It's the the physics of of this entrainment process are not well constrained so we basically have a model where we just turn an ob and get the entrainment rate that we want. It's not well physically constrained but just to just just trying to drain the lake basically and come up with here's another view of the same simulation and you can see downstream of the dam it's essentially a debris flow. The colors here are the solid volume fractions so I think of this brown as a debris flow and dilute flows below that. It's essentially a debris flow below the lake and eventually the entire lake drains. We've simulated different landslides with different mobilities to come up with a range of possibilities. So this this shows the domain of the simulation and you can see it's quite large if you look in this box here you can probably barely make out Carver Lake here and the landslide here. The community of sisters is down here. We use in our codes we use adaptive mesh refinement to track the flood so we can resolve the moving flood or debris flow on on grids of enough resolution but not have to waste computational cost for areas where nothing's happening. So we have coarse grids where nothing's happening and fine grids tracking the flow. Now the result we got was was quite different than the original study in the 80s. So in the 80s they used 1D modeling and they had to assume that they made these they could only track essentially the flood going down the main channel of Whitejust Creek. But you can see our results here there's a bifurcation of the flood as it hits this alluvial fan near sisters. So upstream of this in Whitejust Creek it's a fairly steep channelized canyon. The flow comes down there and then and then hits this flat plane and bifurcates. So when this first happened it was so different than what was expected we thought we better go out and look in the field and see if this seems reasonable. So we drove out and this is a map of Whitejust Creek here that the canyon is sort of down here it's right upstream of this and the alluvial fan starts about here which is right about here on this frame. It's hard to make out but this cross section here shows the elevation of the banks of the channel and it's very low when you get there it's not surprising that this flood would leave the channel and spread out over this fan. This transect here it shows the channels about a meter high over here is where the road is near the channel it's actually lower than the water level in the channel. So we think this is a reasonable expectation of what would happen in the case of a flood. It's hard to make out the scales here but this looks pretty devastating but if you take a closer look at the scales here it's actually quite shallow and it seems more like a flooding nuisance than a real hazard in terms of lives and so forth. And again this was our worst case scenario. So moving on to another application these are somewhat this is much bigger potential events. We're looking at potential lahars off of Mount Rainier particularly coming from the west side of Mount Rainier that could flood the Puyallup River basin or the Paswali River. Most focus on these Mount Rainier lahars has been on the Puyallup River drainage in particular this community of Orting downstream here is quite aware of the potential of lahars and they're very proactive in terms of they have sirens and they would be triggered if there are early warning triggers up in the channels of the Puyallup that would be triggered by a lahar that sets off alarms in Orting. They do drills every month where essentially people leave the town. So the town here is built on lahar deposits and it's sandwiched in between the carbon and Puyallup River. So they're right in the path of the flood and their evacuation strategy is basically to walk across the bridges to leave this flood plain. All of the schools are located in this flood plain here. Now the knowledge of these lahars a lot of it comes from this early study where they've identified old lahars deposits from what's known as the electron mud flow. So this was a so-called unheralded lahar that occurred without not during eruptive activity. It was just a landslide essentially in the Sunsend Amphitheater. About 260 million cubic meters inundated this map here. So we've focused on potential lahars of that size because we're looking at we want to look at unheralded events that wouldn't have eruptive activity or precursory activity that would give an indication. Though those could be much larger but really these could happen spontaneously so that's what we want to look at in terms of potential hazard. We've identified two source areas all on the west side of Rainier but one is the Sunsend Amphitheater here. It's sort of known as okay this is the most likely source of the large landslide. But based on slope stability studies done by Mark Reed there's also an area south of Sunsend Amphitheater here that's the least stable zone so we've looked at those two regions of sources. So this here shows the further north source from the Sunsend Amphitheater. When we construct our landslide failure volume essentially we just look at a map view of what seems like a potential likely failure area and then we construct a basal slip surface to give us essentially the volume we want. That basal slip surface becomes a continuous surface and then we just initialize the granular food material slowly rise the pore pressure until it fails and then let it go. So when we do this rising the pore pressure we manually raise it until failure commences at some point and then once that occurs we let the model equation evolve. We've done different landslides with different mobilities. This is one that's not particularly mobile you can see a lot of material remains up here in the source area. This is a map view of the same simulation. You can see flow goes down this Pealup River. This is the flood plain that Orting is built on and you can see it takes about an hour to get there. Now the community of Orting's very interested in the timing because they design these evacuation strategies and they take roughly 40 to 50 minutes to get everyone evacuated from that flood plain. It takes about an hour to reach there so it's there's not much time to buy so they're very interested in the timing but of course we can't say exactly how long it will take because different simulations of different mobilities give different arrival times. I don't think I can play that. So this is the other source area we've looked at. Same volume landslide. Now this one flows mostly down the Nisqually drainage which is a drainage that really hasn't been looked at as much. Here's a top view of that simulation. You can see it also bifurcates and goes down the Pealup drainage but also goes down this Nisqually drainage south here. Now another issue has come up with this these Nisqually drainage lahars and that's that there's a lake down here, Alder Lake, which is actually a reservoir. There's a dam here down at the downstream side so the dam operators are interested in this. What you know potentially a lahar could inundate the lake, rise the lake level, overtop the dam and cause a flood downstream of the dam. They're also concerned about you know with you're increasing the density of the lake with the lahar material that potentially compromise the dam. That's a question of course we can't answer but we can simulate the lahar entering the lake. Here's a photos of Alder Lake Dam. See rain air in the background here. It's actually about 100 meter dam. It's much more impressive when we drove up and saw this and it was much larger than I had sort of imagined when we first modeled this. So here's a oblique view of the same simulation. You can see the lahar coming down the drainage here. Interacts with the lake. Sort of a mixing region in the front. A map view of the same simulation. This shows the volume fractions mixing and the waves from this have already caused overtopping at the dam. You see the solid material but of course there's water waves in front of that. We thought that it might actually create a fairly large tsunami like some landslide tsunami model but because the lake's fairly shallow here at the upstream side it mostly just rises the lake level. So here's the same animation showing the water level rising and then eventually overtops the dam. Now it turns out this is actually another reservoir downstream of the dam. It's the original reservoir. It's smaller downstream and we're currently looking at how high could that reservoir raise and then that overtop that. So that's all the applications I'll show and just kind of point out some of the future directions we're looking at. We want to improve these models for entrainment. More physically based models maybe include true sediment transport deposition by settling. So in the Mahar model it can deposit simply from the material stopping coming less mobile. We're also looking at more distributed source debris flows. So the debris flows we model are based on these single source landslides which fail and become debris flows but one can also have debris flows that actually begin as rainfall runoff or flood which are in training debris and eventually entrain enough debris as they flow downstream that they become debris flows. Those are a different initiation process that we're looking at where we would drape material over a landscape and have some model for the core pressure increase based on rainfall or some other model. And then eventually we want to develop the same idea of multi-phase two phase models but have multi-layers. So this would be more appropriate for modeling things like submarine landslide generated tsunamis where you really need you can't really shouldn't have a single volume fraction through a depth in one region you'd want maybe a more dense material underlying more fluid material but then also have mass exchange and momentum exchange. So those are the directions we're going in and without all thank you and for any questions. Are there questions for David? So can these be used for areas where there's ice in the landslide essentially? Yeah certainly the ice can contribute to water entrainment. These are developed more for true solids. How applicable it is to a landslide which has ice? Kind of hard to say I mean it wasn't developed for that but the ice might you know it has some properties so it can be used for it is it as good for that? I don't think it's particularly earth-centric. We've modeled landslides on Mars. I mean it changed the gravity and and no no water. Well yeah yeah I don't think I don't see anything earth-centric about this but I think it could be applied to that. I just thank you which is really great great stuff and I just had a quick question on your last case study you showed the debris from approaching the dam and dam over top thing. Yeah. Have you done any dam reaching or dam? Well we're I mean we're that's kind of outside of our expertise studying the structural integrity of the dam but I think this this type of model could be used as an input for geotechnical engineers that could look at that but we haven't. So impressive simulation since you brought up the entrainment question what fraction of the hazards that you just showed us is due to material they're picking up along the way and how are you treating this right now? How much does it matter if this is a dry gravel bedded valley versus wet and so forth? What's how's what's the sensitivity look like? Well so in these cases it actually plays a pretty big role and I mean mostly for a few reasons one just increasing the volume you know we we of course like I said it wasn't real physically constrained about how much volume we're in training we try to look at worst case scenarios but simulations that just empty the lake and don't entrain any any volume are quite a bit less less inundation downstream and sisters for example in that in that example. So I think entrainment plays a big role in the size of the hazard also entrainment can significantly increase the mobility of the flow which is somewhat counter-intuitive but if the material is saturated that it's a training that can beat more mobility through rising core pressures. In fact we've done quite a few experiments at the flume where we have two debris flows one that hits a bed with an entrainment material it becomes much much more mobile so yeah I think it plays a big role and that's one reason we like to better have better physical models for that entrainment process. Okay we should move on but let's uh let's thank David one more time. Thanks