 Okay, terrific. Thank you, Johannes. Okay, our next speaker is Joel Johnson from UT Austin and he's going to be talking about using tsunami sediment transport experiments to improve heliohydraulic inverse models. All right. Okay. I can certainly hear myself. I'm sure you can hear me too. So I think that I'm going to be talking about something a little different. First of all, it's physical experiments. It's physical tsunami models. But you know, depending on that point of the challenges of sediment transport, I think was a nice segue. The other thing is that when Robert Weiss ended his talk this morning, you know, it was talking about field data, but really about the needs and the challenges of, you know, getting the high quality data on sediments. And what I heard him describing the whole time was flume experiments. So this is what, you know, basically, you know, the set of experiments that I'm going to talk about, I think, can in a sense do for you. This is work that I've done with a variety of people. Katie Delbeck did a master's project, collecting the data and doing some of the analysis. Also work with one suck him and David Morrig. I will say that most of what I'm talking about is published in a couple of papers that you see here. Now, motivation is to be able to take sediment deposits and of course do the inverse problem. That's where I'm coming at this from. To basically, you know, see how well we can predict the flow depth and velocity, for example. I put both tsunamis and storm surges on there because at least the models, and it's a particular advection settling model that I'm going to be talking about, you know, the model is simple enough that it really doesn't distinguish in terms of its inner physics between tsunamis and storm surges. So, you know, that right then and there, the fact that they're both there kind of say there's still work to be done to refine and improve our understanding. And doing this, you know, the benefit is to improve risk estimates for coastal areas because we're talking about extreme events that happen rarely. Most people are familiar with that idea. Now, there's a variety of papers. This is certainly not complete, but taking sediment deposits and quantitatively trying to infer the way of characteristics done in a variety of different ways. Certainly what I'm talking about is relevant to, I would say, all of these. And the big challenge is with field data. The big challenge with field data, while field data is great, is that you're chronically under-constrained and especially for the extreme events, both the rarity and just the difficulty to impossibility of making the measurements that you really want. And because of that, it's really hard to actually validate your models. You want to validate them independently of the, you know, exact events that you're trying to study. And flume experiments can let you do that. One other thing that I will say is, you know, I do think of these as physical models, but there are, I think, distinctions that are sometimes underappreciated between laboratory experiments and the models that, you know, this community and myself included, develop because, you know, in the physical experiments, you know, we simplify the geometries, we simplify the boundary conditions, but we don't simplify the underlying physical processes and interactions that go on between the flow and the sediment, whereas with equations that we develop, you know, those, that underlying physics is also simplified. So basically, from these flume experiments, we're going to compare to one example of an advection settling model to look at both the accuracy and uncertainty of that model because uncertainties are critically important here. And, you know, finally, a lot of what I'll talk about is, is not model validation or anything like that, but is just understanding and still incomplete understanding of the transport physics of suspended sediments in these experiments. In particular, you know, we'll be looking at grain size distributions along the flume and what controls them. And, you know, kind of a, you know, simple way to state the results are that, proximally, so over short transport distances, it's all about the source grain size distribution. That controls what the deposit is. As you go farther away, as you transport inland, farther in the case of a tsunami, your sediment deposits increasingly reflect the flow characteristics that we're trying to get out. So there is a spatial change from source to, to flow, I think. And I'm also going to focus on dispersion, which was left out of some models. And, you know, I believe from these experiments that dispersion caused by turbulence is fundamentally important and needs to be included in models where it's not. So what were the experiments? They were done in an outdoor flume at UT Austin. It's a, you know, 32 meter long flume downstream of the lift gate. So there's a computer-controlled lift gate where we can basically instantaneously release about six cubic meters of water at the upstream end of the flume. And it, it then, you know, moves down the flume as a bore, you see there. And the first thing it hits is a source dune of sediment. So basically the flow entrains sediment from that location. This would be like a, you know, barrier island coastal dune, basically a, you know, idealized localized source of sediment for the, for the experiment. Yet, you know, the sediment is entrained into the flow as the bore moves down the flume. And at the end of these experiments, you know, there is a, there is a barrier perforated to reduce return waves. But basically water gets to the end. Some sediment certainly goes out of the flume. And, you know, then the sediment settles, finishes settling, I should say, and the flume slowly drains. I'm going to present results from six experiments. All of the sediment was sand between about 100 to 900 microns. We used, basically, you know, some experiments used a finer grain size distribution within that range. Some used a somewhat coarser grain size distribution within that range. And we also varied the ponded water depth. So, basically, you know, five of the six experiments started with ponded water. And, you know, this would represent a case like a shallow lagoon, you know, a place where you're most likely to preserve a tsunami or storm surge deposit. We collected data along the flume in terms of, you know, flow depths, and of course how they changed with time. ADVs measured the flow velocities and turbulence within the flow. And then afterwards we measured the thickness and grain size distribution of the deposited sediment along the flume. And this, hopefully, is what it looks like. So, this is real time. So, it's an I plus. Now it's going again. This will go by pretty fast. So, that's a bore. We give it about 20 seconds. And we have a tsunami experiment. You can see right now some billows of suspended sediment in that flow as well. I'm going to go back one thing, but I'll just grab this. It's handy to actually look at this, you know, the wave actually coming in. You can see that there's actually, you know, a lot of aerated water at the beginning, but because of the ponded water, it takes a little bit of time, you know, again, fractions of seconds, but for the flow velocities to pick up near the bottom. And, you know, again, a little bit more time, but for a lot of the suspended sediment really comes into this. So, these are overlapping data from four different experiments that used basically the same initial ponded water depth and other characteristics were the same. They really show that the flows are pretty reproducible. You know, we get velocities that overlap for these. You can see turbulence in there. Higher velocities at first gradually decrease. You know, flow depth, you know, similarly, you know, goes up at first, gradually decreases certainly some waves on the surface. We can calculate fruit numbers and the fruit numbers were, you know, around one for most of the experiments. Other, under some slightly different conditions here, we had fruit numbers a little lower, a little higher, but that's the, that's the range that the flow, that. Let's start looking at the sediment deposits. So, here we're looking at the upstream end and moving along the flume. So, flume so proximal to distal. And, you know, again, there is an initial source dune at the upstream end, which was actually a fair bit deeper than four centimeters. It was about, you know, 10, 12 centimeters deep at the, at the upstream end. I think that's right, that initial number. In any case, that, the deposits for the six different experiments looked fairly consistent. You can see that there's, you know, a, a thicker part, you know, two to four centimeters in the upstream roughly third of the flume. And then you get measurable deposits, but fairly thin, you know, measured in the, the several millimeter range at the downstream distal two thirds. And we can take these flow conditions and calculate a shear velocity and then look at bed load velocities that you can predict from that using a relation from Raleigh-Martin, what I'm using here. And basically, in these different experiments run under some different conditions, you actually predict different distances that bed load should transport over the, over the duration of the, of the flow in the flume. So these errors that I've drawn on here aren't just eyeballed to the data, they're actually independent calculations based on the shear velocity and the characteristics of the flow. And so, you know, where we basically see this thicker pile of sediment to the upstream end corresponds quite nicely to the distance that you would expect bed load to transport and be transported along the flume. And so we're interpreting this upstream coarser part as bed load transport. You can also think of it as just kind of shearing of the, shearing of that initial source dune smearing out in that way. That said, there's also entrainment going on from the source dune and probably from here as well. And so this upstream deposit I think is really caused by a combination of bed load transport and deposition and suspended load deposition. Whereas we interpret the distal thin part as dominantly suspended load on deposition. So let's look at what the deposit, grain size distributions look like. This axis is transport distance, so moving down the flume. The green is actually thickness again for this one particular deposit. And, you know, this axis is grain diameter. So we have d50, the median size. It's relatively, you know, flat over the bed load zone. And then you see a decrease in grain size along the flume. And basically you see that pattern for both finer and coarser grain size per centile. At this point, let me step back and say, you know, are these flumes realistic? Do they actually apply to natural systems? Well, here's a comparison to some field data from a 2006 event. And, you know, basically I've normalized grain size and normalized transport distance inland. But for the different grain size percentiles here, you know, we get reasonably consistent trends. So, you know, at least that makes me feel good about the grain size trends that we see. Of course, we've looked at, you know, a formal number scaling. And I think of these experiments as, you know, one-tenth to 100th scale models of natural events. And so if you use those scaling factors, you basically would say these experiments are representative of, you know, flow depths during a tsunami of three to 30 meters. You can see velocities. The durations probably are a little short compared to some field cases. But, you know, one to four minutes of corresponding time. But I think that the scaling is good for the flow. It's also good for suspended sediment. So, we can calculate Rouse numbers that indicate that the whole range of sediment sizes we're using is at least suspendable. Some of it still moves as bed load, but it is suspendable sizes. What's now compared to a model? Well, more at all, and then John Woodruff at all, you know, with some modifications, presented an advection settling model. And the key here is that the combination of the grain size and the depth of velocity control how far grains move down, you know, downstream in the case of the flume or inland. And then basically we can infer the depth of velocity. A fundamental variable in the model is called the advection length scale. You calculate it from the average velocity, the average flow depth, and the average settling velocity. So basically the advection length scale is how far you'd expect a grain to be transported along the flume if it started from the surface and settled down. I noted this a little different than is often used to say that it's a still water settling velocity or really an average settling velocity. I'll come back to that later, but turbulence will give you a distribution of settling velocities in reality. That's not in this model. Really, the model is both of these equations, just crude number and that advection length scale. Now, an important assumption is that that distance that grains are transported by the flow corresponds to that advection length scale. They're not the same one as a measured distance from the flume. One tells you about the flow depth of velocity and settling. Basically, from these equations, we have two equations to unknowns. We assume a crude number or in the case of these experiments calculate it and go from there. The one other thing that we need to do is take deposits along the flume and pick a representative grain size with which to calculate a settling velocity to then go into the model. In this initial model, the way that that was done was the fairly reasonable assumption was made that the coarsest grain at a given location, let's say this grain, is the representative, is going to be most representative for the flow characteristics. The idea is that for grain of a given size, the ones starting out highest in the flow, get transported the farthest and that basically those grains are going to be at a given location, the coarsest ones in the deposit. So these are some model results. Again, the initial model was assumed to work for D95, or size fraction, and these diamonds are the D95 size along the flume. The green line is a fit to those data. You can see that it fits part of the data quite nicely, but it predicts a flow depth that's almost double, but was actually the experimental flow depth. Here you see a calculation based on the experimental flow depth, and that actually matches the distal portion, the suspended portion of the flume quite nicely, again, when using the D50 grain size. Now upstream you can see that the models don't really fit the data, and you can also see that the models are predicting a grain size, and the predictions of the grain size of the models are really a lot larger than and quickly get to be larger than the actual grain sizes that were used. And so the short of it is that upstream data not matching the model is a source grain size distribution effect. There just aren't the grains there that would settle fast enough to meet the assumption. Now another thing that we can do here is basically go along the flume, so they get upstream downstream, and figure out what grain size percentile, what percentile within the deposits best predicts the experimental flow conditions, the depth and velocity, and it varies for some experiments, and particularly these two, at the end of the flume you're still at a relatively high percentile, D80, and these were the experiments that used a finer grain size distribution as a source. Down here you can see several experiments that used coarser grain size distributions, and those are met more by intermediate grain size percentiles around D50. Now there were different advection length scales for those different sources, basically because the finer grains settle more slowly, that gives a longer advection length scale there. And so comparing the results at the same distance along the flume isn't necessarily ideal. So basically I'm going to do a non-dimensional transport distance, normalizing the actual transport distance by the source advection length scale, and that does give a not perfect, not terrible collapse of the data, so this would be a non-dimensional transport distance. And so basically, you know, interpretation here is that you have to be greater than an advection length scale for your deposits to really be sensitive to the actual flow conditions, and upstream you're basically much more representing the source grain size. We can predict uncertainties for the different grain sizes from this. It's perhaps a little confusing, but basically for different grain size percentiles I calculated since 95% confidence intervals based on these six experiments. And so, you know, for an intermediate percentile, you know, a value of 0.5 here would say that we're predicting within plus or minus 50% the flow doubts, and actually predicting velocities a little bit better. So we can look at uncertainties compared to the model with these data. You also see in the middle, you know, you can pick a broad range of grain size percentiles and not have that much change in general answer. So I am gonna have a little bit more to say, well, why is the D50 or an intermediate grain size better? And as I've hinted and said, I think it basically is coming down to turbulence. And so, you know, the turbulent velocity fluctuations mean that it's not just whether particles start higher or low in the flow. It's whether they, you know, happen to be dispersed farther or due to dispersion transport a shorter distance along the. And so, you know, the assumption that the transport distance matches the convection length scale is basically going to be true for the average blood transport distance of the grains, but not for the part influenced by dispersion. And the thing is that it's actually the tails of the distribution, say, of transport distances. So it's tails that represent the dispersion. And so basically, the grains of a given size class that go far this down the slope represent the ones that happened within spurs farther down the stream. You basically have to make a mental calculation, you're a person to go from single grain size and how it's spread along the flume to vertically integrating, but a grain size distribution would be. But, you know, basically, your finer grains at a given location, you know, the smaller percentiles are going to represent, you know, the grains that happened to disperse shorter distances or courses for the farther distances. Your intermediate grains, roughly the 50, are the ones that are going to represent your average blood length. I'm going to zoom through these quickly. But conceptually, we can look at advection length scales versus the actual transport distance along the flume. And, you know, now on an advection length scale, you basically would have coarser grains that got dispersed, you know, finer grains that got dispersed at this end. And again, if you take the same axes, but now think in terms of at a given location along the flume, what is the grain size distribution made up of these different sizes going to look at? Basically, the finer percentiles are going to, you know, your advection length scale will increase faster than your measured transport distance and vice versa for coarser grains. Well, we can actually, you know, test, test relations like this based on the flume experiments, because again, we know, you know, we actually, because it's an experiment, independently know UH, we calculate the settling velocity, and we know the grain sizes, we know the transport distance. And so, you know, for the 50, and we find these are just two example experiments, we find that, yes, in fact, the measured transport distance is a good predictor of the advection length scales. And that we also have the expected trends that dispersion should give you for both coarser and finer grain size classes. So, I am wrapping up. Honestly, didn't get as far with this as I wanted, because it's a work in progress. But the next step that I'm doing with a variety of people who have contributed in different ways is, you know, putting together a simple particle tracking model, basically where we can put in a grain size distribution, put in synthetic turbulence or, you know, real-time series of turbulence, and look at how grains get transported and dispersed down a flume and make deposits. And I'm really excited about this, because I have interpretations about dispersion, also about why resuspension and such should be relatively unimportant in these experiments, but they are interpretations. And so with a particle tracking model that makes deposits, I'm planning on testing the sensitivity to resuspension and various other factors. So, to conclude, I am honestly surprised that more work hasn't been done using flume experiments to do tsunamis, storm surges, rapidly changing flows, because the scaling actually works out quite well. And I do think of this as a benchmark data set that I hope that people will use, please do. I'm happy that Peng and Weiss have used it for some of their model validation. And, you know, another thing is just some models oversimplify the physics. We all kind of know that, but leaving dispersion out was kind of a simplification too far, I think, for some inversion models that have been picked up and used, basically because they weren't progressively tested ahead of time based on flume experiments, for example. And then a couple of details specific to these. Thank you very much. Thank you, Joel. Are there a couple of questions? I'm just curious, are you assuming basically that the different particle size classes don't interact with each other and treating them? So, your answer is yes, absolutely. The Infection Settling Model assumes that in what I've done with this, I am assuming that as well. I certainly appreciate that that isn't necessarily true. And, you know, smaller grains can damp turbulence, larger grains can enhance turbulence. I would say that the understanding of applying that to deposits like this isn't yet at that level of the underlying physics. Yeah, yeah. So, I think that you can find many papers when you go looking for this that make the comments that we don't understand entrainment into suspension really well, and that it's hard to predict. And so, I didn't show them. There's actually some hints in these data that the entrainment into suspension is grain size dependent even for these sands. That is the opposite of what my intuition would say. I'm not totally sure it's true. We're way above the thresholds of motion for these grains. I would think that they were just being trained independently of size. There's some trends in the grain size data that suggest otherwise what research is all about. So, figuring out new questions. As you know, many traditional sediment transport study, they try to quantify the amount of bed load and suspended load using the full parameter, which I think is the ratio of settling velocity versus shear velocity. So, I think one intuitive intuition regarding the scale effect is that in the real field situation, your shear velocity, bottom shear velocity. So, you may not even have that much. I wonder what is your Yeah. So, thank you for saying that. So, the Rouse number that I briefly showed up