 Let me find some place to put this. All right, so, oh, OK, good. All right, so I'm going to, since I'm day three, I've sort of modified a few things. So maybe like the first talk we had, title isn't as appropriate to what we're going to talk about here, but it's generally good. Now, if you've all been watching any of the local TV in your rooms, you've seen an awful lot of political ads. And, but I heard a couple of things that caught my ear. One of them was a public service announcement. Those are the kind of things we put out when there's a storm coming, right? Let the public know what's going on. It's essentially a public service announcement. And it's about a store your marijuana in a place where your kids can't get it. OK, makes sense. And at the end of that commercial, a woman, the mom, is walking off the screen. And in the background, her kid says, hey, mom, I want you to do something. And she goes, whatever. It struck me kind of. First of all, I never thought I'd heard of marijuana commercial on TV. But that's the same kind of attitude that the public takes a lot of times. With this information we're trying to give them, they're either overloaded or we've talked about this on a number of occasions about the false positives. Oh, yeah, last time they said it was because it didn't. But the other one that caught my ear was an all-state commercial. I haven't seen this one in Louisiana either. The guy says, a 500-year storm comes every 500 years, right? And then he says, in the last decade, we've had 26 500-year storms. But again, I don't think the public gets that either. So I don't know how we communicate some of this stuff. Is this working? This one here? All right, sorry. OK, all right, good. All right, there's forward, there's back, there's forward. Got it. All right, so I thought I'd start by just talking about, so what is risk? A risk is expected loss. And statistically, you can calculate this. The top equation is the way a frequent statistician would calculate it. You have some decision rule, and that determines the difference between an observation and your decision point. And you take the expectation over your data and pick the optimal, the thing that minimizes your error or loss. And in a Bayesian format, you do the same kind of thing, only you have a prior. And I think in our case, an interesting thing here is that prior could be like satellite data that's constantly monitoring the seasonal, monthly, whatever time step you want to look at, whatever the hydrologic parameter is. It might be soil and moisture, it might be precipitation, it might be potential of aperture transpiration. There's a lot of these things that are available from satellite platforms. So that's your prior. And then you have some data that's sampled now, and you can update the prior and come up with a posterior distribution. And in the Bayesian framework, you have a loss function that you integrate over the parameters of that posterior distribution, and you pick the optimal action. Now, one of the interesting things, and we kind of did this the other day in our little DICE experiment, loss function counts. If your loss function is a squared error loss, like ordinary least squares, the least squares kind of paradigm that many statistical parameters are sampled over, then the optimal decision is going to be made at the mean of the data or the posterior distribution. If your loss function is linear, then it's at the median. And so why is that important? Well, a lot of hydrologic things aren't like nice, normal distributions with the mean and the median of the same place. And you've got large tails. You might want to be looking at the median of the response, which gives you 50% of time I'm going to be below this, 50% of time I'm going to be above it, and frame your decisions and actions based on that. The loss function we had the other day, by the way, was an asymmetric loss function. If you incurred a flood, it cost you four beans. If you didn't incur a flood, it didn't cost you anything. So you can actually work through the calculus of all this stuff and figure out what the optimal action would take on any one annual throw. In fact, you could even work in the nine-sided dice thing in that period of 10 years and figure out what the optimal thing is to do every time. So the point is the guys in the back just got lucky. Lucky beats whatever every time, I guess. So all this is models. So what's a model? And a model is an abstraction of reality. We like to think our model captures reality, but we do our best. So you create this model to sort of understand how things work and to capture the essence of whatever it is that you're modeling. And the better you are, the more credit you get for doing it right. But another important part about models, and we've talked a lot about this over the last few days, is that model has to be in the context of some decision support role. So that model has to deliver me information that I can use to take my action, determine what I'm going to do. And otherwise, it's just a model. So models are important to decisions. It has to give me things in the right time steps. It's got to give me things in the right spatial resolution and those kind of parameters. And again, in our case, it implies a spatial domain. So we're talking about a watershed. We're talking about a base in some place, or a coastline, or a lake. Whatever we're trying to model a flood risk in has a spatial kind of domain implied to it. We don't care about the things that don't drain in there, just the things that do. And then, of course, the time element is important. So do I need this information? Well, first of all, the data that I'm putting in my model, how is that provided to me? Is that an annual number, a monthly number, a daily, an hourly, every 15 minutes, whatever. Models operate all these, or data is delivered in all these different scales. Then the models operate in whatever scale they operate in could be any of those. And then the information that comes out of them has to be in a scale that provides me in the time frame that I need that information. So here's just sort of the generic model. You have the output. You have the model itself, which is this F thing. You've got an XYZ, in this case representing sort of maybe the characteristics of the basin. The spatial extent, the soils, whatever it is that are important in your hydrologic model. T, the time thing, again, this is a daily time step. Is it an annual time set monthly? What's important for making my decisions? The part that gets left out a lot of times is the thing on the end. As we focus on the model itself, and the error is just like this nuisance thing that we just kind of want to ignore. So we know our model isn't perfect, but it's OK. But that nuisance thing is comprised of a couple of different components. The error term encompasses the things we don't know, the things that our model doesn't account for in providing the output. But it also includes sort of intrinsic variability in that system. So for example, here, I've got three simple models. Let's just say the flow on the y-axis is a function of the area of the watershed. We can kind of all agree that that's kind of an increasing function. Bigger watersheds have bigger runoff, that kind of stuff. But you can take three watersheds that have exactly the same area on the x-axis. But they're comprised of different things. They have different combinations of soils. They have different slope regimes. By the way, different slope regimes. This is something that I think is fascinating. We do a lot of watershed modeling based on the area of the watershed. Well, you throw a slope in there. If you have a slope in that watershed, then the actual surface area in that watershed is higher. So we never account for that. And maybe that's an important factor in our modeling. So in this case, we've got three relatively identical watersheds. But their intrinsic variability is different because of how they're composed, at least with respect to the relationship between runoff and area. So in the upper left, you've got a relatively low error. You've got pretty good models. You can see the line there. The regression line in the second one is in the same place. But the stuff scattered around a little bit more. So we need to understand, and of course, in the third case, eventually, if your model doesn't do very well, variability can overwhelm it if you need to do more. So let's talk about geospatial models. And geospatial models are important in the context of what we've been talking about all week. Because they're the way we're going to get the things we need, the data we need, to answer our questions, to run our models, whatever. And we've sort of implied a lot of this stuff. The geospatial model is something that the model elements are spatial objects. And those spatial objects can take advantage of geographic information systems, concepts, and data, and technology. And the most important things here are the relationship between these objects, the spatial relationship. So the juxtaposition of elements, what's next to what overlaps what. And that's really important because when we're talking about data, and I'm not sensing guys understand this, right? You've got things that give you data at 30 meters, things that give you data at 10 meters, things that give you data at 1 meter. So you're always dealing with different scales of information. And in the geographic information system, geospatial kind of format, you can combine that information because you can handle juxtaposition. So what overlaps what is known, what doesn't overlap what for that matter then, is also known. And then topology. That's another really important thing, right? I don't care what happens in the next basin over, but I do worry about what's upstream for me. So connecting upstream and downstream, that's what topology is all about. So two kinds of basic ways that things flood. You've got rising water that just inundates stuff. And it causes chemical reactions and whatever. It causes infrastructure to have problems. If you flood a road for a couple of days, it's pretty well known that the foundation of that road is going to deteriorate over time. So that road's going to start buckling and cracking. And so, for example, in Katrina, we made sure we knew where every flooded road was. And we're able to put together our damage assessments and talk to the FEMA people and the government about that. So that's one kind of damage. And that comes from storm surge. We had a lot of backwater flooding in Louisiana. So when the wind is blowing from the east to the west across Lake Pontchartrain, the water level on the west end of the lake could be four or five feet higher than the east end. And that pushes across that little isthmus. We'll see a few maps later that separates it from Lake Marapa. And most of the things that drain into that basin come through Lake Marapa. So all of those basins, all of those drainages that are going to Lake Marapa now have higher water at the bottom. So the water that's coming down the drainage is backing up. And you can get backwater flooding clear into Baton Rouge, which is tens of miles away from there. So that's one of the ways we flood things is this rising water kind of deal. And flowing water, that's a whole other deal, right? Now you've got the force of the water in here. So you've got flash floods, tsunamis. You've got riverine flooding. And I'll talk about this, I guess, a little bit later when we talk about the flood of 2016. But I remember my guys upstairs, hey, we need to know what's going to flood in the Amite River. Can't do that without a model, right? You can do rising water flooding pretty easy. I don't know if anything's catching up here. So rising water flooding is fairly easy to do because the surface is well-behaved. It's got a little bit of play in it, but it comes up from the bottom of the watershed. But the flowing riverine channel type flooding has a slope to it. So I just can't say that if I got a gauge height up here, that everything that's below the elevation of that gauge height is going to flood, is contained in the channel. So let's talk about rising water flooding, specifically storm surge. So actually, on the right-hand side here, I've got a map of the slosh grid, right? The Sea Lake Overland surge from hurricanes. By the way, they now include tropical storms. So maybe they need to call it slots or something now. So that's a category zero storm when you get the data. So you can see this model starts way up in the New England area. It goes all the way down the coast. Actually, I chopped off Texas. But anyways, in that model grid are all the bathymetry and land surfaces and obstructions and things. We actually went out, by the way, the data that was available in that regard, particularly the land forms, even the bathymetry. Some of the bathymetry we had in South Louisiana was last collected in the 1800s. We've had a lot of sediment delivered into the coastal areas. So those are all gone. I think I mentioned something about this the other day about some of the lakes in the Piedmont. The lakes have all filled up from the bottom with sediment. So all of our flood models that are thinking that's a nice sink to hold water, they don't have the capacity they used to have. Anyway, I digress. So your grid has all this stuff built into it. We even went out, by the way, and measured the elevations of those Jersey barriers going down the middle of the interstate so that Hurricane Center guys could work that stuff into their models. This is a very important thing. All these little things become barriers to storm surge, either moving in or out, for that matter, and the effect of timing and ultimate water levels and all those kind of things. I've also included on the right-hand side here some details, the Chesapeake Bay, South Carolina, Coastline, North Carolina, and the New Orleans area. So they do this slosh model thing, and it's a geospatial model. It takes geospatial data. It has a geospatial framework within it, and it gives me data in a grid of polygons that tell me what the water level is going to be within that particular polygon. And it does that in actually several different products. It gives me a set of polygons with a 10% exceedance, 20%, 30%, 40%, and 50% exceedance. So how do they get this, right? So they take a storm and they run it into the coast, and when it hits the coast, it does all of the hydrologic processes that affect storm surge, and they do that a couple thousand times. And so the 10% exceedance is the 90th percentile of all of those outcomes. And then the 20 would be 80 and the 30. So the other important one to me is the 50. How do you interpret what a 10% exceedance means, how does the public do that? How do you interpret the 50%? Well, 50% is pretty easy, right? There's a 50-50 chance this thing is going to happen. I can make a decision on that a lot easier than, well, there's a 10% chance it might be this deep or a 90% chance, which is actually pretty good, that it's not going to be that deep. So that's an issue. So here's Hurricane Nate from last year. In fact, it was about a year ago at this time that Nate was wandering around the Caribbean and the Gulf. Oh, and by the way, I noticed yesterday that the Hurricane Center is watching something in the Western Caribbean right now that they're expecting to form somewhere, maybe down in the area between the Yucatan and Cuba. Got 30% on that one right now. So here we are looking at one of the things I look at as I'm collecting information. This is a Huravak, which is a decision support system. It's right now it's on a Windows platform. They're moving it all into the web now, but it'll be the same kinds of stuff. This is the Winswath representation of Nate at this particular advisory. And you can see that Louisiana is sort of at the edge of the tropical storm force winds, the blue stuff. But Storm Surge, it didn't follow those lines. The Storm Surge is this what they like to call it, a dome of water. It's being pushed by the winds. It's usually pretty heavy on the right-hand side of the storm. So we're in a good spot in that regard and it's gonna be Mississippi and Alabama to get the worst kind of storm surge from this. Another interesting thing, and I'm not sure how well it was modeled, but what was it, Maria? The storm that went up the middle of Flora, on the right-hand side of that storm, was pushing water on shore and flooding things. On the left-hand side of the storm, it was pulling water away and there were boats sitting on the bottom of the bay kind of stuff. And that's not just like you're talking about the pleasure boats kind of stuff. The guys that are trying to run freighters in and out, they gotta worry about the depth of that channel so they don't get grounded. So this is the grid I get. And so remember that big grid that goes from Maine to Texas? I just clip out the part that's important in Louisiana. And this is sort of, you can see kind of the extent of what I get. It's kind of hard to show things, but you can see underneath the landforms, you can see the places where there is no grid cell, it tells you to expect that to be dry. But in the top center of this, you're clear up in Baton Rouge. So I-10 and 12 go across here. They come through Slidell on the right-hand side of the map and run across to Baton Rouge, cross the Mississippi River and go into Houston. That's the major east-west trucking route between the west coast and the east coast and on the southern tier, right? You gotta go up to I-20 before there's a next east-west. And that goes across the top of the state. That's our ultimate route. So now I got these cells and they tell me what the water level's gonna be. In that location. And that's how I do my analysis. So let's talk about what the products are though, because we talked yesterday about how in a real storm event, I don't get operational forecast. So an operational forecast is a forecast that's based on the actual storm conditions, right? So they come up with a advisory every six hours and that sets what the Hurricane Center thinks will be the direction, the forward speed, the strength of the storm. And then those parameters are put into the model and run. And they're varied a little bit to get the sort of stochastic element of all of those things. The storm track isn't gonna be exactly where they said it was in two days. The strength might be a little bit more, a little bit less, those kinds of things. So they run all these simulations, a couple thousand simulations and then they deliver these grids to me. But in that three days, right? So most emergency management, a timeframe start five days out. So at 120 hours. In those three days before that 48 hour first actual operational forecast, I've got what's called moms and meows. By the way, I'll point out that for Katrina, tropical depression 12 formed at 117 hours out. So we were already three hours behind our timeline and it was just a tropical depression. Not even a hurricane. Anyway, and the track came across Southern Florida and it went up the west coast of Florida and it stayed over there. That was Tuesday when it started and it stayed over there until Thursday and all of a sudden it jumped over to New Orleans. So you got to account for track and certain things there. So let's talk about what the moms and meows are because that's how you're making your decision. So the meow is again, a series of simulations given a storm direction, a storm strength, forward speed and tide. They use a mean tide and they use a high tide. And so they run thousands of those simulations. So you've got for every combination of those factors, you've got a cell in the model. And they cut the, to make this a little bit simpler is they cut the, that long big model into basins. So there's a Louisiana, there's a New Orleans basin which runs from Mississippi to almost all the way across the coast. There's a Vermillion basin which is kind of the south central part runs from maybe the east side of the Mississippi River to about Texas. And then there's the Sabine basin which kind of picks up in the middle of Louisiana and runs into Texas. So that helps them reduce the delivery issues for if you did that for the entire coastal model, it would be huge. And people are not interested in anything but their own stuff, right? Anyway, so the meow is, I know the storm direction, I know the storm strength, I know how fast it's gonna be going and I can get a better estimate than just a wild ass gas. Well, what they did then was they said, well, we may not know all these things so let's take all of them meows and we'll stack them up and we'll pick for each of the cells the worst, worst case or a given category and title regime. So regardless, basically what we're saying here is regardless of storm direction, regardless of storm speed, the mom is a worst case scenario or what we expect to happen. And that may be all you've got for a couple of days. In fact, see if I can do this right. So this colored windswath, that's 72 hours. So you're not even gonna know a real storm direction or forward speed until 72 hours out and that's still 24 hours before you're gonna get an operational forecast. So those are the moms and meows that you use for mitigation purposes maybe, for drills for working without perfect information or maybe just informing the public, you wanna get out of the way because this is what could happen. And then the operational forecast come out and again, the operational forecast is the same model. They take the actual storm parameters, read them in the model and calculate what the actual damage is going to be. So what I do then is I get that grid and this is a little piece of State Highway 23 down very near the bottom of the Birdfoot Delta. The end of the road is probably not too many miles south of here which would be to the right. And I can subtract the water level elevation in that slosh cell. I can subtract from that the elevation of the road. So on a two year annual cycle, by annual cycle, but on a two year cycle, we go out and we collect road inventory information. As part of that, we get a point every four million miles. That's 21.12 feet. So we have a location and an elevation. We're using real time network, GPS to pull this down. The RTN guys tell, well, it's like kind of golf ball accuracy. Maybe softball for three dimensions. So we've got a pretty good idea what the road elevation is at these points. And then that gives us a basis to calculate the potential flood here. And I don't know how well you can actually see it, but there are red dots on there which are the high end of flooding. It might be a feet, for example. The lighter dots maybe something, I think I put these in quarter foot increments. So the lighter dots are slightly flooded kind of areas. The black dots, by the way, are places that don't show up as a flood. So the storm surge elevation was below the road elevation at those points. And the top map is the 10% exceedance slosh. The bottom map is the 50% exceedance. So you got a 50-50 chance of seeing this. You got a 10% chance of seeing this other one. And a couple of interesting things. I finally figured this out last week, talking to the road inventory guys. I kept looking at, I kept getting values that had like minus 40. There's no place in Louisiana that's 40 feet below sea level. And there's things that are maybe 20 feet. But I realized, I actually have three tunnels. And those are tunnel road elevations. We got a tunnel underneath the Mississippi River south of Louisiana, one just upstream and one over in Holma. And I just thought those were errors in the data. So now I can start telling people, well, we got to close it. It doesn't matter, by the way, where it floods on this road. If it floods anywhere, we can't get people through it, right? Oh, and what is a flood? So a flood is, or a road closure criteria, our decision rule. Three inches, right? Three inches. And you start getting more than that and cars start going off the road and you have other problem with getting traffic through. Another really interesting thing is that slosh now includes tide, but it doesn't include wave action, all right? So, you know, wave action, what kind of wave action do you think you get with a storm? Three inches, six inches of foot? So it really means almost any flooding is gonna be closing that road when you throw the wave action thing in. By the way, there's a huge area of research on that topic. So anyway, so that's what I'm looking at, you know, at the detail level. This is the same data on the left, 10 at the top, 50 at the bottom, with our district O2, the entire district, which is our Southeast Louisiana group. On the right-hand side is Buris. My sample came from up on the left, right-hand side of that. So this is kind of the local community Buris during Katrina was an urban cluster in the US Census data. That was what, 2005? In 2010, it wasn't anymore. I mean, it just wiped through there and took things out. I drove through that area in maybe October of 2005, and there were trucks up in trees and boats and houses in the middle of the road. You had to go around them out in the field someplace. It was pretty amazing to see what happened from the storm surge. All right, so what are some of the issues? Timing is a real big issue. So I know how high it's gonna flood in this six-hour window, right? But I don't know when that's gonna happen. And two, three, four, five hours of lead time on when I have to close a road and evacuation route in particular, could be very important. This is just not in the data. It's not available. So timing is an issue. Again, the forecast comes out every six hours from the hurricane center, and then the storm surge guys start working on that forecast and in about an hour, they deliver the slosh products. So there's a little lag in that. There are some other issues with the data. So the way they communicate the data to the public, so when you're listening to the weather channel or your local weather guys, they're all reading the same data by the way, right, from the weather service. They may employ some commercial models too, who knows. But they're looking at all this stuff, and what they're being told is there's this above ground thing, okay? So the hurricane center decided a few years ago that people don't understand of storm surge flooding from water levels, right? And, but they would understand above ground level. Well, the problem with above ground level is you're reporting depth, okay? So the water surface is fairly well behaved over fairly large area. These cells could be thousands of meters, but the land underneath them have a lot of terrain, right? The second derivative of elevation underneath them is fairly high. And that causes in these AGL above ground level numbers situation where if you're in a local high spot, it way overestimates your risk. If you're in a local low spot, it way underestirates your risk. So you still gotta know where you are sort of in a relative sense in whatever grid cell you're in, which the public doesn't know that. The public can know, by the way, and a lot of them in Louisiana do, what their base flood elevation is, what their slab elevation, what their house elevation is. So giving them elevation in my opinion is more useful. And it's in my case, I can't use a depth. There's no way to back out the water level elevation and run it against my road points. So I pull down the water level elevation directly and I use that. Another issue I think is this concept of 10% exceedance. I don't know how the public really interprets that. It really does mean that 90% of their simulations were below that. And that's, how does the public understand that? Now, 50-50, and I talked to our engineers about this when we decided how we're gonna use these data. I said, well, let's look at the 50-50 because depending on your loss function and depending on your risk aversion and whatever kind of stuff, if there's a 50-50 chance that this thing is gonna be at this level, then people understand making that decision. The engineers kind of got that. And we look at both. The Hurricane Center guys call that 90%, the 10% exceedance value, maximum regret. And their interest is in trying to get people to be motivated to evacuate. So they wanna give them a high value. I understand that. But these are just some of the issues that happened with the storm surge data. So let's talk about the flood of 2016. And I'll apologize for not really getting what I would have liked to with this, but this is the Lake Pontchartrain Basin. It's got this kind of conehead top over on the northwest side. It's bordered on the west by the Mississippi River. Okay, so the Mississippi River levee starts in Baton Rouge and goes all the way to the south. From Baton Rouge north, there's a bluff, actually, that everything in this diagram on the Mississippi River flows to the east and into Lake Marpar Lake Pontchartrain and then out to the Gulf. A lot of people don't understand that. We're looking at hydrologic unit code level 12 basin delineations, local watersheds. They're called sub-watersheds. And in those sub-watersheds, the HUC system is kind of a hierarchy of things. So each of these belong to a watershed, which belongs to a sub-basin, which belongs to a basin, which belongs to a sub-region, belongs to a region. So the region here actually starts at the Mississippi River and I think goes all the way to Virginia or something. That's the south region. The region on the other side of the Mississippi River goes all the way to Mexico. That's the Texas Gulf region. And the rest of you guys, the 42% of the United States that is upstream from Louisiana, that just comes down between those two little levees and you pretty much keep them in there. There's no interaction. Actually, they're trying to use Mississippi water through coastal restoration processes. They'll siphon out water with sediment in it and try to re-nourish the marsh with it. That little, thin piece of thing between the levees really doesn't interact at all with the local hydrology. So let's talk about this watershed. So Baton Rouge is kind of over here on the left-hand side in the middle. I-10 and I-12 kind of come across the middle of this thing. They built I-10 and I-12, mostly to stay way above like coastal flooding elevations. Although, because interstates were required to go through major cities, they had to dip I-10 down into New Orleans and back up. Most of I-10 is actually on a bridge in the New Orleans area. Not hardly anything's that great. The ones that are get flooded. So this is the basic hydrologic layout. You really can't see what's going on here very well. So let's go to the next slide. So in this slide, I've gone to the next level up in the hydrologic units, the HUC-10 level. So each of these HUC-12 sub-watersheds are now colored by the watershed that they belong to. And you can kind of see here now that there's a sort of a drainage pattern on the far left side. And we kind of stripe our way down through that sort of conical upper part of the watershed in a North-South sort of fashion. And then we start at some point moving to the East. And this is part of what caused the flood. Okay, the 2016 flood, I don't know how many people said, oh, they never flooded here. We, well, I've lived here for 50 years and we never had a flood. We lost something like, you know, 60,000 vehicles because people saw water in their yards. By the time they got the stuff together, the car was flooded and they couldn't get out. It was a fast-moving flood. It came from all directions. One of the most amazing things is on one of these basin boundaries, one of my friends lives right below the road that he lives on is on the ridge between two basins. It's down towards the Amy River. And this house got flooded downhill. I don't think a flooding coming like that, right? It came over the top of the basin boundary just north of him. His house sits maybe six or eight feet below the road level and it cascaded. He had white water coming over the edge of that into his house and, of course, now he's got flowing type damage. So anyway, so, and this is the sort of geospatial things. Like the arrangement of these things is important. So what happened with the inputs? The inputs were like 35 inches of rain over night kind of stuff. We can't even, we can't even design for a storm that's that big. You couldn't afford to build the highway high enough or the building strong enough or whatever. But what happened is over here on the right-hand side in the Big Bend area of Florida, there was a low pressure system that kind of wandered around the Gulf. Never got a name. It wasn't strong enough to become a tropical storm, even a depression. And then it just kind of sat over there. And my original thought, and this is again where the spatial and temporal stuff comes in. My original thought was when I look at the rainfall patterns on this, what I'm gonna see is this counterclockwise circulation centered around that low that's off sitting over Alabama and just constantly whacking us. There's a really cool site at University, I guess it's Iowa State University, where you can download historical data rainfall in five minute time increments. And then again, we're gonna talk now here about some time things. So what I found when I started animating the rainfall across this basin is the rainfall actually pretty much came in from the Northwest and across to the Southeast. And so it's, because it was so big, it was filling up the entire watershed simultaneously. But remember this drainage pattern. Drainage pattern is these sort of North-South troughs that contain the water within their boundaries until they got down to where they made a left turn to the East and met the backwater flooding coming out of the Amy River and stacked up and then went over the lip of the basin. We just don't see that kind of flooding. I mean, that's just, it was very, very, very odd. The other thing that I thought was really important when I started looking at the precipitation animations was a time step. So if I did an hourly time step, I saw one thing. But if I cranked that baby up and looked at what happened every five minutes, what I was seeing was these big red things that popped up and maybe only lasted for five or 10 minutes. But those things were raining at just incredible intensities. I went to work on Friday morning, which would have been, I think, the 12th of August and almost got flooded out, getting to work in downtown Baton Rouge. They eventually told us, don't come to work. You know, when you started 6.30, you're at work for everybody else as to make that decision, I guess. So I got a call that evening from the Secretary of the Department of Transportation saying the Weather Service wants to know how high the bridges are on the Amy River and the major rivers on I-12 that come from Slidell. It's kind of a shortcut, so you don't have to do it down to New Orleans on the interstate system, because they think that those things are gonna flood. So I got out my road points and now that's the elevation of the bridge deck. So I had to find an engineer that could tell me, you know, what's the bottom? Take about six feet off of those and that'll be the bottom of the bridge. Well, there's a gauge right upstream on the Amy River from where I-12 crosses. That gauge, I think, got up to 46 feet. The bridge, the bottom of the bridge was 29 feet, okay? But remember, we're talking about a riverine flood. That bridge never flooded. And the Under Secretary, the decision maker people at DOT kept saying, you need to do me an inundation map. I can't do an inundation map in a riverine flooding situation without a model. I have to contain that stuff within the basin boundary and I have to know the elevations along the sides and it's not just a 2D thing. So I guess I don't talk with those guys much anymore. So here's a cool thing I did. There's a product in the ESRI suite of products called Schematics. And I actually did this a number of years ago. The watershed boundary data set tells you what watershed you flow into. So I can build this flow network and I can actually visualize things this way that you can't by looking at this kind of a representation and certainly not over here, right? You can't tell what goes where. You can kind of get an idea this way. But when I build that schematic, that flow diagram, if you will, I can know where everything flows. In fact, it was amazing to me. I lived in Baton Rouge for 20 years and I didn't realize that almost all of the things that go into Lake Pontchartrain go through Lake Marlpov first. Nice little settling basin. But you can also then create a schematic that's more of like an engineering diagram of this network. And you can see where these points are correlated in between the geographic schematic on the map and the engineering diagram, let's call it on the right, and where those basic choke points are. And those are the kind of, we talked about visualization. These are maybe some of the tools that we could use to identify places that are gonna be important in a hydrologic event like we had in Louisiana. So time was a big issue. And again, just visualizing things. I didn't see the red spots like I saw. I didn't see them at one hour time steps like I saw them at five minute time steps. So, we need to understand some of those concepts. So this is actually the diagram in a little better, little bigger picture. So you can, for purposes of say flood routing or maybe the local floodplain manager or something or other, you can get an idea. And of course, actually you can put the gauges so you can identify where gauges are along this network. One of the cool things on the network is within the ArcGIS platform, you can actually trace things upstream and downstream. So it has the topology built into it. And I believe that is it. Thank you, Jim, for this very nice and technical account. Somehow it was very important to link it to modeling and everything because there's a lot of stuff we just assume in our models but we can't really test them in a context like this. But I probably should have had a concluding slide that said, and this occurred to me in one of the breakouts yesterday is, you know what, we've got a whole lot of hind casting we can do, right? We can go back to these different events and look at the historical record and say, doing an expected value of perfect information kind of analysis. If I'd have known at this point in time what it really did other than what I knew before we got to that point in time but I made a different decision. And then talk about the things we need to make those decisions. I think something like this could be a use case almost like a benchmark study to see with different models, do we get this right or not? I would be very surprised to see that we get this right but we could learn a lot of lessons. For instance, how is, when you get this above ground elevation on the roads you actually need to count also for maybe the rack points are moving to the black points and you can only do that as Jim pointed out to the, or 3D, I don't know, model. You know, another, in the GIS business I've developed some concepts over the years which was three rules of GIS. The first one is everything you know is wrong, right? And that's because everything we do is a model, right? It's not reality. So we have to accept the fact that we've modeled reality and that there's some fuzz in there that we can't account for. So don't think of your boundary lines, your grid, your soil polygons, all these things as being perfectly located because they're not, don't even think about it. Any quick questions like one or two we can take before others introduce crack? Oh, no, we get another one. Nope, yes? Specifically for the storm surge. Well, okay, I'm monitoring gauges all the time. I've put together some, actually there are RGS online story maps that you can look at current gauge information. So I'm looking at those where that comes in handy. We have had issues and I think this is important to bring out because we've talked about the credibility of what the information we're trying to communicate. We've had some issues in past storms where like the LSU guys that run an ADCERC model they wanna get their model on the table and they wanna make decisions. It turned out that their forecasts were really pretty bad. But, and you don't want people doing unauthoritative forecasts and sending them out to the public. In this case it was even worse because it was going to the governor and the emergency operations guys. And it just confuses them. And I think another thing is we talk about storm surge, expected flooding above normally dry land. So what does that mean? I think people would really rather hear a decision than a data point. It get out now, because we think you're gonna flood really bad. Rather than, well the water's gonna be 10 feet there. Coming back to the communication issue that was pointed out so many times doing this. Any more question? Yes, the last one, yes. That's absolutely true. Keep in mind that NOAA does the storm surge stuff and the USGS or the weather service. So there's some of those kind of issues. If you know anything about NOAA it's a very compartmentalized thing, right? There's guys that do the track. There's guys that do the strength. There's the guys that do whatever. And they're different people. And they kind of, I don't know I have this vision that when they set an advisor out they all kind of come together and say, well let's do this, let's do this. But trying to deal with flooding coming out of the gauging system and combining that with the storm surge that's a real difficult problem. Always figured it out. Thank you. I think we better move on to keep it in time. So thank you, Jim. Albert, do you want to? So it's the important thing to think about.