#51 Dam Breach Modelling

subscribe to this youtube channel by clicking the subscription button below for future webinars and online short courses please visit our website at australianwaterschool.com welcome to this australian water school webinar brought to you by ice warm this webinar series aims to bring you the many innovative ideas and projects coming from the global water sector and today we're going to look at one very important idea and by the look of the number of people here it's pretty important it's called dam breach modelling by three professionals in this area bill syme chris goodell and craig price today's presenters and uh they're bill sime chris goodell and craig price um all three deeply involved with modeling two flow in in bill simon's case and uh chris goodell uh and craig have been certainly into hec-ras um bill's had 30 years experience i won't read the whole thing riverine estuarine and coastal studies um and most have been in the field of flood hydraulics and chris gordell is consultant for hydraulics and hydrology at client associates specializes in water resources hydraulic engineering and craig price educated university of california berkeley is a civil engineer project manager with international experience it is going to be a feast i can tell you that right now are you there guys uh let's see your great faces on screen there we go good i'm bill up in brisbane queensland craig in wa and chris goodell on the west coast of the us thank you so much for joining us gentlemen and taking the time to do this together so i'm going to let craig kick this off and moderate and then we'll go straight to um to chris and then straight to bill so over to you gentlemen thank you so much for taking the time today uh we are looking forward to this uh time together great okay um thanks i'll kick this off all right so um so maybe quite not not quite as heroic as um as these guys all introduce our dam busters for today um this will be um over to um chris uh myself and bill we've divided this geographically um i am curious can you see my mouse when i move my mouse that's so good to highlight things with my pointer um basically we're dividing this geographically into three reaches i will be talking about a couple of basic um items about the reservoir drawdown how we simulate a reservoir chris will talk about the dam itself how we model the breach parameters and bill will talk about how that flood wave is routed downstream um this is just a basic profile but um if we take a look at a movie screenshot if we could render things this well i will be talking about the reservoir how it draws down uh first we'll talk about the dam breach itself and bill we'll talk about the flood routing um now these are just a few uh shots just so we can get a feel for what we are talking about it should look this exciting but when we do 1d and 2d models um you know we are depth averaged we won't quite be able to do this unless you're getting into full um computational fluid dynamics models which um you can see some pretty cool videos of some of those but you can see the reservoir drawdown right there that's what i'll be uh i'll be discussing um we will um then get into some of the 1d items um which is like you see at the top there the reservoir or the dam breech formation um and then when we uh get down into the flood wave we'll talk about the things specifically like you see superman here uh trying to uh outrun the flood wave there how fast does that flood wave propagate downstream how long do you have uh before you need to evacuate what are the different modeling parameters that feed into that now i'll kick this off then just for a couple of minutes um i'm going to keep my bit fairly short because i'm going to just talk about one thing that's a little bit boring on the damn breach modeling side of things and that's the reservoir so if you look at homer here the question is do his toes get wet before his belly if they're at the same level if you pull the plug what happens and if you think about this in a modeling sense you know what are the parameters that drive this what is it sensitive to does this drop as a level pool does everything drop everybody you know does he get uh get his toes and his belly exposed at the same time or because his toes are closer to the drain do they get exposed first and that's going to depend of course on the size of the plug how you open the plug you open it slowly or quickly it's a big drain a little drain how is the bathtub configured is it long or is it narrow or is it just straight if it's just dropping as a level pool it becomes a very simple exercise you define this with a stage versus discharge rating curve basically and then take that and turn it into a stage versus storage curve it has your bathymetry built into it and you can make it actually into a zero d no dimension no time dimension at all uh item that can be a stage storage curve that you can run into one model basically the reservoir if you're discharging a certain rate of discharge out of it or into it and you multiply that by the time step you get a volume and that volume what happens if you're dropping as a level pool is that it occurs all at once meaning you know you could have a 10 kilometer long reservoir you draw a drop of water into one end of it the other end reacts instantaneously the stage storage curve reacts instantaneously and we talked about a little bit about how the shape makes a difference of homer's bathtub well the shapes are actually very complicated um in our exercise this is one of the things we're going to use in the upcoming course in march uh is grand canyon we've got lake powell up at the top lake mead at the other end and glen canyon dam and hoover dam and we're gonna put a dam right in the middle and look at the complex shape of this reservoir um you know how do you define that uh and of course we'll go ahead and talk about the reservoir shape and put a few of these in there and these are just some terrain files that i've built very simply to show the same volume configured a couple of different ways and then we'll run that through a few different ways and see what happens to the the flood wave as it goes through if it is a long reservoir and everything is in line you get a quite different reaction to the case where you've got a you know kind of more of a square or a circular reservoir and on profile you'll notice a few things that you can model now again our question is does this behave and drop as a level pool meaning every time you take a certain amount of flow out of it the pool drops together or does it drop dynamically and here's a case where this is where i've got the reservoir that i just showed you and it drops fairly evenly you look at that pool in the upstream end and the level on one end is pretty much the same as the level on the other end um turn that around though and here i've got um lake powell falling into this new dam we've built in the middle of the grand canyon and you can see this shock wave going back and forth as the uh the breach of uh of the glen canyon dam sent shockwave up against our new dam and you can actually watch it oscillate back and forth and this is one that i've modeled again as a big linear function it's basically like a flume in a lab but you can see initially you'll bounce back and forth and there are dynamic functions that can only be modeled when you take some of the inertial terms and momentum into effect and watch that energy go back and forth you wouldn't see those in some of the simplifications and you obviously wouldn't see those in a 1d model that's just based on stage and storage but then as it drops and as this breach forms you'll notice then all of a sudden you get to a case where it flattens out in that case it doesn't matter you could save yourself all the computational time in the world you could model it with one meter grids or 10 centimeter grids or 10 you know even one kilometer grids in some cases and get the same answer because it's going to drop according to these equations that you could just put into a spreadsheet basically so that's the bit that i wanted to talk about with the upstream reservoir modeling level pool versus dynamic we'll get into much more of that in our upcoming course um and chris actually has a paper i'll try and uh while we're talking here i will uh pop that into uh um into one of the chat windows so that you can download chris's paper which is available online that talks a little bit about that so chris feel free to mention and go back into any of these things that you like but um i will turn this then back over to the center part here which is um chris uh talking about the dam breach modeling parameters so thanks for that yeah thank you craig um yeah let me start uh my presentation here and uh first i want to uh thank you craig for that great introduction on uh level pool versus dynamic routing a very important piece of this uh overall dam breach uh modeling venture that we're talking about today um but what i'm gonna focus on now is um the dam itself how do we determine the amount of flow that's going to come out of the dam what uh what what do we use to determine the size of the breech and how fast does it form and um the overall shape of it um so i'm going to get into that and and i hope if nothing else that you get out of this talk here my my part of the talk is that these parameters are very uncertain we really don't know what we're gonna get and these are just means for us to estimate um so conservativeness is a big deal here uh we like looking at worst case scenarios but i also want to introduce at the end the concept of probabilistic dam breach modeling as well so what's the reason what is the purpose for determining breach parameters well what we're trying to do is develop a breach outflow hydrograph that we're then going to ultimately route downstream and that's going to help us determine flood inundation how fast the flood wave gets to diff different parts of the downstream reach and ultimately what are the consequences what are the damages now there are several options we can use to do this uh from simple to more detailed the most simplest approach is to use a parametric equation for peak discharge and there's several of these that are available out there just takes a quick google search to find some of them but you can use that to determine the peak discharge coming out of the dam and use a formation time equation to determine the duration of the hydrograph now that's going to work uh primarily in more level pool type drawdowns smaller reservoirs more compact reservoirs longer reservoirs the formation time that you get from the the breach parameter equations is not going to work so well but regardless you come up with a peak discharge a duration and that provides you with a nice triangular simplistic breech hydrograph that you can route downstream there are also physically based breach models out there actual software that you can run and it will determine the breach hydrograph i'll show you a few of those in a little bit and then finally this is probably the most commonly used is parametric equations to determine the size shape and formation time of the breach itself and then you can input that breach into your model into your routing model and you can watch it grow over time and and the model itself will compute the discharge hydrograph coming out of the breach most importantly you need to know what type of dam you're breaching uh there's several different types of dams out there and they will breach differently for sure when you're talking about a concrete gravity dam you've got a dam that's held in place purely by its own weight it's just a massive concrete that's sitting there um and that's all that's holding back the waters the mass of the of the dam itself on the other hand uh there are concrete archdams archdams are typically very thin and uh they're held in place by the arching effects and translating the force of the reservoir into the abutments embankment dams are made out of erodible material could be earth it could be rock it could be a combination of the two they may have an internal clay core several different ways you can construct these but these dams are going to erode when they're over topped or you might have a combination dam may have elements of two or more of these uh different types of dams in there here's an example at the bottom of the dallas dam in oregon and it's got an embankment dam um on the lower portion of the photograph followed by a gravity dam powerhouse and then uh going transverse across the river is a gravity dam spillway and so they're a combination of dams so you have to sometimes look at multiple different failure locations and see which one is going to be the controlling failure now with concrete gravity dams as i mentioned before they're held in place sheerly by their own weight and they're typically composed of multiple monoliths okay so it's not a single mass of concrete it's actually a bunch of blocks held together now typically what you're going to have are either foundational failures or failure from extreme over topping and i say extreme over topping because a properly designed gravity dam can take quite a bit of over topping several meters or more and some historical dam failures have shown dams to withstand uh way more than that in fact but if it's under underdesigned like you see in the photograph on the bottom left that's austin dam in pennsylvania uh it doesn't even need to over top before it may shift or slide or even topple a little bit for archdams again i mentioned before that uh the characteristic the defining characteristic of an archdam is that it translates the force of the reservoir into the abutments so as a result it can be very thin and arch dams typically are very thin structures all the forces translated to the abutment and what that means when we're talking about a dam failure is that if you compromise the strength of that arch if one of those monoliths is displaced then the entire uh strength properties of that dam go away and it can lead to a complete and total failure of the dam as you can see malpa said damn in france uh almost the entire section is gone just a little bit of the concrete around the foundation was left behind embankment dams are going to be primarily erosional um typically these are earth or rock filled dams and the failure mechanism for an embankment dam is either going to be over topping erosion from the top down or piping from some weakness inside of the dam where you get seepage through the dam itself and it opens up a hole in the dam which would then eventually collapse the dam and form the breech over on the top right you can see a schematic that's provided in the htc's using heck grabs for dam break studies it shows you an example of how embankment dams typically erode and what's interesting is when it starts to over top you'll first see the erosion at the toe of the dam and it typically works its way up towards the top or towards the crest in um more of a kind of a mass wasting um type failure as opposed to typical sediment transport kind of layer erosion there are several guidelines out there these are just a few of the ones that i use every country every state every province probably has its own or at least adopts others that are out there these again are some of the ones that i use but these are just broad ranges of different types of breach parameters for example average breach with the side slopes failure time for different types of dams and typically what i'll do is i'll use these to guide how i set up my failures of concrete dams but what i'll use for earth embankment dams i'll show you on the next slide are breach parametric equations then i come to the guidelines just to make sure that i'm in the right range and so here's some parametric equations just a few and there's several more out there uh available for you to use but these allow you to determine or to estimate the size shape and formation time of the breech opening for an embankment dam based on simple characteristics of the dam in the reservoir i.e the height of dam or the volume of the reservoir and maybe a few coefficients describing the types of dams here and there again several of these equations exist it's very important that you know how they were developed and if they fit the dam that you're actually breaching okay because they don't not all these equations fit every single dam out there and every dam is unique so it's important that you review these equations before you use them tony wall's paper is an excellent resource for this because he summarizes uh many of the commonly used breach parametric equations that are out there and it'd be a great way to determine if this is going to be an equation that's going to work for your dam also the hydrologic engineering centers um td 39 technical document is another good resource that summarizes breach parameters out there now it's also important to know your site conditions just because the breach parameter equation tells you a you have a certain width or you'll have a certain width opening doesn't mean you're gonna get that you need to know uh what's what do you have at your dam uh what does the foundation look like is there non-erodible uh rock below the dam itself is that going to inhibit the degree to which you have erosion or breech opening so here's an example here the yellow trapezoid shows you a a breach opening that was calculated with reach parameters but if you look closely you can see there's an unerotable or erosion resistant rock line that's going to limit how big the breach can be so it's important to check that before you adopt breach parameters now a physically based computer models out there um uh nws breach national weather service breach has been around for a while um and uh it's no longer maintained by the weather service but there's a third party that has kept it going and even developed a gui around it wind dam and hr breach are two other popular breech models that will determine your breach hydrograph for you finally i want to touch on probabilistic dam breach uh this acknowledges the extreme uncertainty in dambreach parameters and determining or coming up with a breach size shape and formation time we really don't know what it is but we can have a pretty good idea of ranges and so i'd like to just uh introduce this just briefly as a concept i think this is the direction we're going but instead of coming up with a deterministic dam breach single breach outflow hydrograph you can actually come up with hydrographs that represent different exceedance probabilities so the one percent chance exceeds probability five percent chance etc and um this is going to require automation of your software because it requires multiple thousands of runs but it can be done and is being done already mcbreach is software out there that does this with hec-ras and hr breach which i mentioned previously also does this uh in their breach model so with that i want to turn it over to bill and bill is going to talk to us about routing this breach hydrograph in the downstream direction can i can i just make a break bill just before we begin thanks so much chris and just wanted to see if we could just touch on a couple of these questions that are coming through is that all right gentlemen uh craig as well um uh i have uh before i start those questions though i just put up in the chat line uh some of chris's work his book breaking the hecras code his um uh raz solutions uh blog page you'll see all that on the chat line grab that and that'll be useful but there's a couple of questions one uh at the very bottom there you might see uh where has it gone non-newtonian uh done that thank you yeah i tried i tried to answer that one we may get into that a little later because uh specifically talking about any software package here yeah that's all right that's all right items we'll we'll we'll can cover that in the q a it's a definitely a good question right there's a couple here uh yes uh tahira yes man from canada saying i'd like to learn about bridge over topping modelling from this discussion is that is that probably is that going to be happening in this discussion yeah i mean as far as bridges yes uh definitely what what happens with the bridges downstream um you know normal floods might just go right through a bridge and back up against it but a dam a dam breech wave can just take the whole thing out and and there may be some uh opportunity to uh take your bridge and give it a time function where the bridge is there for a certain amount of time or to a certain depth and then dynamically you remove the bridge chris or bill you may have some some comments about that as well but sometimes you've got an obstruction that's only there temporarily and once you get to a certain level the bridge is gone and you need to be able to change your geometry um to simulate the removal or the failure of that bridge because it will behave very similar to a dam or a weird yeah it's it's a it's a good question um bridges downstream uh certainly nobody knows for sure if it's gonna fail if it's gonna wash away and if it does win during the event will it will it happen but it does have an impact on the attenuation of the flood wave as well as the amount of flooding upstream of the bridge and so a lot of times we'll model both if we're looking at maximum attenuation we'll leave the bridge in place and that's going to cause more flooding upstream but then we're looking for damages downstream of the bridge we may remove the bridge and look for maximum damages that way great look um should we press on the time is burning away there are a lot more questions we'll come to them why don't we move straight on now to you bill and um and 15 minutes and then we'll crack into that q a uh straight after that so thanks for hanging in everybody see you all there that's great thank you very much um over to you bill thanks uh chris and craig and trevor there's an excellent um excellent introduction so i'll just bring up the um presentation hopefully everyone can see that coming on there we go yep all good that's great yeah yep well thanks everyone for darling and this is a pretty exciting topic um today i'm going to be talking about um the the solver or the processes that we're trying to solve and how we might go about solving that and also touch on some important issues such as mesh convergence uh and then briefly have a little case study at the end to finish off with so that equation you can see on the screen there is effectively the nasty one the momentum equation and all those different processes uh what are often required by coastal models and a lot of these 2d schemes originated out of the coastal world and they're just trying to solve velocity over time and all these other things are influencing how the water moves around um for dam breach routing we can pretty well rule out coriolis force which is earth rotation atmospheric pressure external forces things like wind won't have much effect on a dam breach and so we're left with those other ones and you definitely need inertia otherwise you won't get that sachin as crave showing earlier you need gravity otherwise water won't go downhill which is a fairly fundamental part of down bridge routing uh bed resistance of course but that's people mostly think about that as men exam but it can also include other energy losses as well and our turbulence or any viscosity is often needed because we're talking about very complex flows more benign flows more gentle flows you can possibly get away without that last term so how do we know that we need all those terms well it's just simply through benchmarking to theory and to flume tests and if you have a calibration data set often helps so this is a good little exercise this is a phlegm model out of belgium simulating a reservoir where they remove a gate and they're looking at the water flowing against a building and so what happens is that a hydraulic jump forms in front of the building you get a wake behind the building you have supercritical flow in that red region and it's yeah it's a pretty nice little study they they had um a number of gauges um around the building and we're just showing here an animation from one of the two flow simulations of this flume test it's about three and a half meters wide you see that hydraulic jump slowly propagates upstream which is an interesting um a bit of phenomena so i'd like to show this one um so looking at two identical simulations except for one thing that's different and i tend to ask the audience who which one is more which one is least wrong which one is more correct and i won't impose that on you today but by and large most people actually say the bottom one is um is the more correct one but if you thought that you'd be wrong so the difference between the two simply the turbulence term is switched on in the top one and switched off in the bottom one so the bottom one may look pretty um pretty amazing but it's actually not close to what was measured in any shape or form and how do we know that we just simply compare with the measurements so here we have one of the gauges like gauge one what you're seeing on the screen there is water level on the top left velocity of the water on the bottom left that the black line is what was actually measured at that gauge and the red one is the one with the eddy viscosity turned off so you can see it's it's it's quite jumpy as you would expect there's no viscous forces in play um and yeah very sensitive but not at all close to what was measured um the blue line is the one with eddie viscosity term on and you can see it gives a much better production we can look at location two this is an interesting one because we have the hydraulic jump passing through this gauge that's where i get that big increase in water level around about eight to nine seconds and you can see the one with eddie viscosity turned off didn't come close to reproducing that that situation and location three is another example there so it's through this benchmarking and so forth that we know whether these schemes work well or what sort of parameters we need to use etc etc and what this exercise does does tell us is that you do need that turbulence term in this sort of complex hydraulics the other strong message here that i really like to stress to everyone is that you really have to be aware of the cool animation because it can be completely wrong um yeah so don't always leave a pretty picture [Music] i'll just talk briefly about the mathematical solution um what we're effectively trying to do is solve velocity over time uh that was our last calculation shown there in green and we don't know the exact answer if we knew the exact answer we wouldn't have such um complicated mathematics going on and we wouldn't have so many variations between solution schemes so we don't know the next answer in one time step when we go from one time set from the green to the orange a first order scheme is the most simple approach and this the analogy is effectively is making a linear solution or linear attempt to get to that point in time and you know it will never hit the mark exactly unless things are very very like in steady state conditions or something like that um second order schemes tend to the analogy is like solving a polynomial fit to what's going on and you'll tend to be much closer to the mark each time the consequences of this is that that error that's created can cause numerical diffusion which you know if or whatever lay person's uh term or that it would be a rounding or blurring of results so first order solutions tend to exhibit that numerical diffusion um because on the bottom left there a snapshot of some of the that last animation i showed without the turbulence term um and as you would have noticed it's got quite sharp features um yeah there's no viscosity forces going on exactly the same simulation using our first order scheme you can see that there's been a blurring of results and the viscosity of that that numerical diffusion is sort of distorting the uh the answers and what you'll tend to see in the time series is the blue line is what from the second order scheme and the pink line is from a first order scheme and you see once again that they're more blurring a rounding of results i'll just touch now on mesh size convergence what i'm talking about here is that especially what resolution should our 2d mesh be to you know get a good answer and we have an image on the screen there now some of you who are very astute may well have uh worked out what it is already and as we make that resolution of that image finer and finer we start to head to where we think we're going and to get the right answer but i should just say sometimes things don't always work out how you would like them to be let's look at an example of this so here we have test five from uk 2d benchmarking exercise it's actually a dam breach you see the hydrograph there on the left so 30-hour simulation the model there is on the right the inflow the hydrograph is applied at the bottom left um there's a number of reporting locations we'll just focus on the more downstream ones at four and five and uh should you say that there's no downstream boundary for this test they just said let the water or pond at the bottom end so this model is run through a number of different resolutions and let's see how the results look so this is your 10 meter case so it's pretty fine you see the mesh here on the left or it's probably a bit hard to see but it's a very very fine mesh for this size model you can see the time hydrographs there are at location four on the right top right top and right bottom is location five let's jump to 20 meter resolution it's a little bit coarser and you can see the 10 meter and 20 meter almost give identical results which is which is actually great that's what we're actually looking for and what that means is that you could easily run this and study at a 20 meter resolution with no fear of having any distorted results let's move to 100 meters so you'll see 50 and 100 on the graph on the right but you on the left you can see that that 100 meter resolution is starting to get a bit coarse but you know it still gives a reasonably good match compared to the 10 20 and the 50. so you could possibly even do your study at 100 meter resolution um move to 150 and we're starting to see some distortion at location four and as you can see on the top left the um you know the grid is starting to get very very coarse and so probably again going too far and we go through 200 and 250 meters and at 250 you know it's clearly the mesh is way too coarse and you can start to see that in your results so that red line which is a 250 is starting to deviate strongly from the others at location 5 in particular so just to summarize it's a good exercise to do for not maybe every study but at times if you're trying to decide on what cell size is appropriate we tend to like to have a minimum four to five cells elements across your main flow path as a minimum anything less than that you can start to have um maybe a loss of accuracy and from this little exercise you could probably conclude that i could do my study at 10 20 50 or 100. and obviously just the larger the cell size the the faster the run time so that obviously kicks into the decision making as well yeah i'll just now touch on a case study uh this is a fred haye dam in queensland australia and uh for this study it was quite interesting exercise they applied ten different methods as to what sort of flow uh would come through the breach of that dam and as chris mentioned earlier this this is where probably nearly all your uncertainty is i mean which one do you run with the ones on the left there tend to be um just estimating a peak flow or some you can generate a hydrograph from there's a couple of high ones there um yeah so you know if you if you're wondering where the uncertainty is in this this type of work is probably mostly in how you derive your hydrograph and i think the problem ballistic monte carlo style approach is going to be increasingly the way to go um so i'd probably feel like that that kid there if i was trying to work out which one to run with so but what we'll do we'll just focus on the last couple there because as they're the two though actually finally used for this study and just compare the results of those so the first one was effectively a method used by the queensland government to derive a hydrograph using a level pool type spreadsheet analysis like um chris mentioned earlier and so that was run and that that hydrograph was applied downstream of the wall um the other option which was sort of increasingly people are wanting to model the whole thing in 2d or 1d and actually breach the actual damn wall as part of that simulation so this this was set up in two flow software to re-form or reshape that damn wall as it had eroded uh and we've effectively followed the same breech arrangement as for approach i uh there's also a spillway which flowed until the water fell below the top of the spillway um and that was just simulated using a 1d element inserted in the model so this is a reach hydrograph that came just downstream of the dam so the spreadsheet approaches the red curve and the black line is what came out of the the 2d simulation so actually remarkably um similar in some ways which is which is good to know given all the uncertainties uh there's a few assumptions of course as craig mentioned earlier the um the spreadsheet approach just assumes it's still pool there's no gradient or um there's no um you know satin oscillating of the waves that will happen in the dam whereas the 2d simulation has this reverse wave propagating back upstream and then oscillating throughout the dam also approach the second approach will take you out taking check out knee tail water effects which might not kick in until the lower part of the hydrograph but that will also be accounted for just looking at some comparisons downstream so you can see the dam there on your left and we have a cup those two yellow dots we'll just have a look at the hydrographs that came out of the modeling so the top black line is inside the dam so you can see that's from the 2d simulation um the first set of dashed lines is um at the top on the left there and the second on the right so you see both approaches aren't giving an overly different um peak water level and importantly for dam bridge modeling the arrival time is it is different but it's not demonstrably different so when you have a lot of uncertainty in your modeling it is a really good approach to model things with different approaches um and just get a comfort zone that what you're doing is is going to fit with other ways of solving the problem um and the really nice thing about the uh the full 2d solution is that you can get some pretty cool animations which of course are absolutely right as i mentioned earlier no but you while that may look pretty impressive on the screen you still need to um you need to make sure that it's doing the right thing so there's there's the ten methods um the other method the queensland government uses this cube reach one which is actually the highest one i think they've got to be scared by that number and they went and did a bit more in-depth analysis so as you can see large uncertainty yeah where do we go you can probably rule those two out the rest are actually probably surprisingly consistent which is nice to see so just to conclude um dambridge flood wave propagation is actually a very complex free source hydraulics problem you must take the view that no numerical model is right and that's absolutely the case no numerical model is right but you must always try and minimize the wrongs and you know the big ticket items are good topographic data you think garbage into one of these models you'll get garbage out if you you know perfectly if you are going to model that um propagation downstream that you know the 2d full shallow water equation solutions are the way to go you probably need the full suite of terms in there from a dam bridge point of view but you won't need all of the terms from a coastal point of view second order spatial solution strongly preferred to pick up complex hydraulic situations to avoid numerical diffusion we also tend to find first order schemes overestimate the energy grade line yeah they benchmark these solvers you know there's theoretical um comparisons uh there's fluid models as i showed earlier and probably hard to find in the dam breach area but if you've got any good calibration yeah make sure you use it um check for mesh convergence issues it's very important that you've got a sufficient uh resolution in your 2d model if you're going to model these things compare different methods don't just run with one method in this sort of business because as chris mentioned earlier there's a lot of uncertainty in that breach hydrograph and watch out for the outliers and finally the um be very critical of those cool animations we're all showing you thank you brilliant animations bill that was an absolutely fantastic um presentation from yourself and uh from chris and from cray the animations for a non-professional model like myself not a model at all actually they say so much um before you go before we go into a whole lot of other questions there's one short one here from uh daniel adria is this a subgrid model i think she's talking about the last uh he's talking about the last um sub grid i'm not entirely sure this it's a [Music] those models you saw there effectively a fixed grid so it's the same cell size over the whole model and um this one elevation cell we have uh next year we're actually putting out the ability to sample at a finer resolution so you basically have um fine features and also some of you may have seen out there on linkedin the quad tree approach where you can actually put finer cells down your main river so that's actually often working and we'll be releasing that next year so pretty exciting space for studies of this nature from our point of view right there's a there's a few things happening here so let's just get right into this craig chris if you come join us as well that'd be great uh there's uh a person who could come on screen right now uh that's um uh praveen got another question let's let's let's let's move straight on uh uh the top question here top buddy question here is i'm interested to find out more about whether modeling pipe failure in two flow is possible and what is the best way yeah sounds like it uh yeah so actually pipe flow um uh we're actually building that in and that should be in for next year's release yeah that sounds good all right that's an easy one uh cameron hall thank you for your questions and uh look i should say before i even go any anymore with these questions is that some of the questions thank you chris and craig you've been answering uh written answers are offline but if you if you have a follow-up question i want to re-raise that question again by all means let's do that cameron um hi do you have any comment on relative size of breach for small reservoirs i feel that the empirical approach is used in freolitch 1995 and more recent publications can have issues modelling peaks discharges when small volumes are involved and i think chris has been giving that one a shot so chris you want to just uh sum that up let's see what's going on there chris it's cameron halls yeah yeah that's cameron that's a really good question um and all right yeah so cameron good question um and you're right the breach parameter equations are based on historic dam failures uh most of them are on the larger size the smaller ones we don't really hear about too much and so nobody does measurements or forensics afterwards so we don't really know a whole lot about it and that really touches on the overall uncertainty in what we're dealing with um what i like to tell people is if you're not comfortable with the breach primary equations if they're not working well for you i would certainly encourage trying out those uh theoretical models i mentioned three of them but there are several more out there as well do a little research and several of them are free as well free to download so you might want to give that a try no that's all right if you could type your question out and send it through in a q a proven that would be really good thank you all right we'll head straight on to martin fernandez hi uh from spain we use the parameters of the laws i realize the values are not always the same but there's a spanish software that has already been already studied these kinds of errors any comments on that grey i'm not sure about the parameter of laws any uh bill or chris you guys uh uh no no i haven't heard of that um i'm not quite sure all right we'll come back to that offline that's okay adrian uh emilio uh hi from columbia uh i would like to ask one does the type of fluid related somehow to the type of dam failure uh thinking about the the the massive landslide at the vergant dam in italy yeah and i think this one actually you know with the time left i think we can categorize some of these questions and there are quite a few questions coming through about the uh the fluid and the properties of the fluid um and so maybe if bill and chris you may want to each comment on this one as well we're talking generally about water and most software packages um everything um is newtonian and uh if you want to look into non-newtonian fluids and what happens when a tailings dam fails what happens you saw some of those movie animations when the material that's coming downstream uh includes trees and rocks and bridges and cars and other material you know how does it actually behave is it really water and we might use some parameters to fake in how it behaves maybe with the roughness maybe with some other parameters uh turbulence coefficient or something like that um but most software packages that model dam breach assume that it's water and and that that's the material coming down there are some others um that will involve some sediment transport some mud slurries um and how that might behave coming downstream um but that's a whole different ball game so we typically uh try to simulate things using parameters that we can adjust to reflect the type of fluid that's moving downstream uh bill chris you guys want to comment on your own on that one it covers about four or five questions that have already been raised and answered yeah i think you covered a lot of that cray what i was going to say i i guess i would add that modeling it as water as opposed to a debris flow is typically going to be more conservative so if you're looking for a worst case scenario then you might go ahead and assume it's water flow versus the the high more viscous debris flow type events but if you're curious to see the difference in how that water moves i mean it's very extreme it's very visible you can google debris flows and see any number of videos online of mountain debris flows and they move a lot differently than water that's for sure so yeah thank you chris and craig probably answered that pretty well actually and um there are situations where you did move into non-newtonian flow and yes the the package i was increasingly allowing that to do that's something we've built in and releasing next year as well um but i think there's some very interesting complications for trying to represent that and it just adds more uncertainty to the whole process as chris was saying just run lots of different tests get a feel for what parameters affect your results and try and research that to try and hone in on a more accurate answer but you just got to be very cautious of of your modeling okay that top question there now from nathan young in the us i see westchester philadelphia is can you please comment on the reason as to why the diffusion wave equation set is often not appropriate not appropriate for danbridge modeling and why the full momentum equation set is the better choice and i think this one uh that's a specific question to hecariz and there there's an equivalent i think um twofold's got its own equivalence here essentially we're looking at dynamic flows i mean it's changing very rapidly and uh being able to zoom in on the dam failure itself and maybe adjust your time steps so that when the dam fails you've got some really fine times up and then they get coarser as you go on that's something that's now available as well to help mitigate for some of the the the instabilities that would otherwise occur it's just because of the inertial terms that need to be accounted for with such a dynamic uh dynamic situation uh but um you know where your uncertainties are and you see this other question maybe i'll try and hit this at the same time somebody asked the question you know are your uncertainties worse right at the dam um you know and your uncertainties can be defined determined by running some sensitivities to equation sets and like bill showed a 1d just in a spreadsheet-based equation can sometimes do the trick it and it might be close enough might be within the same zone of zone of confidence and that's not the the question uh that was asked here online was um on on the chat line was whether your uncertainties are worse near the dam or farther away from the dam and and that's a that's a very good question uh what ends up happening is you might have certain uncertainties at the dam and and they look you know like in the breach parameters uh that the bill was showing they look like they're they could be really far off but downstream you could be even further off the propagation of your flood wave and the amount of attenuation that that peak hydrograph gets as it moves downstream you know you could have orders of magnitude difference in the amount of flow that comes down or in your arrival times based on you know whether that even ever gets to you that peak flow may attenuate long before it gets to you in your village or your town or your building may actually be safe from this flow depending on how it's routed to the stream so that's a very tough question to answer i'd say there's uncertainty everywhere test it everywhere test your breach parameters test your downstream wave propagation test the reservoir you know level pool versus dynamic test it all and uh and and get your get your uncertainties um chris bill i want to comment on those yeah so certainly craig yeah you can't use the diffusion wave in this situation just because of the um the uh the dominance of the inertia and turbulence terms really kick in uh when you have uh very what you might call deep flowing low resistance systems um yeah you really have to have those terms well well represented i agree definitely with bill and cray on that um i mean the defining characteristic of diffusion wave with respect to the full momentum is its lack of acceleration terms and if you don't think acceleration is an important characteristic of a dam breach flood wave uh you might want to think again so it's definitely important i always recommend people use full momentum when doing dam break studies certainly in the primary flood path at least um and then as far as the uh crate kind of touched on the uncertainty at the dam versus further away uh one thing i do want to mention is there have been some studies that have looked at the coalescing of floodways so they move further away from a dam and if you look at maybe two different sets of breach parameters that produce two different peak flows the further you move downstream the closer those peak flows will come to each other to a point you'll eventually get to a point where you'll see no difference between the peak flows but that could be several kilometers downstream before you get to that there's one question here from bonnie bear which has got the highest vote at this point is is there a rule of thumb for how far down street stream you should build your hydraulic model there's a biggie yeah the the rule of thumb is run your sensitivity and if it's still sensitive to uh if if your area of interest is still sensitive to the parameters there extend it so i mean that that's my rule of thumb you guys may have others as well for me it's all it's it's all experience i mean i've never found a a an equation a rule of thumb or a you know magic bullet that says this is how far downstream you go and i think the more experience you have in doing these the better your guess is going to be it certainly is a little bit painful to have to come back and extend it a little bit further down because you didn't do it far enough the first time but with practice you'll you'll get better and better at estimating that just based on the characteristics of your stream and the size of your dam and reservoir and so forth yeah yeah i'll just add in there i agree entirely with all that the um the it a lot depends on where your area of interest is if it's if you're looking at urban populations you clearly need to be well downstream of them with downstream boundaries the main thing is that the boundary is not affecting your results in your area of interest so if you're really focused just downstream of the dam breach you might not have to go very far but if you're trying to represent the fud wave down through a series of towns and populations um you need to make sure you you probably end up with a stage discharge boundary and you need to make sure that's well removed from those areas of interest so that's been great look we are almost reaching the hour mark here and we are nowhere near the list of questions you can see them there gentlemen as you fire down this q a you can pick one if you wanted to to do a last one but craig could if you picked the question out what would you pick out first up oh i think um i do see that it's getting some votes on the um the terrain resolution again um watch your grid size watch your terrain run your sensitivities um that that may be important and and i guess wrapping it all up um you know this is something that we look at hypothetically but um you know i saw an ankle presentation about the um the ten worst tailings damn failures and in each case a engineer went to prison over these things and there has been loss of life so i mean this is something that we do um so that we can prevent it we watch these things fail and we simulate them so we can avoid the actual occurrence of these things and it is very scary when you look out there at some of the dams and the condition of the dams out there you know if this happens in a big way it can be a massive uh you know massive loss of life and that's what we're trying to prevent the better we can simulate these things the better we can assess the risks and warn people downstream you know this is something that the world needs and we need to be able to do it better and take advantage of the hardware and the software out there i love seeing those animations bill it's great to be online with you and with chris with uh experts in the field i'm happy to be involved in this and um yeah let everybody do their closing comments because it looks like we're out of time so thanks any comments and then we'll leave it yeah no we'll go with me i'll just uh emphasize that importance of the terrain data and i think uh if you're after an accurate simulation that question refers to a 90 meter dem resolution you'd need to probably be finer than that unless it's a massive system you'll need much better data than that to be honest yeah yeah i just want to say thanks to everybody for joining in and if you remember nothing else remember the uncertainty extreme uncertainty with doing this kind and take appropriate uh action to uh to deal with that thanks very much uh everyone for being here today thank you for to bill to chris and to craig uh i do want to remind everyone that of the um uh of the book from chris goodell you see it on the chat line there chris goodell's raz solutions blog page uh and that we have the danbridge modeling course coming up in march but also a flood flood modeling the cloud webinar coming up in february two three gentlemen so much appreciate your time and effort here today uh and for everyone who's joined us all across the world uh it's been an absolutely fantastic time i hope we can see you again but for now we'll say goodbye bye bye chris hi bill [Music] subscribe to this youtube channel by clicking the subscription button for future webinars and online short courses please visit our website at australianwaterschool.com

Channel: Australian Water School

Views: 10,641

Rating: 4.8285713 out of 5

Keywords: hydrology, water modelling, dam breach, water planning, flood management, Trevor Pillar, ICE WaRM, Water Modelling & GIS

Id: iRCYmQsWTXA

Length: 53min 23sec (3203 seconds)

Published: Tue Dec 11 2018