Reservoir Sediment Modeling with HEC-RAS: (Part 3) Reservoir Modeling Demo

I'm teaching a reservoir segment modeling class next week with the course Reservoir sediment specialist John Shelley and Paul Boyd and we developed a workshop that's a simplified Reservoir model with local data and so I was about to sit down and do a dry run of the workshop just to make sure everything works and I thought hey why don't I turn on the camera and capture this both for the students in the class but potentially for anyone else because even though it's a simplified model and not really ready for production work it does illustrate a number of the best practices and like assumptions you need to make and maybe not so best practices that you would encounter in a reservoir segment model so what I'm going to do is just as I kind of do my dry run of this Workshop I'm going to turn on the camera this will be a longer video because I'm going to go all the way through a model starting with the geometry and ending with a run and so we will hit all of the major steps of building a simple Reservoir sedimentation model and so here is my domain I've got a relatively truncated geometry that's built off of a contour shape file so it's not the highest quality but it's good enough for what we're doing and then we also have about 25 years of flow data and Reservoir stage data and so you can go look at some of the other videos where we talk about different ways of modeling a reservoir but we're just going to do the simplest approach here we've put our Downstream cross section on the Upstream side of the dam which means that if we use that as a downstream stage boundary condition then we will operate the dam essentially historically and basically follow these steps I'm going to add the Quasi and steady flow data I'm going to add the sediment initial conditions I'm going to add the sediment boundary conditions and then in the piece that I think is probably going to be most valuable I'm going to walk through the options menu in the sediment data and choose the options I think are best for a cohesive Reservoir sediment model and then we'll run the model and look at some results and I'll put time markers in the comments below so the first thing we're going to do is add the Quasi and steady flow data and of course you can do this as an unsteady model but causing study is a lot more stable and we're building like initial simplified models so we're going to do as quasi and study so you'll either go to edit quasi and steady flow or you'll press the Quasi and steady flow button and you'll get this editor you'll notice that the editor automatically populates the upstream and the downstream boundary condition and when you click on these only the boundary conditions that are appropriate for that show up so let's start with the Upstream boundary condition we have 25 years of data in CMS and you'll notice that that's about 9670 flows so we will click the flow series because that's the only one that's available and the first thing we want to do is fix the start time these data have an actual start time we know it it's the first of January in 1997 so I'm going to go switch to thick start time and put in 01 January 1997. if I do this that means that you know I don't actually have to start at time step one I could actually run like two years in the middle of this if I wanted to if I wanted to run like a flushing model on a high flow or something like that and so I'm going to fix this start time and then what you'll notice is we actually only have 100 rows by default in here and so you have to go and change the number of ordinance and so we're going to call this 96.70 so that we have enough rows for all of our flows and then I'll just come in here and get my my flows Excel trick Ctrl shift down and it'll get everything in that row Ctrl C copies and then you come up here just click on the heading control V and that copies our Flows In CMS all right so then we have these other two Fields you know Raz has these two time steps the computational increment is actually the time step that's what the model that's what the model is going to use to compute the flow duration is just how long are each of these flows that you're putting in you know if they were just daily flows we could do it like the and stay flow editor but they don't have to be regular you could put in irregular intervals in this case the regular and you'll see that these are in fact daily flows which is appropriate for a system of this size and so what I'm going to do is you know that means that they're coming in at a 24 hour time step so I'll just put a 24 up there at 24 down here and I'll hit interplay values and now we have daily data then a little bit more sensitive is this computation increment now the computation increment is generally the biggest problem with a lot of models I see this is a reservoir model so if you choose the correct options a lot of times it won't be as sensitive but in general people run computation increments they're too large the problem is you actually kind of want to run large computation increments when flows are small and there's not a lot of sediment load and you really want to focus your runtime on when flows are large and you've got a lot of morphological movements and so what you should do is kind of start with your best estimate and then make the time steps bigger and smaller it's called a convergence analysis until you see that there's no change I'm not going to do that I'm going to you know make a first estimate but before I delivered a model like this I would do a convergence analysis in order to make sure that by changing the time step I don't change the answers and so I want to have these computation increments or that the account computational time step change with the flow but I've got 9000 of these I don't want to go in and type them in so I'm going to use this option down here the compute computation increment based on flow it says hey this is going to wipe out all of your computation increments that's okay I don't have any and so I'll say yes and so now I can go in and put a range of flows and so you want to start with zero and then the other thing you want to look at is you want to make sure you complete the whole range and here the max flow in CMS is about 500 and so we're going to kind of make sure that the range includes 500. so I'll come here and I'll say from 0 to say like 50 CMS those are low flows I'll go ahead and use my 24 hours I probably wouldn't do that in a project I probably would have a Max of six or something like that I'd go in and compute a Quran condition actually but we'll just do that for uh for the workshop then let's say up to a hundred I'll drop it to 12 again that's higher than I'd probably start with but we'll we'll do that and then you know up to 200 now we're going to get down to a three hour time step up to 400 we'll go to a one hour time step and anything above 400 let's go to a thousand because you know this model could be used for you know Future Extreme events since we want to make sure that we capture those we'll make that 0.5 hours okay so now you'll see that basically all of these have computation increments associated with them and so if we press OK and then reopen it now we can plot it and you'll see that we have our hydrograph this is a histograph right it's a quasi and stay flow so it doesn't look like a curve like you might be used to an unsteady flow these are constant flows over the period of time okay so that's our Upstream boundary condition our Downstream boundary condition is going to be the stage at this location and we have the pool elevation again daily over that period of time so I'm going to click on here I'm going to go stage series I'm going to do the same thing I'm going to fix the start time 01 January 1997. this is a step a lot of people miss and I think a lot of models run into problems with that next I want to set my number of ordinances we are at 9 670 again then I'm going to come back to my data Ctrl shift down control C and then Ctrl V to paste those data over here and then again these are coming in as daily data and the downstream stages don't have computation time steps with them if you had like a tributary flow coming in that could have its own computational time step and then the model would find kind of the least common denominator but stages we just you know put those in at whatever they were defined at and then we sample those so we'll just put 24 at the top and 24 at the bottom interpolate those hit OK and we're good to go all right so we have causing stay flow data well we have the flow data but we actually aren't done you see I I put this little required down here because people kept forgetting the set temperature data and so you do need to set temperature data so we're going to open that up but what you're going to find is that I actually don't have any temperature data and so my opening offer is I don't know 15 degrees Centigrade we will improve this model as more data come to the table but for now I just need some sort of like reasonable constant temperature I think the next thing I'll do talking to the local experts is yeah but what's the range so we can kind of put the low in the month that it is in the high and the month it is and then create like a cycle something like that that's something I do pretty often but for now I have no information so I'm just going to start with um 15 degrees centigrade and a you know a good way to put in a constant is just to put in a big number I like to put in 240 000. because actually there's a limit to how big this number can be but that's divisible by 24 and uh and it's a pretty big number so I'll put that at the top and then put it again in the bottom and Triple Eight and so now basically we've got this huge time step of just 15 degrees Centigrade you could just leave that at simulation time but I always like to fix the time even if it's a synthetic data so I'm going to make that 01 January 1997. okay so now we have our quasi and steady flow data I should have saved this already but we'll save it now and we'll just call this 1997 to 2022. all right we have our flow data so the next thing we're going to do is enter our sediment data so to enter your sediment data you'll go edit sediment data or you just press the sediment button here and you get your sediment data editor okay so the first thing you'll notice there's a bunch of tabs up here we only really care about two of them we care about the initial conditions and boundary conditions we're not doing any bank failure analysis so we'll just leave that and so we'll start with our initial conditions well the first thing we need to do is Define our control volumes and the control volumes are your movable bed limits you know what part of your model can erode or deposit but specifically a road and then how deep can a road and that's kind of the reservoir of sediment that's available to be eroded now we're doing a reservoir deposition model and so there's a little trick here because you know deposition is actually a lot easier than erosion and one of the hardest parts is to get the Upstream reach of your Reservoir stable and not to erode below where it actually erodes and so if you're really just looking at a depositional model or if you're looking at a like a flushing model where you would only erode what has been deposited one thing you can do to simplify your modeling a lot is just say well with my initial bed you know at the closing of the reservoir because I'm modeling this reservoir from close to present we're just going to say my control volume is zero zero depth I have no sediment which means that whatever that initial bed is that's like a flume it's not going to erode below that you will be able to deposit sediment on top of it and then erode that sediment but it's not going to erode into the materials beneath and since I actually don't know what the materials are beneath it would take me a lot of time to choose a like artificial gradation that would make those stable and so this is actually really helpful simplification for models like this you could also go in and Define those elevations if you wanted um but you know this this is just as easy so now what I'm going to do is if I hover over this the bottom right corner of that I get a plus and that means that I can drag so I'm just going to drag that to every cross section so now every cross section has a zero maximum depth which means that we're going to be able to deposit and then erode the deposition but we aren't going to erode into whatever's beneath that and that'll simplify a lot of things as you'll see then the next thing we need to know is what are the left and right limits of this control volume or our movable bed limits now you might think well we don't have a control volume it's zero thickness so why does this matter well this matters for a lot of other things you know a lot of times we multiply the transport functions the width that we multiply those by is the you know the distance between these movable bed limits and we also only allow erosion or deposition Within These mobile bed limits unless you make a different decision which we will but for now the move bed limits end up being pretty important but because later on I'll show you that we're actually going to deposit outside of the movable bed limits and limit erosion inside the movable bed limits then setting the movable bed limits to the channel especially in a reservoir setting where you're only going to row deposited material isn't a bad estimate and so we have this button right here you don't always use this button but in this case use the banks for the mobile bed limits goes in and you know these were red dots because these were the red dots were the banks but we're putting the yellow dots over top of them to show hey these movable bed limits are actually on top of your Banks now if you wanted to say move those we have these buttons up here and so you can see if I move the left movable bed limit over to now we've got a different movable bed limit and Bank station but for now we're just going to use the bank stations as a little bit limits then we need bed gradations at each of these cross sections and that's a big lift because no one hardly ever has bed gradations at every cross section unless you're like on the Missouri and uh we actually have very few begradation measurements and the migration measurements we have are not particularly useful for this because they're measurements of the deposited sediment in 2022 and so it's not reflective of what was in the bed back when the reservoir closed back in 1997. so we're gonna need to make it up but here's the thing it's not going to be a sensitive parameter because we're not going to be able to erode into this material because we set this as a zero and even if we were to set this as in rotable bed then it would only matter in the Upstream portion in the downstream portion where you have the reservoir pool you're never going to erode anyway and so it would just deposit on top of it and so this doesn't really matter but you need it so I'm going to go in and I'm going to call this synthetic bed gradation just so I know just so the I and future users know I made this up and then basically I'm going to say 10 wash load so I want my 10 percent to be you know at the fine sand and we'll kind of distribute this pretty randomly and then I'll probably want everything to be smaller than coarse gravel and of course I could actually compress this more to save runtime but the runtimes are pretty short so I won't worry about it too much and then usually there's kind of a jump around here so I'll make this 30 and I'll make this 70. 20 something like that again doesn't really matter because it's not sensitive this parameter okay so then I'll go in and I'll just select that here and then if I hover over this and pull this down then it that populates now this is the only kind of model where that doesn't matter most other models this is one of the most sensitive parameters so I don't want you to get from this video hey that doesn't matter you can make it up but because we're doing a a a reservoir model which is mostly deposition but we're only allowing it to erode the sediment deposits it's only that case where this is not a sensitive parameter Okay so we've defined our initial conditions up here we're going to select a transport function now we're going to have almost entirely cohesive sediment coming into this so the transport function also is not going to be very sensitive again one of the only cases is where that's true but I'm going to choose Larson Copeland because Larson Copeland does extend down into the silt range so that if this transport function was to be invoked it would be at least the most appropriate one we're going to leave it for the defaults for sorting method and the fall velocity method Okay so those are initial conditions let's move on to your boundary conditions when I save it save sediment data all right so next I need to define the sediment flux upstream and so there actually are some sediment measurements at this substitute boundary Condition it's complicated because it's not the whole Watershed but because it's a flow load relationship we're going to apply it to the flows from the whole Watershed and I think that it's going to be pretty appropriate and so we're going to define a rating curve here but let's actually look at how I developed that rating curve because I use tools in Raz to do that and so what we're going to do is we're actually going to close out of this and what you'll see here is these are the data I was provided I have dates of flow load Pairs and you'll see that I actually have about 600 measurements of flow load between you know 2001 and 2020 that's not bad and so what I'm going to do is I'm going to open the tool that we've developed in Raz to do this and so if you go to the HD editor the hydraulic design editor so you can go run hydraulic design function or you can press the HD editor and that'll bring up the bridge scour analysis tool which is not what we need but if you go to type and you go to rating curve calculator it'll open up the tool that we use to develop the statistics to develop a flow load rating curve from these data so we have a lot of videos on this tool if you are interested in how to use this tool we have extensive videos on that but we're just going to kind of run through it quickly the uh generally this tool is set up to download data from the USGS data site but this Reservoir is not in the United States and so the data aren't available just from the USGS portable we have them right here and so what we're going to do is we're just going to import the data and to do that all you have to do is have your data in this format one column is date one column is time one column is flow and one column is load or concentration now we don't have time is that okay that's totally fine in fact if you don't have date or time you can just leave those two blank as well but you do need all four columns and so in this case having a date without a Time is fine we're just going to assume that it kind of happens on noon of each day and so what I'm going to do is I'm just going to go Ctrl C copy all these data and then I'll come over here into the rating curve calculator and go file import and then we'll go all the way down to the bottom from clipboard and it says Hey are these loads or concentrations now because this is an SI we have to kind of keep we kind of have to do this right because in 6.4.1 the SI conversions aren't great you have to bring it in an SI and kind of keep it there we'll have that fixed for uh 6.5 but we're going to say yeah it's loads import and we import the data and you can see that it Imports all these load relationships and then starts as a kind of an opening offer fits a power function through them with the bias correction functions you know all log transform power function fits are you know fundamentally and systematically biased and so you have to unbias them but with either a Duan or Ferguson correction factor I have videos on that if you want more information on that but that's what we've done okay so we could just run with this and you know in in one of the simulations we do but we want to look at some other things too like you can look at stationarity and these data are actually pretty stationary you know they don't change significantly with time you can look at hysteresis and look how the kind of rising limb and falling limb of the annual hydrograph changes but what I'm interested in is the question does one single line fit these data well and so we've added this lowest this local regression to kind of evaluate that and I'm going to change the span to 0.5 which just means that for each of the local regressions it's going to use about half the data to look at what if we just fit a regression to this part of the curve and so then if you go in and you calculate the low s you'll see that you know it fits a curve and this is not bias corrected and so it plots lower but it's the shape that we care about if it's a curve that's relatively pretty straight locally through here but then really tails off at the end which suggests that maybe we're going to over correct on the high end well we have a fix for that and what we could do is we could fit a piecewise linear model to that and so I'll go in here and say hey go ahead and fit a piecewise linear model and so what it'll do is it'll actually fit two curves to minimize the residual of the regression and what you see is actually it finds the statistical minimum the the inflection point with the minimum residual and a lower residual than this straight line is around 125 CMS which is about where we see the lowest turning over and so I actually like this piecewise linear fit you know obviously I wish I had more data on the upper end of this curve but our resolution War models is going to be more sensitive to this like you know fifth 50 to 500 CFS range and so we want to make sure that we're kind of getting that as close as possible so what we'll do is we'll come over here to the result you see it gives us the coefficients here but we actually want to just type this render as and so I'm going to go 0.01 as kind of my low bound and then I want to put in the actual inflection point because I want to get a load for that as well and then again I'm going to kind of look for a high end so I'll just say a thousand CFS and what we have get here is now we have the loads associated with this piecewise linear regression with the the bias correction that we can put right in the model so now I'm going to go back and open up my sediment data and I'm going to open up the boundary condition and the rating curve and now I want how many sets do I want well in general you want as few sets as possible because you're going to be adjusting this in the calibration process so I if I can get away with two I do but in this case I need to capture the inflection point so I'm going to need three sets and so now my flows are 0.01 I'm just going to kind of transpose these data in here 125 and 1000 and what loads do those go with well again about 0.1 I it's totally appropriate to round these because these numbers are you know kind of estimates anyway and so laughs those are our data that's our flow load relationship and if we uh if we open it up and we plot this you'll see we have a piecewise linear curve that looks like this okay so we have our loads so what's all the rest of this white space here well Razz like most sediment transport models is a grain class specific model and so it's not enough just to tell us how much sediment is coming into the model we need to know how much sediment is coming in each grain class now unfortunately in this model like many models that we work on we don't have any of this data we only have the loads we don't have any idea how those loads are broken down into different gradations but I've already shown you the data we have which is kind of the interpretive key to this if we go back to the bed gradations the bed gradations weren't useful for initializing the bed but these bed gradations these are taken after a couple of Decades of deposition so this material that has deposited in the reservoir is actually characteristic of the flux that's coming in the reservoir that you want to model and so if we look at this there isn't a lot of spatial variation that's predictable but and so if you kind of look at the overall averages we're looking at about 60 clay about 30 silt and about 10 sand and so that would be a pretty good opening offer and so we could start with that and we just said okay 60 percent clay and then we want 30 silt but we need to you know split that up over four silt grain classes because what you don't want to do is you don't want to just go 60 30 8 because it's a reservoir model so it's gonna try to you're going to get much more even deposition if you spread that out and so then we'll go ahead and put eight there that so that would be kind of the best use of the data but from this model a couple times I know how it behaves and I also know how sediment models behave and generally you want to reduce clay a couple of reasons for that one is that when the clay is measured basically it it's de-aggregated and so when you're actually measuring the gradation of the clay you're measuring the individual clay particles but when Clay is you're actually measuring the individual clay particles but when when Clay is transporting it's often transporting in flocks or clods or even with silt too so in general these gradations are going to be coarser the finer end of these gradations are going to run coarser in your depositional model than in your actual samples and so kind of as an opening offer I'm going to just drop that to 50. I might drop that to 40 or 30 eventually to get it to run and in general too these are the materials that aren't going to fully deposit and so depending on what your trap efficiency should be you might need to reduce this number more is that fake no it's not because a lot of clay is transported in silt-sized Grain classes and actually a lot of silts are transported and sand sign grain classes because they aggregate with the clay and so then I also know that in general a large sand content is going to you know over predict the amount of deposition Upstream because of the low energy up there and I tend to think that a lot of times these samples collect ambient material and so actually gonna just wipe out the sand for now and just run you know 50 clay and we'll just say 20 10 10. now we have the opportunity to make that change with flow as flow increases does this get coarser or finer well that's a big question that I've asked in Publications and if you're interested I'll put a link below but we actually just made this up so do we actually have any information about how this changes we don't we could we could coarsen it you know but in general you don't want to make your input more specific than your information now again I just kind of made that up how comfortable are we with that it doesn't totally matter because this is our least certain data this is the data we have kind of released information about which means that when we go to evaluate and calibrate the model these are going to be the first ones changed so we're going to put our best guess in there to run the model but then we will revisit these data so we have finished our sediment initial conditions in our sediment boundary conditions but we're not actually done with our sediment data and this is the real reason I wanted to run this video because a lot of you have kind of seen me work through these online before but the main reason things I wanted to show you were the options that you should use for a reservoir model because if you don't use some of these options you're going to get a terrible result so we're going to go to options and we're just going to start here with user-defined grain class we'll just kind of work our way through down through and so in these most people probably have never changed their user defined grain classes and that's fine you know the these are kind of set to Industry standards so it's fine if you don't touch these you could actually go in and change the name of some of these and sometimes like if someone has a lot of medium sand they want to break medium sand into sub grain classes but the one thing that I do go in here and change pretty often is the unit weight which somewhere along the line got labeled BD and Si which I will fix for the next version but in si this is density in U.S customary units it's unit weight so they're actually not entirely the same thing but they convert and so you'll notice that for the sand and gravel and Boulder grain classes you know our default unit weight is 14.89 which is the conversion of the US customary default but that's about right you know if you think what is the dry density of sand or gravel or something like that you're going to be in the like 1400 and 1600 range so that's okay but then look at the silt silt's about a thousand and clay in particular is 480. now that's not totally wrong this is actually based on observations in the 40s and 50s of what the actual density of clay in reservoirs was and clay and reservoirs when it deposits it can be what we call fluffy which is kind of a funny technical term which means that you know early clay deposits have a lot of void space and so they found that the actual dry density of a lot of deposits in reservoirs can be as low as 500 but that's a low bound and so as sediment consolidates then it act the density increases and so if you run it with these low densities your mass will translate into more volume change which means more depositional you know distance and so I having no other data I've found that you know a thousand is the low bound that I've interacted with that's not universally true I know a lot of folks that work on reservoirs that use 500 but a thousand is kind of the low bound that I've worked at and I would probably go with 1200 here and again these are things you would rather measure and or calibrate but I'm going to kind of sit bring those lower bounds up to kind of the median range where I've worked at before okay so then we go to options set cohesive options let's leave that for the end let's go to bed change options 1D and so Raz uses the veneer method which we call the peanut butter method which you know we if we're going to deposit our road we're going to evenly spread that deposition or erosion over all of the wet nodes within the movable bed limits and you can see that for our Global bed change options we basically have for deposition and erosion what are you going to do in the channel well by default we're going to use the veneer method for both deposition and erosion in the channel but you can change these and by default on the overbank outside of the movable bed limits we're going to do nothing we're just going to leave it fixed but I often and particularly with Reservoir models in the overbank will select veneer and so what we're going to do now is we're going to deposit in erode in the Channel with the veneer method but we're also going to allow deposition in the overbanks and this is important in a reservoir model because the overbanks are always wet if you think about what is this cross section going to look like when the water surface elevations are at like 2 260 to 270. well it's gonna be completely inundated for the next for the whole simulation you can see I actually went in and put in an effective flow areas here um that's pretty important but that means that if you've got cohesive sediment traveling Downstream this cohesive sediment is gonna basically go bank to bank and so you're gonna allow deposition outside of the channel but then in the parts of the model that are Upstream where the water surface elevation will actually drop so that it is channel flow then you're not going to erode in the overbanks but you will erode in the channel so you want to allow erosion in the channel and also if you're going to do some sort of flushing analysis probably what will happen is that the river will reoccupy its Relic Channel especially in the early years well while erosion deposition in the channel but only deposition of the overbanks and so that's all we'll do here we're doing this globally we did add this like local ability to do this cross-section by cross-section I don't love this and uh probably in future versions I'm not sure we're going to have this okay so then we're going to go to options transport methods now this is mostly 2D transport methods you'll see that this is all 2D stuff down here that you use with the 2D model but there really is just this one 1D routing method that we use particularly for Reservoir models and the issue is is that we use the extern equation and the external equation is temporally naive it doesn't actually track how far sediment can go in a Time series so it'll just move sediment from control volume to control volume without kind of asking yeah but what was the resonance time of that sediment and did we give it enough time to deposit and so you know by default that's what we do and it's computationally efficient but for Reservoir particularly for silt and Clay what really matters is how long does the this sediment stay in there and how much time does it have to deposit and so we have this option which says limit the sediment to the water velocity and what that does is in each time step it looks at how much flow actually leaves the control volume associated with the cross section and then it says hey we're only actually going to let that much sediment go and we're going to keep the rest of the sediment in that control volume to give it another shot at depositing next time now interestingly it's still vertically mixed the whole time and is that the way that sediment Works no a sediment moves down the reservoir you know the center of mass drops which is another bias which is another reason why if you kind of think about these gradations that they often maybe should be coarser than they are in reality because you want to induce a little bit more deposition but okay so we went to options transport methods we chose limit to water velocity and so that leads us to our final method and that is the cohesive options now by default we use transport functions to move cohesives because we don't want to give you default parameters for this and the basic gist there is that if you're doing a river model you know you could just leave the wash load out of it unless you're looking at concentration so tmdls if you're looking at actual bed change but if you are doing a river model you know your cohesives are generally going to be wash loads so if you use the transport function they're wrong they're going to be wrong by orders of magnitude but you know it's still going to like move all of the cell and clay and so you know that's not a terrible assumption if you're doing a river model but in a reservoir model that is like 90 Silvan clay it's really important to get this right particularly if you're going to erode and so we have two different cohesive transport options here they're both a combination of the Crone and parthenetis models for silt and clay and Crone is deposition probably in 80s is erosion but this is one of the things that we did differently than a predecessor model HTC 6 and so there are two different ways to do this but we're just going to choose the Chrome method and then I'll do a whole another video on what the M form versus the KD form of this is It's a dimensional and dimensionless form but let's leave it um and the idea here is that below a critical shear stress you're going to deposit and above a critical shear stress you're going to erode and then kind of above another threshold you're going to erode at a faster rate it's what we call the mass wasting Zone and so this is a depositional model so really what we want to get right is what's the shear stress at which we start to deposit and this is again going to be another calibration Factor we're kind of collecting more than we want at this point but we're not actually going to worry too much about this now if this was a flushing model which eventually it will be then these things become more important but you would probably want to measure these with like a said Flume or a jet test but what I'm really looking at is zeroing in on this deposition a one Now the default in you know us customary is 0.02 pounds per cubic foot which is one pascal so let's just start with that so obviously the model is going to be sensitive to this but that that's where I'm going to start what I'm going to do is I'm going to make the mass wasting threshold kind of large I'm going to make it 10 and then I'm going to put a very low erosion rate in between those so that really we have a Zone in which neither erosion or deposition occurs and it is until these high erosion rates that then I'm going to go in and say well then we're actually going to get some erosion but this is all fantasy here I don't know any of this yet these are things that I these are things that I'm going to need to figure out with some data but in order to get the model to run basically all I've done here is I have a depositional threshold okay so now I have all the data that I need I'm going to save it all right so now I'm I'm going to run the simulation so I'm going to go to run quasi and study analysis sediment or I'm going to choose the sediment running person I'm going to go to file save sediment plan as and let's call this 1997 to 2022 97 to 22. and then I'm going to choose that 01 January 1997 as my start date and I will end before the you know the data went up to about June 2023 but I'm going to run the 25 years here and say you know 31 December 2022 at 2400. all right so that's great but before we run let's go look at the sediment output options and so you see that we're running at level three we're also writing a lot of data we don't need I'm gonna turn off this binary output and then I want to run at level four and I also want to get cross-section output but what I've noticed is that in the in the recent version if I get output every day for 25 years the profile is actually pretty sluggish in the in the output so for now just a triage it let's go ahead and kind of get data every month and see the results every month because we're doing a 25 year analysis that'll give us plenty to look at for now all right so press OK file save and I'm going to compute and so that took about seven minutes to run which isn't bad which means we can definitely do better with you know more Crossings if we need them or definitely um shorter computation times but I want to go and actually look at the messages we got you know it gives us a write out of the different methods we're using and then it gives us warnings and notes here it's just saying that hey the boundary condition actually had to extrapolate below the lowest flow in the flow load curve which is fine because it's just going to extrapolate to zero it's a this this comes in red if it actually is over the uh the top of the curve because that doesn't extrapolate that just flattens out and then we have a little bit of a mass budget report which I find really useful we added this because I'd really like people to start evaluating their Mass budget before they even look at results and you can get this both either in mass and volume the difference obviously will be those densities right and so depending on how different those densities are these numbers can be different but you know for just looking at because Raz actually works in Mass if we have like 7.8 tons coming in and we've got 2.5 tons going out well then the Trap efficiency you know how much of material is trapped well it's this minus this and so we got about 5.3 million tons trapped so the Trap efficiency is going to be 5.3 over 7.8 so we're looking at like 65 to 70 percent trap efficiency which means we're still passing quite a bit of material through this pretty large Reservoir so let's actually go look at some results right now we're going to go to view sediment output and let's start with the profile it does take a little bit of time for the profile to load because it the way the hdf5 data are structured but once you load it once it should be pretty Snappy we'll go and we'll just let's start looking at invert change if we look at the change in the invert if we look at the invert change it is pretty jagged but it also is relatively predictable I would probably go in and look at these peaky cross sections to see if there's something about my cross-section geometry usually might affect the floor area that's causing that then I'm going to turn on the max because the final is not always the max sometimes it deposits and then erodes and so this actually lets you know you can look at the Min too but the Min will be zero for change but then what I prefer to look at is actually the mat the cumulative Mass bed change and you'll see that that tells a different story because the mass bed chain doesn't just show you what's happening in the channel it shows what's happened to you in the whole cross section and see if you look at a couple of these cross sections you know these cross sections are in the 11 000 you can see down at the bottom there and the next version we'll actually show you which cross sections these are and so we actually go look at those cross sections I'm going to switch to the cross section plot and if we look at say some of those some of the ones Upstream we'll turn on the first and the final you can see that's you know it's a near deposition in a reservoir that's not outrageous we you know especially when you're kind of Downstream of any sort of bed low you don't even actually expect the channel to fill first unless you're getting turbidity currents but if we go down into the 11s and early tens where we're getting some of the massive deposition now we're looking at something weird right we have these kind of overbank depressions which are lower than the thaw egg itself and are filling with sediment that's where we're getting those Peaks and you need to ask yourself the question is that real and so if we go in and we look at what that cross section looks like you know these cross sections I drew kind of to include this portion but this portion is actually a tributary and so that's actually a channel and we weren't modeling that specifically but this Channel's lower than this channel you see this channel actually has built some natural Levee walls and so I think it's a legitimate question as to whether it's appropriate to include that in the model I might actually truncate these cross sections and try to do something else with that and so one of the things that I think you I want you to get from this is my first run and so my first run is exploratory we're nowhere near being really being ready to answer questions or deliver this model yet we're going to evaluate the results to see you know how is the model performing where do the one-dimensional assumptions break down and you know what sort of updates do we need to make to our model in order for this to be strong another thing I like to look at is the grain class percent profile and so that's going to actually show us a gradation in the downstream Direction and so I'm going to look at the cumulative Mass change that's what com means the cumulative Mass change that means the mass change in the bed at all time up to whichever time you select and we will turn off the first one and then turn on the final one and what this is going to show us is you know the red line is that mass change that we were just looking at but then it shows us what percentage of the mass change in this case it's all deposition is associated with each grain class and black is clay gray is silt and then the Browns are the Sands and so you can see kind of As you move Downstream it gets finer because you have more and more clay and less and less of the coarser silk classes and you can actually go in and turn on the D50 as well and it'll show you the gradation dropping in the downstream Direction one other thing I like to look at let's go back to the profile and at this level output level I'm only doing the mass in teach control volume I'm not finding the mass out but that's a good enough surrogate and so what we can do is we can look at the mass flux into each control volume at any time step and this is cumulative so it's over the course of the whole simulation if you choose the last time step so this is the mass flux that comes into each cross section for the whole model run for all 25 years and so you can see as the purple is the total and so this is the mass that's coming into the Upstream cross section but we don't have any other sources or sinks other than deposition in the model and so this is the mass that's leaving the con the the whole model or at least you know minus the downstream cross section and so that is actually your trap efficiency that shows you everything between those two numbers is deposition and the this number and this number you know divided by this number will give you the Trap efficiency but then the thing I like about this plot is then it breaks down by grain class and you can see these darker Grays are the coarser silk grain classes you can see that they all go to zero which means that they all have 100 crap efficiency all of the coarser grain classes basically go to zero which means they're completely deposited within the reservoir but finest silt and Clay don't go to zero which means that they're actually passing flux through the reservoir they don't have 100 trap efficiency and so if you need more trap efficiency if you feel like you're passing too much sediment if you're getting let's say you have Downstream concentration observations that are there in your results are too high then you can move material from your fine silt and Clay grain classes and you know assume oh actually there's more flocculation going on they they thought or the vertical mixing bias is stronger than I thought and you can move those into the coarser grain classes and so that's just a very kind of brief primer on a very simple Reservoir model to wrap up let's actually look at the reservoir profile if we go in and we say hey I want the water surface and profile and I want the bed elevation and profile and we're going to turn on the front the first and the last you'll see kind of what our deposition profile looks like but you'll notice that our high water surface actually you know Daylights at the top end of the model this is not ideal because reservoirs do tend to build Deltas kind of Upstream of where the water surface Daylights and so this is actually just all of the geometry I had but one of the things that I'd want to do to make this model better is actually extend this geometry I would say another you know five kilometers Upstream so we could get the course Delta that builds in that part one of the reasons why you know putting course material into this model doesn't really work is because the course Delta often develops Upstream of that where that water surface Daylights we don't really have capacity for that in our model and so you know that's a my first shot at a very simple model of the system but I did think that it provided some lessons learned along the way that if you're building a reservoir model uh that it would give you some insight on you know where people often make mistakes and some of the options and methods that are most useful for Reservoir modeling my name is Danver Gibson I'm the sediment transport specialist on the HC Ras team and this video was funded by the HH and C set program

Channel: Stanford Gibson

Views: 1,507

Keywords:

Id: ZuhK_gTcz1g

Length: 49min 40sec (2980 seconds)

Published: Wed Aug 16 2023