[Jarvus]

It is hard to know for sure. Plot the energy grade line. It may be that the water surface is dropping because the velocity head is increasing where it “drastically narrows”.

[Jarvus]

That approach sounds reasonable to me. If the large town is the main focus and it is ~90km downstream, then the exact mechanics of the dam breach are less important. The farther downstream it is, the less the dam breach itself matters. If the town was immediately below the dam, then it matters a lot.

So doing a triangular hydrograph seems reasonable. Nobody knows exactly how a glacial lake dam is going to fail anyway. If you get your model up and running, you can run several different shaped hydrographs of different durations.

If neither you nor your supervisor have a background in this, I’m not sure what suggestions I can really offer outside of trying to find someone who does have a background who would be willing to review your model.

There are lots of different things that could be wrong with your model that is causing bad results. Some of them may be very easy fixes and some of them may not.

For instance, if the terrain file has not previously been used for a river flood study, it may need minor various tweaks. Or it could need major work. Even if it has been used for a river flood study, if the glacial outburst flood is on a much bigger scale and the inundation is much higher, you may need to make adjustments to parts of the terrain that were not adjusted for previous studies.

[Jarvus]

Could HEC-RAS be used to model a glacial lake outburst. Yes. The software won’t, of course, tell you how fast the ice dam will collapse, but if you enter appropriate data, it could model everything else. You can enter a trapezoidal breach and have it grow at various, user entered rates.

There are many techniques for modeling flood hydrographs that involve various levels of effort, require different data, and have different accuracies. But if you want a flood hydrograph and inundation map, I would think a dam model like RAS would be a good choice.

However, for someone completely new to RAS, that sounds to me like a lot of work for a final step to complete a thesis. It is also not clear what your background is in river hydraulics and flood modelling. Which again, isn’t to say that one can’t learn how to do it, but it could take quite a bit of time, especially if you are tying to basically do this on your own.

I’ve read a little about glacier lake outburst, mostly from articles in the magazine New Scientist. Are you trying to model a huge historical event, say from thousands of years ago? Or is this a modern, smaller scale event? I’m wondering what you are going to use for your terrain model. If you can download a current terrain model, that would save a lot of work. If you are trying to recreate conditions that existed in the past, that sounds like a lot of work.

I’m not an expert on the terrain side of things, but dealing with the terrain could be more work than the river hydraulics.

RAS is free to download. The example data sets include a dam break model. You could play around with that to get a sense of what you might be getting into.

[Jarvus]

Flow is total for the entire structure.
Weir flow is the flow over the top of the structure.
Culvert flow is flow through the culvert.
Outlet RC is flow from the outlet rating curve.
Negative flow usually means you have reversed flow from the tailwater side to the headwater side. For a long connection, it is possible for flow to be going one way over one part of the structure and the other way over a different part. It is also possible that the rating curve has not been entered correctly.

[Jarvus]

This is confusing. And I agree that it needs better terminology. My best guess is:

RAS is using the term ‘inlet control’ to refer to the situation where the inlet equations are controlling the flow. These are the equations that are based on the shape of the culvert (ie, square edge, corrugated metal etc) that came from the work initially done by Federal highways.

If these equations do not control, then it is labeled as outlet control even if the inlet passes through critical depth and the culvert is supercritical the entire way. But I would tend to agree that this is also inlet control. Although it is not the ‘inlet control equation’ as described above.

[Jarvus]

If you select the 1D weir equation*, I think technically the flow is computed in the same manner, however, you can get differences.

I’ve commented on this before.

Short answer: I generally recommend trying to do everything in a single 2D area because it tends to be more stable.

Longer answer: If the dam break is inside of the 2D area, the flow through the dam break gets updated for each 2D iteration. If the the dam is between two different 2D areas, the weir flow computation is either lagged (if 1D/2D iterations is off) or the iterations are on the time step (if 1D/2D iterations are on). But in either case, it tends to have more stability problems.

Yes, if the dam is inside of the 2D area you have to fill it (RAS needs to let you specify a starting WSE for interior dams). You can add a hydrograph boundary condition line inside of the 2D area immediately upstream of the dam and give it an extremely large flow just to fill the reservoir up quickly at the start of the run. Writing out a restart file, after the reservoir is filled can also save some time for future runs.

* If the dam is inside of the 2D area, you also have the option of using the 2D domain instead of the 1D weir equation which is not an option if using two, 2D areas. However, this equation assumes open channel flow, so it is not generally a good choice at the start of the dambreak where the weir flow assumptions are more valid. Using the 2D domain equation when the flow is actually weir flow seems, in my experience, to produce too high of velocities.

[Jarvus]

I created a simple data set that was 5000 meters long, 100 meter drop and cross sections every 100 meters, with a Manning N of 0.010. With an upstream boundary of critical depth. At the downstream end, I got a water depth of 1.389, fairly close to the 1.4 m you report.

1:50 gradient and a Manning N of 0.01 gives a normal depth that is very supercritical. I do not believe the RAS steady flow culvert routines will compute a supercritical flow going into the culvert. The flow inside of the culvert can be supercritical and the flow leaving the culvert can be supercritical.

But the water surface immediately upstream of the culvert is, I believe, going to be assumed to be subcritical.

I didn’t try modeling the culvert itself. However, if the water surface upstream of the culvert you describe is subcritical and the downstream boundary is low enough, then the culvert is probably going to be under inlet control. So the ability of the flow to get into the entrance of the culvert is going to be the limiting factor.

If you have a 3.5m by 1.4m channel flowing into a 3.5m by 1.4m box culvert and the flow doesn’t hit the lid, then there may not be any entrance losses. But if the flow touches the lid or there is any change in shape between the channel and culvert, there are going to be some type of entrance losses especially given the very high velocity.

Also, the capacity of a box culvert flowing full is not the same as the open channel capacity. The top of the culvert adds another 3.5 meters of wetted perimeter. So this culvert, or the equivalent cross section with lid would not have a capacity of 58 m3/sec at normal depth.

If you have a situation where there is a supercritical channel flowing into a culvert and the flow stays supercritical the entire way, in order to model that with RAS, I think using cross sections with lids is the only viable option. As this has the capability of staying supercritical. Depending on the transition from the channel to the culvert entrance, you would have to give a lot of thought to what the entrance loss should be. The entrance losses would probably be higher than the normal contraction/expansion loss that is computed. Since the normal contraction/expansion loss assumes a gradual change over the entire distance, not a sudden change like the entrance to a culvert. RAS has an option to enter an additional energy loss that is a coefficient multiplied by the velocity head. This could be used in order to model the entrance loss. Although I’m not sure how would be the best way to determine the value of the coefficient itself.

[Jarvus]

I don’t think there is any reason to be concerned because there is an adverse slope where the ground gets higher as you go downstream. For 1D, the exact location and depth of the invert doesn’t matter. The important thing is for a given water surface, how much area and wetted perimeter there is.

The RAS hydraulic reference manual goes into great detail about how this is computed. Or any basic open channel text book/reference book would also.

[Jarvus]

I don’t know what is going on with the RASMapper not being an exact minute. Unless the model is set to output every time step (even the reduced time steps after halving), it looks like some type of rounding error problem.

[Jarvus]

”
From what I can see, I deduced that the maximum time step is actually 1 minute, that is equal to the mapping output interval. So the software, even if you provide meaningless data, automatically corrects it in order to obtain the right time steps, in reference to the mapping interval.

Could someone confirm my assumptions?”

Yes, it will not use a time step that is longer than the mapping output interval. It will also make sure that any time step that gets doubled will still go into the mapping interval. So for instance if the output is 1 minute and you choose a ten second starting time with doubling: ten seconds and 20 seconds are okay but 40 seconds doesn’t go into one minute. So it would shorten the starting time step to 7.5 seconds (15 seconds, 30 seconds, one minute).

[Jarvus]

If you model a road crossing as a culvert and weir, I believe RAS computes the culvert flow and the weir flow separately. It iterates back and forth until it gets the correct flow going through the culvert and the correct flow over the road to balance. If you run enough data sets for enough years, you will probably eventually see a message along the lines of, ‘the weir/culvert flow didn’t converge’.

Cross sections with lids are not like this. I believe that the wetted area above the lid is combined with the wetted area below the below the lid (and the same thing for wetted perimeter). It then does a steady flow backwater on this combined hydraulic area.

If you have a very long culvert you model with cross sections with lids, you could start out with flow above the road at the upstream end but as you go downstream and the energy drops below the road, all the flow disappears from the road and is transferred into the culvert. This obviously isn’t how most culvert work.

So for the culvert part of the flow being modeled with lids, the lid needs to be thick enough to not have flow above it.

If your model has a 4X6 with a couple bends and also a 36″ pipe, I would presume this is not a short, simple road crossing but instead the culvert is probably underground for some distance.

So upstream of the culvert you could add a junction. And you would have a second junction downstream of the culvert. Yes, there would be two parallel reaches that contain cross sections. The reach that represents the flow over the road would be normal cross sections. The reach that represents the flow in the culvert would have cross sections with lids. It is a little awkward to set up. And it can be even more awkward to explain to a client. However, given the limitations and assumptions that are already built into a 1D, steady flow model, I think you can get perfectly fine answers.

You would also want to turn split flow optimization on for the upstream junction in order to get Steady flow to iterate back and forth and balance the flow/energy between the two reaches.

[Jarvus]

If you are modeling a culvert as a cross section with lid and the culvert is pressurized, the profile plot will show the water above the top of the culvert. However, if you look at the cross section specific output, you can see that the area and velocity is correct. It is just a confusing output issue where the water surface elevation is being reported as the hydraulic grade line instead of being reported as the top of culvert.

However, as I mentioned several times before, it is important that the lid is thick enough that the water does not get above the lid. For cross sections with lids, the area above the lid is connected to the area below the lid essentially allowing flow to move back and forth. So this is why you need a thick enough lid to prevent this from happening.

If you are modeling a culvert where the road is over topped and you want to use cross sections with lids for the culvert, it is a little more awkward. The best I have come up with is to create a junction on either side of the road with two separate reaches. One reach for modeling the flow over the road and the other reach for modeling the flow through the cross sections with lids.

[Jarvus]

I haven’t tried the cross section with lids between culverts for unsteady flow. Because of the bridge/culvert HTAB curves, you might need three cross sections between the culverts instead of two. But if you put them close together, I think that would be fine.

[Jarvus]

I don’t know what it would take to satisfy FEMA. However, if you are sure the system is under inlet control, then hydraulically speaking you are done. If it is under inlet control, the two bends and the 36″ pipe at the end will not have any effect on the WSE and EG upstream of the culvert. Just model it as the 4/6 box with inlet control.

If you aren’t sure about inlet control, or FEMA requires more:

Yes, you could try and model it all as cross sections with lids but that won’t compute an inlet control answer. However, you could compare the answer with lids to the inlet control of the 4X6 box and go with the higher of the two.

You could also try and model it as a combination of culverts and cross sections with lids. If this is a steady flow FEMA floodway analysis, that is probably what I would try. You mention instability, so I’m presuming this is unsteady? Doing a combination of culverts and XSs with unsteady is possible, but instability can be an issue.

For your “Model 2″, you say there is a spout of water. I’m not sure what you mean by this. For the cross sections with lids, the lid needs to be thick enough (that is, the top of weir elevation needs to be high enough) that you don’t get water showing up on top of the lid. So maybe you just need to make the top of the lid higher. Otherwise, I’m not sure what you mean.

For steady flow, you don’t need the Preissman slot. For Unsteady flow, yes you definitely need it.

For a combination XS and culvert:

To connect the 4X6 to the 36″, add a culvert for the 4X6 box and add a second culvert downstream for the 36”. You will need two cross sections in between the culverts. I would probably make one cross section approximate the 4X6 box and the other cross section approximately a 36″ circle. Make sure the top of the lids are high enough that water doesn’t appear above these lidded cross sections. The upstream 4X6 will still compute an inlet answer and the dual 4X6-36″ culvert will compute an outlet control answer.

There are various ways to handle the energy loss between the connection. If you want to handle it as a minor loss: set the exit loss on the 4X6 to zero, set the entrance loss on the 36″ to zero, and set the contraction/expansion loss on the two lidded cross sections to zero. Then add an additional energy loss to one of the two cross sections using the coefficient multiplied by the velocity head. This would be the minor loss coefficient. Alternately you could allow the lidded cross sections to compute the contraction/expansion loss if the transition is gradual enough.

For the bends, the 4X6 box could be broken into 3 different culverts with lidded cross sections between them and then add a minor loss for each bend as described above.

I think that should be straight forward for a steady flow answer. But all this could be tricky if you are having unsteady stability problems.

But running a steady flow model with a range of tailwaters and a range of flows might give you a better understanding of how the culvert system is behaving and what needs to be taken into account for unsteady flow. And again, if you can demonstrate that it really is inlet control over the flows and tailwaters you are trying to model, you only need to deal with the upstream part of the 4X6 box.

[Jarvus]

For velocity mapping of a 1D model in RASMapper: if you are asking if it takes into account the underlying terrain between two cross sections, I do not believe so.

Each 1D cross sections gets a velocity distribution along the cross section. If you open up the geometry editor and click the HTAB button, there are columns for the number of horizontal velocity slices. So by default the channel would be sliced into five equal top widths and a velocity computed for each slice using the conveyance computed from Manning’s equation.

RASMapper knows the location of the stream center line, the location of the channel banks and the location of the ends of the cross section. I believe it uses this information to curve the velocity for the static velocity arrow or for the dynamic velocity tracing.

So I think the magnitude of the velocity is basically a straight interpolation along a stream path. For instance, the stream path that is just inside the left channel station. The location of this stream path and the direction of the velocity arrows is based on some combination of channel center line and bank locations.

Depth I think it does look at the underlying terrain, which as you point out is pretty straight forward. Although I could quibble slightly with calling it flat. I would suppose that the water surface elevation is linear interpolated along a stream path.

I have not tried it, but I was not aware that you could get RASMapper plots like depth without a terrain file.

[Author]

Posts