Do tiny mountain streams make spiky plots?

[LorenAmelang]

Our little association owns an 89 year old dam that is facing the new California modeling requirements. We can’t afford the studies to keep the dam, or the cost of removing it. I think DSOD has vastly overrated the danger we present, so I’m trying to model possible failures to justify reclassification. I have a model that runs with no errors, drawn in RAS Mapper 5.0.5 over a one-meter LiDAR DEM. The Results in RAS Mapper look great, and confirm my guess that our little lake is no danger to anyone downstream.

But I worry that I may have some deadly errors… Up to flows that would dump the lake in six hours, the spread a mile downstream increases with increased flow from the lake. Above that flow, the spread in the first 1000′ of rough, winding narrow canyon increases greatly, but the width in the wide, flat floodplains a mile downstream actually declines a bit. Can that be right?

The View -> Summary Err… for my project is full of “energy loss” and “conveyance ratio” warnings – but I notice the BaldEagle example shows just as many of those.

The example Flows are smooth, and share the same shapes as if they are responding to actual terrain:

But mine are spiky, and seemingly random:

The curves to the right of the cursor label and below 400 cfs are the initial ramp up. All the lines that cluster around 400 for the right third of the plot happen while the flow is already declining. They go much higher where the front gets to the wide, flat floodplains and can spread out. But they don’t show the matching shapes of the flows in the example plot. Does that suggest they aren’t actually following the terrain?

Here’s the input to that model:

And here’s what the View -> Stage and Flow Hydrographs plot makes of it:

That’s the first cross section, where the smoothly ramping flow is supposedly input. How can it have those spikes? The BaldEagle plot is smooth…

Here’s my Rating Curve – also spiky:

Maybe those spikes lead to the randomly spiky downstream results? I compared the station numbers of extreme positive and negative flows, and they are unique to each calculation run – even with the same input hydrograph and terrain! Seems very suspicious…

[Anonymous]

Decrease your time step and your mapping interval and hit re-run. 1 second will work. If that helps but not completley go to the cross section with the spikes and give them maximum headwaters and tailwaters in the htba parameters.

[LorenAmelang]

I’ve now extended a few cross sections it occasionally complained about, changed my input Flow Hydrographs to 1-minute intervals (instead of 10-minutes or 1 hour), tweaked the lower end Normal Depth to avoid any slant on the last segment of plots, speeded up my calculation time interval to 1 second and all reporting times to 1 minute – and everything is as spiky as before – but in randomly different places.

I had been setting the input Flow Hydrograph – Critical Boundary Condition – Max Change in Flow value to 1, and it had obviously slowed down calculations in the parts where the flow was changing rapidly. I tried it as low as 0.1, which turned a 1-minute calc into 8 minutes, but didn’t reduce the spikiness.

I still see occasional “EG Max WS” values that are 200000 feet above my real surface – never in the same places twice. The “WS Max WS” values are sometimes a bit odd, but always at least physically possible.

The only explanation Google finds:
“EG Max WS: elevación máxima de la superficie de agua. Crit. Max WS: elevación crítica de superficie de agua. WS Max WS: elevación de la superficie de agua.”

I guess that means “EG” is the maximum ever seen in calculations? Do we care, if the average value settles on a more reasonable “WS”?

As for “maximum headwaters and tailwaters in the htba parameters”, I’m completely lost. I can’t find anything like that in any of my Htab Parm or Cross Section dialogs. But I suspect it doesn’t matter – my wild spikes are never at the same cross section twice.

To compare with the earlier Stage and Flow Hydrographs plot, here’s essentially the same thing with all the factors speeded up:

Actually puts less water up on the flat bank areas, as I’d guess from less area under the curve. How can the flow be so wrong at the very first cross section where the hydrograph is dumping it into the system?

Thanks, Luis, for giving me things to try, and someone to try to explain this to…

[Anonymous]

Your issues are a common instability problem solving exercise. Htab parameters for cross sections are VERY important for stability. If its a 1D cross sections it is located in the geometry data editor on the left tool bar just above the “View Picture” Button. All minimum WSELs or starting points should be at the cross sectional minimum and max out everything else with small intervals both vertically and horizontally. Change your mapping interval to only double your computational time step.

How many cross sections is this model? When you said you changed your input flow hydrographs you changed it in HEC-HMS? or interpolated down to 1-minute and re-entered the data in the flow editor? Since there are 1D cross sections whats your initial flow? Sometimes an unsteady model is a game of finding the right matching numbers for stability in playing with multiple parameters, as you learn the software more and more it becomes less of a guessing game and more of mathematical and dynamic approach to a solution. You green flow hydrograph looked very odd as it is showing data points every 6 hours or so. Stage should never vary that much though so you are correct in assuming your model is unstable to say the least

[LorenAmelang]

Luis, thanks again for giving me things to think about!

> Your issues are a common instability problem solving exercise. Htab parameters for cross sections are VERY important for stability. If its a 1D cross sections it is located in the geometry data editor on the left tool bar just above the “View Picture” Button. All minimum WSELs or starting points should be at the cross sectional minimum and max out everything else with small intervals both vertically and horizontally.

I’ve done the “Copy Invert” button there. Shows 30 to 80 vertical points, I guess RAS Mapper has chosen those numbers individually as I drew them? Can I tell it to use more?

But the “Horizontal” columns there are all zero. In the screenshots in the manual, those columns don’t seem to exist… I’ve tried to find out what they should show. A count of horiz points? In the Cross Section editor, each line has between 250 and 500 horizontal points.

Image search found me http://www.hec.usace.army.mil/software/hec-ras/documentation/HTab_Parameters_Error.pdf which shows those columns. Doesn’t talk about them, but theirs are all “5”…

> Change your mapping interval to only double your computational time step.

So Computation Interval 1 second, Mapping Output Interval 2 Second?
…
That may be better… Stage and Flow Hydrograph looks just as spiky, but the surface profile spikes are at possible levels instead of stratospheric. Will have to try more plans.

> How many cross sections is this model?

There are 122, in 5715 feet of channel. Some as far as 80′ apart, but much closer in “interesting” spots. If my spikes matched certain spots, I’d definitely add some, but the spikes seem unrelated to cross section spacing.

> When you said you changed your input flow hydrographs you changed it in HEC-HMS? or interpolated down to 1-minute and re-entered the data in the flow editor?

I changed the interval in the flow editor, and used its interpolate button. I realize now that doesn’t smooth out the flow corners in the Stage and Flow Hydrograph plot… But it (or something…) certainly changed the detail of the stage part.

> Since there are 1D cross sections what’s your initial flow?

This little stream spends most of the year under 1 cfs. Average flow at the first gauge, 19 miles downstream after four other rivers join this one, is around 16 cfs. But this is California, so a few winter storms can make huge spikes… I start each model at 1 cfs, and I’ve tried peaks of 20, 200, 400 (dump the dam in six hours), 800, and 5000 cfs (dump the dam in one hour). Somehow 400 spread most across the wider floodplains downstream.

I’m thinking maybe I need to put way more detail into Manning’s n for the wildly varying terrains here. If I draw the 2D layer or import a land use file, can the 1D model use that info, or are they totally separate?

> You green flow hydrograph looked very odd as it is showing data points every 6 hours or so.

Yes, I realize now my “interpolation” strategy just draws the same straight lines.

> Stage should never vary that much though so you are correct in assuming your model is unstable to say the least

Something big and obvious must be wrong, but I certainly haven’t spotted it yet.

[Anonymous]

Ok now we are getting somewhere here

uis, thanks again for giving me things to think about!

**
I’ve done the “Copy Invert” button there. Shows 30 to 80 vertical points, I guess RAS Mapper has chosen those numbers individually as I drew them? Can I tell it to use more?

But the “Horizontal” columns there are all zero. In the screenshots in the manual, those columns don’t seem to exist… I’ve tried to find out what they should show. A count of horiz points? In the Cross Section editor, each line has between 250 and 500 horizontal points.

**
Copy inverts is your starting point yes. Now for vertical select all the values and set them all at 100, and also set the Increment to “0.1”

Horizontal mapping set them all as well to “500”

The settings of these just help the solution get its number faster based on the parameters you have set here. The less availability for data within this table the harder the software will have to work to find a parameter that may lay outside of the default data that HEC-RAS initially sets. Try that #1

**

So Computation Interval 1 second, Mapping Output Interval 2 Second?

**

Yes try this for all future iterations.

**
There are 122, in 5715 feet of channel. Some as far as 80′ apart, but much closer in “interesting” spots. If my spikes matched certain spots, I’d definitely add some, but the spikes seem unrelated to cross section spacing.
**

In your failed runs, or runs with erroneous stage elevations, click the profile button and select the “plot initial conditions” button at the top right. This should show the solution of your model after intial conditions have finished computing. Are there any issues with this?

Whats the maximum slope or elevation change between cross sections? Whats the total slope from the first to last cross section?

**

I changed the interval in the flow editor, and used its interpolate button. I realize now that doesn’t smooth out the flow corners in the Stage and Flow Hydrograph plot… But it (or something…) certainly changed the detail of the stage part.
**

Yea, if you can smooth out that hydrograph in HMS if possible or within a data editor like excel to give HEC-RAS more computation points to use

**

This little stream spends most of the year under 1 cfs. Average flow at the first gauge, 19 miles downstream after four other rivers join this one, is around 16 cfs. But this is California, so a few winter storms can make huge spikes… I start each model at 1 cfs, and I’ve tried peaks of 20, 200, 400 (dump the dam in six hours), 800, and 5000 cfs (dump the dam in one hour). Somehow 400 spread most across the wider floodplains downstream.
**

Let your inital flow be a game of chance at first. I usually start my inital flow with 10% of my peak flow and work accordingly, Add flows at large drops in elevation between cross sections, the point here is to keep the model from ever drying out.

**
I’m thinking maybe I need to put way more detail into Manning’s n for the wildly varying terrains here. If I draw the 2D layer or import a land use file, can the 1D model use that info, or are they totally separate?

**

The 2D layer or computational mesh will use the assigned mannings value in the data entry location in RAS mapper. The 1D cross sections will always use the data from the geometry data editor n table only. VERY IMPORTANT. Your N values MUST match at the interface of your 2D and 1D connection. It is very important and written in the RAS manual.

[LorenAmelang]

Thank you Luis for another day’s suggestions!

> Copy inverts is your starting point yes. Now for vertical select all the values and set them all at 100, and also set the Increment to “0.1” Horizontal mapping set them all as well to “500”

Looks like 100 is the maximum Horizontal it can save, even though it says in the header that the vertical count can go to 500.

> In your failed runs, or runs with erroneous stage elevations, click the profile button and select the “plot initial conditions” button at the top right. This should show the solution of your model after initial conditions have finished computing. Are there any issues with this?

The very first cross section had a second dip way out past the bank line, and apparently it was dumping about 40% of the flow over there instead of in the spillway channel. It would end up in the main channel at the third cross section, but fixing that has made all the plots a bit smoother.

> What’s the maximum slope or elevation change between cross sections? What’s the total slope from the first to last cross section?

The worst drop near the top was 30′, a few more around 20′. Below the spillway there is way less than one foot drop between cross sections. I added some at the top, to keep the drop under ten feet. Some results are smoother, but now the stage jumps up 70′ in the 4′ drop of the first XS distance, instead of up 9′ in the previous 25′ drop distance. After that first extreme, the rest of the stages and elevations looked reasonable.

Total drop from the bottom of the spillway to end of model is 90′ in 5600′. Above there the spillway drops 30′ in 75′.

But that brings me to a new insight… In the YouTube videos I’ve seen, people just dump a flow hydrograph at the top of a river and watch what happens. But in a dam failure that starts down from the top, the elevation where the water is pouring out will keep dropping as the dam fails. In order to dump 5000 cfs, the dam will have to be washed away almost down to the bottom of the spillway.

So maybe trying to model dumping huge flows at the top of the spillway is just wrong? But it seems dumping the same flows down at the beginning of the shallow slope would lead to even higher initial spikes… I see people saying to create a storage area at the top of the stream, but it seems that would also require a fixed elevation for the connection between storage and stream. How do you model a failure that washes away your top cross sections?

> Yea, if you can smooth out that hydrograph in HMS if possible or within a data editor like excel to give HEC-RAS more computation points to use

Haven’t tried that yet. Actually there are lots of points, they just fall in straight lines with corners instead of a smooth curve.

> the point here is to keep the model from ever drying out.

Difficult when the stage seems to be mostly tiny spikes, and the flow bounces from positive to negative constantly.

> The 2D layer or computational mesh will use the assigned mannings value in the data entry location in RAS mapper. The 1D cross sections will always use the data from the geometry data editor n table only.

That’s what I thought. I haven’t begun to think about 2D yet…

So here are today’s whines about probable bugs:

Tons of errors about left bank Manning’s n not set. Totally set in Geometric Data Editor table. In cross section profiles, channel and right n was shown, left was either missing or to right of right n width. Looked in RAS Mapper, noticed those cross section Edge Lines were not at the XS ends.
-> Cross Section -> Edge Lines -> [right click] Edit -> Compute Edge Lines at XS Limits -> Stop Editing (Save) fixed this, and it ran!

But then I added some cross sections in the steep initial spillway…

Bank Lines, Edge Lines, and Flow Paths would zoom to feature properly, but Bank Stations, Cross sections, and Rivers would zoom way out to show coordinate 0,0 to the southwest. Mapper’s Recalculate didn’t fix this… Flow Hydrograph entry dialog showed an unnamed reach of unnamed river…
Geometric Data Editor -> Edit -> Delete -> Reaches …

Also must manually fix Htab Param and Manning’s n tables after adding XS in Mapper.

And once again Mapper Recalculate has added a bunch of those “zero width walls” to be manually fixed.

Definitely still deserves the “Beta” label.

[Anonymous]

**The worst drop near the top was 30′, a few more around 20′. Below the spillway there is way less than one foot drop between cross sections. I added some at the top, to keep the drop under ten feet. Some results are smoother, but now the stage jumps up 70′ in the 4′ drop of the first XS distance, instead of up 9′ in the previous 25′ drop distance. After that first extreme, the rest of the stages and elevations looked reasonable.

Total drop from the bottom of the spillway to end of model is 90′ in 5600′. Above there the spillway drops 30′ in 75′. **

This is your issue. HEC-RAS and the equations that govern solutions have slope limitations (if i remember its due to the cos(X) portion of the eqaution) When you have these large drops in short distances you have to trick HEC-RAS. Ways to do this include modeling the drops under virtually placed structures that can handle that flow transition much better than the actual channel. These virtually placed structures are bridges with HIGH decks that represent the large drop over around 10-ft max between 4 cross sections. If you isolate your model into 3 separate branches separating the drops, i bet your model would solve very easily with no error.

[LorenAmelang]

> This is your issue. HEC-RAS and the equations that govern solutions have slope limitations (if i remember its due to the cos(X) portion of the eqaution)

I’ve been wondering if the steep part was a problem. So I saved a modified Geometry, where I deleted the XS above 5640 – the steep part is gone from “-spillway” plans.
Compute is not slowing down as it hits the serious flows – maybe the wild spiking is gone? Still took 5 minutes, though…

Now the initial pileup is down on the more level part, and it is taller as I expected:

***** Warning! Extrapolated above Cross Section Table at: *****
Walker Creek Reach 1 from R.S. 5640 to 5493
The maximum xsec wsel error was 4.259
Walker Creek Reach 1 1437 at 13SEP2018 01:13:51

The Stage and Flow Hydrograph plot shows the new plan and geometry names, but the only top plot option is the old 5715 top XS. All other plots have the new 5640 top. Reselecting Plan doesn’t help. Bottom stage looks just as spiky as before. Flow at bottom never gets above 2 cfs – something is way wrong! Large Flow at top stops between 5512 and 5349… Hydraulic Depth seems to have flow all the way down???

So I tried dividing the Flow Hydrograph by 10 – for 40 cfs max…

No extrapolation with the lower flow:
The maximum xsec wsel error was 2.754
Walker Creek Reach 1 25 at 13SEP2018 04:02:44

Guess I need a different Normal Depth at the bottom, it is sloping way up in the last segment.

Stage and Flow Hydrograph is still stuck showing 5715. Bottom flow never goes above 3 cfs, stage still just as spiky.

Flow plot goes -380 to +200 cfs in spikes, Initial Conditions plot exactly 1 as expected.

Depth plot peaks at 3.2′, never below 0.3′, follows terrain, looks good! Depth Initial Conditions plot also follows the terrain.

But the Flow Initial Conditions plot is a straight line… I guess overall flow has to be the same all along, for a steady input flow… But the Flow plot during simulation changes so wildly + and -, even in the Profiles for the last two hours of steady 1 cfs input. It never settles back down to the “Initial Conditions” straight line. I don’t understand…

Have to quit for now, another project needs attention today!

[Author]

Posts