Another Reason for Interpolated Cross Sections

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2010. All rights reserved.
Here’s a classic case for interpolation of cross sections in a RAS model.



Notice the drop in water surface to near critical depth at the grade break. Then, at the next cross section upstream of the break, RAS seems to overestimate the head loss. Sometimes this phenomenon can be much more dramatic, occasionally initiating oscillations upstream of the break that can lead to instabilities.

When scanning the results of your model run, or when trying to diagnose sources of instability, keep an eye out for this phenomenon. Interpolation of cross sections around this point will provide a much more accurate and stable answer. Notice in the figure below, that further interpolation upstream of the newly interpolated reach is probably warranted.

Steady versus Unsteady Flow

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2010. All rights reserved.
I was recently emailed this question regarding steady flow versus unsteady flow results:

“My question is fairly basic. I have heard from a couple of other engineers that when flood levels are generated with usteady flows that it is typical to then run a steady simulation with the routed flows from the unsteady run as unsteady results may be considered less accurate? Is this standard or conventional practice to your knowledge? “

While it is common to have a companion steady flow version, I would disagree that this is done because unsteady results are considered less accurate. Two equally well constructed models, one steady, one unsteady could certainly produce different results, but the unsteady will be more accurate. The unsteady flow St. Venant equations are more physically correct than the energy equation which is used to compute steady flow. From a numerical standpoint, the energy equation has an advantage in that it can be analytically solved, yielding an exact solution. The unsteady flow equations must be reduced to a discrete form (finite difference approximation) to solve, so there is some built-in error. However, if done correctly, this error is inconsequential for river models. RAS does a good job of this. The biggest difference is that steady flow models do not take into account the effects of in-channel and off-line storage in the attenuation of the flood wave. This can be quite significant. Some modelers take the tact of constructing a steady flow model with the computed routed flows from the unsteady flow model. While this is a “forced” way of accounting for the storage effect, it is no more accurate than its unsteady counterpart. And it would be a significant effort to set up a steady flow model with a flow change at every cross section. I don’t recommend it.

In short, given two equally well constructed models, one steady and one unsteady, of the same reach, I would take the results from the unsteady flow model every time.

Any other thoughts out there on this subject?

Downstream Boundary-Normal Depth?

Written by Chris Goodell, P.E., D. WRE
Copyright © RASModel.com. 2010. All rights reserved.
Probably the most commonly used downstream boundary condition in both steady and unsteady RAS is the Normal Depth assumption.  You make an assumption that your river flows under normal flow (uniform flow) conditions at the downstream boundary of your model.  This allow you to provide an energy slope, and then RAS will automatically back-calculate the depth using Manning’s Equation.  This method’s popularity stems from its ease of use and it’s semi dynamic properties (i.e. as the flow changes, so will the downstream boundary depth). 
The most common question I get about using the Normal Depth option is what slope to use.  To be true to Manning’s equation, the correct answer is the discrete energy slope at the downstream cross section.  Well, without computing, this is impossible to come by, and in order to compute, you have to assume an energy slope.  There are a few ways you can assume an energy slope for your downstream boundary:

1. Measure the average bed slope of your stream in the profile plot. You can do this by holding the ctrl key and drawing a line to represent the average slope as shown above.  The pop-up window gives you a dy/dx term, which is the slope. 
2. Measure the bed slope of the last two cross sections at the downstream boundary.  This will better approximate the energy slope at the downstream boundary, but you can run into problems if your reach ends on a horizontal or adverse slope. 
3. Use a USGS Quad map or other contour file or terrain model of the reach downstream of your modeled reach. Measure the distance between the first two contour lines that cross the your stream. 

Regardless of the method you use to get an energy slope, there will be some error associated with what you use.  Besides the fact that you have to more-or-less guess at an energy slope, natural rivers and streams rarely-if ever exhibit uniform flow conditions.  But, if we can manage the error associated with using normal depth, it can serve as a very convenient tool for estimating the downstream boundary condition. 
To manage the error, I do a couple of things:

1. Move the downstream boundary location as far away from my area of interest as possible.  For example, if I’m working on a bridge hydraulics study, I try to get the downstream boundary at least  5 cross section widths downstream of the bridge. 
2. Perform a sensitivity study.  Bracket a range of realistic energy slopes and then run both the high and low value.  Plot both results on the profile plot and see where the solutions for both energy slopes converge.  If it is below your area of interest than the boundary is far enough downstream.  Caution…for you steep slope, make sure you do not force the results to critical depth by selecting a slope that is too steep.

Here, you can see with the sensitivity study, the three profiles (with three different downstream boundary energy slopes) converge at the bridge.  It might be wise to move the downstream boundary downstream a little further. 

Latest DSS-Vue Plug In

Posted by Bill McDavitt:


"(DSS-Vue) now allows users to injest 1 hour data (sometimes the gage even has 15-minute data) from the USGS Instantaneous Data Archive using the latest plug-in, just released in November. I've been able to work on some projects that happen to have gages very nearby, which I realize is a bit of a rarity. Nonetheless, with 15 mintue data in a DSS file, making an unsteady flow model using data for a particular storm flow can be done with relative ease and minimal additional time spent.
If a user is comforatable scaling/adjusting the gage data for their non-gaged site, HEC has an Excel tool whereby the flows could be modified in Excel and then repackaged into a DSS file."

2009-A look back in HEC-RAS

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2009. All rights reserved.
2009 was a very interesting year for HEC-RAS. A beta version of 4.0.1 came out which included a new Floodplain Mapper (RAS Mapper). This looks like a great start to realizing a fully functional HEC-RAS model with built-in GIS capabilities. Also, Modified Puls Routing for steep streams and a new and improved method for computing at junctions in unsteady flow were added. Sediment Transport continues to improve, and I know that HEC is currently working on a two-dimensional component to HEC-RAS (both in series and in parallel). More water quality functionality was added (though I admit to not having worked with this too much). All great stuff. From what I hear, version 4.1 will be out soon. Keep checking back on the HEC website often.

As for the first full year of the HEC-RAS Bloggery, I was pleased with the response and readership. We talked about the limitations of RAS, failing bridges, Cross section spacing, minimum flow requirements, Htab strategies, cross section interpolation, RAS precision, bridges and culvert strategies, sediment transport issues, and possibly the most popular topic of late...n values in steep streams.

Thanks to all for reading and responding. Please let me know if there are ways to improve the HEC-RAS Bloggery, or any topics you would like me to write/comment about. Also, I want this to be more of a public blog posting site. So if any of you have articles you've written that you would like to post in the HEC-RAS bloggery, feel free to email to me (Chris G.) and I'll post them up. I would like to hear about other's experience with Water Quality in RAS, Sediment in RAS, Channel Modification, more on n values in steep streams, or just something you've done with HEC-RAS that was really cool.

Thanks and all the best in 2010!

n-values in steep streams

Written by Chris Goodell, P.E., D. WRE
Copyright © RASModel.com. 2009. All rights reserved.
What is a good range of n values for a typical river or stream bed? 0.03?0.045? What about a mountain stream? 0.05? 0.07? Jarrett has a very simple formula that serves as a good check on n values in mountain streams. He developed his equation from 75 observations of streams in Colorado. His streams were composed of bed material ranging from cobbles to small boulders. Range of energy slopes were 0.002 ft/ft to 0.09 ft/ft and range of hydraulic radii were 0.5 to 7 ft.

Jarretts equation is: n = 0.39*(S^0.38)*(R^-0.16), where S is the energy slope and R is the hydraulic radius of the stream (in US Customary length units of feet).  (*Note-the original post had mistakenly listed the equation as 0.47*(S^0.38)*(R^-0.16).  That was incorrect.  The correct equation, as published in "DETERMINATION OF ROUGHNESS COEFFICIENTS FOR STREAMS IN COLORADO" by Robert Jarrett is n = 0.39*(S^0.38)*(R^-0.16).  Sorry about the mistake!)

Using his range of energy slopes and hydraulic radii, you could compute n values from 0.032 to 0.21. Yes, that's 0.21!!! I have had discussions with many class participants of mine who indicate that indeed they are finding n values much higher than traditionally what have been used. We're talking as high as 0.12 to 0.15 in some cases. This definately fits within Jarrett's confines. Partly to blame in this underestimation of n values in steep mountain streams would be the very popular table of n values in Chow. Chow lists mountain streams as having n values from 0.03 (gravels, cobbles, and a few boulders) to 0.07 (cobbles with large boulders). Also, another popular n value predictor, Barnes (USGS), lists it's highest n value stream as Rock Creek near Darby Montana, with an n value of 0.075. Rock Creek is composed of boulders with a d50 of about 220 mm. However, this was measured during a flood. It is likely that the n value is much higher at lower discharges where the bed irregularities have a greater impact on the overall roughness.

I like to use Jarrett's equation whenever I'm dealing with steep mountain streams that fall within (or close to) his experimental range. A little secret here: higher n values helps to stabilize unsteady flow models!

I'm curious to know if anyone out there has comment on this topic. I'd like to know what kinds of n values you all are coming up with for steep streams.

Dam Breach class-Portland, ME

I'm heading out to the HEC-RAS Dam Breach class in Portland, Maine tomorrow. Glad to see that enrollment in our classes is making a rebound. Mt last 2 classes were cancelled due to low attendance.