## Dont use an elephant gun to shoot flies

Proper stormwater design "engine" selection guarantees better use of time

"Don’t use an elephant gun to shoot flies," said Eric Thompson, P.E., a senior water resources engineer at MSA Professional Services in Madison, Wis. He’s referring to one factor to consider when choosing software "engines" (the underlying mathematics) for stormwater design and modeling. It’s an important topic; getting the model right means that designers can correctly size pipes, channels, and ponds to account for peak flows, while avoiding undersizing—which can cause flooding—and oversizing—which wastes money.

Most stormwater modeling is based largely on Saint-Venant equations, but not necessarily on the full set of equations. Fully dynamic wave and simplified dynamic methods produce hydrographs showing flow rate over time at each point in the system. They also account for attenuation of flow as constant inflows move through the stormwater structure. Steady state methods are concerned only with peak flow, and can only be used when sizing is based solely on peak flow.

Three tiers
According to Thompson and his colleague, Uriah Monday, P.E., it’s helpful to think of stormwater modeling methods as falling into one of three "tiers," which are based on the fully dynamic wave method, hydrologic routing, and the rational method.

Tier one—The top tier, which includes the model that addresses the most design variables, is also dubbed fully valid hydraulically dynamic models. Software in this class would be the "elephant gun" that Thompson referred to.

Links, such as ditches with consistent cross-sections, connect one stormwater feature (node) with another.

Designers think of stormwater systems in terms of nodes and links. Nodes are points of change in the system such as inputs, gates, manholes, weirs, grates, diameter changes, and other interfaces between one set of conditions and another. Nodes are connected by links, reaches of consistent cross-section (except in natural streams) between nodes, usually pipes or ditches, that connect one system feature with another. Thompson explained one generalization that applies to nodes and links, "At a node you’re worrying about depth, and in links you’re worrying about flow rate and velocity."

When analyzing peak-flow performance at various points in a complex system, both upstream and downstream flows and volumes must be considered. And that’s where the "fully valid" comes from in Thompson’s scheme. It means that the model is based on all relevant data available, and that all available mathematical analysis is being applied. This means the model can handle flow splits, overflows, and backups automatically. Moreover, these models can report flow and volume over time and show what’s happening at individual nodes as peak flows move through the system. Dynamic methods are "usually but not always more accurate," according to Thompson.

But accuracy and completeness come at a price. Working with software that uses fully valid hydraulically dynamic methods is a complex business that requires very experienced designers, and it takes time. The resulting models are sensitive and can destabilize. Moreover, the resulting reports provide a great deal of data and are necessarily complex. Practically speaking, they are overkill for some projects.

But for some applications they are essential. "Modifications to existing systems to prevent overflows and backups often require these methods," said Thompson. "Without full analysis, it can be hard to say what effects a change to the system might have."

Tier two—The second tier can be referred to as a hydrologic, or simplified, routing model. In the link and node terms described above, this simplified method considers only upstream nodes when looking at any given link. Modelers use terms such as "Muskingum," "kinematic wave," "Puls," and "convex" to describe these methods. They are still dynamic, reporting on change and storage over time.

A hydrologic, or simplified, stormwater routing model considers only upstream nodes such as detention basins.

"Designers of detention basins tend to use this level of performance most often," according to Monday. "Inherently, detention basins require an inflow hydrograph to see what happens as the basin fills. And since downstream boundary conditions are usually known, or at least well-estimated, it’s acceptable to look only at upstream conditions. In a rural subdivision, for example, pond discharge may simply be to an open channel; so the downstream condition is a known, non-variable design factor."

Because fewer factors are considered, resulting reports and analyses summarize less information, and are more readily understood by review agencies. Put simply, this intermediate level of complexity is often just right for particular projects that require dynamic analysis, but not full analysis of all conditions.

Tier three—Third-tier methods use steady-state rational method hydrology; they don’t calculate flow attenuation over time, and therefore aren’t dynamic. Links and nodes are still the modeling elements, but links are considered solely for connectivity. A single flow rate, based on rainfall, is calculated.

"Steady state approaches use the rational method to determine peak flow rate quickly," said Monday. "They assume that the peak occurs everywhere at the same time, whereas dynamic methods can describe different peaks in different parts of the system. The tradeoff is that you generally get a ’worst case’ scenario."

 The right choice for complex projects B. Finley Vinson III, an E.I.T., and stormwater designer at Development Consultants, Inc., in Little Rock, Ark., doesn’t always reach for the biggest hammer in his toolbox when contemplating models for stormwater design. "Fully dynamic methods are definitely not essential for every project. For a ’traditional’ project, with a detention pond and an underground storage system, for example, they might be overly complex: The work required wouldn’t pay off. I have simplified methods I use for that kind of project." But there are times when a fully dynamic wave method, such as Bentley’s CivilStorm, is the right—even the essential—choice. As an example, Vinson described a recent bank building project. "What made this site tricky is that it was a corner lot that was essentially a hole with inflow and outflow culverts—and the outflow pipe was a smaller diameter than the inflow. Basically, our clients had bought themselves a detention pond! The new detention structure had to be underground, and there was no disconnect from inflow to outflow that would allow me to model drainage and detention separately." This fact alone, that the drainage structure would definitely be storing water during storm events, made a fully dynamic model necessary. "In addition, the upstream and downstream systems were included in the model, to reduce the risk of flooding," said Vinson. "CivilStorm helped me design this system with more accuracy and less expense." Like many powerful tools, software based on a fully dynamic model is not used for all, or even most, projects. But often enough, software like Bentley CivilStorm is not only the right tool, it’s the only tool that will get the job done right.

Steady state methods are much simpler to implement, but that doesn’t mean they’re necessarily inferior; on the contrary, for some jobs they’re the best choice. For one thing, as Thompson suggests above, in certain applications they provide the most conservative analysis. "Steady state methods compare peak flow to system capacity," he explained, "and do a good job providing a ’go, no go’ answer when trying to decide if a system works. But using steady state doesn’t necessarily indicate over capacity." In other words, steady state methods are superb at providing a highly reliable answer to the simple question, "Will this system fail?" But they may or may not tell the designer where he or she’s providing more structure than is actually needed; much depends on the software implementation.

Thompson feels that steady state methods are best suited to "new systems or independent systems, because the designer has control over more variables." Also, systems with balanced inflows and outflows lend themselves to steady state methods, because there is little in-system storage and attenuation is not significant. And in some areas, such as some aspects of floodplain modeling, steady state is the accepted standard of practice.

Choosing models
Thompson applies various principles when selecting an approach to particular projects. For one thing, he believes, "Don’t use an elephant gun to shoot a fly. Always choose the quickest and easiest method that will solve the problem."

Thompson also says that the audience for subsequent reports should be considered. Some agencies will not accept certain models, and compliance may require particular methods.

He explained, "I once did a study for a developer using a second tier model, with the understanding that the project would only be subject to local review. But as it turned out, FEMA got involved, and they wouldn’t accept the modeling software. So the project had to be reworked by a different method, solely to comply with FEMA requirements."

But pitfalls like this aside, Thompson said, "It’s a good idea to choose models that generate user-friendly reports, especially when dealing with less sophisticated reviewing agencies." Streamlined reports can mean shorter, less problematic reviews because agency staff can understand and approve them quickly, with greater confidence.

Also, the modeling choices of other designers—up and downstream—are something to think about. "One thing that makes me nervous is when people use steady state to design storm sewer systems, but then use other methods to design ponds," said Thompson. "There’s a potential for mismatch there, and yet it happens more often than not."

And there is at least one more factor to think about: Does your firm have the necessary skill to work with a particular method? Plenty of designers are competent with one approach, but not others. If improperly set up, a powerful model can give seriously poor results, just as a novice carpenter can do more damage with a table saw than sandpaper. Be realistic about skill levels, and get training when needed.

So ultimately, the choice of methods depends on the task at hand, just as it does for any engineering project; or for that matter, any kind of project. The choice of design tool should not be automatic, it should be carefully considered, along with the "downstream" implications. Use of the right tool leads to better answers, more useful reports and, most importantly, projects that work.

Angus W. Stocking, L.S., is a freelance writer (www.coloradowriting.com).