May 2007, Vol. 19, No.5

Sleuthing Collection Systems


Urban underground collection systems often are so complex that simulating the flow path of stormwater runoff can be difficult in large or older cities. Researchers are working on a novel approach that would give stormwater managers an
edge in designing and reconfiguring collection systems for meeting new conditions.

“In a perfect world, with unlimited manpower, time, and money, each city would map out every inch of its sewer network,” said Ferdi Hellweger, assistant professor of civil and environmental engineering and associate director of Northeastern’s Center for Urban Environmental Studies. “[But] when time and money are factors, artificial networks can be quite useful. We believe that the amazing technology that makes video games fun to play can make urban hydrologic models more realistic and powerful, and thus can help engineers and scientists better understand urban hydrology and manage our urban environment.”

The researchers’ goal is to develop a statistically equivalent way of generating artificial collection systems that cover large areas by trying to understand in test cases what is occurring in small areas.

“This kind of approach could enable us to explore the hydrologic response of a very large watershed to one of these urban stormwater systems in ways we couldn’t otherwise,” pointed out Richard Vogel, a professor of civil and environmental engineering at Tufts University (Medford, Mass.).

Virtual Systems
Currently, hydrologic analyses that model water flow through a city require a detailed description of the collection system, according to Hellweger. The problem is that in many situations, data on pipe location or size, for example, are unavailable or inaccurate, especially in older cities. “Even if you had detailed information on where every single pipe is in a city, you can’t drill down into the ground and inspect the state of every pipe,” Hellweger explained. Similarly, when city planners are trying to analyze different land-use scenarios, such as separating collection systems or developing a new area, such data do not yet exist, forcing engineers to make their best estimates.

Generating an equivalent system mathematically, on the other hand, could enable the development of efficient, large-scale analyses even in the absence of detailed information, Hellweger maintained. He and his colleagues are developing algorithms for simulating the effects of artificial collection system networks that are comparable to actual ones based on certain attributes, such as an area’s pipe density and average flow length to outlets. They recently presented their first model results developed for a highly urbanized area of Boston at the annual meeting of the Geological Society of America (Boulder, Colo.) in Philadelphia.

Using the watershed area as input, the researchers created the artificial networks using fractal technology. They then plugged the data from the actual networks and the artificial networks into the U.S. Environmental Protection Agency’s stormwater management model, making output comparisons based on the flood hydrograph and various other parameters, such as peak flow rates. So far, they’re getting comparable results. What they’re still working out is the basis for comparison, Hellweger noted.

“We can get reasonable results using the artificial network,” Hellweger explained, “[but] there are a number of different ways you can compare them.” For example, when looking at peak flow, the model may yield a different result than for total loading or total flow over the year. While the output for the artificial network will never be exactly the same as the actual network, Hellweger acknowledged, the question is whether it’s good enough. “Hopefully,” he said, “we can identify some rules for generating these artificial networks that are robust and work in a number of different areas.”

The idea is that while an artificial network may look different from what’s actually underground, maybe that’s not what’s important, Hellweger explained. Instead, “you may be able to generate a network artificially that has the same attributes [such as pipe size, density, and configuration] as the actual sewer network,” he said, adding, “the bottom line is that while it may look different, it functions the same,” producing the same results when applied in a model.

Large-Scale Applications
The tool that Hellweger’s team has developed can be implemented within geographic information system software, so “you can take that tool and use it with any data set you want,” said Andrew Miller, an associate professor at the University of Maryland–Baltimore County (UMBC). Moreover, they’re working to build in enough flexibility to be able to simulate nearly every aspect of a network design. “For trying to predict realistic consequences of future development, future upgrades, or just changes to the infrastructure, this is a very valuable tool,” Miller added.

Vogel agreed. “The key thing is that it enables you to look at the effects of sewer network expansions and design changes at larger scales,” he said. In this way, “you can look at a whole system rather than just focusing on all the little individual pieces,” he added.

A major potential application lies in helping city planners determine various land-use and zoning laws, Miller noted. Under current stormwater regulations, for example, numerous detention basins are being built and designed to hold back a certain amount of runoff over a specified period of time, but it’s rare for anybody to look at how a whole network of detention basins behaves and their impact downstream, Miller pointed out. “You design detention basins thinking about the individual consequences of the individual basins,” he said, but as it turns out, “the aggregate effect may be very different from the intended effect of any single basin.” In fact, a single detention basin may limit the amount of stormwater flow during peak discharges that occur immediately downstream, “but the aggregate effect of the whole network may not mitigate things at all, and may even make things worse,” Miller said.

Tough Sell?
The use of artificial networks is common practice in rural settings where hydrology is driven by the natural topography, but they could be a tough sell for engineers in urban settings where natural flow paths have been dramatically altered, Hellweger acknowledged. “It’s so new and radical that practicing engineers may not be able to appreciate it,” he said.

Vogel agreed. Hellweger is “talking about using characteristics of existing systems to try to simulate similar systems,” Vogel said. “It’s the kind of thing civil engineers probably are not going to adjust to that easily, because they’re so comfortable designing systems with real systems that they’ve taken measurements for,” he explained. The advantages of Hellweger’s approach, on the other hand, “could enable them to do things they couldn’t [do] now with the limitations of existing networks,” he noted.

Using a statistical method such as this as a replacement for real systems “holds a lot of merit,” said Claire Welty, a UMBC professor of civil and environmental engineering. She noted that many natural systems are modeled in such a way, one example being the subsurface structure in aquifers.

“We don’t know precisely what’s there, because we can’t see it under layers and layers of clays and sand,” Welty explained. “So what we do is sample the aquifer in places and then generate statistically what we think is between the wells and come up with a sort of virtual aquifer. Generating this kind of equivalent for underground sewer systems is exactly what [Hellweger] is trying to do.”

Hellweger and his colleagues are now working on applying and testing their methodology on other Boston-area watersheds. Ultimately, they plan to generate an artificial network for the entire Boston area. “We’ll have a representation of every single pipe in the system, including as many linear feet of pipe as there are under Boston, so it’ll be a really large model,” Hellweger said.

— Kris Christen, WE&T