The three systems had originally been held in separate models, with the foul and surface networks each modeled in InfoWorks CS and the rivers in ISIS. “There has been a lot of careful work to bring the models together, but it does provide benefits,” says Tony Bamford of MWH. Appreciation of how all the conveyance systems work together is the key to understanding, and hence preventing, the area’s flooding. Use of an integrated model also saves time during the analysis.
The study covers a 29km2 (11 square miles) catchment in North Brent in West London. The area is within London’s M25 orbital motorway and has seen huge amounts of development over the last 50 to 70 years. “Large parts have suffered quite extensive and repeated flooding,” he explains.
Integrating models is more complex than working with individual ones, says Mr Bamford, but it has proved worthwhile, particularly as there are many links between the foul, surface and river systems. “We can get much more of a feel for the catchment by bringing them all together. We can understand how each is interacting.”
Bringing the models together introduces more aspects to consider, including the level of detail to adopt. “We’ve opted for a quite a lot of detail in linking them all together,” he says. “There are significant benefits in knowing how it all works together.”
Careful data verification and calibration has been carried out for the combined model. “We’ve achieved better correlation with the recorded flooding and have improved on the standing water flooding mechanisms,” says Mr Bamford.
Another benefit of a single model is that there is no need to export or import data sets once it has been set up. Use of inflow and outflow hydrographs would have become very complicated, given the large number of links.
The number of links meant that MWH felt it was quicker to do everything within InfoWorks CS, rather than using InfoWorks RS river modeling software too. Further work will include studies of the overland flood pathways and the intention is to make use of InfoWorks 2D, Wallingford Software’s new 2D solution for surface flood modeling.
The catchment being studied is very impermeable and prone to flash flooding, with the high volumes of surface water leading to spillages into roads and blown manhole covers. Short durations of high intensity rain can cause rapid rises, even when the overall amount of rainfall is not particularly large, says Mr Bamford. He showed an example of a river channel which was running at a depth of just 200mm (8”) at 11am but was almost over its 2m-high bank within an hour. By 3pm the skies were blue, and the water course level had returned to 200mm.
The original foul water model covers a population of about 170,000. There are some 4,000 links in the model and 213km (132 miles) of sewers. About 30% to 40% of the runoff from the roof area goes into the foul system, says Mr Bamford, which causes some problems. “The system was designed for the foul flow, but it wasn’t really designed to cope with all the rainfall as well.” A surface water model had also been built in InfoWorks CS, but unlike the foul water model, it had not been fully verified.
The fact that both had been built in InfoWorks CS aided their integration. Furthermore, they had been designed from the outset to be carefully joined. It was important to avoid treating them in complete isolation, he says. “Before you know it, you could end up with more than a 100% contribution to the system.”
The cross connections are extensive, to provide relief in times of high flow. “In this system, there are foul-to-foul cross connections, and surface-to-surface as well as foul-to-surface and surface-to-foul,” says Mr Bamford. There is often no clear indicator of which way the flow would go. “It could be the surface water going into the foul or vice versa - basically they are level pipes.” One of the more unusual features of the area’s network is the inclusion of overflows for air raid protection, used to redirect flows from sewers damaged by World War II air raids.
Historical information was used in verifying the model at each stage. Quality assurance is important, stresses Mr Bamford, and enough time must be allowed for checking. “If you are not careful with the asset data it is very easy to miss out a cross connection where there should be one, or put one in where there isn’t one,” he says.
There were more than 6,000 nodes and links once the two models had been connected together. “I doubt if anyone could guarantee that every single one would be perfect.” Care was therefore taken to perform checks. There were challenges in modeling the culverts and their links with the rivers, particularly as adjoining sections could be the responsibilities of different organizations. There were some gaps in the data, such as breaks in the model where a water course enters a stretch of culvert just downstream. “All these things have to be looked at and picked up,” he adds.
Radar rainfall data and historical information were revisited during the checking and calibrating process, and the verifications showed that a good fit had been achieved for the catchment.
Adding in the rivers
The complexity of the integrated network led to the team’s decision to bring the ISIS river model into InfoWorks CS. Although not designed for rivers, InfoWorks CS can accommodate river sections. The first step was to run the river model independently in CS to ensure that comparable results were achieved. “Some adjustments were required to fine-tune the results to achieve close matching, particularly in areas where there are structures,” says Mr Bamford.
Apart from an inflow at the top of the watercourse, there is no need for inflow hydrographs for the river as the integrated model provides links from the surface water system. But it was still essential to check the numbers, and some details still remain to be solved. “Because we’re bringing the three models together, it would be very easy for us to end up double-counting flows at some point,” he says.
The analysis is much more manageable with all three models integrated into the same InfoWorks CS platform, particularly as there is such extensive interaction between the foul, surface and river systems. “In a way it’s really just one big system,” he says. “Up to 40% of the run-off from the roof area ends up in the foul system, the foul system connects to the surface water system, and the surface water connects to the river system.”
Attention to detail
Using a single integrated model is also enabling more detailed modeling to be carried out. This will prove particularly useful in aspects such as overland flood routing to clarify where the water goes when it escapes the banks. Modeling of the overland pathways benefits from the advantages offered by the flexibility of features within InfoWorks CS, he says.
Representation of the structures is one of the more difficult aspects because of the sheer number. “There are nearly 300 cross sections,” says Mr Bamford. “To put them all in and get them right does take time.” Individual aspects are carefully examined, to decide whether any simplification is possible. Full detail is essential in situations such as where a surface water outfall is positioned low down within a river bank, as rise in the river which remains in bank can affect the surface water flow. Discharges will be incorrect if details such as these aren’t modeled in the surface water calculations, he stresses.
Culverts must also be modeled accurately, particularly as they may have unusual features. In common with other areas in London, there are extensive transport links within the catchment, including bridges. A number of old bridges had been given additional support which reduces the cross section considerably. Another twin culvert only used one half during normal operation - the other was overgrown, again reducing conveyance capacity. Getting feedback from site visits is essential for ensuring proper representation, advises Mr Bamford.
The checking and validation process is often complex. The team has collated and compared a wide range of information, including recorded flooding logged by the council. Most of the different types of flooding have already been successfully replicated in the model.
Some interesting findings are emerging, including a large area where the model shows a significant amount of flooding even though there were no records of flooding. “Speaking to the general public, we have found that flooding does happen and causes quite a lot of highway issues in that area,” he says. Communication is a vital part of the process. “When you speak to people in the area, they can be very helpful in pointing out things you wouldn’t necessarily know.”
This article is based on a paper presented by Tony Bamford of MWH at the Wallingford Software International User Conference, 2007.