A model approach to flooding
With tackling sewer flooding now a significant component of AMP4, Saul Bradford of HartFair looks into the use of sewerage modelling tools to devise remediation strategies
Undertaking a project for verification and optioneering of sewer models where infiltration and overland flows played a significant role in property flooding, consultants HartFair used InfoWorks CS and its Infiltration Module to understand and reduce the flooding risks.
The study area covered 140 predominantly flat hectares of urban dwellings in the midlands. The housing development contains 9,000 residents and was built in the 1950s and 1960s on an area containing both clay and sandy soil. Before the development was built there was a brook running through the site with a number of very small tributaries.
The shallow valleys of these tributaries were retained as features of the development, with the foul sewers running in the bottom of them. The catchment drains eastward along the line of the main brook. The specific area of study for the project was the separately drained, north-west corner of the site. In this sector there had been four major flood events in the past 14 months during rainfall events with return periods of between 30 and 100 years.
Flooding was from the foul system – the properties at risk of both internal and external flooding lie in a straight line along one of the tributary routes. The source data for the model included an existing macro model, plus a new flow survey, an existing and a new impermeable area study, plus a questionnaire for residents.
The existing model used HydroWorks and was a macro model. Although verified as part of a wider study, this was not sufficiently detailed near the flooding area to assess prospective design options. The new short-term flow survey placed seven monitors in the area of interest, monitoring both the foul and stormwater systems. The impermeable area information came from both a standard IAS sample survey, plus an older study that showed some interesting connections from some properties into the foul system.
The residents’ survey gathered data on the historical occurrence, types and sources of flooding events as residents remembered them. The decision to monitor the stormwater system proved key to the study. During storm verification the first feature noted was a storm response that took 12h to return to its base flow in what was supposed to be a separate foul sewer. This slow drain-down suggested infiltration, so the InfoWorks Infiltration Module (IIM) was used to match the characteristics of the tail-down. The IIM uses as an input any rainfall that falls on a subcatchment and is not taken up in depression storage or lost from the surface as run-off or evaporation. This remaining rainfall is directed into the soil storage reservoir.
When the soil reaches a given saturation (the percolation threshold), water starts to percolate out of the reservoir. A proportion of this percolation flow infiltrates directly into the sewer network, while the remainder penetrates deeper to feed the groundwater storage reservoir. When the groundwater storage reservoir reaches a particular threshold water loss due to baseflow occurs.
When the groundwater level reaches a further infiltration threshold, groundwater infiltration into the sewer network occurs. Because the IIM is a simplified representation of physical processes, it requires some degree of calibration – various coefficients can be changed to modify the volumes and rates of these various flow processes. To verify the model at each of the flow monitor locations, infiltration flow from the soil reservoir was included in the model. Coefficients were chosen to calibrate the model to the observed tail-down while remaining self-consistent, for example, for a clay soil both the porosity and the rate of percolation should probably be lower than for sandy soils. A further complication in a separate or partially separate system is the need to ensure areas are not double-counted.
Once a good match for the tail-down was achieved against the data from the flow monitor, the study team sought verification against the historical data from the questionnaires. The model was predicting minimal volumes of foul flooding – 28m3 at one property and less than 25m3 in total elsewhere.
To put this in context, the level of 25m3 is defined in WaPUG’s Code of Practice for the Hydraulic Modelling of sewer systems as the threshold level at which modelling results need to be substantiated by historical facts. Below this level is effectively considered as within acceptable error limits of modelling software. The results in the stormwater system were higher – significant flooding was predicted, including 140m3 at a location near three properties at risk.
When information from the questionnaires and the old connectivity survey was used to produce a model including overland flood routing (site visits being used to check ground levels and the flood path), the model matched the real flood events much more closely. Sensitivity analysis showed it was an overland connection from the storm into the foul system that was causing the problem, rather than run-off from an adjacent, poorly drained, permeable area finding its way into the foul system.
The optioneering process included the evaluation of a client-suggested possibility – foul storage. Of the two shaft tanks suggested, the model showed one was ineffective and the second retained the overland flow as part of the solution (designing out the flooding rather than addressing the cause of the problem). A smaller ‘diversion and storage’ solution was proposed.
Saul Bradford noted: “The new flow survey really paid off in terms of understanding the flooding mechanism and the software did a great job in providing a platform for an accurate model we could use for decision-making.”
© Faversham House Ltd 2022 edie news articles may be copied or forwarded for individual use only. No other reproduction or distribution is permitted without prior written consent.