39 Modelling flow that shouldn’t get into drainage systems
Martin Osborne
Water industry strategic advisor, asset planner and drainage expert. Winner of the 2023 WaPUG Prize for contributions to the development of urban drainage practice
A geeky modelling topic for this episode of the blog, mostly to do with InfoWorks ICM.
Background
The question is; how do we model the flow that shouldn’t be getting into our drainage systems and we therefore don’t really know how it is getting there?
In the UK we tend to call this “slow response” flow.?Elsewhere it is often termed Inflow and Infiltration (I&I) or Rainfall Derived I&I (RDII).?There is a slight distinction between slow response and I&I.?The UK has many combined sewerage systems that are intended to take direct runoff from roads, roofs and other impervious surfaces.?Some inflow is therefore already accounted for.?The slow response is the rest, from sources that we don’t know, which comes in sometime after the rainfall.?In jurisdictions with mostly separate sewers, I&I includes the direct runoff from impervious surfaces that shouldn’t be getting into the system.?Some I&I is therefore a quick response to rainfall whereas some is again slow response.
There will be two main sources; runoff across the ground, particularly across permeable surfaces, and infiltration through the ground into the system.?The infiltration can be either intentional through land drains and foundation drains that connect to the system or unintentional through cracks and other defects in the system.?Foundation drains are uncommon in the UK, even for properties with basements, but they are common in North America and elsewhere.
Both types of sources will tend to have an initial delay before there is any flow into the system.?For runoff this is to fill up depressions on the ground surface, for infiltration it is to saturate the soil.?After that they will behave differently.?For runoff the flow will tend to increase linearly with the rainfall.?Double the rainfall will give double the flow or even more as more areas start to contribute.?For infiltration the depth of saturated soil will increase roughly linearly with the rainfall, or perhaps a bit less as more drains away through the ground.?However, the flow into the system is akin to the flow through an orifice and so increases with the square root of the depth of water in the soil.?In the real world we probably get a mixture of mechanisms with the relationship with rainfall somewhere between the extremes.
As we do not really know where this flow is coming from we cannot model it from first principles based on the characteristics of the catchment.?We therefore have to calibrate the slow response model against measured flows.
The combination of not knowing the relationship with rainfall and having to calibrate the flows causes a problem as shown in the graph below.
We are carrying out the calibration using fairly small rainfall events (Point A) and can calibrate all three relationships to give us the same value.?However, we want to use the model for large events (Point B), where different assumptions give very different answers.?So how do we know which relationship to use?
There are three things that we can do to help us.
Available models
What models do we have available and which relationship does each of them follow?
RTK model
The RTK model is used a lot in North America.?The flow volume is a constant proportion of the rainfall (defined by R) and the attenuation and runoff speed is a simple triangular hydrograph defined by a time to peak T and a recession time T*K.
It is often used as a group of 3 models with different parameters to represent different flow speeds:
领英推荐
It should be obvious that R1 + R2 + R3 should be less than 1 otherwise we will generate more flow than there is rainfall.
A similar model can be created by using a fixed runoff coefficient to represent R and using the standard Wallingford runoff routing model with a range of runoff routing values for the different components.?I recommend using relative values to multiply the default time factor, with the rule of thumb that a value of 1 gives a time to peak of about 4 minutes.
If the infiltration flows do have a characteristic similar to orifice flow, then these models will overestimate the flow from the infiltration components in large events.
New UK runoff model
One common practice is to represent slow response inflow by adding large permeable areas and using the New UK runoff model but with a high flow routing factor to give much slower response.?This model increases the inflow more than linearly with rainfall as the catchment gets wetter and is therefore likely to overestimate runoff in large storms.?This has been a constant problem with this approach and all sorts of fudges have been developed including putting the inflow through an orifice control to limit the maximum flow or discarding the extra runoff in larger events.?All of these fudges are based on guesswork and are unlikely to be realistic.
Ground infiltration model (GIM)
The GIM gives two component models that can be used separately or together.?The model concept is shown in the diagram below.
Unfortunately this diagram is wrong and the model concept is actually as the diagram below.
There are several things wrong with this model:
The good news is that the ground store model can give a representation of orifice flow and so avoid the problem of too much runoff in large events.
Recommendations
Senior Technical Support Engineer at Innovyze
2 年Just in time to add a few pictures to show how to model GIM in #InfoWorksICM, https://www.dhirubhai.net/posts/melmeng_infoworks-icm-ground-infiltration-cheatsheet-activity-7031712329599778816-1gC2?utm_source=share&utm_medium=member_desktop
Cycling obsessed writer in Amsterdam
2 年Thanks for another great article. As Bob noted, we'll be changing the image that is associated with that help file. Thanks for the great feedback. ??
Deputy Executive Manager - Planning, Design & Construction Administration- TRA
2 年RTK was developed to discect flow survey data. It’s not a bad approach for that purpose. As a model input to create wet weather flows it’s iffy at best. There’s much better approaches possible inside ICM.
Client Services Manager at Autodesk
2 年This publication: https://ukwir.org/water-research-reports-publications-viewer/5429e7c0-b732-4af2-9648-e0a53eb7ac3d covers a lot of the same ground and answers many of the issues raised, including how to effectively switch off either the soil store or ground store. It includes the revised diagram. I'll put my hand up to that one. I drew it in around 2002, so it's had a good run. I'll ask for the ICM help to be changed to reflect the fact the flow is split after leaving the soil store. Any changes to the GIM would have to be additional options as we would want to maintain the results from existing models.
at
2 年Got to love the modelling geek discussions! Great article Martin. We face similar discussions about the pros and cons of the various slow response and GI runoff representations. As you say, the key is to keep it simple and currently all tools are essentially a tool to calibrate against observed tails from flow data. The real questions come about the ‘capping’ of this inflow when utilising hydraulic models for higher return period design storms. Interestingly I know many WaSCs currently lean towards only utilising the soil store, purely for simplicity and the ability of achieving a ‘fit’ before adding additional complexity.