The DWMP blog – Episode 4. Do we need models?

If you haven’t already seen the earlier episodes in this series, I suggest that you start from Episode 1 (https://tinyurl.com/DWMP-blog).

The guidance for how to deliver a DWMP set out to avoid creating “a cottage industry around modelling”.?Ironically thanks to covid 19 that is exactly what we got, with modellers working at their kitchen tables and in their spare bedrooms.?I think what the guidance really meant was to avoid modelling for the sake of modelling that did not add value to the process.?In practice this encouraged some companies to try to deliver catchment plans without using models.?

The late, great Professor Ben Yen used to say, of urban drainage models that, “all models are wrong, but some of them are useful”.?He was quoting the economist George Box, who wasn’t talking about urban drainage models.

This episode will consider the circumstances where a model is useful and where is it not.?A future episode will consider how we avoid doing modelling that doesn’t add value.?I am discussing here the use of hydraulic and water quality simulation models.?Asset health and deterioration models are a different case and may be the subject of a future episode.

The alternative to using a model is to base your decisions on observations of what has happened to the system in the past.?There are two situations where this does not work.

First, if the process has random components so that the situation or combination of situations that you want to design for hasn’t been observed.?Rainfall is inherently random, so for any system taking significant rainfall runoff you are unlikely to have observed the 1:30 annual probability (or larger) event that you want to design for.

Second, if the future is going to be different from the past.?So, any system affected by climate change or significant development or increased paved area will need a model.

So, the situations where we do not need a model are foul drainage systems taking little or no runoff, without significant existing problems and with little future development.?That is likely to be small rural communities.?Everywhere else a model is not just useful but probably essential.

The DWMP guidance included a Risk Based Screening step to determine and document if a DWMP (and associated modelling) was required for each catchment.?This was scheduled to take 7 months to come to the same conclusion; that it was required for all except small rural communities with separate foul drainage.?The time and effort spent on working out how to cope without models would have been better spent on developing methods of quickly and efficiently building models of those catchments that did not have them.

So how do we build models more quickly and efficiently.?This was covered in an UKWIR report that sadly seems to be no longer available (see reference below).?

This considered the four steps of current practice to build an urban drainage simulation model.

• Model build – collating and checking existing data to build the initial draft model.
• Short term verification – carrying out a blanket survey of sewer flows over about 12 weeks to record the response to small rainfall events and adjusting the model to give the correct response.
• Long term verification – comparing the model response to existing data of flows at monitor points, pumping stations and treatment works.
• Historical verification – comparison of the predictions of flooding and overflow spill with past experience.

The report quoted the typical percentage of the costs for each of these stages to produce a robust model as shown below.

Short term verification is the most expensive part of producing a robust model because of its blanket coverage.?However, much of the benefit that it brings is identifying errors in the asset and catchment data that could have been identified beforehand.?

Improved practice would be to put spend more effort into improving the basic data to eliminate these errors and to carry out short term verification as the final step for those areas of the model that give poor results for the long term and historical verification.

The four steps then become:

• Data improvement
• Historical verification
• Long term verification
• Short term verification

Automation of the data improvement step using AI tools could give further savings in time and cost and further improve model robustness.

Summary

Drainage planners need to fully accept the benefits of models and focus on how best to produce them rather than on how to manage without them.?Even with a new approach to modelling it takes time to produce a stock of robust models – so work needs to start now to be ready for the next cycle of DWMPs, which are due for completion in 2027.

Reference

UKWIR SW870 Overcoming the reliance on short-term flow surveys to develop and verify sewer network models.?2015