The DWMP blog – Episode 9. Minimising modelling effort

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”.?I think what the guidance meant by this was to avoid modelling for the sake of modelling that did not add value to the process.

So did the guidance deliver on this ambition? ?Or did it recommend unnecessary modelling that doesn’t contribute to developing long term investment plans.?

Fully modelling the performance of a sewerage system is computationally intensive.?For each future scenario we need to analyse a range of probabilities of events to define the risk envelope.?For each probability we need to look at a range of event durations to fully test all parts of the system – those that fail to perform under short high intensity events and those that fail under long low intensity high volume events.

To predict storm overflow performance, we normally analyse a long period of typical rainfall with a duration of one year to ten years or more.

So, the questions are:

• Can we minimise the number of times that we have to analyse each scenario?
• Can we minimise the number of event probabilities that we analyse?
• Is there a simpler way to analyse overflow performance?

Planning scenarios

For the Baseline Risk and Vulnerability Assessment (BRAVA) stage of the DWMP the guidance recommends that scenarios are analysed for; current day, 5 years’ time, 10 years’ time and 25 years’ time.?In Episode 6 I considered the medium-term and long-term scenarios and how to deal with uncertainty using at most 3 scenarios.?But what about the short-term current day and 5-year scenarios?

The current and past situation does need to be assessed as part of the testing of models, so that we know that they are valid in predicting the performance of the system.?However, this should be done when preparing the models and using conditions appropriate for this test, not necessarily those appropriate for defining performance.

Analysis of the current day situation does not contribute to developing options for future planning.?Even if the current day situation shows service level failures, those failures will have changed by the time we get to implement any options in a few years’ time.?The one exception would be failures caused by maintenance requirements; but these would only be included in the model if we already new about them, so it does not require model results to show that action is required.

There is a potential purpose of assessing the current day situation if we need to implement a policy of no-deterioration of service level from the present day even if this would not breach the target level.?However, this is not explicitly part of the DWMP requirements.?I may explore no-deterioration policies more in a future blog.?

A five-year planning horizon is appropriate for the time to deliver most investments to improve the system.?However, these improvements would not just be designed for the time that they are implemented but for a longer term of at least ten years from now.?So the 5 year horizon is also not useful.

I therefore suggest that we skip the current day and five-year horizons and go straight to the medium-term and long-term horizons that I set out in Episode 6.

We can make a further reduction as we are unlikely to get to the 25-year plus planning horizon still with the current sewerage network; improvements will have been made in the interim.?There is therefor little point in assessing this combination of conditions.?We should first develop options for the medium to long term and then assess how they perform in the longer term.

These changes would significantly reduce the amount of modelling required to deliver the initial assessment of risks where investment is required.

Probabilities / return periods for flooding

The DWMP guidance refers to the Capacity Assessment Framework for details of what annual probabilities of events should be assessed for flooding.?This recommends 1:2, 1:5, 1:10 and 1:30 annual probabilities.?The DWMP guidance adds 1:50 to meet the Ofwat property flood risk measure.?But how many of these are really needed?

The Ofwat measure of 1:50 will become the standard aspirational target and should be used.?1:30 is typically the current target standards and so is also required.?1:10 is typically the trigger for unacceptable performance that requires improvement action.?The more frequent events of 1:2 and 1:5 do not add anything.?If a property with a 1:10 annual probability of flooding requires short term investment action to resolve, then knowing that it has a more frequent probability does not add urgency or importance to the solution.?One exception could be for property level protection measures; but these should probably be considered for all properties that have already suffered flooding.?A property floods not only because of lack of capacity in the sewerage system, but also because it is low lying compared to the landform and its neighbours.?Sewerage improvements will reduce its risk but not eliminate it; the property will still flood given a large enough storm.?So, protection measures are a sensible precaution.

Some companies have also assessed rarer events including 1:100 and 1:200 annual probability events.?This is useful to demonstrate resilience, but for these events the overland flow paths are as important as the sewerage system as restrictions on inlet capacity prevent all of the flow getting into the real system but this is generally not represented in the modelling.?The existing maps of flood risk from surface water that use a simplified representation of the sewerage system are therefore a more appropriate tool for this.

Spill frequency assessment

Minimising the modelling effort for assessing spill frequency is more of a challenge.?The standard approach is to use a continuous record of realistic rainfall with a duration of 1 year, 10 years or 20 years.?This is computationally intensive but is appropriate to demonstrate the suitability of potentially expensive investment plans.?However simpler approaches using a few synthetic storms representative of a long timeseries may be more appropriate for identifying potential level of service failures and initial development of options.

Summary

The DWMP guidance set out to avoid modelling for the sake of modelling, but inadvertently recommended modelling that although it might add to our understanding of current or short-term system performance, does not contribute to developing long term plans.?A reduction in required modelling is not only possible but required to make full coverage of robust DWMPs possible.