The DWMP blog 34 – Revisiting present and future uncertainty

UKWIR has just published a report called Future Asset Planning - Scenarios, Frameworks & Measures: Final Report.?You can find a copy here https://ukwir.org/future-asset-planning-scenarios-framework-and-measures-final-report.

The report focusses on what it calls “Asset Health” and defines this as the probability of asset failure but without full consideration of the consequences of that failure.?It could therefore be described as a pre-risk-based approach.?I am therefore not a fan.

However, the report does include an interesting discussion of planning for current and future uncertainty.?I discussed this topic in relation to drainage and wastewater management planning in Episode 6 of the blog (https://tinyurl.com/yc3m8a23) but I thought that it would be useful to revisit it with these new ideas.

The UKWIR report identifies that there are a range of approaches applicable to different combinations of uncertainty and complexity of the problem.?(I want to say “of the risk”, but the report is not risk based).?Unfortunately, it does not define “complexity” but I think that it refers to the number of variables where the values are uncertain.?The report then represents this range of uncertainty and complexity as a 2 by 2 matrix.?

• Low complexity-low uncertainty
• Low complexity-high uncertainty
• High complexity-low uncertainty
• High complexity-high uncertainty

However, the detail in the report does not align with this grouping.?The report actually defines three levels of uncertainty:

1. Low uncertainty – which can be ignored and fixed values used.
2. Probabilistic uncertainty – which can be represented by a probability distribution where central values have a high probability and extreme values have low probabilities.
3. Deep uncertainty – where the likely range of values can be estimated but the probabilities of each value cannot.

The approaches to these three levels of uncertainty are different.?

The report set out two different approaches to the probabilistic uncertainty group, but I think that there are three.

1. Sample a large number of random values from the probability distribution and average the results from each.
2. Sample a smaller number of semi random values from the probability distribution with the selection controlled to have an equal number from each part of the distribution.
3. Select a small number of non-random values from the probability distribution and combine the results for each with the probability of the value.

To put this into the context of uncertainty in current rainfall events, 1) is like running a 50 year timeseries; 2) is like running events selected from that timeseries and 3) is like running selected design storms.

For deep uncertainty the recommended approach is to develop plausible scenarios of the range of values that could occur.?I would turn the analysis on its head and use the Sherlock Homes approach of excluding the implausible range of values that could not occur.?What is left is plausible, no matter how extreme it appears.?Having defined the plausible range, a simple standard probability distribution (triangular or normal) could be fitted to it and then treated as a probabilistic uncertainty.

An extra check on the vulnerability of the system to suffering catastrophic impact in the worst-case scenario is a useful safeguard.?Similarly check that we are not wasting too much money if the best case scenario pans out.

For selecting the appropriate solutions to the problems identified by the probability analysis the report introduces a similar choice between what I have called a stepwise approach of pre-ranking the likely solutions to the problem and an optimised approach of considering all possible combinations of solutions.?(In keeping with the style of the report these are called “a priori” and “a posteriori”.?Thankfully I did study Latin for a couple of years at school so could manage to keep up.)

The report does lean towards the optimisation approach and expects this to become more widespread in the future.?I think that this is correct but there will always be a place for a simpler approach, especially for an initial assessment.

No doubt this blog has wetted your appetite for reading the full UKWIR report, but I think that I have delivered a public service by reading the report and extracting the useful bits so that you don’t have to.