Dummy's Guide to Apparent Water Losses - The Southern African Edition (1 of a few)

This week some of us had the pleasure of being at the refreshing IWA Water Loss 2018 Conference in Cape Town, South Africa. Congratulations to the organizing committee for putting together an informative and well organised conference! The conference and various conversations I had inspired me to share my thoughts, perspectives and work on apparent water losses, all in a bid to educate all of us, myself included. These will be shared over the coming weeks and I invite one and all to contribute to the discussion.

What are apparent losses? In simple terms, its all water that is consumed but is not recorded. This therefore implies a connection, with apparent losses being all losses after the connection that are not recorded by a metering device (present or absent) . These losses include;

1. Unauthorized/illegal Consumption - over and above the obvious, these unfortunately include water consumed by yet to registered residents of housing projects of developers and the Department of Human Settlements, unless of course you find innovative ways of measuring or dealing with such usage. My interactions with various people who have taken the liberty to do connection verifications in South Africa is that there is an average occurrence of between 2 - 5% of properties surveyed, with the exception of where developers and Human Settlements are involved as an entire suburb can be "unauthorized".
2. Data Acquisition Errors like when your most trusted meter readers have finger trouble (including literally manufacturing readings) and when your billing department/system has the habit of inserting random figures and factors into your readings. Estimates depend on how creative these lovely human beings can be and are therefore varied. In the long run, these tend to "self-correct" as at some point as the correct meter reading may be eventually found.
3. Errors on Estimates of Unmetered Consumption - this is a relatively new term that was introduced with the new guidelines that can be found here. To be honest, I still can't get my head around the estimate of estimates concept, for that what it is to me. If I don't know a value and I estimate it, how do I know what the error of the estimate is? Educate me someone...
4. Water generated in pipes - this is obviously not true but all who claim to have NRW in the single digits (2%, 3%, 4%ish) could possibly be having this phenomenon. I was thoroughly delighted to meet an esteemed professor this week who shares a similar view.
5. Meter Under-registration Errors - literature estimates this to be the highest contributor of apparent losses and shall be the main focus of these series. Let's dive in...

Now that the basics are out of the way, the rest of our conversations will focus on the metering device, and in particular mechanical meters that are largely used in the region, if not almost exclusively.

Meters are not 100% accurate, even when brand new. And this inaccuracy gets worse with age and volume.

Below is an error curve (blue curve from actual test results) for a relatively new meter, together with the error envelope (black lines). The error curve represents the accuracy of the meter at specific flow rates- the accuracy is clearly not constant, which is the case with some meters (non-mechanical). The error envelope represents the allowable limit of the error at each flow rate when the meter is new. When the meter is in use, the error envelope moves from 5% to 8% and from 2% to 3.5% which is not minor (and it is the law). Keep that in mind! You will also note that the error curve is made from more points than the legislated minimum flow rate, qmin, the transitional flow rate, qt, the permanent flow rate qp and overload flow rate, qs.

The actual accuracy of the meter cannot be read off the error curve, the specification chart or the error envelope. It all depends on the consumption of the consumer - not the average consumption. Consumers do not consume in averages but at specific volumes and flow rates from the various water fixtures. If the flow is at 7.5kl/hr, the accuracy above will be about -7% while if the flow rate is about 50kl/hr, the accuracy is about +2% (yes, over reading) etc. Without this information, any estimates are meaningless. While I have met people who think they can guess the consumption profiles by pondering how long it takes to fill a kettle, it is folly to believe too much in your abilities to do that quite well.

Consumption profiles/patterns are another prerequisite to determining water meter accuracy

Shown below is the consumption profile of domestic consumers in Johannesburg. We can pretend, for now, that these are for just one consumer. What is shown here is the proportion of the consumption at particular ranges of flow rates. For example, 5% of the flow happens between 0 - 7l/hr, 9% between 7 - 15l/hr and so forth.

Given both the meter error curve (which is not static but changes with meter age and registered volume) and the consumption profile (which should not be that dynamic) one can then estimate meter accuracy using what is known as the weighted accuracy method. This basically adds the products of the proportion of the flow and the accuracy at that particular flow rate. In our example above the simplified calculation will be 5% * -7% + 9% * -2% + ... = -4.2% (with a good measure of reasonable assumptions). You may download a worksheet that was shared at the conference to calculate the weighted error from here.

In essence, this "newish" meter under reads the volume of water by 4% and your apparent loss due to metering would be 4% of the billed volume. And yes percentages are good here. Remember that this example is for a particular meter model and a particular consumer... If your consumer only consumed at an average of 120l/hr the error of the meter would be a positive 1.52%, which would be nice for the utility but its rarely the case. I might not have mentioned that the meter above is a Class C 15mm meter, for those using Class B (not that it's wrong), your error is higher. We will not talk about those of you who use 20mm and 25mm meters for this same consumer...

This concludes the first post. In summary, the weighted error methodology is the best way of calculating apparent water losses - that doesn't mean its easy though and we will unravel some of the intricacies later. It also provides some insight as to what is required to be able to talk meaningfully about meters, meter selection and all things metering. You need to know your meters (think error curves) and your consumers (think consumption profiles). If you don't have these, you are probably still on the periphery of the real stuff.