Your Water Network Loses 35% of Its Supply. GIS Can Find Where, If the Valve Records Are Right

Your Water Network Loses 35% of Its Supply. GIS Can Find Where, If the Valve Records Are Right

The annual report for the utility indicates that 35 percent of all water pumped into the distribution system is lost, unbilled. The Board of Directors acknowledges this statement. One director then asks what should be the next logical question – where is the water disappearing? And the meeting falls silent.

The number is based on reality. There are several studies that put the average loss of water globally around 29.5 percent; there are countries with losses as low as four percent (Singapore) and others as high as eighty-three percent (Armenia). However, a percentage is an overall total. It gives you an idea of the size of the issue, but it does not tell you where to begin digging up streets.

This is when GIS is typically called upon to help solve a problem, and when it typically fails to meet expectations. People want to see a map showing each of the leaks; what they receive instead is a map showing the pipe system.

Coming back to the main topic: GIS doesn't "measure" water. It links existing measurements (e.g., those taken by flow meters, pressure loggers and field crews) with locations. However, this is only possible when the network data below is accurate. Unfortunately, in many networks it isn't; the single greatest reason why it isn't is a valve.

A Loss Percentage Is Not One Number

Real losses and apparent losses

The loss of water by water systems can be broken down into two types. One type is referred to as "real" losses. These are actual losses due to physical events such as leaks, breaks and overflows. The second type of loss is referred to as "apparent" losses. Apparent losses are paper-based losses. The water was consumed, however it did not result in revenue for the water system due to issues related to customer meter inaccuracies, data processing errors or theft.

The amount of non-revenue water is greater than both real and apparent losses. This category includes water that was used for a purpose other than billing (such as washing down fire hydrants), but it was authorized to be used in this manner. These types of usage are typically a very small portion of the overall water usage.

Why is it important to make a distinction between "real" losses and "apparent" losses when creating a map? The reason is that the solution for each type of loss is entirely different. For example, if a utility has actual water losses (real), then they would need to hire a crew and repair the problem. On the other hand, if a utility is experiencing apparent water losses, then they may need to replace the meters and/or perform an audit of their billing system. In addition, if a large percentage of a utility's total water loss (such as 35 percent) is due to meters under-reading, then conducting a campaign to detect leaks would yield little results, and people would be left wondering why.

Why the percentage is a weak yardstick

The percentage will move when consumption moves, even if the pipes haven't changed. This is why the AWWA water audit method asks for volume, value and validity instead: how much was lost, what it cost and how much confidence do you have in the input? Keep that third one in mind. It's the common thread throughout the remainder of this article.

The Map Shows Pipes, Not Where the Water Goes

Valves: the records nobody trusts

Now we get to the fun stuff. The assumption behind a water network model is that all valves are located exactly as shown on the GIS system and are in the same condition as indicated by the GIS. However, many field crews have learned that this is rarely the case.

One such UK instance illustrates this point well. In a sluice-valve survey of one UK water zone, 3,433 valves were physically surveyed, and 487 of them were not listed in the GIS system. In addition to those 487 missing valves, there were 1,229 total anomalies in either the GIS or hydraulic systems; approximately 42% of the surveyed items. As a result of correcting the data records and redesigning the meter-based areas within the zone, the reported loss of water due to leakage decreased from 5.40 megaliters per day to 3.46 megaliters per day. Please note that while the information provided in the previous paragraph is based on actual events, it is based on a project completed in 2007 by the contractor that performed the work, therefore, use it as an illustration rather than a benchmark.

The current state of a valve is as important as the position it occupies. According to HOFOR in Copenhagen, boundary valves can be left open after being used for repairs. If a valve remains open while a water balance is performed, water flows through the boundary and the estimated loss due to leakage will be incorrect. HOFOR has resolved this issue by monitoring certain valves and adding their status to the Geographic Information System (GIS). When a boundary valve is found to be open, HOFOR will temporarily suspend the calculation until the boundary valve has been closed. This example was provided by a valve manufacturer, however, this problem does exist.

A Dutch pilot study also stumbled upon this discovery. When researchers were conducting tests in a small isolated area, they realized that the primary supply line valve was closed, and two other boundary valves had leaks.

Missing pipes, wrong connections and elevations

Valves aren’t the only issue. The way we build our hydraulic models is based upon how we think we know what is out there – what we can see on the computer (i.e. valves, elevation, connections) and what we don’t have complete information for – such as missing pipes or incorrect pipe connections. If a pipe appears to be connected in your model, yet it isn’t in reality, then your analysis will follow the incorrect path.

We have previously addressed the issue of positional confidence through "The Pipe Isn't Where the Map Says" and connectivity issues through "The Map Says the Road Connects." Elevational data is equally relevant here. Since elevational data is used to calculate pressures, a poor quality elevational model will create problems further down the line (as we explained in DEM quality and flood risk accuracy).

Cut the Network into DMAs and Read the Night

What is a district metered area?

The concept of a district metered area (DMA), which is an area that is sealed off from the rest of the network and water enters into the area through one or more metered entry points, allows for the measurement of water entering the area. This measurement can then be compared to customer usage within that same area. Any discrepancies between these measurements are indicative of potential problems. In addition, how a system is designed can have a significant impact on its performance. For example, the redesign shown above in the United Kingdom, separated the main trunks of the system from the smaller distribution networks. The redesign also limited the number of recording meters used within each zone, as well as capping the maximum number of properties that could be included within a single zone. Zones that include too many properties are very difficult to read.

What is minimum night flow?

Think about 3:00 am. At this time almost no one has their water turned on. Therefore, what continues to flow through the DMA is mostly leak, with some small portion being legitimate use of water by consumers at night. During the "minimum flow" period, the amount of leakage equals the total amount of water that enters the DMA minus the total amount of water used by customers during the night.

Here's a simple example, with hypothetical values, to illustrate how this works. Let's assume a DMA has 600 connections, each of which is drawing approximately 0.002 liters/second during the night. Thus, customers are consuming roughly 1.2 L/s. At the same time, the DMA inlet meter shows a reading of 8 L/s. This indicates there is an excess of about 6.8 L/s, which appears to be due to leakage. If this excess remained constant throughout the entire day (24 hours), it would total approximately 590 cubic meters per day. However, it doesn't, as we'll explain next.

What makes this scenario different? Suppose a boundary valve has been left partially open and there is 3 L/s flowing out of the current DMA and into the neighboring DMA. In this case, the actual leak rate within the current DMA is closer to 3.8 L/s. Therefore, the first value that we determined using the previous formula (6.8 L/s) was an over-estimation of approximately 80% of the true leak rate, and our survey teams will likely use a great deal of time looking for leaks that do not exist.

A second weakness exists as well. Your results will depend upon how accurately you can determine what constitutes legitimate night-time usage. A published study involving 14 DMAs (district metered areas) reached the same conclusion; your estimate of the amount of leakage will be dependent upon this determination. Where possible, use smart meter data to assist you in making this assessment.

Pressure changes the answer

Pressure drives leakage, and therefore leakage varies over time. The engineer uses the FAVAD approach to define this variation using an exponent designated by N1. In theory, N1 will vary from approximately 0.5 (for a fixed area leak) to as high as 1.5 (for a leak that opens wider with increasing pressure).

The N1 value of South Africa's Pressure Management Zones was approximately -0.34 to +2.20, with other possible causes for unusual readings being an open or leaking boundary valve. Therefore, if a Zone’s pressure-leakage performance appears unusual, examine its valves first, rather than assume that the problem lies with the model.

Where GIS Earns Its Place

Joining the layers

At this point, we can now discuss GIS appropriately. To achieve this, you will need to combine the DMA boundary; a validated valve layer that includes the recorded condition of every valve; inflow time series; connection count and customer meter point; pipe age, material, and burst history; as well as pressure logger locations into a single location. After combining these layers, you can then assess zone ranking based upon night flow rate per connection or per kilometer of mains. Similar units are used during water audits since raw volume alone provides limited information regarding the size of the associated network.

Can satellites find water leaks?

Sometimes, as a first step. Satellite radar can point out areas where underground pipe leaks could exist, and it is considered a pre-location technique to use in conjunction with acoustic methods, but not as a replacement for them. Vendor claims about high efficiency should be taken with caution. An independent cost analysis of acoustic and satellite surveys indicated that survey expenses exceeded water savings after the initial survey was completed. Since money is an issue, we can rank the techniques based upon their ability to identify issues at a reasonable expense. For additional information regarding satellite data, please refer to our remote sensing service options.

From Percentage to Repair List

Leak run time

A leak has three stages: awareness of the problem, finding the exact location of the leak and then fixing it. The longer a leak runs the greater the loss experienced by the company over a year. A tiny leak that runs for two years may cause more damage than a large burst that was repaired within 24 hours. Therefore, make sure you know exactly where your leaks are, but also try to estimate how long each of those leaks have been running.

Not every leak is worth fixing first

The minimum total cost of water loss through leakage (i.e. leak detection and repair) plus the value of the lost water is called the economic level of leakage. If you go beyond this level, then you will be spending more money on the process of detecting/repairing leaks than you are saving by reducing the amount of water being lost. Similarly, regulators in California have used an economic model which compares the cost of detecting/repairing leaks to the value of water saved, as well.

Ranking a suspected area based on four criteria: amount of product lost, how long the problem existed, cost of repair, and the impact of waiting to repair or turning off supply to repair. An intermediate size leak in a pipe that feeds a hospital could be ranked higher than a larger leak in a quiet, dead-end street.

A Monthly Loop You Can Actually Run

The above is an example of a workflow that most utility companies could use. It's simple, and that's exactly what I want to achieve with it.

  1. Verify the boundary valves and update the GIS with their real status.
  2. Pull the minimum night flow for each DMA and subtract legitimate night use.
  3. Check pressure behaviour. Treat strange results as a valve check first.
  4. Select the roads in the worst areas by overlaying the pipes' age, composition, burst histories, and pressures.
  5. Locate in advance using an acoustic survey, or satellite data if that is more economically feasible. Then confirm in the field.
  6. Rank the leaks that have been verified in order of size, duration of operation, expense to fix, and how they affected operations; then fix each leak.
  7. Update the records and run it all again next month.

Regulators have the same idea. The Commission of Europe is expected to establish an Infrastructure Leakage Index (or some other type of) leakage threshold on or before January 12th, 2028, and California's Water Loss Performance Standards were implemented in 2023. Keep up to date with your regional regulations, as they can alter. The silent risk is the quality of data: a study done in Slovenia showed that just measurement errors can potentially move more than half of all evaluated systems into a different performance category.

What to Take Away

A loss percentage is just the beginning of a process, not the final product. It is important to break this down into actual and perceived losses prior to sending personnel out to perform work.

Valve records determine whether your DMA and night-flow numbers are of value. Make sure you have verified these numbers, and that they are always up to date.

After that, use volume, run-time, cost and service impact to create a repair list based on your actual budget.

My advice for beginners is straightforward: Pick one DMA, locate all the boundary valves, and watch how the numbers vary. Technology follows. If you need your data cleaned up, our GIS team can do just that type of work.

Back to blog

Leave a comment