The previous two articles built the theory: divide the network into District Metered Areas, read the loss from minimum night flow, and account for every drop with the water balance and the ILI. But a water balance built once a year, by hand, is a post-mortem. To actually control Non-Revenue Water you need the loss measured every night, every DMA, automatically — and a system that turns that torrent of data into a short, ranked list of where to send a crew tomorrow morning. That system is a Water Management System (WMS) such as NETBASE, and this is how it works, end to end.

1 · From a yearly audit to a living network

Manual leakage control has a fatal latency. A burst that starts the day after your annual survey can run for months before anyone notices the bill. Smart NRW management closes that gap by instrumenting the network and letting software watch it continuously. The shift is from knowing how much you lost last year to knowing which zone started leaking last night. Three capabilities make that possible[1][2]:

2 · The smart-water data chain

A WMS sits in the middle of a layered architecture. Each layer has a job, and the value of the system is only as good as the weakest one — bad data in, bad decisions out. Read it left to right: the field measures, communications carry, the WMS stores and analyses, and people and valves act — then the loop repeats[2][5].

The smart-water data chain — measure → carry → analyse → act → repeat 1 · Field flow + pressure loggers, DMA meters, AMI 2 · Comms cellular / NB-IoT SCADA / RTU scheduled + alarm 3 · WMS NETBASE historian + DMA engine 4 · Analytics night flow, alarms, leakage, pressure 5 · Action crews locate & repair; PRV control closed loop — fixing a leak and trimming pressure changes what the field measures next
Original schematic. The WMS (NETBASE) is the historian and analytics core; it is fed by field loggers/SCADA over scheduled and alarm-driven communications, and its output drives field crews and pressure control — which in turn change the next measurements.

The data sources differ in what they see and how often, and a good WMS fuses all of them onto one time-base:

SourceWhat it givesTypical interval
DMA inlet flow loggerZone inflow → night flow, the core leakage signalLog 15 min; send 1–24 h (or on alarm)
Pressure logger (AZP / critical point)Average zone pressure, for the leakage–pressure correctionLog 1–15 min
SCADA / RTU (reservoirs, pumps, PRVs)Levels, pump status, valve positionNear real-time
Customer meters / AMIBilled consumption → apparent loss, legitimate night useDaily–hourly
GISMains length, connections, DMA boundaries, asset attributesStatic, periodically synced

3 · Step one — acquisition and the unglamorous art of clean data

Every leakage figure the WMS ever produces rests on the quality of the raw signal, so the first job inside the software is data validation, not analysis. Loggers drop out, modems miss a call, a meter ices up, a clock drifts. Before night flow means anything, NETBASE-class platforms run each channel through validation, gap-filling and correction[3][5]:

Garbage in, alarms out An un-validated network throws false alarms (a logger gap looks like a burst) and misses real ones (a leaking boundary valve inflates the baseline until a true burst hides inside it). Most "the software is wrong" complaints are really data or DMA-integrity problems. Budget as much effort for data governance as for analysis.

4 · Step two — the night-flow engine

This is the analytical heart of the WMS, and it is worth understanding exactly what the software does between the raw inlet flow and the leakage number on the dashboard. It runs every DMA, every night[2][3][4].

4.1 Find the minimum night flow

Between roughly 02:00 and 04:00 legitimate demand bottoms out, so the inlet flow is dominated by leakage. The WMS picks the minimum sustained flow in that window — the minimum night flow (MNF).

4.2 Subtract legitimate night use

Not all of the MNF is loss. A residual of genuine night consumption remains — toilet cisterns, a few night-shift users, the occasional large customer. The standard estimate is an allowance per property plus any metered exceptional night users[3]:

\[ \text{NNF} = \text{MNF} - \big(N_p \times q_{\text{legit}} + Q_{\text{exceptional}}\big) \]

where NNF is the net night flow (the leakage component at night), \(N_p\) the number of properties, and \(q_{\text{legit}}\) a per-property night allowance (a common default is ≈ 1.7 L/property/hour). Exceptional users (a hospital, a factory) are measured separately and removed explicitly, not averaged in.

4.3 Convert night leakage to a daily volume

Leakage at 3 a.m. is not the daily average leakage, because pressure is highest at night and — by the FAVAD law — leakage rises with pressure. To turn the net night flow into a daily volume you apply the Night-Day Factor (NDF): the number of "night-equivalent" hours in the day once you account for the lower daytime pressure[3][4]:

\[ \text{Daily leakage} = \text{NNF} \times \text{NDF}, \qquad \text{NDF} = \sum_{h=1}^{24}\left(\frac{P_h}{P_{\text{night}}}\right)^{N_1} \]

With constant pressure the NDF would be 24; because daytime pressure is lower, it is typically 20–22. This pressure correction is exactly why a co-located pressure logger matters: without it, the WMS over-states daily leakage by assuming night conditions all day.

Worked example — one DMA, one night A 1,500-property DMA logs MNF = 12.0 m³/h. Legitimate night use = 1,500 × 1.7 L/prop·h = 2.55 m³/h, so NNF = 9.45 m³/h. With an AZP-derived NDF = 20, daily leakage ≈ 9.45 × 20 = 189 m³/day — about 126 L/property/day. Track that number nightly and a sudden jump is a new burst; a slow creep is rising background leakage telling you the zone is due for attention.

5 · Interactive: nightly monitoring & the burst alarm

This is the view an engineer actually lives in: one DMA's minimum night flow, night after night. The WMS learns the zone's baseline and sets an alarm threshold above it; when a burst steps the night flow up across that line, the zone alarms and lands on tomorrow's work list. Set the background leakage, drop in a burst, and tune how sensitive the alarm is — then watch how fast it fires and how much water escapes before it does.

One DMA's minimum night flow over 30 nights
Each point is a night's MNF. The dashed line is the learned baseline; the red line is the alarm threshold (baseline + sensitivity). A burst starts on night 15 and steps every later night up. The alarm fires the first night MNF crosses the threshold.
The DMA's baseline minimum night flow before the burst.
Extra night flow added from night 15 onward.
Threshold above baseline. Tighter = earlier alarms but more false ones.
Alarm threshold
13 m³/h
Time to alarm
1 night
Lost before alarm
Status

A 6 m³/h burst on a 10 m³/h baseline with a 3 m³/h threshold trips the alarm the very next night — days, not months, after it started. Now shrink the burst toward the threshold: it takes longer to stand out from the nightly scatter, and more water is lost before detection. Loosen the sensitivity too far and a real burst can hide in the noise; tighten it too far and ordinary variation cries wolf. Tuning that line per DMA is the daily craft of running a WMS.

6 · Interactive: how often you read the data decides how much you lose

Detecting a burst quickly is worth real water, and detection speed is capped by how often the loggers actually deliver data. A logger that records every 15 minutes but only dials home once a day cannot warn you any faster than once a day. This chart traces the water lost from the moment a burst starts until the WMS first sees it — for a traditional daily read versus the transmission interval you choose.

Cumulative water lost before a burst is detected
A burst starts at hour 0 and loses water at a constant rate. The WMS only sees it at the next data delivery. The blue marker is detection on a once-daily read; the green marker is detection at your chosen transmission interval.
Continuous loss rate of the new burst.
How often loggers deliver data (15 min … 24 h). Logging can be finer, but detection waits for delivery.
Detection delay
6 h
Lost before detection
108
Lost on daily read
432
Water saved

A 5 L/s burst loses 18 m³ every hour. Found on the next day's read, that is up to ~432 m³ gone; moved to a 6-hour delivery, ~108 m³; on 15-minute near-real-time telemetry, almost nothing. This is the business case for more frequent communications and event-driven alarming — and the reason "fast logging on alarm" modes exist on modern loggers. More data is not vanity; it is recovered water.

7 · Step three — closing the loop with pressure

Monitoring tells you where the loss is; the cheapest control acts on all of it at once. Because the WMS already holds each zone's flow and pressure, it is the natural place to drive pressure management. Modern PRV controllers accept a flow signal and modulate the outlet so the critical point sits at target for the actual demand — high at peak, low at night — and the WMS both supplies that flow signal and verifies the result by watching night flow fall[2][6]. The full mechanics — fixed, time- and flow-modulated control, and the FAVAD payback — are covered in pressure management & DMAs; here the point is that the same platform that detects the loss also commands the valve and confirms the saving, all on one set of data.

8 · Step four — deciding where to dig first

A city has hundreds of DMAs and a handful of leakage crews. The WMS turns the night-flow results into a ranked work list so the limited teams attack the zones with the most recoverable water first — the difference between current leakage and that zone's achievable level. It is a classic Pareto problem: a small number of DMAs usually hold most of the recoverable loss.

9 · Interactive: prioritising the crews

Twelve DMAs, ranked by recoverable leakage. You have only so many crews this cycle — drag the count and the WMS assigns them to the worst zones first. Watch how quickly the cumulative line climbs: tackling the top few zones captures the majority of the recoverable water, which is why ranking beats working the network in map order.

Recoverable leakage by DMA — where the crews should go
Bars are each DMA's recoverable leakage (m³/day), sorted worst-first. The green bars are the zones your crews reach this cycle. The navy line is the cumulative share of total recoverable loss captured.
Each crew tackles one DMA (worst first).
Marginal value of a recovered m³ — sets the annual benefit.
DMAs addressed
4 of 12
Water recovered
1,930 m³/d
Share of total
62 %
Annual value
k$/yr

With four crews on the worst four of twelve zones you capture about 62% of all recoverable water — the Pareto pay-off. Adding crews five and six helps far less per crew: the curve flattens. The WMS makes this visible so management funds the crews that matter and stops spreading effort evenly across zones that barely leak.

10 · Don't forget apparent loss — AMI and the meter

Everything so far attacks real losses. The other half of NRW — apparent loss — lives in the same data platform. Advanced Metering Infrastructure (AMI) streams customer consumption into the WMS, which lets it: size the legitimate-night-use allowance from real night consumption instead of a default; flag stopped or under-registering meters by their profile; detect zero-consumption accounts that should not be zero; and tighten the water balance because billed volume is now measured continuously, not estimated. Real and apparent loss share one system of record[1][7].

11 · Inside the software — getting accurate results

The platform is only as good as how it is set up. The settings and habits that separate a trustworthy WMS deployment from a noisy one[3][5]:

The operating rhythm Each morning the WMS presents the overnight exceptions: DMAs whose night flow stepped up (probable bursts) and those creeping up (rising background leakage). Crews are dispatched worst-first; pressure is modulated where the night over-pressure earns it; repairs and pressure changes feed back into the next night's data. NRW control becomes a daily loop, not an annual campaign.

12 · The implementation checklist

The one-line summary A WMS like NETBASE turns a network of loggers into a nightly verdict: it validates the data, computes each DMA's leakage from the net night flow with a measured pressure correction, alarms the zones that stepped up, ranks them by recoverable water, and drives the pressure that suppresses all of it — converting raw telemetry into the shortest possible list of where to dig tomorrow.

References & further reading

  1. American Water Works Association. Manual M36 — Water Audits and Loss Control Programs (water balance, NRW components, real vs apparent loss).
  2. Farley, M. & Trow, S. Losses in Water Distribution Networks: A Practitioner's Guide to Assessment, Monitoring and Control. IWA Publishing — DMA monitoring and night-flow analysis.
  3. UK Water Industry / WRc. Managing Leakage series — esp. Report E: Interpreting Measured Night Flows (net night flow, legitimate night use, hour-to-day / night-day factor).
  4. Lambert, A.O. & Morrison, J.A.E. Recent developments in pressure–leakage relationships and night-day factors. IWA Water Loss conferences (FAVAD / N1).
  5. Ovarro (formerly Servelec Technologies / Technolog). NETBASE Water Network Management Software — product and application documentation (DMA management, night-line analysis, alarms, pressure management, reporting).
  6. Thornton, J., Sturm, R. & Kunkel, G. Water Loss Control. McGraw-Hill — active leakage control, intervention and pressure management.
  7. IWA / smart-water references on AMI, telemetry and data validation for water utilities; AWWA standards for metering accuracy.
  8. EPANET (US EPA); Bentley OpenFlows WaterGEMS; Autodesk/Innovyze InfoWorks WS Pro — hydraulic model integration and demand calibration; Esri ArcGIS for GIS/asset data.
#NonRevenueWater #NRW #SmartWater #NETBASE #WMS #WaterManagementSystem #LeakageControl #DMA #MinimumNightFlow #NetNightFlow #NightDayFactor #BurstDetection #SCADA #Telemetry #DataLoggers #AMI #PressureManagement #GIS #HydraulicModel #WaterLoss #PipelineDesign #WaterInfrastructure