Argus EMS · 05.10.2026 · 6 min read
A water leak makes no noise. A leaking underground joint, a constantly running cistern or a garden valve left open can run unnoticed for weeks and usually reveals itself only when the bill arrives. A monthly index reading tells you how much water was lost, but not when the loss started. The minimum night flow (MNF) method looks at the hours when consumption is naturally at its lowest and makes a leak visible long before the invoice.
MNF is the lowest flow a meter records during the quiet hours of the night. The concept comes from the DMA (District Metered Area) approach, in which water utilities divide the network into zones measured by inlet meters. After midnight most users in a DMA are asleep, so the flow measured in those hours is the sum of legitimate night use and leakage. Once legitimate use is estimated and subtracted from MNF, the remainder is the loss indicator for that zone. The same logic applies to a building, a residential estate or a hotel: the main meter is the zone inlet meter and the sub-meters are the sub-zones.
The strength of the method comes from comparison. A single night's MNF says little on its own; the real information is whether the night baseline has risen persistently relative to the meter's own history. A developing leak persists night after night at the same or a growing level, while one-off night use disappears the following night.
| Component | Typical example | How to tell it apart |
|---|---|---|
| Scheduled legitimate use | Irrigation, tank and booster set filling, night cleaning | Starts and stops at set times; matches a schedule |
| Continuous legitimate use | Cooling tower make-up, circulation loop top-up, process water | Varies with load or season |
| Cistern and fixture leaks | Toilet cistern, dripping tap, faulty float valve | Low but uninterrupted; clear at sub-meter level |
| Pipework leaks | Underground line, joint or riser crack | Lifts the night baseline permanently |
| Valve or tap left open | Forgotten garden valve, open bypass | Long, uninterrupted and steady flow |
The most practical way to tell them apart is to write down every legitimate night use: once irrigation timers, tank filling hours and cooling tower make-up are recorded, each peak on the night chart can be matched to a cause. A baseline flow that has no cause and repeats every night is a suspected leak.
The quality of MNF analysis depends on the meter's ability to see low flows. On a meter with pulse output, every pulse represents a fixed volume; if the pulse value is Vp and the time between two pulses is Δt, the average flow is Q = Vp / Δt. On a meter with a large pulse value, a small leak may not produce a single pulse within the logging interval, and the chart shows zero. Mechanical meters also have a starting (minimum) flow; below it the meter does not turn at all, so a dripping fixture may never be measured.
On a residential estate the main meter never truly drops to zero at night: across a large number of apartments someone is always using water. Instead of expecting absolute zero on the main meter, you track the night's own baseline and look for a persistent rise against history. Block or riser sub-meters, on the other hand, can genuinely rest at night, so uninterrupted flow there is directly suspicious. The hours of common-area irrigation and pool top-up should be clearly defined.
In a hotel the night profile is more complex. Laundry, kitchen preparation, pool and spa top-up, night cleaning and the cooling tower can all fall into the same hours, and occupancy shifts the baseline from night to night. In these facilities the main meter alone is not enough. Monitoring kitchen, laundry and landscape lines with separate sub-meters makes it possible both to separate legitimate use and to locate the leaking zone quickly.
Argus EMS reads water meters remotely over M-Bus and Modbus and calculates consumption from the index difference. Every morning at 06:00 a night evaluation runs for each active water meter: the minimum flow in the 02:00–05:00 quiet window is compared with the baseline learned from the meter's last 21 days. The two most recent nights are excluded from learning so that a leak cannot mask its own baseline, and no alarm is raised until at least seven nights of history exist. During learning, meters fall into two groups: resting meters, whose flow drops below the noise floor in most night samples, and continuous meters such as main meters, whose flow keeps running at night.
On a resting meter, a night counts as a violation when the minimum night flow stays above the noise floor and the meter barely rests at all during the night; two consecutive nights raise a warning and three raise a CRITICAL alarm. On a continuous meter, the system calculates how many standard deviations the night minimum sits above the learned median; the default thresholds are 3σ for warning and 4σ for critical, and the threshold must be exceeded on at least two consecutive nights. A single-night spike does not raise an alarm, and an open alarm closes when the signal disappears. Resting meters are additionally checked for long uninterrupted flow within the last day of data, which points to a valve or tap left open.
Related Content
Explore the system with your own data in a demo session.