Today’s Water Meters (Mechanical) Can Sometimes Greatly Overestimate Domestic Consumption Due to Air in Pipelines: A Field Evidence

1. Introduction

In cities, households are supplied with drinking water via pipelines; this supply can be: (1) continuous (i.e., the pipe network is always filled with water and pressurized so that water is available all the time), or (2) intermittent (i.e., the network is sometimes empty, meaning that water is only available for a few hours per day, or even less often) (Totsuka et al., 2004). Admittedly, intermittent supply poses problems, including: it becomes more difficult for water agencies to control the water quality and leaks inside the network, and it becomes necessary for households to cope with shortages by storing water at the domestic level (e.g., in an underground cistern and/or a roof tank) (Totsuka et al., 2004). Nonetheless, intermittent supply is a reality in many cities with scarce water resources (Totsuka et al., 2004) (Charalambous et al., 2007) (Abu-Madi and Trifunovic, 2013) (Kumpel et al., 2017).

Urban water agencies would like every domestic intake to be equipped with a water meter, in order to optimize water supply (from the knowledge of households’ habits) and bill for this service (Totsuka et al., 2004) (Kumpel et al., 2017) (Taylor, 2018). In this case, it is important to note that the vast majority of today’s domestic water meters are mechanical, i.e.,: they have a rotating element (rotor, piston or disc) that is supposed to be propelled only by the flow of (liquid) water, and a gear system that converts this movement into a registered water volume (Arregui et al., 2006). One condition for these meters to work properly is that the pipe on which they are installed is “completely filled with water”, as stressed by the OIML (OIML, 2024). But contrary to what happens in calibration facilities (AWWA, 2012), this condition is not always met in the field. Indeed, the pipelines of a water supply network may contain air under certain circumstances (Escarameia et al., 2005) (Pozos et al., 2010), including: (1) when a network is successively emptied and filled (which is usual with intermittent water supply, and sometimes necessary with maintenance); (2) if air is sucked by a pump (and entrained downstream in the form of air bubbles, pockets or slugs), or (3) if a hydraulic jump develops inside an inclined pipeline (which generally results in the formation of a large air pocket upstream of the jump, and a flow of air bubbles downstream). Among other issues (Escarameia et al., 2005) (Pozos et al., 2010), air in pipelines poses a metrological problem: if this air flows through the domestic intakes, it may propel the rotating element of mechanical meters, resulting in an overestimation of water consumption; at least, this concern exists since more than a century in the case of intermittent water supply (Jopling, 1857) (Foufeas and Petroulias, 2017) (admittedly, a reverse flow from domestic intakes to pipelines is also conceivable (Totsuka et al., 2004); but as will be argued, it should generally be less pronounced).

Compared to other causes of systematic error (Arregui et al., 2006) (Arregui de la Cruz, 2016) (for a short review, see Appendix B), there is little quantitative information in the literature about the possible overestimation of domestic consumption by mechanical water meters due to air in pipelines: although a growing number of tests conducted in the laboratory suggests that this could be an important issue (Miranda, 2011) (Grando-Sirtoli et al., 2016), particularly in areas with intermittent supply (Cornejo Cid, 2013) (Walter et al., 2017) (Walter et al., 2018) (Taylor, 2018) (Tavares, 2018) (Ferrante et al., 2022) (Ferrante et al., 2025) (Ferrante, 2025), there is virtually no field study on the topic. Thus, the case of one house is presented (Section 2): its mechanical water meter was estimating a huge consumption (Section 3), and this is analyzed (Section 4).

2. Materials and Methods

2.1. The Case Studied

This study was conducted in a residential house located in a city (> 100,000 inhabitants) of central Mexico. The house is: (1) in an area of intermittent water supply (the supply being typically of a few hours per day); (2) almost at the top of a steep street (average slope ≈ 5.5°, as measured with an inclinometer app for smartphone, and verified using a smartphone barometer and Google Earth), and (3) in a sparsely occupied area (≈ 10 domestic intakes over a distance of ≈ 250 m). It is useless to specify the exact location of the house, as we did not obtain detailed information on how intermittent water supply takes place there (only partial information was obtained, which will be provided in Section 4.2). Furthermore, it must be recognized that, during the study, the house’s domestic intake was modified without the permission of a water agency: consequently, certain details cannot be shared for privacy reasons (in this respect, an attempt to justify why the study was conducted in such a manner is given in Appendix A).

To cope with shortages, the studied house has two facilities (Figure 1): an underground cistern (with a capacity of ≈ 5,000 L) and a roof tank (with a capacity of ≈ 500 L). Both have a float valve to prevent them from receiving too much water from the domestic intake. Furthermore, a pump installed into the cistern allows to refill the roof tank automatically, if it does not receive water directly from the domestic intake (i.e., when pressure of the supply network is too low). Finally, all the water from the cistern is supposed to be directed to the roof tank (unless there is a leak between both).

The owner of the studied house lives there alone for two years. This person contacted us at the end of 2024, because he/she was receiving bills for a high water consumption (considering that the daily consumption of one person is normally < 250 L/d) and always for the same amount. In fact, the staff of the urban water agency was not coming to read the house’s water meter (probably because it is not visible from the street, and the house is left alone for most of the day) and, as a result, the agency was only “estimating” domestic consumption. So, the house’s owner began to take readings of the meter. And then, he/she realized that it was suggesting a very high water consumption (equivalent to ≈ 800 L/d).

2.2. Monitoring Campaign

To explain the very high consumption estimated by the water meter of the studied house, three hypotheses were considered from the beginning: (1) there was a significant water leak in the house (of ≈ 30 L/h, which is not impossible, though quite high for a residential house (Tamari and Ploquet, 2012)); (2) the water meter was damaged, and/or (3) air coming from the supply network was affecting the meter. But none of these possibilities was clearly apparent to the naked eye. So, to try to understand the situation, the owner of the studied house was asked to regularly take photos of his/her water meter and send them. It quickly became important to take photos every day (as will be emphasized in Section 4.3). As a result, a monitoring campaign began, which lasted for a year. This monitoring took place in four stages (Table 1):

• First stage of monitoring - The water meter of the studied house was a mechanical meter of the multi-jet type (it will be named “MJ” below). Figure 2a shows how this meter was originally installed. The first stage of the monitoring campaign simply consisted in reading and visualizing the daily data from this meter (based on the photos sent by the house’s owner). Then, four months of observations led to the preliminary conclusion that meter “MJ” was probably affected by air coming from the supply network (as will be argued in Section 4.1).
• Second stage of monitoring - The second stage of monitoring was an attempt to avoid the possible air flow through meter “MJ”. It began when the owner of the studied house announced that he/she had decided to hire a plumber for installing an air-release valve “AV” (model “Foset 40012” from Truper (Mexico), $DN$ = 13 mm) just upstream of the meter (Figure 2b). This choice (based on commercials) was somewhat imprudent, because it can generate water hammer (Lingireddy et al., 2004) (Escarameia et al., 2005) (Zhou et al., 2020) (Ferrante et al., 2023). However, it did not cause any apparent damage to the domestic intake. On the other hand, after two months of monitoring, it became clear that the air-release valve alone was not effective (as will be discussed in Section 4.2). So, we made a suggestion, which the house’s owner agreed to implement (Figure 2c): modify the shape of the domestic intake so as to form a small inverted siphon “IS” (as suggested by a few authors (Walter et al., 2017) (Walter et al., 2018) (Taylor, 2018)). Nevertheless, it became clear after some weeks that this modification was neither effective.
• Third stage of monitoring - The third stage of monitoring began when we suggested to make an inter-comparison between meter “MJ” and two others installed at the same domestic intake: it was hoped that this would help to distinguish between the possibility of a damaged meter and that of air coming from the supply network. For this purpose, a meter of the single-jet type (named “SJ”) was first installed in series (Figure 3a): since it was mechanical, it was expected that it would behave similarly to meter “MJ” in the event of air flows, or would otherwise provide reasonable estimation of water consumption. Then, a non-mechanical meter of the ultrasonic type (named “US”) was also installed in series (Figure 3b): since this type of meter (when designed for water (Lynnworth, 2013)) is a priori not (Ferrante, 2025) or little (Yao et al., 2021) affected by air flow, it was preliminarily considered as a reference for estimating the domestic water consumption.
• Fourth stage of monitoring - Finally, the fourth stage of monitoring consisted in verifying the data provided by the non-mechanical meter “US” installed at the domestic intake. For this purpose, a mechanical meter of the multi-jet type (named “MJ-T”) was installed at the roof tank’s outlet of the studied house (as described in Section C.1). This last meter was considered as the best reference for estimating domestic water consumption, because: (1) it proved reasonable to assume that all the water supplied at the domestic intake was reaching the roof tank (since there was no significant leak between the domestic intake and the roof tank, as will be demonstrated in Section 4.1); (2) there should be no problem of air flow at the outlet of a roof tank (Charalambous et al., 2007), and (3) it was possible to carefully verify the meter accuracy in situ (as shown in Section C.2).

2.3. Used Water Meters

The accuracy class of all meters used in this study is “2” (OIML, 2024), which is common for today’s water meters (Arregui et al., 2006). It means that the maximum permissible error (MPE) for estimating a water volume should be ± 5% at flow rates between the minimum ($Q_1$) and transitional ($Q_2$) flow rates, and ± 2% at flow rates between the transitional ($Q_2$) and overload ($Q_4$) flow rates (when water temperature < 30 °C) (OIML, 2024). Otherwise, the meter error is not specified for too low flow rates (below $Q_1$); but for this condition, mechanical meters tend to underestimate the volume of water passing through (Arregui et al., 2006) (see Figure C2 for an example), and ultrasonic meters can no longer detect water flow (so they stop counting the volume of water) (Lynnworth, 2013).

The technical specifications of the four water meters installed in the studied house are given in Table 2:

• Meter “MJ” - The mechanical meter “MJ” officially installed was quite new (≈ 2 years old). Its model is not specified for privacy reasons. Nevertheless, it complies with standard ISO 4064: 2014, which is similar to recommendation OIML R-49: 2013.
• Meter “SJ” - According to its data sheet (Siemens, 2021), mechanical meter “SJ” complies with recommendation OIML R-49: 2013. The best practice for installing water meters with a rotor (i.e., single-jet and multi-jet meters) is to mount them on a straight pipe that is horizontal and long enough (Arregui et al., 2006) (AWWA, 2012). But for lack of space, meter “SJ” was installed on a vertical pipe (so that water flows upwards) and with no concern for pipe length: in fact, this is possible according to the meter’s manufacturer (Siemens, 2021), but the meter then has a higher minimum flow rate, i.e.,: $Q_1$ = 40 L/h (which is attributable to the fact that the meter’s rotor is subject to greater friction (Arregui et al., 2006)). Also note that it was not possible to purchase a meter compatible with the original diameter of the domestic intake ($DN$ = 13 mm), but with a larger diameter ($DN$ = 20 mm).
• Meter “US” - According to its certificate (SLM, 2025), ultrasonic meter “US” complies with recommendation OIML R-49: 2013. And for identification purposes, the same certificate (SLM, 2025) specifies that the meter has one pair of “wet” transducers, which operate according to the transit-time principle (Arregui et al., 2006) (Lynnworth, 2013) (Yao et al., 2021). The best practice for installing ultrasonic water meters is to mount them on a straight pipe that is horizontal and long enough (Arregui et al., 2006) (Lynnworth, 2013). But as for meter “SJ”, meter “US” was installed on a short vertical pipe: in fact, this is possible according to the meter’s manufacturer (SLM, 2025). Note that, compared to the used mechanical meters, meter “US” can reliably measure at very low flow rates (i.e., $Q_1$ = 3 L/h).
• Meter “MJ-T” - According to its certificate (ANCE, 2025), mechanical meter “MJ-T” complies with Mexican standard NOM-012-SCFI: 1994 (DGN, 1994), which is similar to standard ISO 4064-1: 1993 (ISO, 1993). This meter was installed at the outlet of the house’s roof tank, ensuring that it was on a sufficiently long horizontal pipe portion (see Section C.1 for details). Also note that the diameter of the pipe at the roof tank’s outlet ($DN$ = 25 mm) was greater than that of the pipe at the domestic inlet ($DN$ = 13 mm): in this case, special care was taken to select a meter with a quite low minimum flow rate (i.e., similar to that of meter “MJ”).

3. Results

3.1. Data from the Officially Installed Water Meter (and from that of a Nearby House)

Before presenting the data from the water meter “MJ” of the studied house, it is appropriate to present those from another meter officially installed in a nearby residential house. This house was occupied by three persons, who agreed to take photos of their meter every day and send them (they also agreed to take photos of meter “MJ” when its owner was away). The meter was mechanical (multi-jet) and rather old (≈ 20 years). Its data reveal two trends (Figure 4):

• A first trend that looks normal - During the first 8 months of monitoring, the data of the water meter looked normal: (1) estimated daily consumptions fluctuated, but were on average of ≈ 425 L/d (Figure 4b), which is a typical value for the domestic consumption of three persons (Mazzoni et al., 2023) (please note that the slight decrease in June 2025 is attributable to the fact that one person was away), and (2) the estimated total consumption was increasing regularly (see dotted line on Figure 4a). Furthermore, daily water volumes estimated by the meter were sometimes very low (i.e., < 100 L/d: see the grey rectangle on Figure 4b): considering that the residents of the house were keeping the same lifestyle, this does not reflect a sudden decrease in their water consumption, but an intermittent supply with some days without any supply at all. So, these observations suggest that a mechanical water meter can behave normally, at least apparently, even under intermittent supply: this finding will need to be discussed together with the data from the water meter “MJ” of the studied house.
• A second trend that reflects a deterioration of the meter - Unexpectedly, from a certain date (around September 9th: see red arrow on Figure 4a), the water meter estimated a gradually lower, and finally null consumption: considering that the residents of the house were keeping the same lifestyle, this reflects a slow deterioration of the mechanical meter. Is this normal wear (Arregui et al., 2006), or accelerated damage caused by air flow in pipes during intermittent water supply (Totsuka et al., 2004) (Foufeas and Petroulias, 2017) (Ferrante et al., 2022)? Even if we have no answer to this question, the example presented shows that in a study like this, the possibility of a subtle deterioration of mechanical water meters cannot be ignored.

Compared to the data from the water meter of a nearby house (as long as it seemed to work normally), those of the meter “MJ” officially installed at the domestic intake of the studied house were quite different. Basically, this meter was estimating a very high and erratic consumption (Figure 5):

• Water consumption estimated by meter “MJ” was very high - On average, the daily water consumption estimated by meter “MJ” during the monitoring campaign (from 14/12/2024 to 09/02/2026) was 1,040 L/d: in line with the concern of the house’s owner when he/she contacted us, this is very high (≈ 6.5 times more than the usual domestic consumption of one person (Mazzoni et al., 2023)).
• The estimated water consumption was erratic - Whereas the total consumption estimated by the meter of a nearby house was increasing regularly (see Figure 4a until ≈ September 9th), that estimated by meter “MJ” was characterized by jumps (Figure 5a). These jumps correspond to huge estimations of daily consumption, some of which even exceeding the built storage capacity (i.e., ≈ 5,500 L): this is shown on Figure 5b (see peaks in the red zone).
• On occasion, the meter was estimating negative daily consumptions - Interestingly, meter “MJ” has sometimes estimated negative water consumptions: (1) the most notable case occurred on April 11th, with an estimated consumption of ≈ -1,900 L/d (one day before the installation of an air-release valve, which is a coincidence); (2) but this was not the only event (indeed, meter “MJ” estimated a daily consumption below -50 L/d on 5 occasions during the monitoring campaign).
• The meter’s behavior did not change after the installation of an air-release valve - Following the installation of an air-release valve and a small inverted siphon at the domestic intake (see yellow circle and triangle on Figure 5b), the meter’s behavior did not appear to change noticeably: indeed, meter “MJ” was still estimating huge water consumptions for certain days.
• The meter did not estimate high water consumptions when the house was unoccupied - On four occasions, the owner of the studied house was away for several days (see green lines on Figure 5a). During these periods, meter “MJ” did not estimate high water consumption: this finding will allow to discuss the hypothesis of water leaks in the studied house (Section 4.1).

3.2. Comparison with Two More Meters Installed at the Domestic Intake

By comparing the daily water consumptions estimated by the three meters installed at the domestic intake of the studied house, it was found that:

• Meter “MJ” vs. meter “SJ” - The comparison of data from mechanical meters “MJ” and “SJ” revealed that (Figure 6a): (1) meter “MJ” was tending to estimate larger consumptions; (2) however, meter “SJ” was also sometimes estimating huge consumptions (i.e., > 5,500 L/d on three occasions); (3) there was a significant nonlinear correlation between the data from both meters (Spearman correlation coefficient: $R_S$ = 0.95 ***), meaning that the data were tending to vary in the same way; (4) many estimates of daily consumption, at least those that were not too high (i.e., < 500 L/d), were similar (difference < 10%) for both meters, and (5) on two occasions during the inter-comparison, meter “MJ” estimated a negative consumption, whereas meter “SJ” never did (see the red dotted ellipse on Figure 6a).
• Meter “SJ” vs. meter “US” - The comparison of data from mechanical meter “SJ” and ultrasonic meter “US” revealed that (Figure 6b): (1) meter “SJ” was tending to estimate larger consumptions; (2) meter “US” was estimating a rather low daily consumption (i.e., < 500 L/d; admittedly, it estimated a consumption of ≈ 1,300 - 1,500 L/d on 18/11/2025 and 02/02/2026, but for these days, the house’s owner explained that he/she had consumed more water than usual to wash clothes and irrigate flowers); (3) there was a significant nonlinear correlation between the data from both meters (Spearman correlation coefficient: $R_S$ = 0.88 ***), and (4) many estimates of daily consumption, at least those that were not too high, were similar for both meters.

Figure 6. Inter-comparison of daily consumptions estimated by the water meters installed at the domestic intake of the studied house: (a) meter “SJ” (single-jet) vs. meter “MJ” (multi-jet); (b) meter “US” (ultrasonic) vs. meter “SJ” (single-jet). The dotted green line is the first bisector.

Figure 6. Inter-comparison of daily consumptions estimated by the water meters installed at the domestic intake of the studied house: (a) meter “SJ” (single-jet) vs. meter “MJ” (multi-jet); (b) meter “US” (ultrasonic) vs. meter “SJ” (single-jet). The dotted green line is the first bisector.

Figure 7. Comparison of the total volumes estimated by the three water meters installed at the domestic intake of the studied house, during two periods of continuous supply: (a) at the middle of the monitoring campaign (since 27/09/2025 at 13:28); (b) at the end (since 30/11/2025 at 11:38).

Figure 7. Comparison of the total volumes estimated by the three water meters installed at the domestic intake of the studied house, during two periods of continuous supply: (a) at the middle of the monitoring campaign (since 27/09/2025 at 13:28); (b) at the end (since 30/11/2025 at 11:38).

Interestingly, just after the installation of meter “US” (i.e., at the middle of the monitoring campaign), water supply at the domestic intake turned out to be continuous for some time; this was evident because: (1) the underground cistern of the studied house was seen filling up; (2) no air was heard coming out from the air-release valve “AV”, and (3) the ultrasonic meter “US” was estimating a flow rate of ≈ 700 L/h with no warning about air in pipes. This provided an opportunity to compare changes in volume displayed by the three meters “MJ”, “SJ” and “US” under continuous supply (until the cistern has filled, i.e., for ≈ 30 min). The result shows a satisfactory agreement (Figure 7a): in fact, differences between the estimation by each meter and the average value remained < 1 L during the whole test (i.e., < 0.3% of the total volume recorded at the end). A similar test was performed at the end of the monitoring campaign (on 30/11/2025), which also yielded satisfactory results (Figure 7b): in fact, differences between the estimation by each meter and the average value remained < 10 L (i.e., < 4% of the total volume recorded at the end).

3.3. Comparison with Another Meter Installed at the Roof Tank’s Outlet

For the reasons mentioned above (Section 2.2), two water meters were a priori considered as a reference during this study: (1) the ultrasonic meter “US” installed at the domestic intake (and thus strictly intended to estimate water supplied to the studied house), and (2) the mechanical meter “MJ-T” installed at the roof tank’s outlet (and thus strictly intended to estimate domestic water consumption). In this regard, the results obtained are as follows:

• Once installed, meter “MJ-T” proved to be accurate enough - According to a field verification done with a volumetric container (described in Section C.2), meter “MJ-T” proved to be accurate enough: indeed, its error was found to be within ± 5% for flow rates above the minimum flow rate (Figure C2).
• Good agreement was obtained between meters “MJ-T” and “US” - The comparison between meters “MJ-T” and “US” began once the underground cistern and the roof tank of the studied house were full. During this comparison, a satisfactory agreement was obtained (Figure 8); indeed, the difference between the readings of meters “MJ-T” and “US” was always < 560 L, which is: (1) realistic (i.e., not more than about the roof tank capacity), and (2) < 2.5% of the total volume recorded at the end of the comparison (which lasted ≈ 4 months).

Finally, Table 3 shows the domestic consumption and water leak estimated by the four meters installed in the studied house:

• Estimated domestic consumption - For a fair comparison, the average daily consumption has been computed over the same period of ≈ 4 months: (1) data from meters “US” and “MJ-T” are close and small (≈ 200 L/d), whereas (2) the data from meter “SJ” is substantial (≈ 2.5 times more), and (3) that from meter “MJ” is huge (≈ 7 times more).
• Estimated water leak - Water leak inside the studied house has been estimated over the periods of > 5 days during which the owner was away (these periods are shown on Figure 5a). Overall, the four meters estimated a low leak (even for a residential house (Tamari and Ploquet, 2012)), i.e.,: on average < 24 L/d. Admittedly, if we consider that a leak occurs at an almost constant flow rate (Tamari and Ploquet, 2012), it is worth noting that the estimated leaks (equivalent to < 1 L/h) are much lower than the minimum flow rate ($Q_1$) from which the meters start to measure reliably.

4. Discussion

4.1. Interpretation of the Main Results

So, what would be the general cause of the very high domestic consumption estimated by mechanical meters “MJ” and “SJ”? Playing devil’s advocate, it could be argued that a tap left accidentally open is enough to explain water losses as large as ≈ 5,500 - 18,000 L in a day (see Table 3 of Mazzoni et al. (Mazzoni et al., 2023)). However, reviewing the results in reverse order leads to a different conclusion:

• Meter “MJ-T” was reliable - First, the mechanical meter “MJ-T” proved reliable for estimating the amount of water consumed from the house’s roof tank: (1) the meter was installed in an atypical but a priori correct way (i.e., on a pipe that should be always filled with water, and according to manufacturer’s specifications (ANCE, 2025), as described in Section C.1), and (2) it was verified in situ that the meter’s accuracy was acceptable (error within ± 5% for flow rates above Q₁ = 35 L/h: see Figure C2).
• Meter “US” proved reliable - Next, the ultrasonic meter “US” installed at the domestic intake proved reliable for estimating the amount of water supplied to the studied house: indeed, there was a satisfactory agreement with meter “MJ-T” (Figure 8).
• There was no significant leak in the house - Since the first stage of the monitoring campaign, it was suspected that there was no significant leak in the studied house: assuming that meter “MJ” was not completely damaged (see next point), its erratic behavior (Figure 5a) was not compatible with that expected for a leak (which would ultimately have resulted in a rather regular increase in the recorded total volume) (Tamari and Ploquet, 2012). The absence of significant leak was confirmed by subsequent measurements: (1) when the house was unoccupied (Table 3), none of the used water meters estimated a large leak (at least, it is certain that the leak was < 3 L/h, i.e.,: less than the minimum flow rate $Q_1$ of meter “US”), and (2) over the entire monitoring campaign, both meters “MJ-T” and “US” estimated a water consumption of ≈ 200 L/d on average (Table 3), which is realistic for one person (Mazzoni et al., 2023) (for more details, see Section C.3) and therefore contradicts the hypothesis of a significant leak.
• Meter “MJ” did not appear to be damaged - Contrary to the meter of a nearby house (Figure 4), meter “MJ” did not appear to be damaged over the course of the measuring campaign: (1) it was found to work properly during two periods of continuous water supply (Figure 7), and (2) it was often estimating daily volumes that were consistent with those from ultrasonic meter “US” (Figure 6). So, meter “MJ” was at least not completely destroyed.
• Air was most likely the cause of abnormal behavior of meter “MJ” - Having discarded the hypothesis of a significant leak (even hidden) and of that of a damaged meter (at least, clearly damaged), it turns out that the huge and erratic daily volumes estimated by meter “MJ” (Figure 5b) were most likely due to air sometimes coming from the supply network.
• Huge daily volumes estimated by meter “MJ” were most likely not due to a malfunction - Although it was not clearly damaged, it must be recognized that meter “MJ” has been exposed to severe conditions of use: when air was passing through it, its rotor may have sometimes been spinning extremely fast (Totsuka et al., 2004) (Foufeas and Petroulias, 2017) (for more details, see Sections B.3 and D.2), which may have caused the meter to malfunction (Sherman et al., 1923). However, meter “MJ” has a magnetic coupling between its rotor and gear system (according to its data sheet, and like most modern mechanical water meters (Arregui de la Cruz, 2016)): this coupling tends to slip when the rotor spins too fast, leading to under-registration (Arregui de la Cruz, 2016). Consequently, the huge daily volumes sometimes estimated by meter “MJ” were most likely not due to an eventual malfunction of the meter.
• Similarly, air must have been the cause of abnormal behavior of meter “SJ” - Like for meter “MJ”, the large daily volumes sometimes estimated by meter “SJ” (Figure 6) were most likely due to air in the supply network (admittedly, meters “MJ” and “SJ” did not behave in exactly the same way: this will be discussed in next Section).
• Direct observations support the idea that meters were affected by air flows - Finally, the fact that mechanical meters “MJ” and “SJ” were affected by air flows is supported by two direct observations: (1) following the installation of air-release valve “AV”, the owner of the studied house reported that he/she was often hearing air escaping from this valve (we have heard this too), and (2) on two brief occasions during which air was heard passing through the underground cistern’s inlet (whereas the roof tank was full), he/she could film the gear system of meters “MJ” and “SJ” turning in the normal direction, while meter “US” was displaying no flow (and a warning about “air in pipes”).

So, it has been shown that two mechanical water meters that do not seem to be damaged can greatly overestimate domestic consumption due to air in pipelines. To our best knowledge, this is the first time that a clear field evidence is published on this issue.

4.2. Further Discussion (with Some Numbers)

At this point, the question arises as to whether the abnormal data from mechanical meters “MJ” and “SJ” can be explained not only qualitatively (as just done), but also quantitatively (as far as possible). To this end, a few more details were obtained about the location and water supply of the studied house:

• Location of the studied house - As mentioned (Section 2.1), the studied house is almost at the top of a steep with an average slope ≈ 5.5°. More specifically, it turns out that: (1) the upper part of the street (over ≈ 50 m) has a gentle slope (≈ 1°), whereas (2) the lower part (over ≈ 200 m) has a much steeper slope (≈ 7°). Furthermore, the street is not a dead end, in the sense that it connects: (1) at the top to a street with a gentle slope, and (2) at the bottom to a larger street that is almost horizontal (according to an inclinometer app for smartphone and Google Earth).
• Water supply - According to a technician, the studied house is supplied with water as follows: (1) from the top of the street; (2) by a pipeline made of plastic (HDPE) and with diameter $DN$ = 50 mm (which is quite plausible (Abu-Madi and Trifunovic, 2013)); (3) almost every day (which is consistent with the data shown on Figure 4), and (4) typically from morning until late afternoon (during the night, the underground valve that controls the supply to the street pipeline is generally closed, while the pipelines of other streets are supplied with water).

Next, three hypotheses are made; they greatly simplify the analysis that follows, and despite appearances, are not completely absurd:

• Hypothesis 1: sometimes, only air passes through a domestic intake - Admittedly, a flow of air mixed with water (unless dissolved (Kell, 1977)) should affect a mechanical water meter (Arregui de la Cruz, 2016): at least, this has been verified for water with air bubbles (Miranda, 2011) (Grando-Sirtoli et al., 2016). However, describing how a mixture of water and air can flow and affect water meters is out of the scope of this study, because it is very complicated (at least, there are different air/water regimes (Escarameia et al., 2005) (Zhou et al., 2020)). In what follows, only the case where pure air passes through a domestic intake is considered: this is sometimes plausible (Totsuka et al., 2004) (Foufeas and Petroulias, 2017) (Taylor, 2018) (Ferrante et al., 2022), and so it is appropriate to analyze this situation at a minimum.
• Hypothesis 2: sometimes, air behaves like an incompressible fluid - Furthermore, it is assumed that air can flows at an intermediate velocity: (1) fast enough, so that air easily spins the rotor of a device like meter “MJ” (Schubauer and Mason, 1937) (Alfonso-Corcuera et al., 2024), but (2) not too fast, so that air behaves almost like an incompressible fluid (Houghton and Carpenter, 2003). If so, a mechanical water meter should record the volume of air passing through it almost as if it were a volume of water; at least, this situation has been observed (to a first approximation) during certain laboratory tests (Miranda, 2011) (Grando-Sirtoli et al., 2016) (Walter et al., 2017) (Tavares, 2018) (for more details on the validity of this hypothesis, see Section D.1).
• Hypothesis 3: sometimes, the studied house was the only one receiving air from the street pipeline - Finally, it is assumed that at certain times, the studied house was virtually the only one of the street able to receive air from the supply pipeline (i.e., not the other ≈ 10 houses, if their storage facilities were full and with the float valves closed).

Based on the above considerations, three reasons that could explain how air in the street pipeline was affecting meter “MJ” are discussed:

• Possible reason 1: air flow due to fillings/emptyings of the water supply network - During the transient fillings/emptyings of a water supply network (Figure 9a), air is pushed/sucked and should therefore flow through the domestic intakes if it has no other way to escape (i.e., in the absence of air-release valves installed by the water agency at strategic locations along the network (Escarameia et al., 2005)). So, during network emptyings, there should be a reverse flow from houses to pipelines (Totsuka et al., 2004) (unless there is a check valve (ISO, 2005) (AWWA, 2012) (Arregui de la Cruz, 2016), which was not the case in the studied house); but this flow should be rather small, because the emptyings of a water supply network are usually more gradual than its fillings (Foufeas and Petroulias, 2017): (1) this seems in line with the finding that the (net) daily volumes estimated by meter “MJ” were sometimes negative, but not often (i.e., below -50 L/d on five occasions in a year; please, note that this cannot be due to a reverse flow of water, since the volume of the bypass shown on Figure 1 is < 40 L), although (2) this does not seem enough to explain why meter “MJ” has once estimated a daily volume as low as ≈ -1,900 L/d (indeed, the downstream street pipeline, which is ≈ 200 m long and ≈ 50 mm in diameter, has a volume of only ≈ 390 L; and if we consider that the emptying of the supply network was not limited to just the street pipeline, this would imply an additional length of ≈ 100 m for a pipeline with a diameter as large as 150 mm). On the other hand, during network fillings, there should be an air flow from pipelines to houses: (1) this seems in line with the fact that meter “MJ” was tending to overestimate domestic water consumption (Figure 5b), although (2) this is not enough to explain why it was sometimes estimating daily volumes > 5,500 L/d (as this would involve a considerable length of pipeline, even if considering a large diameter of 150 mm).
• Possible reason 2: reverse air flow caused by a hydraulic jump during periods of continuous water supply - During a period of continuous water supply, a hydraulic jump may develop in the lower part of an inclined pipeline (Figure 9b), and if this occurs, air can be sucked by the jump (Escarameia et al., 2005) (Pozos et al., 2010). This possibility was examined for the pipeline of the studied house: (1) a hydraulic jump can indeed take place, and (2) it could suck air at a rate of up to ≈ 180 L/h (according to the numerical application provided in Section D.3). If such a hydraulic jump remained stable for ≥ 12 hours, this could explain the (net) negative volume of ≈ -1,900 L recorded by meter “MJ” during one day.
• Possible reason 3: air flow when the street pipeline is not supplied with water - During periods when the domestic intakes of the studied street are not supplied with water (typically at night), those of the street below might be: (1) if so, air bubbles transported by water should tend to accumulate in the pipeline of the studied street (because it is sloped), and then (2) its domestic intakes could act like air-release valves (Figure 9c). In fact, a typical domestic intake (if considered as a pipe of diameter: $DN$ o 13 mm) can easily release thousands of liters of air in a few hours (according to the numerical application provided in Section D.2). Unfortunately, due to the lack of details about the water supply network, we cannot estimate the amount of air that could be transported and released by the pipeline located in the street below.
• Summary - To sum up, the fact that meter “MJ” was affected by air in pipelines does not appear to be due to a single mechanism, but to at least three: (1) transient fillings/emptyings of the street pipeline (but this does not explain everything); (2) hydraulic jump in the street pipeline during periods of continuous supply (which can explain the lowest negative daily volume recorded once), and (3) accumulation of air during periods with no water supply of the street pipeline (which could potentially explain the huge positive daily volumes recorded quite often).

It is now necessary to try explaining the data from the two other studied mechanical meters:

• Meter “SJ” (single-jet): why were its data less dramatic? - Despite the fact that both meters “MJ” and “SJ” were affected by the presence of air in the street pipeline (Section 4.1), they did not behave in exactly the same way: (1) the overestimation of consumption by meter “SJ” was less pronounced, and (2) on some days, meter “MJ” estimated a negative consumption, whereas meter “SJ” never did (Figure 6a). Considering that both meters were installed at the same domestic intake, how can these two differences be explained? At least, this could be due to three reasons: (1) air had less chance to spin the rotor of meter “SJ” because it has a larger diameter (i.e., $DN$= 20 mm instead of$DN$= 13 mm, in which case air should pass through at lower velocity due to Bernoulli’s principle); (2) air also had more difficulty to spin the rotor of meter “SJ” because it was installed vertically (in which case the rotor’s pivot was exposed to greater friction (Arregui et al., 2006), resulting in a rather high minimum flow rate: $Q_1$ = 40 L/h (Siemens, 2021)), and (3) unlike most single-jet meters (including “SJ”), multi-jet meters (like “MJ”) have a bypass (Arregui et al., 2006), which would make them more sensitive to air flow at rather low velocity (Walter et al., 2018) (for more details, see Section D.1).
• Meter of a nearby house (multi-jet): why was it apparently unaffected by air flows? - Unlike meters “MJ” and “SJ”, the meter from a nearby house did not seem to overestimate consumption during the monitoring campaign (Figure 4b). But if we consider that there was sometimes air coming from the street pipeline, why was no effect noticed on the meter during the measuring campaign? At least, this could be due to two reasons: (1) the device was worn out (in fact, it ended up completely stopped) (Arregui de la Cruz, 2016), and (2) a different location of the domestic intake relative to the street pipeline (see Figure 9). In any case, one observation suggests that the meter of the nearby house has been affected by air flows in the past: in the spring of 2025, its owner noticed once (at ≈ 3 o’clock in the morning) that the meter was making a lot of noise and that its display was changing very fast.

Finally, it remains to try explaining why the small air-release valve “AV” and inverted siphon “IS” installed at the domestic intake of the studied house proved ineffective:

• Why was the air-release valve ineffective? - An air-release valve installed just upstream of a water meter can let air escape to the atmosphere, but without preventing flow toward the meter. In this case, the air-release valve “AV” installed by the owner of the studied house (Figure 10) was virtually ineffective because its orifice (diameter ≈ 1 mm) is small compared to that of meter “MJ” ($DN$ = 13 mm). The ineffectiveness of small air-release valves installed at domestic intakes is commented by Arregui (Arregui de la Cruz, 2016) and Taylor (Taylor, 2018). It is true that a valve with a larger orifice can be effective, but then there will be more risk of water hammer during valve closures (Lingireddy et al., 2004) (Zhou et al., 2020) (Ferrante et al., 2023) (for more details on air-release valves, see Section D.2).
• Why was the inverted siphon ineffective? - By modifying the shape of a domestic intake so as to form a small inverted siphon “IS” (Figure 2c), it was hoped that air would have more difficulty to pass through; this proposal was based on laboratory tests (Walter et al., 2017) (Walter et al., 2018) and a suggestion from an author who confesses: “despite the apparent efficacy of this simple solution, the author has never seen the recommended configuration used” (Taylor, 2018). But it turned out that the inverted siphon was ineffective: it must then be admitted that air pressure in the supply pipeline was usually high enough to overcome the small hydrostatic pressure in the siphon (of ≈ 300 Pa, which is equivalent to 30 cm of water column) (Weston et al., 2023). In fact, according to the owner of the studied house, pressure in the supply pipeline was sometimes high enough for water to rise directly to the roof tank.

4.3. Significance of the Case Reported

To assess the significance of this study, it is worth recalling the context in which it was conducted:

• Concern of citizens about air in pipelines - In many areas with intermittent supply, citizens worry that their water meter overestimates domestic consumption due to air in pipelines (Totsuka et al., 2004); this concern is reflected by: (1) comments in the non-academic literature (in the form of newspaper articles, blogs on the Internet, and commercials for supposedly simple remedies) (Tavares, 2018), and (2) air-release valves installed at domestic intakes (Figure 11).
• Previous studies about air in pipelines - It is known for a long time (at least since Woltmann, at the end of the 18th century (AWWA, 2012)) that meters with a light rotor designed to measure water velocity in open channels (Schubauer and Mason, 1937) or in closed pipes (AWWA, 2012) are sensitive to an air flow: in short, these meters can behave like a cup anemometer. But surprisingly, the fact that this could be an issue for domestic water meters (Arregui de la Cruz, 2016), especially in areas with intermittent supply (Jopling, 1857) (Totsuka et al., 2004) (Foufeas and Petroulias, 2017), has been little studied (according to the published literature). Yet, the concern of citizens tends to be supported by a growing number of laboratory studies (Miranda, 2011) (Cornejo Cid, 2013) (Grando-Sirtoli et al., 2016) (Walter et al., 2017) (Walter et al., 2018) (Tavares, 2018) (Ferrante et al., 2022) (Ferrante et al., 2025): indeed, they have shown that when a volume of air (defined at normal conditions for temperature and pressure) passes through a mechanical water meter (at least, single-jet or multi-jet), it can record ≈ 0 - 150% of this volume (depending mainly on air velocity, as argued in Section D1). But then, if we consider, for example, that the filling of a water supply network pushes towards a domestic intake an air volume of ≈ 100 L (i.e., the volume of the pipeline upstream of the studied house; for other houses, this volume can be very different (Taylor, 2018) (Ferrante et al., 2025)), it should produce a meter error of ≈ 0 - 150 L (equivalent to ≈ 0 - 0.9 time the typical daily water consumption of one person (Mazzoni et al., 2023)).
• This study - Compared to previous laboratory studies, this field study reports a very high overestimation for at least two mechanical water meters installed at the same house: on average ≈ 2.5 times the real water consumption in the case of meter “SJ”, and ≈ 7 times in the case of meter “MJ” (Table 3). Admittedly, the meters have most likely been affected by air flows for several reasons, due to: (1) intermittent supply; (2) sloppy street, and (3) sparsely occupied area. Furthermore, the city’s water agency may have operated the supply network in an unusual manner: we have no information on this (which is a major limitation of the study). So, it could be that the case reported is uncommon, but to find out, further studies are needed.

Otherwise, the study shows that in areas with intermittent water supply, it is not so easy to diagnose an effect of air flow on mechanical water meters:

• The effect of air in pipelines can vary greatly over time - Compared to other reasons that could explain a high water bill (i.e., hidden leak inside a house (Tamari and Ploquet, 2012), or bypass of a multi-jet meter that gets clogged (Arregui et al., 2006)), the effect of air in pipelines on a mechanical water meter should vary greatly over time (e.g., Figure 5): indeed, it will at least depend on the status of houses’ storage facilities (i.e., with float valves open or closed). But to what extent can this temporal variability be distinguished from that of normal water consumption (e.g., Figure 4)?
• To detect an effect of air in pipelines, water meters must be read often - A high temporal variability implies that to understand the abnormal behavior of a meter, it is necessary to take readings often (at least, every day). But when they have doubts about their water meter, citizens generally take readings from time to time, which may not be enough: for example, if readings of meter “MJ” were taken only once a month (Figure 12), it would not be possible to understand what was happening (apart a high estimated consumption, as noted by the owner of the studied house).
• The effect of air in pipelines should vary greatly in space - Furthermore, mechanical water meters should be affected by air in pipelines quite differently from one house to another: (1) not only because air tends to accumulate in the upper parts of water supply networks (Totsuka et al., 2004) (Pozos et al., 2010) (Weston et al., 2023), but also (2) because the rotating element of meters can spin more or less easily (as argued in Section 4.2, it depends on the meter’s minimum flow rate $Q_1$).
• Air in pipelines can flow through water meters for different reasons - Finally, it is worth noting that most previous studies about the effect of air on water meters (Cornejo Cid, 2013) (Walter et al., 2017) (Walter et al., 2018) (Tavares, 2018) (Ferrante et al., 2022) (Ferrante et al., 2023) (Ferrante et al., 2025) (Ferrante, 2025) focus on a specific reason: the fact that air in pipelines is pushed/sucked during the fillings/emptyings of a supply network (Figure 9a). However, this study suggests that air in pipelines can flow through meters for at least two other reasons, i.e.,: (1) a hydraulic jump that may develop in a steep pipeline during periods of continuous supply (Figure 9b), and (2) a pipeline that remains a dead end during the continuous supply of downstream pipelines (Figure 9c); although these phenomena have been studied extensively (both the hydraulic jump (Escarameia et al., 2005) (Pozos et al., 2010) and the dead end (Totsuka et al., 2004) (Weston et al., 2023)), their practical importance from a metrological point of view has not yet been examined in detail.

5. Recommendations

A condition for conventional water meters (mechanical) to work properly is that the pipe on which they are installed is completely filled with water (OIML, 2024). But this condition is not always met in the field, especially in areas with intermittent water supply. In this context, the case reported allows to make three recommendations:

• Do not fully trust the cheap devices supposed to avoid air passing through water meters - When there is air in pipelines (which depends on the operation by water agencies), there does not seem to be a simple way for citizens to efficiently prevent it from passing through domestic intakes (Arregui de la Cruz, 2016): as seen, installing a small air-release valve and an inverted siphon did not solve the problem of the studied house.
• Study more how air affects mechanical water meters - So, there is a need to study more how air in pipelines can affect mechanical water meters; it should depend on: (1) not only the characteristics of the water supply network (including its topography, density of domestic intakes, and filling speed) (Totsuka et al., 2004) (Weston et al., 2023), but also (2) the consumption habits of households (which impact the status of their storage facilities, i.e.,: with float valves open or closed), and (3) the actual minimum flow rate $Q_1$ of meters (which depends on the device’s installation and wear).
• Consider alternatives for measuring water consumption - Otherwise, some non-mechanical meters can be considered as an alternative for measuring domestic water consumption: (1) electromagnetic meters are now formally accepted by the OIML (OIML, 2024), and (2) ultrasonic meters might soon become popular (Arregui et al., 2020) (Ferrante, 2025). Admittedly, these new water meters can be affected by air in pipelines (Arregui de la Cruz, 2016), but to a lesser extent (Miranda, 2011) (Yao et al., 2021). And although they are more expensive than conventional meters (≈ 2 to 4 times more) (Foufeas and Petroulias, 2017), this is negligible when considering the benefits in situations like the one just described.