*2.1. The District Metering Area*

The area under investigation is "North Soccavo" (Figure 1), which is one of the administrative neighbourhoods of the Municipality of Naples (Italy), counting about 20,000 inhabitants. The area was selected by the local water company as the pilot case for the implementation of a District Metering Area. The reason for this particular choice lies in the observation that this area is connected by a single branch to the water network of the City, which makes it particularly prone to be distrectualized. In recent years, the local water company connected all the existing water meters with a telemetry system, allowing for the automatic radio collection of consumed water volumes at a 1-h time step. The connection involved water meters related to both single and multiple households, as well as commercial activities and public buildings.

**Figure 1.** Pilot area: Water distribution network, water meters, and census particles.

Data referring to the year 2016 have recently been the subject of a multi-purpose research focusing on the prediction of water demand patterns, useful for the local water company to define management and leakage detection strategies [39,40,42]. The findings, based on hourly consumption data collected from 1 January to 31 December 2016, highlighted the following issues:


August consumption, no further seasonal cycle can be observed, and the five patterns in Figure 2 show no significant differences in the remaining months. As a consequence, if August data is discarded, daily discharges can be considered a random variable with no deterministic dependence. As far as the daily cycle is concerned, three different non-dimensional patterns were identified corresponding to Sundays, Saturdays, and Mondays–Fridays, respectively [40].

4. For residential connections, scaling laws were proposed [40] providing the mean hourly discharges and related standard deviations as a function of the number of aggregated households. The regression parameters depend on the characteristics of the specific dataset in terms of single-user behaviour and cross-correlation structure.

**Figure 2.** Average nondimensional patterns for the five main different clusters of consumption behaviours observed in the pilot area (*Qm* and *Qa* are the monthly and annual discharges, respectively).

The scaling laws proposed for the pilot area [40] are a function of the number of aggregated households, instead of the number of consumers, because the information about the number of people "hiding" behind each water meter was known only for a few cases. In the present paper, however, this information was derived by intersecting the spatial distribution of residential water meters (Figure 1), with the number of inhabitants at the census scale. As a result, the average number of users per connection ranges from 2.8 to 3. This uncertainty is caused by the delay between the date of the census survey (2011) and that of data collection (2016), and to a small number of water meters that still missed their connection with the telemetry system by the end of 2016.
