Urban water managers need to track the water source in the pipe network and to monitor and modify its quality between the points of treatment and use. It is necessary to track the sources of tap water, as the associated quantity (e.g., water source depletion) and quality risks (e.g., source pollution) differ between the various sources (Wet et al., 2020). For example, groundwater is believed to be more resilient to short-term droughts and less vulnerable to pollution than surface water (Foster et al., 2020), which has a potential knock-on effect on the pipes they supply. The water quality at the point of use may differ from the quality at the source because of the modification processes within the distribution network (Liu et al., 2018), making monitoring the water quality modifications imperative. The International Standards Organization (ISO 5667-5:2006) and WHO recommend municipal water monitoring to detect these changes.

In cities with centralized water sourcing and well-engineered pipe networks, it may be feasible to determine the water source using pipe network maps and the pressure distribution within the system. In rapidly-growing urban centers in developing countries, multiple and decentralized water sources (reservoirs, well-fields, solitary wells, springs, etc.) contribute to the piped network (Price et al., 2019). This makes it challenging to backtrack the water source and determine the residence time of different water sources at the point of use.

The water quality may change along the flow path within the pipe network (Zlatanović et al., 2017; Wee et al., 2021; Zhang et al., 2021). The residence time of pipe water, the infiltration of exogenous substances, in the case of negative pressure in the pipes, as well as the mixing of water from different sources could lead to changes in the water quality. A longer residence time often leads to a greater probability of water contamination within the pipe network (Tinker et al., 2009; Sekozawa et al., 2013; Zlatanović et al., 2017; Geng et al., 2021; Kennedy et al., 2021). Thus, water residence time in pipe network is a variable of interest for urban water utilities; it can be determined by using methods such as hydraulic models (Tinker et al., 2009), chlorine decay model (Bhadula et al., 2021; Geng et al., 2021), or, to a lesser extent, by measuring trihalomethane, a by-product of the chlorine-organic matter interaction (Sekozawa et al., 2013).

Another characteristic of interest in piped networks is the irregular changes in its quality. Irregular changes in the quality of the source water cause microbiological and physicochemical destabilization of pipe materials, which subsequently lead to the release of substances into the drinking water through biofilm detachment and resuspension, which cause health and esthetic issues (Liu et al., 2017, 2018). Thus, monitoring the stability of water quality of a piped network is of interest in urban piped-water quality management. Theoretically, pipes sourced from surface water reservoirs should show seasonal variations in the water quality, and borehole-sourced pipes should show a stable chemical composition, due to the general stability of the groundwater chemistry, compared to that of the surface water. However, this may not be the case if boreholes source their water from different layers of the aquifer, which have different chemistries. The water quality variations in borehole water can occur in response to the degree of pumping and to seasonality (Yager and Heywood, 2014; Piscopo et al., 2020). The systematic and comparative investigation of the stability of the water quality for differently-sourced (surface water, groundwater, etc.) pipe networks is lacking in the literature. As more and more cities in developing nations shift from centralized to multiple decentralized sources (Chávez García Silva et al., 2020), an understanding of the stability of the water quality of piped waters, may be of interest.

This paper aims to test the applicability of environmental tracers as tools for backtracking the pipe-water sources, monitoring the composition stability of pipe-water, and determining its residence time in the pipes. We will demonstrate the capability of water isotopes (δ¹⁸O and δ²H) and Electrical Conductivity (EC) in backtracking the sources of water. We aim to demonstrate the potential use of ²²²Rn (Radon isotope), something that has not previously been considered, to determine the residence time of pipe water. Furthermore, we will test the potential of ²²²Rn for discriminating between surface water and groundwater sources in pipe networks. We have used Addis Ababa, the capital city of Ethiopia, as our case study site.

Environmental tracers (δ¹⁸O–δ²H, ²²²Rn, and EC) are widely used in hydrology to trace the origin and movement of water in the hydrological cycle (Gat, 1996; Bresciani et al., 2018; Kebede et al., 2021). Water isotopes (δ¹⁸O–δ²H) have recently been employed in urban water systems to differentiate between the sources of water in pipe networks and to investigate the reliance of the urban water supply on one source or another (Bowen et al., 2007; Jameel et al., 2016; Nagode et al., 2022). The advantage of water isotopes is that once the source water enters the pipe network, its δ¹⁸O–δ²H composition does not change within the pipes. Thus, δ¹⁸O and δ²H are powerful conservative tracers for backtracking the sources of pipe water, given that the different sources show different compositions.

²²²Rn is the other isotope that can be used to trace pipe water sources and estimate water residence time. ²²²Rn is a naturally-occurring and inert radioactive gas with a half-life of 3.82 days. It is produced in rocks and is abundant in groundwater, but non-existent in surface water. Depending on their residence time and degassing rate, water sources connected to groundwater should show a measurable ²²²Rn content. Since the half-life of ²²²Rn is short, it is an ideal isotope for determining the residence time of groundwater-sourced pipe water, as the typical age of water in pipes ranges from a few hours to a few days. Furthermore, ²²²Rn is an ideal tracer for discriminating between groundwater-sourced pipe water from surface-sourced (reservoir) ones. Given that the residence of water in the pipe network is sufficiently short (e.g., <10 days) and that the source water contains sufficient initial ²²²Rn, pipe water sourced from groundwater must show detectable levels of ²²²Rn. All surface-sourced pipe water must show non-detectable levels, as the ²²²Rn in this water is near zero. No published literature is available on the use of ²²²Rn for determining the residence time of pipe water and for backtracking the water sources. However, a number of studies employ ²²²Rn measurements in pipe networks to assess the radioactivity hazard of the water (Thabayneh, 2015; Büyükuslu et al., 2018). The current study is apparently the first in using ²²²Rn to determine the water source and residence time in pipe networks.

Electrical Conductivity (EC) is a vital water quality parameter that can be measured easily (Bresciani et al., 2018). It correlates with most other dissolved substances in water and is thus widely used as a proxy for other dissolved substances in water (USGS, 2019). Thus change in the EC signifies a change in the water quality, implying EC can be used as a tracer of the water quality changes within a pipe network. Groundwater has a higher EC than surface water due to its chemical interaction with rocks (USGS, 2019). Generally, the shallow layers of an aquifer have a lower EC, compared to the deeper layers (Hem, 1989). Thus, EC can be used to backtrack pipe water to its shallow aquifer, deep aquifer, or surface water sources.

The study was conducted in the water-challenged city of Addis Ababa. The challenges include complexity of the pipe network, intermittent water supply, and inadequate water quality at the point of use. The city has a complex urban water supply system. Many other cities in developing nations face similar challenges (Oluka et al., 2013; Chakava et al., 2014; Beard and Mitlin, 2021). Firstly, the pipe water is sourced from multiple origins (Figure 1A), including two large surface water reservoirs (Legedadi R and Gfersa R), multiple well-fields (Akaki 1, Akaki 2, Akaki 3, Old Akaki, and Legedadi), solitary decentralized boreholes within the premises of the urban area, and dispersed springs (Figure 1A). Secondly, tens of thousands of households have installed secondary water storage reservoirs (with a 1–5 m³ capacity) to overcome the intermittent water supply. The reservoirs could alter the pipe water residence time and quality because of the potential backflow from the storage to the pipe network. Thirdly, untreated groundwater from hundreds of boreholes (BHs) is haphazardly injected into the pipe network from multiple decentralized solitary boreholes. The borehole water may be injected directly into the primary or higher-order pipes or singularly, where they are used to supply a specific area (e.g., the high-rise condos) of the city. This increases the risk of adding unaccounted-for pollutants into the pipe networks. Studies have demonstrated that tap water in Addis Ababa is polluted with trace elements, including As, Cr, Ni, Pb, Fe, and Mn (Dessie et al., 2021), turbidity, E. Coli bacteria (Mekonnen, 2015; Dessie et al., 2021), and coliforms (Wolde et al., 2020). The causes of pollution are unknown; it could be because the contaminated source or modification of the water quality within the pipes. A systematic investigation has been lacking into whether the above-normal concentrations are due to the water quality characteristics of the source, the infiltration of contaminated water into the pipe network, or the water quality changes within the pipe network.

There is limited literature on the use of environmental tracers for backtracking the water sources, estimating residence time, and understanding the water quality stability. As a result, we designed our sampling and measurement strategy as a reconnaissance survey and thus conducted a number of snapshot measurements at multiple spatial and temporal scales.

We sampled 48 groundwater sources (springs, shallow wells, and deep wells) and ten samples from the reservoirs and stream waters that feed the reservoirs (Gfersa and Legedadi) for δ¹⁸O and δ²H measurements. We measured the EC of tap water at 510 locations (Figure 1B), which are fairly distributed across the water source map produced by the Addis Ababa Urban Water Utility (Figure 1B). We sampled the tap water at 220 locations (Figure 1B) for δ¹⁸O and δ²H, primarily before the pipes entered the household storage reservoirs. ²²²Rn samples were measured at 27 selected locations. For the ²²²Rn measurements, we took samples along the pre-defined flow direction of the pipe water.

For isotopic and EC measurement, grab samples were taken from water sources (water wells, reservoirs, springs and streams) and tap waters. We allowed the tap water to run for 15 s before taking the samples. In all cases, leak-proof doubly capped plastic bottles were used for sample collection for δ¹⁸O and δ²H analysis. Locations for the spatial sampling of tap waters were chosen to represent the entire city and the water sources. The sites for temporal monitoring of water quality in tap waters were chosen based on a priori knowledge of the water source at the point of monitoring.The Spatial Analyst tool in ArcGIS version 10.3 was used to produce the maps in this work.

Furthermore, on-site EC (25°C) monitoring was conducted hourly at five selected households at the five locations (Sites RS1, RS2, BH1, BH2, SP1 in Figure 1A) for a minimum of 72 h and a maximum of 4 months. Hand-held devices (Hanna^TM 98312) with automatic temperature compensation and daily calibration were used to measure the EC of the water. The monitored sites included two predominantly surface water-sourced sites (RS1 and RS2), two solitary borehole-sourced sites (BH1 and BH2), and a site with mixed surface water and spring water source (Sp 1). The purpose of the long-term hourly measurement was to monitor the instability of the water quality within the pipe network accompanying the various sources.

A RAD7 radon electronic detector, fitted with a Big Bottle System for high sensitivity (Durrige Company Inc., 2020), was used to analyze the in situ ²²²Rn content on grab samples of the waters. The detection limit of the device is in the order of 30 Bq.m⁻³. The counting was based on the principle of liquid–gas-membrane extraction (Schmidt et al., 2008). An extraction module, consisting of hollow vinyl fibers allows radon stripping from the water sample into a connected closed-air loop. The water temperature was measured to convert the ²²²Rn in the air to ²²²Rn in the water. The overall standard deviation varied between 15 and 20%, which is higher for a low radon content. ²²²Rn counting was conducted at each site over a period of 2 h in four cycles of 30 min each. The first reading was discarded, and the average of the latter three readings was taken as the mean ²²²Rn composition of the specific water point. Liquid Water Isotope Analyzer version DLT 100 at Addis Ababa University was used to determine the δ¹⁸O and δ²H results in per mil (‰) notation calibrated against Vienna Standard Mean Ocean Water (VSMOW). The laboratory standard deviations are 0.2‰ for δ¹⁸O and 3‰ for δ²H.

The current work demonstrates the potential of environmental tracers, such as water isotopes (δ¹⁸O–δ²H), radon (²²²Rn), and Electrical Conductivity (EC), for backtracking the sources of water, for estimating the residence time of water and for monitoring the instability of the water quality in an urban pipe network. The tracers have effectively discriminated between surface water-sourced pipe water from those sourced from groundwater. The groundwater-sourced sites have been further discriminated into shallow groundwater vs. deep groundwater. The use of the ²²²Rn decay equation provided reasonable age values for the selected known sections of the pipe. To avoid complications from unaccounted radon rise, storage, and sink, we have isolated a specific pipe section with a known pipe layout with no storage, rise, and sink in order to estimate the residence time of water flow from the upstream end to the downstream point. While the non-reactive nature of ²²²Rn and its half-life is suitable for the estimation of water residence time in the urban pipe network, the impact of (a) degassing of ²²²Rn, (b) ingress of gasses into the pipe networks and, (c) radioactivity that may arise from the pipe wall, must be the target of future research. Sediment deposition inside pipe walls serving as a source of ²²²Rn is the other complicating factor that needs further research.

We established that groundwater was the predominant water source for the case study site in the southern part of the city. In contrast, surface water sources dominated the northern part of the city. By using EC and ²²²Rn, the presence of water originating from solitary boreholes was identified in the surface water-dominated pipe network in the northern part of the city. Water residence time varied from a few hours to 1–6 days. Surface water-sourced (reservoir) pipe networks show a seasonal trend and have relatively high water quality stability, except when it shows outlier values following pipe water interruptions and reconnections. Small diurnal variations in chemistry are noted for the surface water-sourced networks. Solitary borehole-sourced pipe waters show a high water quality variation, probably in response to the degree of pumping and pump stoppages and restarts, following electricity disruptions. This reveals the potential link between the energy supply and water quality in urban settings.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The authors confirm contribution to the paper as follows: SK: study conception, design, analysis and interpretation of results, and manuscript writing. BB: GIS and production of figures and data collection. KH: data collection (δ¹⁸O–δ²H, ²²²Rn) and contribution to writing. AS: data collection (EC). All authors reviewed the results and approved the final version of the manuscript.