FAGFELLEVURDERTE ARTIKLER Demonstrating the potential of salt tracer studies to improve Norwegian drinking water network models and water age estimates By Jon Kristian Rakstang, Michael B. Waak, Marius M. Rokstad and Cynthia Hallé Jon Kristian Rakstang er Master of science innen vann og miljøteknikk og ansatt hos COWI AS. Michael B. Waak er Ph.D. innen vann og miljøteknikk og forsker hos SINTEF AS. Marius M. Rokstad er Ph.D. innen vann og miljøteknikk og ansatt hos Asplan Viak AS. Cynthia Hallé er Ph.D. innen vann og miljøteknikk og førsteamanuensis ved NTNU. Sammendrag Summary «Tracer»-studier med salt kan potensielt forbedre Hygienic drinking water is a fundamental norske vannettsmodeller og vannalder-estimater. human need. While traversing the drinking Hygienisk sikkert drikkevann er et grunn­ water distribution network, water age underlies leggende menneskelig behov. Drikkevanns kvali­ many water­quality degradation processes, and tet påvirkes av råvannet, vannbehandling, samt water with extended age may differ from initial prosesser og hendelser ute på vannettet. Det er post­treatment quality. In this study, water age vist at drikkevann med høyere alder, eller opp­ was estimated in the Trondheim municipal dis­ holdstid, kan ha forskjellig kvalitet fra drikke­ tribution network using a water network model vann som nylig har gjennomgått vann behand­ and simulations. A full­scale sodium chloride ling. I denne studien ble vannalder estimert i (NaCl) tracer study was used to assess the model. Trondheim kommunes vannett ved bruk av data­ Empirical water age estimates at monitoring simulering, samt vannettsmodellen til kommu­ sites were consistently longer than the simulated nen. Modellens nøyaktighet ble vurdert gjen nom tracer peak arrival times. Nonetheless, simula­ en «tracer»­studie med bruk av natrium­­klorid ted and empirical water age correlated well, (NaCl). Empiriske estimater av vannalder var indicating that additional adjustments to the konsekvent høyere enn simulert ankomsttid for water network model may improve accuracy. «traceren» ved alle målestasjonene. Likevel ble This work demonstrated that salt tracer studies det funnet god korrelasjon mellom simulert og are a cost­effective, simple, and safe method to empirisk vannalder, noe som indikerer at ytter­ directly estimate drinking water age and calibrate ligere justering av vannetts modellen kan for­ municipal water network models. bedre nøyaktigheten. Dette arbeidet demon­ strerte at «tracer»­studier med salt er en Introduction kostnadseffektiv, enkel, og trygg metode for å The United Nations Sustainable Development skaffe direkte estimater av vann alder, samt kali­ Goals envision, among other things, good health, brering av kommunale vannettsmodeller. clean water and innovative infrastructure for VANN I 01 2021 61 FAGFELLEVURDERTE ARTIKLER society (United Nations, 2015). Drinking water full­scale tracer study, in which water age was infrastructure in Norway will require significant directly measured at six monitoring sites in the investment, approximately 220 billion Norwe­ distribution network using conductivity to detect gian kroner, to update all existing infrastructure salt plugs. Finally, empirical observations of to a standard that satisfies these goals (RIF, water age were compared to model predictions, 2019). Furthermore, after the enteric illness out­ revealing some of the challenges of this method break in Askøy, Norway in summer 2019 due to but also opportunities for improving water age fecal Escherichia coli contamination, drinking estimation in Norwegian distribution networks. water quality and water infrastructure have re­ ceived more critical public interest, and many Materials and Methods municipalities have given potential or existing Description of the municipal drinking vulnerabilities more consideration (Bruaset, 2008). water system During the days or weeks that water travels Trondheim, Norway, has a population of appro­ through a municipal drinking water distribu­ ximately 205 000 (Statistics Norway, 2020). tion network, water quality may degrade due to Municipal drinking water originates from two bacterial growth in the water, interactions of the nearby lakes, Jonsvatnet to the east (primary water with pipe materials or biofilms, or with supply) and Benna to the southwest (secondary intrusion of external contaminants via breaks, and reserve supply) (City of Trondheim, 2017). leaks or planned maintenance activity (van der Jonsvatnet, by way of Vikelvdalen water treat­ Kooij, 2000; Makris et al., 2014; Chan et al., ment plant (VIVA), provides water to about 2019). Hydraulic characteristics like flow velo­ 99 % of the city population and is also the city as well as water residence time (or ‘water reserve supply for the nearby municipality of age’) are often critical factors in these events in Melhus (population 16 700). Raw water from addition to the subsequent propagation of con­ Jonsvatnet is withdrawn from a lake depth of taminated water in the distribution network 50 m and then travels 4 km by tunnel to VIVA. (Douterelo et al., 2013; Haig et al., 2018). Despite Water first passes through a granular limestone its importance, however, water age is difficult to bed to increase water hardness for corrosion measure or infer directly, especially in the control. Disinfection includes 0.1 mg/L free Norwe gian context where there are usually little chlorine (HOCl), produced by electrolysis of a or no chemical additives in finished waterproduct. NaCl brine to NaOCl (i.e., the chloralkali pro­ In contrast, practices common in some other cess), and ultraviolet (UV) irradiation (40 mJ/ countries, such as residual disinfection or fluori­ cm2). Under normal operation, production at dation, may be utilized to help elucidate how VIVA is about 750 L/s (23.7 × 106 m3/year). The long water has been in a distribution network in distribution network includes 800 km of pipe, those countries. 12 elevation basins, 20 pump stations and 7000 In this investigation, we aim to demonstrate manholes. Secondary water supply from Benna that a salt tracer, using brine already present at a is withdrawn via two parallel intakes from a water treatment plant (WTP) for onsite chlorine depth of 32 m, travels by tunnel 1.5 km, and is production, is an important assessment tool disinfected at Benna WTP with UV irradiation available for water age model calibration when (40 mJ/cm2) and about 0.1 mg/L free chlorine. other empirical indicators of water age are absent Normal production is 150 L/s to Trondheim or unavailable. First, a water network model (4.7 × 106 m3/year, via 24 km pipeline) and 50 used by the municipal water authority of Trond­ L/s to Melhus (1.6 × 106 m3/year). heim, Norway, was adapted to EPANET, and then average water age was simulated through­ Water network model and simulation studies out the municipal distribution network (Rak­ The City of Trondheim maintains a water stang, 2020). Next, the model was used to plan a network model using MIKE URBAN software 62 VANN I 01 2021 FAGFELLEVURDERTE ARTIKLER (DHI Group), which was exported to EPANET which has a 30 min residence time. Hydraulic 2.0 software (U.S. Environmental Protection and water quality calculations used a 5 min time Agency) for simulation studies. This model con­ step. sisted of approximately 9500 nodes and 10 800 Notably, conductivity predictions in the simu­ links, with 48 different daily demand multiplier lation represented the change in conductivity patterns to represent water consumption varia­ due to the tracer, in contrast to the actual con­ tions in the distribution network. The model ductivity. Therefore, in subsequent comparisons was built with a traditional top­down approach against measurements from the tracer study, (Blokker et al., 2016), meaning water demands simulated values were transformed by either were clustered into the 9500 nodes rather than adding the median observed conductivity at 51000 individual building connections (City of each site (sites 2 to 6) or minimum conductivity Trondheim, 2017). Custom scripts in MATLAB (site 1), which were assumed to represent the software (MathWorks, Inc.) were used to per­ background. form water age and chemical tracer simulations with EPANET, utilizing the ‘epanet­MATLAB’ Tracer study package (Uber, 2013). Water age was estimated in the full­scale distri­ bution network using a salt tracer, as previously Water age simulation demonstrated (Skipworth et al., 2002). Salt­ For water age simulations, both WTPs were saturated brine (NaCl) was already present at assumed to be operational. Reservoir nodes at VIVA for production of free chlorine via electro­ Jonsvatnet and Benna were designated the lysis. Salt was selected as a tracer because it is origin (t = 0). During initial simulations, water safe for consumers, cost effective, and easy to age stabilized at approximately 120 days (2880 h), measure in real­time as conductivity. In addi­ so this duration was used in subsequent runs. tion, the onsite brine system was mostly auto­ Time­step intervals were 10 min. Final simula­ mated and required little personnel time. The tions determined the average and maximum brine was injected directly downstream of filtra­ water age for each node, with averages represen­ tion for 1 h, dosed to achieve approximately 20 ting the mean of the final 48 h in simulation mg/L NaCl and corresponding to a conductivity time. Water ages were visualized using QGIS spike of about 30 µS/cm (at 200C) (Figure 1). By 3.4.15 software (QGIS Development Team, injecting at the