Wireless Sensors for Measuring Drinking Water Quality in Building Plumbing: Deployments and Insights from Continuous and Intermittent Water Supply Systems

Abstract

Despite continued calls to increase the monitoring of drinking water systems, few communities and utilities have adopted modern, distributed, and real-time monitoring systems. Measurements of drinking water quality are often only made at the treatment plant, with limited grab sampling taking place throughout the distribution system. At the building level, where most of the public's exposure to drinking water takes place, the capacity to make continuous measurements to characterize water quality dynamics has been almost impossible. Innovation in sensors, microcontrollers, and data services is underpinning a broader smart cities movement, but their value as a tool in the management of drinking water systems is still unclear. In this paper, we present a new open-source wireless sensor platform, which allows water quality to be measured at the tap. Our internet-connected devices transmit data back to cloud hosted services, where they can be analyzed in real-time. We provide examples of large-scale deployments within buildings in Ann Arbor, Michigan, USA and Mexico City, Mexico. In each of these studies, we demonstrate the detection of phenomena that would have been missed through existing, low-throughput monitoring approaches. The deployment in Ann Arbor emphasizes the importance of real-time measurements in a drinking water distribution system, highlighting shifts in neighborhood-scale electroconductivity (a proxy for total dissolved solids) that would have been missed as part of established sampling procedures. The Mexico City deployment demonstrates highly variable water quality and supply in intermittent systems and characterizes the variability of chlorine concentrations between continuous and intermittent portions of the city.

Identifying water quality variables most strongly influencing Legionella concentrations in building plumbing

“…This degree of variation of water entering the service line has not previously been widely reported in the literature. Though, intensive water sampling as conducted in this study is also rare in the literature (Martinez Paz et al, 2021; Melville‐Shreeve et al, 2021; Saetta et al, 2021). Notwithstanding these challenges, free and total chlorine were strongly correlated (0.791), and correlations with other variables each had the same sign.…”

“…Full‐scale monitoring of a residential home, inclulding online monitoring capabilities, is exceedingly rare in the United States based on plumbing and green building industry discussions. However, three recent studies have been conducted using similar approaches (Martinez Paz et al, 2021; Melville‐Shreeve et al, 2021; Saetta et al, 2021). In addition to the high level of resources expended to collect this rich data set for analyses, the close proximity of the sampling location to the analytical lab facilitated increased data collection by eliminating the logistical requirements associated with sample handling such as long‐term storage and transportation.…”

Machine Learning-Assisted, Process-Based Quality Control for Detecting Compromised Environmental Sensors

“…Figure shows data taken from wireless sensor networks operated by the authors’ research group in Michigan, USA. , Figure a shows a reliable data stream collected by a water level sensor, while Figure b shows an obstructed water level measurement, where the ultrasonic sensor transducer is partially and periodically blocked by a nearby object, resulting in “jump” data faults. Figure c shows another obstructed sensor measurement taken from a pH sensor measuring drinking water quality at household taps in Ann Arbor, MI.…”

“…Open-source plans to build the sensor devices are documented in Bartos et al 2018 14 and Martinez Paz et al 2022. 2 Each of the three data sets were processed in isolation of each other.…”

“…High-resolution, real-time data are poised to play a prominent role in the management of environmental systems, while a new generation of machine learning (ML) and data science techniques promise to lead to new discoveries that would otherwise be obfuscated by purely mechanistic approaches . The past decade has witnessed a boom in the development of low-cost, in situ sensors for applications such as drinking water monitoring, soil pollution tracking, and solid waste management . The resulting data streams have been analyzed with ML algorithms, solving previously intractable problems such as modeling and mapping air quality, , estimating hard-to-measure wastewater variables, and predicting river dissolved oxygen levels across continental scales .…”