Sensors & software detect combined sewer overflows in real time
Accurately quantifying the problem is one of the biggest challenges to successfully controlling combined sewage overflows (CSOs). The most recent infrastructure report card published by the American Society of Civil Engineers gave U.S. wastewater systems a D+, noting that tangible data regarding storm water and combined sewage systems are few and far between. This lack of data exists because most storm water and combined sewage systems currently are planned and evaluated based entirely on modeled data with little to no validation with local empirical data. Jersey City, N.J., sought to monitor its CSOs in real time. The goals of the project were to learn the region’s baseline flow patterns and outline characteristics indicative of CSO events for the ultimate purpose of building analytics capable of detecting CSO events in real time.
Jersey City (population of 250,000) is located across the Hudson River from New York City, and is one of the many cities under a consent decree from the federal government to control CSOs. The current sewerage system is managed by the Jersey City Municipal Utilities Authority (JCMUA) and consists of two wastewater treatment plants (WWTPs) and 21 CSOs that discharge a mix of runoff and raw sewage during rain events. The city is working on a long-term CSO control plan to control, treat and ultimately cease releasing sewage.
Currently, JCMUA relies on modeled estimates to predict and quantify CSO events. JCMUA and the Jersey City Office of Innovation partnered with StormSensor to use continuous data to identify CSO events in real time. The system consists of a combination of hardware and software. Scute sensors, which can be placed at various points within the storm water system and at outfall monitoring points, measure water depth, velocity and temperature. Measurements are taken every five minutes and are sent via LoRaWAN and MachineQ into Terrapin, a web-based software. The software matches incoming data with local weather data. Over the course of the project, a series of machine-learning algorithms were developed to analyze incoming data, learn expected baseline conditions and flag deviations from baseflow that are consistent with established multiparameter indicators of CSO events.
StormSensor installed the first pilot array for CSO monitoring in Jersey City in March 2018. The short study pilot phase yielded promising results. Multiple CSO events were identified, including at least one not predicted by JCMUA’s model. Continued monitoring has led to the accumulation of more valuable data. This case study focuses on a single outfall that discharges into a tidal canal that eventually empties into the Hackensack River. The outfall was monitored at two locations: one upstream of a netting facility, and one directly at the discharge point.
Pilot monitoring efforts in Jersey City were resumed in November 2018. The first step in building an algorithm to detect CSO events is to establish solid baseline conditions at each monitoring point, understanding that each will have unique flow conditions. This site was chosen as a focus because the two monitoring points, while geographically close, are known to have markedly different baseline conditions. The upstream monitoring point is located within a vault just downstream of the combined system mainline and upstream of a tide gate that keeps the river water out of the sewer system during high tides. Upstream flow is uniform in nature, while the downstream flow is influenced by tidal fluctuations. Continuous monitoring has established baseline conditions at both the upstream and downstream sites. The time period between Dec. 11 and 14, 2018, serves as an example of baseline conditions that must be established by machine-learning algorithms before deviations such as CSO events can be identified. There was some light rain on Dec. 14, but it did not greatly impact either system. The baseline tidal flow conditions observed at the downstream outfall shows good agreement with tidal level measurements reported at the nearest National Oceanic and Atmospheric Administration monitoring station.
These baselines have been used as a starting point to build a series of machine-learning algorithms aimed at detecting CSO events. The goal of these algorithms is to detect when flow conditions have differed from the known baseline in a way consistent with observed patterns indicative of a CSO event. The algorithms rely on multiparameter analyses, including water depth, velocity and temperature data captured using sensors, along with local weather data. While still in beta, initial runs have returned some promising results.
The storm event that occurred Dec. 1 to 3, 2018, serves as an excellent example of a typical CSO event patterns that can be detected by machine-learning algorithms. The small to moderate storm event dropped 1.03 in. of rain over the Jersey City area. Data collected at the upstream monitoring site exhibits a temperature trend indicative of a CSO event. The rain event began Dec. 2, shortly after which the water level in the pipe begins to rise, indicating a flow of storm water to the system. Sanitary sewage has an average temperature of 60°F, warmer than the incoming rainwater. Therefore, the distinct rise in temperature occurring on Dec. 2 just after 9:03 a.m. indicated that the wastewater treatment plant capacity had been reached and sewage began flowing along with storm water into the system. The CSO continues to flow until 7:54 a.m., Dec. 3, as indicated by the maintained depth of water in the pipe and extended period of elevated water temperature. Water temperature in the pipe takes longer to return to baseline conditions due
to thermal conditions within the pipe.
The downstream monitoring site serves as an example of how CSO flows can be differentiated from tidal fluctuations. In accordance with upstream data, the CSO event is expected to have begun near 9 a.m., with some apparent dampening of the tidal pattern present before that time likely due to reported easterly winds pushing water upstream. The pattern indicative of CSO flow was maintained over four tidal cycles and returned to expected baseline conditions within minutes of the upstream site at 7:55 a.m., Dec. 4. This example highlights the importance of monitoring at both upstream and downstream locations, so variations in data can be placed in context. As temperature and velocity sensors are added to the downstream site, measurements will be further refined.
As monitoring continues at Jersey City’s CSOs, the next step in this process will be auto-calculating the volume of a CSO event. As this is achieved and more CSOs are set up for continuous monitoring, Jersey City will be able to fully validate its storm water model and better prioritize sub-basins for improvements and retrofits. In doing so, Jersey City can target retrofits in the highest-impact locations first, resulting in more significant overall improvements relative to dollars spent, while improving the health of its aquatic and urban environments.