Augmenting in-situ with mobile sensing for adaptive monitoring of water distribution networks

Instrumenting water distribution networks with sensors for monitoring is critical to maintain adequate levels of water quality and quantity. Existing efforts to detect and localize adverse events in the network have explored either installing in-situ sensors on junctions or deploying a number of mobile sensors through pipes. These approaches have high costs, low sensing accuracy, lack sufficient coverage or provide intermittent monitoring. In this paper, we combine the benefits of in-situ and mobile sensing with various geosocial factors to develop a cost-effective hybrid monitoring architecture that minimizes the impact of adverse water events on the community. The architecture can adaptively adjust sensing resolutions on-demand within the network, determine required sensing capabilities based on the event, and respond to varying event severities. We propose a two-phase planning and deployment approach that first integrates network structure, event, and community information with simulation based analytics to determine locations to install in-situ sensors and mobile sensor insertion infrastructure. We then incorporate network flow information to determine mobile sensor deployment locations and volume to quickly localize detected events to minimize their impact. We evaluate our approach using multiple real-world water networks for adverse water quality and loss events and compare it to existing approaches. Our results show that our proposed approach can achieve upto 79% reduction in impact with upto 68% greater cost efficiency compared to approaches using traditional coverage heuristics, and upto 30% reduction in impact while being upto 52% more cost efficient compared to approaches that attempt to minimize impact.