Smart Sensor Networks for Urban Water Quality Monitoring and Decision-Making

Thesis Abstract

Urban water quality management faces increasing pressure from rapid urbanization, climate variability, and aging infrastructure, which compromise timely detection of contaminant risks and undermine informed decision-making for municipalities. This study develops and evaluates a smart sensor network platform that integrates low-cost in situ water quality sensors, edge computing, and cloud-based analytics to enable near-real-time monitoring, anomaly detection, and decision support for urban water systems. The aim is to improve surveillance accuracy, responsiveness, and governance of drinking and non-potable water streams through automated data fusion and actionable insights. Specific objectives are to (i) design a modular sensor-network architecture with robust communication protocols and energy-efficient sensing schedules; (ii) implement edge- and cloud-based data analytics for quality indicator estimation, trend analysis, and anomaly detection; (iii) evaluate the platform’s performance in a multi-zone urban catchment under varying hydrological conditions; (iv) assess the effectiveness of decision-support tools for operator interventions and policy decisions; and (v) examine stakeholders’ acceptance and organizational readiness for ICT-driven water governance. The methodology adopts a mixed-methods research design comprising quantitative sensor-based experiments and qualitative stakeholder engagement. The population includes three urban catchments within a mid-sized city characterized by heterogeneous land use and episodic contamination events. A stratified sampling approach selects 60 representative monitoring sites across residential, industrial, and mixed-use zones, with 180 sensor deployments over a 12-month pilot. Data collection instruments comprise multi-parameter water-quality sondes measuring pH, temperature, dissolved oxygen, turbidity, electrical conductivity, nitrate, and surrogate microbial indicators; a Raspberry Pi-based edge gateway for local data pre-processing; and a cloud platform for centralized storage, analytics, and visualization. Instrument validation follows calibration against certified reference methods (EPA-approved protocols) with quarterly cross-checks. Data analysis employs time-series modeling, regression analysis for sensor calibration and indicator estimation, anomaly detection via unsupervised learning (Isolation Forest; one-class SVM), and forecasting with ARIMA and Long Short-Term Memory (LSTM) networks. A risk-based decision framework integrates Bayesian decision theory to translate sensor outputs into actionable operator alerts and management actions. The study also employs thematic analysis of semi-structured interviews (n=25) with water-utility engineers, policymakers, and community stakeholders to gauge acceptance, perceived trust, and organizational readiness. Key expected findings include (1) a validated, scalable sensor-network architecture enabling <5-minute data latency and target RMSEs within 5–10% for primary parameters (pH, turbidity, nitrate); (2) robust anomaly detection with precision and recall above 0.85 under noise conditions typical of urban environments; (3) improved prediction of short-term quality excursions by hybrid ARIMA-LSTM models outperforming baseline models by 15–20%; (4) a decision-support interface that reduces time-to-intervention by 25–40% in simulated operational scenarios; and (5) evidence of enhanced stakeholder trust and procedural alignment with ICT-enabled governance, contingent on transparent data policies and user-centered design. The study contributes to knowledge by integrating edge-cloud analytics for urban water quality monitoring, advancing methodologies for data fusion and real-time decision support in municipal water networks, and providing empirical evidence on organizational adoption barriers and enablers for ICT-driven environmental governance. The anticipated conclusion emphasizes the viability of modular smart-sensor deployments coupled with adaptive analytics to strengthen urban resilience against water-quality threats. Recommendations include standardization of sensor calibration protocols, development of open interoperable data schemas, refinement of the Bayesian decision framework for diverse jurisdictional contexts, and strategies for stakeholder engagement to sustain long-term adoption and funding.