An innovative water-monitoring system designed to detect, track, and help manage contamination in urban water networks is being developed by researcher Zabeer Zarif Akhter. Over the past year, Akhter has shifted his focus from solely treating polluted water to creating a comprehensive approach that also identifies sources of contamination and evaluates the effectiveness of treatment in near real-time.
The system, called UrbanSentinel, integrates artificial intelligence with multiple sensing technologies to rapidly assess water quality and predict contamination patterns. It uses a variety of signals, including ultraviolet-visible absorption, infrared spectroscopy, Raman scattering, and electrical properties such as capacitance and resistance. These are analyzed by a lightweight deep-learning model that estimates key water quality parameters like biochemical oxygen demand (BOD) and chemical oxygen demand (COD).
Beyond individual sampling points, UrbanSentinel aims to function as a “nervous system” for the water network by combining monitoring node data with external factors such as rainfall, water flow, and levels. This network-level analysis helps forecast the potential spread of contamination downstream. When detected discrepancies occur between nodes—such as clean water in one area and polluted water downstream—the system uses timing, spectral fingerprints, and hydraulic modeling to generate a probability-ranked list of likely pollution sources.
Akhter emphasizes that the system is not designed to definitively identify polluters or assign blame. Instead, it provides evidence-based insights that can inform further investigation by regulatory authorities through established legal channels.
Currently a Research Assistant at the Institute of Appropriate Technology (IAT) at Bangladesh University of Engineering and Technology (BUET), Akhter has expanded his work to include calibration, control samples, and considerations of repeatability and uncertainty in his measurements. He acknowledges potential limitations and failure points throughout each stage of the system’s operation.
These complexities were highlighted during the World Water Challenge 2026, held in South Korea and co-organized by the Ministry of Climate Energy and Environment and the Korea Water Forum. Akhter presented UrbanSentinel to an audience of researchers and industry experts, focusing on both the system’s demonstrated capabilities and its ongoing development needs.
The sensing subsystem has been tested against laboratory standards, showing an average agreement rate of 96.4%, with nearly 93% of readings within 10% of lab measurements. However, the dataset used for training remains limited, comprising just 28 paired samples, which restrains claims about universal accuracy and scalability.
Preliminary treatment experiments using plasma technology have also shown promising results, reducing water color and chemical oxygen demand substantially in test samples. Nonetheless, these remain early-stage findings requiring further validation.
UrbanSentinel represents a step forward in urban water quality management by linking rapid detection, network-level modeling, and actionable insights, although researchers caution that more data and refinement are necessary before it can be widely deployed.