Detection and classification of anomalous events in water quality datasets within a smart city-smart bay project

Continual measurement is key to understanding sudden and gradual changes in chemical and biological quality of water, and for taking reactive remedial action in the case of contamination. Monitoring of water bodies will increase in future within Europe to comply with legislative requirements such as the Water Framework Directive and globally owing to pressure from climate change. Establishing high quality long-term monitoring programs is regarded as essential if the implementation of pertinent legislation is to be successful. However, conventional discrete sampling programs and laboratory-based analysis techniques can be costly, and are unlikely to provide timely and reliable estimates of true ranges of deterministic physicochemical variability in a water body with marked temporal or spatial variability. \ud Only continual or near continual measurements have the capacity to detect ephemeral or sporadic events, thus providing the potential for on-line event detection and classification. The aim of this work is to demonstrate the potential role of continuous data acquisition in decision support as part of a smart city project. In this work, a multi-modal smart sensor network system framework for large scale estuarine and marine water quality monitoring is proposed. The application of a number of evolving techniques that allow automated detection and clustering of events from data generated by in situ sensors within environmental time series datasets is described. We provide examples of how change in the range of variables such as turbidity and salinity might be detected and clustered to provide the end user with greater ability to detect the onset of environmentally significant events. Finally, we discuss the acquisition of data from in situ water quality sensors and its utility within the framework a smart city-smart bay integrated project