Urban water-quality management often depends on sparse and irregular monitoring data. This is particularly challenging in the context of microbiological hazards, where grab samples may fail to capture short-lived contamination peaks. The AQUAMON project develops an integrated monitoring and decision-support framework combining advanced sensors, automated monitoring, data-driven models and digital tools. Its objective is to transform heterogeneous monitoring data into actionable information for timely, risk-informed water management.

Within AQUAMON, KWB contributes two Bayesian approaches for supporting decisions under limited and uncertain data.

First, a quantitative microbial and chemical risk assessment (QMRA/QCRA) will combine literature-derived knowledge with site-specific observations using Bayesian networks and inferential methods. In the Turin use case, human-health and environmental risks will be assessed along the water-reuse chain, from raw wastewater through treatment and irrigation to exposure via crop consumption and direct contact, including contaminant fate in soil and potential transfer to groundwater. As monitoring data accumulates, the assessment will be updated sequentially, showing how risk estimates and uncertainties evolve and where additional data provide the greatest value. Applications are also planned in Lviv and London.

Second, an enhanced version of KWB’s early-warning model SWIM:AI will support adaptive bathing-water monitoring in Paris. Using rainfall, river-flow and E. coli data, the model predicts bathing-water quality and quantifies predictive uncertainty. Bayesian novelty detection will identify situations where existing information is insufficient and trigger on-demand E. coli measurements using the Fluidion ALERT in-situ sensor. New observations will then be incorporated into subsequent predictions.

Together, these approaches aim to move from fixed monitoring towards uncertainty-aware, adaptive monitoring that collects additional data where they provide the greatest value for risk-informed decisions. As AQUAMON runs until the end of 2028, the poster presents the methodological framework and planned applications rather than project results.

This report delivers a practical manual to support operators with the management of sensors networks in existing water infrastructures. It includes (1) the presentation and assessment of a new easy-to-use sensor for faecal bacteria measurements, (2) methodologies for the validation of online sensors and analysers and (3) best practices for installation, operation, and maintenance.

In DWC, raw data collected from on-line sensors and lab analyses are integrated and analysed to gather conclusive information and early warning to support decisions to deliver safe water reuse and inform about bathing water quality. Three relevant case studies, namely Paris, Berlin and Milan, were investigated in this research. In the case studies of Paris and Berlin, sensors were installed to monitor microbiological contamination in bathing water sites. In the case study of Milan, a real-time sensor network was designed to promote safe water reuse reducing the risk of microbial contamination of soils and crops during irrigation, while assuring compliance of wastewater quality with reuse standard limits.

The technical characteristics of all the installed on-line sensors are reported in section 1, including the innovative ALERT devices manufactured by FLUIDION, which allow the on-line measurements of faecal bacteria indicators. The section also describes in detail measurement characteristics, i.e., static and dynamic characteristics of instrumentation, operational modes, initial measurement accuracy and standards.

The use of real-time data to support health protection and risk management requires primary their validation, in terms of reliability, in order to integrate the standard lab measures with a continuous monitoring system, for control optimization and risk minimization. To date, one of the main lacks on risk management approach is the absence of common procedure on how to treat non-standardized data, such as real-time online data. To answer this question, this report intends to provide practical information about validation, operation and maintenance of on-line sensors for the three representative case studies. Particularly, this report includes

• Return of experience on installation, troubleshooting and maintenance (section 2).

• Data analysis and assessment of the bias, precision and accuracy of the online sensors (section 3).

The conclusions are reported in section 4.