The early warning system is an open source software interface which enables real-time bathing water quality assessment. The model is based on machine learning and/or statistical modelling and predicts bacterial concentration in specific river sections using a set of local data such as rainfall, river flow, temperature and water quality. The EWS will help to (1) manage bathing authorizations in urban bathing sites, (2) monitor the efficiency of the sanitation policy and (3) improve the real-time management of the sewer network and urban bathing sites.


There is an increasing social demand from citizens to reduce environmental impacts in urban rivers and to benefit from urban bathing areas. Several cities like Berlin achieved excellent water quality levels for their urban rivers and propose swimming areas at the core of the city.

A major challenge regarding bathing water management is that concentrations of fecal bacteria may show spatial and temporal variability. In urban rivers, discharges from CSO and stormwater may contain high amounts of fecal bacteria and contaminate bathing water quality.

In many cases, even with sufficient water quality, bathing can be forbidden as current monitoring protocols are not sufficient to protect human health. Even modern rapid monitoring approaches, still need up to 12-14h before results are available and classical grab sampling does not allow to track pollution variability since events may occur between sampling intervals or cannot be collected for logistic reasons (e.g. events might happened at night or during weekends).

Limitations of current practices

If bathing waters are subject to short-term pollution, the current European Bathing Water Directive (BWD Article 12(c)) explicitly demands the implementation of early warning systems in order to prevent bathers from being exposed to contaminated water. However, the BWD neither provides guidance on how to implement early warning systems in practice nor defines water quality alert thresholds.

Bathing water quality is assessed only in the long term by estimating parametric 90th and 95th percentiles based on the monitoring data of the previous four years. The lack of specified thresholds makes it difficult for the responsible authorities to justify and defend short-term decisions about closures of or warnings on bathing sites.


The early warning system is based on an innovative probabilistic approach. The tool translates the current approach of long-term classification according to the European BWD to real-time management for early warning and allow to close a major gap in current European bathing water legislation. The availability of online water quality prediction significantly improves microbial safety and reduce the risk of contamination at bathing waters. It enables to establish bathing waters at challenging location, subject to short term pollution (e.g. urban agglomerations). It also allows to manage nothing authorization considering modelling uncertainty and strengthen bathing water profiles. The tool provides users with a free and user-friendly software to be easily implemented at new bathing waters. A mobile application is proposed to inform key decision makers and citizens of bathing water contamination risks.

City tests

The early warning system integrates two main innovations of DWC: an advanced machine learning model for robust early forecast of the water quality and real-time measurements of bacterial contamination (solution 1). The model will be deployed at selected sites over the urban stretches of the rivers Seine and Marne and coupled with SIAAP’s sewer system real-time control to allow the optimization of sewer management based on river water quality objectives.

It will provide information to communicate with the decision marker who will have to take the decision to close and open the bathing places in relation with the microbial pollution risk and public health issues and the general public who will be informed on the compliance of the water quality to bathing standards, or any other indicators relevant with that stake.

Contact solution

KWB – Wolfgang Seis :
SIAAP – Jean Pierre Tabuchi :
SIAAP – Angelique Goffin :

Follow us on twitter