EGU25-11429, updated on 15 Mar 2025
https://doi.org/10.5194/egusphere-egu25-11429
EGU General Assembly 2025
© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.
Oral | Tuesday, 29 Apr, 15:25–15:35 (CEST)
 
Room 2.31
From rain to health risk: forecasting microbiological contamination in urban rivers using dimension reduction techniques
Arthur Guillot - Le Goff1,2, Yoann Cartier1,2, Brigitte Vinçon-Leite2, Sebastien Boyaval1, Paul Kennouche3, and Rémi Carmigniani2
Arthur Guillot - Le Goff et al.
  • 1LHSV, ENPC, Institut Polytechnique de Paris, EDF R&D, Chatou, France
  • 2LEESU, ENPC, Institut Polytechnique de Paris, Univ Paris Est Creteil, Mane-la-Vallée, France
  • 3Direction de la Propreté et de l’Eau—Service Technique de l’Eau et de l’Assainissement, Paris, France

Urban swimming has re-emerged as a popular activity, especially in France as the Paris 2024 Olympic and Paralympic Games races in the Seine River marked a significant milestone in the revival of open-water swimming. Yet, maintaining water quality in urban areas poses important challenges, especially with the increasing frequency of extreme weather events linked to climate change. Indeed, heavy rainfall leads to sewer system overflow.

This study presents a framework to estimate health risks at urban bathing sites by linking rain intensity to microbial contamination. High-frequency bacteriological timeseries monitoring based on a new monitoring system were collected from the Seine River, between 2021 and 2023.  In parallel, meteorological and hydrological data were collected in the upstream urban watershed.

The timeseries dataset was split into discrete events. An event is defined as a rainfall period, possibly extended to a bacterial peak. Each event was characterised by indicators such as total rainfall, mean flow rate, maximum bacterial concentration, etc. The dimension reduction was first based on a Principal Component Analysis (PCA) and then on a Manifold Isomap technique.

PCA confirmed correlations between rain parameters and bacterial concentrations. Then, Manifold Isomap synthesised selected rain characteristics into a single dimensionless indicator (Global Rain Parameter, GRP). A threshold effect appeared in the relationship between GRP values and bacterial peaks. Below this threshold (GRP = 1.5), no bacterial contamination is observed. Above this threshold contamination increases linearly with GRP. The method was then tested to predict water quality during the Olympic and Paralympic Games. It successfully forecasted future rain events as problematic or not and estimated periods for safe swimming conditions.

The proposed framework opens up new perspectives for the future management of the public bathing sites that will open as a legacy of the Olympic and Paralympic Games in summer 2025. Furthermore, this methodology could be adapted to a wide range of applications when it comes to forecasting surface water quality in urban areas.

How to cite: Guillot - Le Goff, A., Cartier, Y., Vinçon-Leite, B., Boyaval, S., Kennouche, P., and Carmigniani, R.: From rain to health risk: forecasting microbiological contamination in urban rivers using dimension reduction techniques, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11429, https://doi.org/10.5194/egusphere-egu25-11429, 2025.