Incertitudes sur la concentration moyenne d'un événement pluvieux induites par différentes stratégies de prélèvement

The mean concentration of a given pollutant for a rain event may be defined as the ratio of the pollutant mass to the corresponding water volume. It is usually derived by analysing a compound sample cumulating several sub-samples dispatched along the duration of the event to provide a mean concentration value. The representativeness of the compound sample depends on the sampling strategy used to control the automatic sampler. This study uses experimental data to assess the errors on the mean concentration of a given event implied by two usual sampling strategies. Then uncertainty ranges for all events are derived and compared. It relies on discharge and turbidity recordings available on separate and combined drainage systems included in URBIS network. Two sampling strategies controlled either by cumulated discharge or by elapsed time were simulated for several hundred rain events, with different parameter set-ups. Results clearly show that the control by cumulated volume provides a far more representative sample than the control by elapsed time. Indeed the latter implies a weighting error, which may be huge for some sites and some events. Unfortunately, it is quite difficult, or even impossible, to assess the uncertainties implied by both strategies for a particular site, unless flowrate and pollution are continuously monitored while sampling campaigns are performed.