Riverine fluxes of pollutants: Towards predictions of uncertainties by flux duration indicators

Discrete river water quality surveys (sampling frequencies from 2, 3, …, 30 days) are simulated using a Monte Carlo technique on a set of four French stations (Seine, Marne, Oise, Loire Rivers) and two Lake Erie's tributaries (Grand, Cuyahoga Rivers) surveyed daily during at least 10 years, for various riverborne materials (Total Dissolved Solids–TDS, Cl − , NO 3 − , NH 4 + , PO 4 3− , total N, total P and Suspended Particulate Matter–SPM), corresponding to 300 years of daily records. Related interannual and annual fluxes are then calculated using the discharge-weighted concentration method. The comparison between sorted fluxes and reference fluxes, as determined from daily records, gives the errors distribution: imprecisions (ranges between upper and lower deciles of errors) and biases (median of errors). For a given sampling frequency, both errors can be predicted from a single indicator, the percentage of riverine long-term flux transported in 2% of the time ( M 2 ). The nomograph linking imprecisions and biases to M 2 , for a given frequency, has been established and validated previously for SPM. We demonstrate here that the M 2 nomograph is applicable to all other types of river material (major ions, dissolved and total nutrients). Flux errors are also determined on an annual basis and compared to annual flux duration indicators ( m 2 ): the same M 2 – error nomograph is still applicable. Station comparison suggests a possible regionalisation of M 2 indicators – for a given class of river material and for similar stations – and, therefore, of flux errors and/or a possible optimization of river surveys for a given targeted error.