Choosing methods for estimating dissolved and particulate riverine fluxes from monthly sampling

In discrete water quality surveys, riverine fluxes are associated with unknown uncertainties (biases and imprecisions). Annual flux errors have been determined from the generation of discrete surveys by Monte Carlo sorting for monthly sampling, from 10 years of daily records (120 records). Eight calculation methods were tested for suspended particulate matter, dissolved solids and dissolved and total nutrients in medium to large basins (10(3) to 10(6) km(2)) covering a wide range of hydrological conditions and riverine biogeochemistry. The performance of each method was analysed first by type of riverine material, which appeared to be much less pertinent than the flux variability matrix. The latter combines the river flow duration in two percent of time (W-2%) and the truncated exponent (b(50sup)) defining the relationship of concentration vs discharge (C-Q) at higher flows (C = aQ(b50sup)). As flux variability increases (high W-2% and/or high b(50sup)), averaging and rating curve methods become less efficient compared to hydrograph separation methods. Flux biases and imprecisions were plotted in the [W-2%, b(50sup)] matrix for discrete monthly surveys.