River flux uncertainties predicted by hydrological variability and riverine material behaviour

Abstract

Abstract Data on riverine fluxes are essential for calculating element cycles (carbon, nutrients, pollutants) and erosion rates from regional to global scales. At most water‐quality stations throughout the world, riverine fluxes are calculated from continuous flow data ( q ) and discrete concentration data ( C ), the latter being the main cause of sometimes large uncertainties. This article offers a comprehensive approach for predicting the magnitude of these uncertainties for water‐quality stations in medium to large basins (drainage basin area > 1000 km²) based on the commonly used discharge‐weighted method. Uncertainty levels – biases and imprecisions – for sampling intervals of 3 to 60 days are correlated first through a nomograph with a flux variability indicator, the quantity of riverine material discharged in 2% of time ( M 2% ). In turn, M 2% is estimated from the combination of a hydrological reactivity index, W 2% (the cumulative flow volume discharged during the upper 2% of highest daily flow) and the truncated b 50sup exponent, quantifying the concentration versus discharge relationship for the upper half of flow values ( C = a q b 50sup , for q > q 50 , where q 50 is the median flow): M 2% = W 2% + 27.6 b 50sup . W 2% can be calculated from continuous flow measurements, and the b 50sup indicator can be calculated from infrequent sampling, which makes it possible to predict a priori the level of uncertainty at any station, for any type of riverine material either concentrated ( b 50sup > 0) or diluted ( b 50sup > 0) with flow. A large data base of daily surveys, 125 station variables of suspended particulate matter (SPM), total dissolved solids (TDS) and dissolved and particulate nutrients, was used to determine uncertainties from simulated discrete surveys and to establish relationships between indicators. Results show, for example, that for the same relatively reactive basin ( W 2% > 25%), calculated fluxes from monthly sampling would yield uncertainties approaching ±100% for SPM ( b 50sup > 1.4) fluxes and ±10% for TDS ( b 50sup = −0.2). The application to the nitrate survey of the river Seine shows significant trends for the 1972–2009 records. Copyright © 2012 John Wiley & Sons, Ltd.