Uncertainties in selected surface water quality data

Abstract

Monitoring of surface waters is primarily done to detect the status and trends in water quality and to identify whether observed trends arise from natural or anthropogenic causes. Empirical quality of river water quality data is rarely certain and knowledge of their uncertainties is essential to as-sess the reliability of water quality models and their predic-tions. The objective of this paper is to assess the uncertainties in selected river water quality data, i.e. suspended sediment, nitrogen fraction, phosphorus fraction, heavy metals and bi-ological compounds. The methodology used to structure the uncertainty is based on the empirical quality of data and the sources of uncertainty in data (van Loon et al., 2005). A liter-ature review was carried out including additional experimen-tal data of the Elbe river. All data of compounds associated with suspended particulate matter have considerable higher sampling uncertainties than soluble concentrations. This is due to high variability within the cross section of a given river. This variability is positively correlated with total sus-pended particulate matter concentrations. Sampling location has also considerable effect on the representativeness of a water sample. These sampling uncertainties are highly site specific. The estimation of uncertainty in sampling can only be achieved by taking at least a proportion of samples in du-plicates. Compared to sampling uncertainties, measurement and analytical uncertainties are much lower. Instrument qual-ity can be stated well suited for field and laboratory situations for all considered constituents. Analytical errors can con-tribute considerably to the overall uncertainty of river water quality data. Temporal autocorrelation of river water qual-ity data is present but literature on general behaviour of wa-ter quality compounds is rare. For meso scale river catch-ments (500–3000 km 2) reasonable yearly dissolved load cal-culations can be achieved using biweekly sample frequen-cies. For suspended sediments none of the methods inves-Correspondence to: M. Rode (michael.rode@ufz.de) tigated produced very reliable load estimates when weekly concentrations data were used. Uncertainties associated with loads estimates based on infrequent samples will decrease with increasing size of rivers.