Uncertainty Analysis for River Quality Measurements

Abstract

Physico-chemical water quality measurements are frequently applied for river quality management purposes, e.g. for the assessment of current and historical state of surface waters, for water quality risk assessment, for calibration and validation of river quality simulation models, etc. The measurements are, however, subject to errors, which might be considerable in some cases or for some variables. The errors cause uncertainties in the assessments and in the model calibration and validation.
In this paper, river water quality measurement errors are statistically analyzed, making use of the relative comparison of sampling data repeating their analysis three times. Data along the Egyptian Nile and the Nile Delta branches are considered, as well as data along canals and agricultural drains next to these rivers. Quantification is made of the mean relative error for the 44 water quality variables considered, as well as the uncertainty in this quantification due to the limited number of samples available. Investigation is furthermore made of the presence of systematic differences in the mean relative error of water quality samples taken along the rivers, the canals and the drains, of systematic differences in the errors between different variables, and of correlations between the variables. Strong correlations have been found between subsets of variables, which allow the number of variables and laboratory analysis costs to be strongly reduced.