The problem of estimating nutrient transport in large rivers and the u
ncertainty of such load estimates was studied both empirically and the
oretically. In the empirical part of the study, time series of data fr
om the Rhine, Meuse, Vistula and Oder Rivers were examined. The result
s of this data analysis justify the use of linear interpolation to est
imate concentrations prevailing between sampling occasions. A special
study of the spatial variation of concentrations within different cros
s-sections of the Vistula river showed that such variation can contrib
ute substantially to the uncertainty of load estimates. In general, ho
wever, sampling at one point in the cross-section did not result in bi
ased load estimates. In the theoretical part of the study, simple ARMA
(autoregressive-moving average) models were used to derive generally
applicable formulas for the expected mean square error of load estimat
es based on serially dependent concentration data. These formulas were
then used to estimate the uncertainty of calculated nutrient loads in
the Rhine and the Vistula, respectively.