The objective of the paper is to interpret data on water level variati
on in a river affected by overflow from a sewer system during rain. Th
e simplest possible, hydraulic description is combined with stochastic
methods for data analysis and model parameter estimation. This combin
ation of deterministic and stochastic interpretation is called grey bo
x modelling. As a deterministic description the linear reservoir appro
ximation is used. A series of linear reservoirs in sufficient number w
ill approximate a plug how reactor. The choice of number is an empiric
al expression of the longitudinal dispersion in the river. This approx
imation is expected to be a sufficiently good approximation as a tool
for the ultimate aim: the description of pollutant transport in the ri
ver. The grey box modelling involves a statistical tool for estimation
of the parameters in the deterministic model. The advantage is that t
he parameters have physical meaning, as opposed to many other statisti
cally estimated, empirical parameters. The identifiability of each par
ameter, the uncertainty of the parameter estimation and the overall un
certainty of the simulation are determined. (C) 1997 IAWQ. Published b
y Elsevier Science Ltd.