A new methodology is developed for data collection network design. The
approach employs a measure of the information flow between gauging st
ations in the network which is referred to as the directional informat
ion transfer. The information flow measure is based on the entropy of
gauging stations and pairs of gauging stations. Non-parametric estimat
ion is used to approximate the multivariate probability density functi
ons required in the entropy calculations. The potential application of
the approach is illustrated using extreme flow data from a collection
of gauging stations located in southern Manitoba, Canada.