Development of a Methodology for Data Collection Network Design

Abstract

A data collection network design methodology is presented using entropy and a directional information transfer index (DIT). The methodology developed includes three stages. The first stage is to investigate the stochastic relationships amongst the gauging stations. The non-parametric estimation technique is used to estimate the joint frequency distribution for pairs of stations. The distributions are then used in the second stage to compute the directional information transfer (DIT) index. In the last stage, DIT is employed to assist with the network regionalization as a component of the station selection process. A streamflow network in southern Manitoba, Canada is used as a demonstration of the methodology.