Optimal design for dispersion experiment.

Abstract

The dispersion coefficient (D) is a very important parameter for the management of water quality and pollution control. Conventional sampling methods are based on tracer studies. Sampling periods and intervals are often subjective and left to the choice of the experimenter. Sometimes several observations are made to determine D. Yet, making several observations does not necessarily ensure accurate estimation of the parameters. Sampling of tracer concentrations at poor times will result in inaccurate estimates of D and the flow velocity (u). Therefore, the main aim of the study in question was to investigate the optimal sampling times for conducting fixed position variable-time sampling for tracer concentration experiments. This objective was achieved by the minimization of the least square criterion and applying the method proposed by Box and Lucas. The relative efficiency of each experimental design is predicted before data collection and analysis and confidence regions plotted. Results, on optimal sampling times for experiments are presented and illustrated. The application of these findings will harmonize results and reduce cost and labour expended on dispersion experiments.