Optimal sampling network for monitoring the representative water quality of an entire reservoir on the basis of information theory

This study presents a method for optimal selection of sampling stations for representative and effective water quality monitoring of a reservoir on the basis of information theory. We adopt an objective function used for the design of other hydrological monitoring networks such as rain gauge networks. Unlike rain gauges, in which only a single variable of rainfall is of interest, water quality monitoring stations measure multiple water quality variables. To consider this nature, a new concept of multi-variate weighted total information is proposed. This allows us to consider the relative importance between multiple water quality variables in the design of water quality monitoring networks. The proposed methodology is applied to Lake Yongdam, South Korea, where water quality has been extensively observed at several points. The optimal combination of sampling sites selected with the proposed method is found to show little redundancy in comparison with a previous independent study that presents statistical analysis of the same dataset. Further, the water quality data averaged over all stations are very close to those averaged over the selected sites only, implying that the optimal combination of sampling sites is representative of all sites.