Optimizing Ground Water Observation Networks in Irrigation Areas Using Principal Component Analysis by S. Khan, H. F. Chen, and T. Rana Abstract Ground water monitoring networks can provide vital information for sustainable water resources management. This in- volves the measurement of ground water level, solute concentration, or both. This article deals with the former. It optimizes network distribution of piezometer or data sampling wells to effectively monitor ground water levels under an irrigation region while retaining adequate overall measurement accuracy. This article presents a structured process for applying princi- pal component analysis (PCA) in optimizing a ground water monitoring network in an irrigation area of Australia. The PCA functions, distributed with the MATLAB package, were used to determine relative contributions of individual piezometers in capturing the spatiotemporal variation of ground water levels. Kriging gridding interpolation algorithm was used to render the data surface presentations and determine spatial differences in piezometeric surfaces using different number of data sets. The results show that the overall difference of ground water level between the original piezometer network and the opti- mized networks after the PCA process was applied is less than 20%, while the total number of piezometers in the optimized network is reduced by 63%, which will save the time and cost to monitor ground water levels in the irrigation area. Some figures in this paper are available in color in the online version of the paper. Introduction Ground water monitoring networks can provide vital information for sustainable water resources management. This involves measurement of ground water level, solute concentration, or both. The goals of ground water moni- toring can be ambient resource condition monitoring, com- pliance monitoring, risk detection monitoring, research monitoring, or a combination of these (Gangopadhyay et al. 2001). For example, the observation and estimation of ground water levels in irrigation areas play an important role in monitoring and evaluating the sustainability of the land and water management practices. Such monitoring usually involves over- or underinvestment due to poorly planned distribution of piezometers, often chosen ran- domly. Significant reduction in monitoring costs can be achieved if the number of observation piezometers can be reduced rationally without compromising the overall moni- toring goals. The ground water level monitoring goals could include the effect of land and water management practices on recharge and discharge, the hydraulic attributes of ground water systems, and the extent and degree of confinement of aquifers (Foglia et al. 2007; Gangopadhyay et al. 2001; Szidarovszky et al. 2007). Several approaches have been used to optimize ground water monitoring networks and to improve water manage- ment systems. Two general approaches to ground water monitoring network design are the hydrogeologic and statis- tical approaches. The hydrogeologic approach is based on the judgment of quantitative and qualitative hydrologic information without the application of advanced statistical techniques. The statistical approach can be further classified into variance and optimization-based techniques. Of these, the optimization approach has found wider application. The optimization approach considers the design problem as a mathematical programming problem consisting of an objective function subject to constraints. Integer or mixed- integer programming algorithms are commonly used to evaluate the presence or absence of a piezometer at a partic- ular point. Theodossiou and Latinopoulos (2006) used a krig- ing technique to optimize their ground water observation networks in order to increase the accuracy of ground water simulation models. Sadik and Karam (1988) used piezo- metric fluctuation analysis to acquire 15 piezometers out of a 68-piezometer network in the Erbil hydrogeological basin, Copyright ª 2008 The Author(s) Journal compilation ª 2008 National Ground Water Association. Ground Water Monitoring & Remediation 28, no. 3/ Summer 2008/pages 93–100 93