Multicriterion Network Design Using Geostatistics

Publisher Website
Google Scholar
Abstract
Geostatistics and multicriterion decision making are combined to design a regular observation network for several spatially correlated and anisotropic parameters. The decision variables are network density, distance between observation points for the Various parameters, and observation effort. The estimation error is calculated as a function of the decision variables by use of a geostatistical model based on prior variograms. Composite programming, an extension of compromise programming with more than one value of p in the lp distance, makes it possible to account for the analytical characteristics of statistical criteria versus the economic value of observation effort. Thus an l1 metric is applied to statistical criteria and anl2 metric to observation effort criteria. The algorithm for solution uses gradient optimization. The approach is extended to the case when an existing network is to be augmented. The example of a two‐layer aquifer system, where thickness and porosity are the parameters to be identified, illustrates the methodology. The composite solution appears to be quite robust with respect to parameter weight changes.