Network design for the spatial estimation of environmental variables

摘要

A design methodology is developed for observation networks of several spatially correlated and anisotropic environmental variables such as meteorological, hydrological, mineral, and soil data. Geostatistics and multicriterion decision making are combined in order to find a tradeoff between estimation accuracy and observation effort. In general, the decision analysis consists of two steps. First an unconstraint network is sought under the joint consideration of estimation-accuracy criteria and the observation-effort criterion. Next, the above network is optimally adjusted to the existing stations. The methodology assumes that variograms are available for every environmental variable considered. A gradient method of optimazation is used in both steps. A composite form of compromise programming applies an l1 metric for the statistical criteria and an l2 metric for observation effort criteria. Two case examples illustrate the methodology.