Comparative Study of Ground‐Water Mapping Techniques

Mapping of ground-water spatial data is an important part of any geohydrologic investigation. There are three main classes of interpolators used for such mappings. The first group include simple estimators which are commonly used in practice. The second group are least-squares estimators which are basically fitting processes. The last category are Gauss-Markov estimators, such as kriging, which beside being exact interpolators, produce measures for the accuracy of the estimated field. These estimators are compared theoretically and numerically. These studies show that kriging yields relatively robust estimates. However, its suggested statistical inference method may not always produce robust estimates of the covariance function parameters. Simple estimators produce unstable results, while least-squares methods ignore local variations by fitting a single polynomial function over the whole field. For this study, water-table data from northwest Kansas are used.