Variability Structure Estimation: The Role of Sample Support

307 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2000.The characterization of spatial variability in soil properties is a prerequisite to many activities, such as site specific management (i.e., precision farming), groundwater and solute transport modeling, groundwater pollution assessment and remediation, etc. Due to soil heterogeneity, statistical measures are often used for variability description. The experimental point variogram and the experimental histogram are the two most widely used statistical measures of variability structure. For a given spatial domain, it is often difficult to accurately estimate the point variogram due to sampling costs and limited resources, thus we aim to maximize the relative information in our data sampling efforts. The work presented in this document shows in detail the development of a methodology to estimate the point variogram using different types of regularized data, i.e., single-support variograms and mixed-support variograms. The applicability of the method is shown using two different sets of data one is a conditional simulation based on 1650 measurements of phosphorous in a section of one mi2 area (640 acres, or 259 ha). The second data set consists of chloride mass recovery measurements within a small field plot. With these application examples, the relative information content of different measurement methods for characterizing the point variogram are evaluated (in terms of integral scale and sill). The effectiveness and robustness of the methodology are analyzed by means of Monte Carlo analyses, using the conditional simulation data. A cation exchange capacity (CEC) field was obtained by a conditional simulation of 1650 CEC measurements from Williams Field and is used with several sampling schemes to analyze the influence of sampling patterns in three different numerical experiments. The results are shown as confidence intervals of the estimated variogram, the histogram of the data, and the spread of the confidence intervals of the parameters obtained from the data fitting routines.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD