Date of Award

2001

Degree Type

Thesis

Degree Name

Bachelor of Science Honours

Faculty

Faculty of Communication, Health and Science

First Advisor

Lyn Bloom

Second Advisor

Dr Ute Mueller

Abstract

Geostatistics is a field of study that deals with spatially dependent attributes. As information regarding these attributes is usually only available at sample locations, estimates must be made at unsampled locations. Sample data are usually measured on point support within the study region, however in reality decisions are based on small blocks and not on points. A change-of-support model is required to obtain the theoretical distribution of block values given the sample point values. Estimates arc then made for a collection of blocks, referred to as a panel. Kriging is a generic term adopted by geostatisticians for a family of estimators appropriate for spatially distributed data. The main focus of this study is the method of Disjunctive Kriging that employs the use of a family of orthogonal functions known as the Hermite polynomials. This thesis presents comparisons of the results from Disjunctive Kriging with those from the more commonly used methods of Ordinary Kriging and Indicator Kriging. Ordinary Kriging can be used to generate estimates for each small block in the study region. Panel estimates can then be derived from the block estimates within each panel. Indicator Kriging and Disjunctive Kriging use change-of-support models to obtain estimates of functions of the attribute for the panels in the study region based on the chosen block support size. Two sets of isotropic data are analysed, one of which is approximately normally distributed and the other is positively skewed. Exhaustive data is available for both sets of data for comparative purposes.

Included in

Geology Commons

Share

 
COinS