From International Mining and Materials, May 2001, No 41

Practical Geostatistics 2000 by Isobel Clark and William V Harper

This book is a continuation and enlargement of Dr Clark's 1979 Practical Geostatistics; it is not an account of state of the art geostatistics at the millenium. It is also available on CD, which includes the data and software so that the analyses can be reproduced by self-teachers.

Ten data sets, taken from a variety of sources, are used throughout the book in a series of worked examples to illustrate the various applications of statistics and geostatistics. There is a heavy emphasis on ordinary statistics as the book is divided into three parts: 184 pages of statistics, 158 of geostatistics and 39 of tables. The ten data sets are listed also.

The first section begins with the simple statistics of means, variances, skewness and kurtosis. The next covers normal, lognormal, discrete, eg Poisson, and compound distributions which include Sichel's t-distribution. The usual hypothesis tests follow and the final section covers regression and trend surfaces. What distinguishes the first section from the usual dry text of a statistics primer is a wealth of pointed comment, which many may well find useful. Mathematics are not avoided. They are, instead, explained in the simplest possible terms and, as stated, illustrated by reference to worked examples based on the data sets. There may be more statistics here than are needed for everyday usage, but that is probably better than neglecting the finer points completely.

The second section on geostatistics begins with a discussion on some aspects of spatial estimation, eg inverse distance weighting, anisotropy, clustered data, skewed distributions, non-stationarity, and it ends with a number of maps produced by inverse distance weighting. This is followed by details of how an experimental semi-variogram is calculated, and then by descriptions of the semi-variogram models in common use. This section concludes with more examples on trend surfaces in the case of non-stationary data.

Estimation and kriging follow, beginning with some further discussion on anisotropy, skewed distributions, non-stationarity and in particular the nugget variance; is gamma (g ) really zero or has it a value of Co? This question is discussed several times in the book and the authors' comments lead one to believe that many users of black box software are unaware of the answer. It is time for them to take up their tin openers. Some 30 pages follow on estimating point values by ordinary kriging and by simple kriging, to show how the kriging equations are set up and kriging variances determined. Model validation by point kriging naturally follows this.

The section on kriging the contents of areas and volumes is fairly short, most of the work having been covered under the point kriging section which preceded it. The reader is, in part, referred to Dr Clark's 1979 book.

The final section introduces universal kriging, lognormal kriging, indicator kriging and rank uniform kriging. The description of the UK matrix is particularly clear, and indicator and uniform kriging are well covered. Probability kriging is not discussed, possibly because the authors wished to avoid having to embark on co-kriging as well.

The authors have written an introductory text that will take its readers up to a good standard of the practical usage of geostatistics, and they both have many years' experience of teaching the subject at different levels. They make no claims of trying to reach higher that that, but undoubtedly there is enough here to keep their readers busy for long enough and by persevering turn themselves into competent estimators of spatial phenomena.

Reviewed by Dr A.G. Royle, CEng, FIMM

Published by Ecosse North America Llc, Columbus, Ohio, USA

21.5 cm x 28 cm

ISBN: 0-9703317-0-3