Goovaerts, P. 1997. Geostatistics for Natural Resources Evaluation. Oxford Univ. Press, New-York, 483 pages.
The book is addressed to students and practitioners with an undergraduate knowledge of statistics who want to gain an understanding of the methodology. The various tools are illustrated using soil contamination data that are provided with the book and were analysed using the public-domain geostatistical software library GSLIB , which makes the book highly suitable for teaching both theory and practice of geostatistics.
The book is subdivided into seven chapters and the presentation follows the typical steps of a geostatistical analysis. After an exploratory analysis of the pollutant data, the random function model is introduced and the problem of inferring its parameters from the sample information is addressed. Theoretical and practical issues related to modeling experimental direct and cross semivariograms are discussed. Two chapters give a detailed description of the multiple variants of kriging algorithms for interpolating values of a continuous attribute using only values of the attribute under study or accounting for secondary information. Prediction performances of the different methods are compared using reestimation scores of pollutant data.
Gaussian and indicator algorithms for modeling local probability distributions are then introduced. Several ways of accounting for this uncertainty assessment in decision-making processes such as delineation of contaminated areas are reviewed. The last chapter is devoted to stochastic simulation and presents algorithms for generating multiple realizations of the spatial distribution of either continuous or categorical attributes, e.g. sequential indicator or Gaussian simulation, p-field algorithm, LU decomposition, simulated annealing. In short, the book aims at bridging the gap between Isaaks and Srivastava's introductory book and GSLIB's more complete user guide.
Chapter 1: Introduction 5
1.1 The Jura Data Set 6
1.2 Plan of The Book 8
1.3 Terminology 9
Chapter 2: Exploratory Data Analysis 11
2.1 Univariate Description 11
2.1.1 Categorical variables 11
2.1.2 Continuous variables 13
2.2 Bivariate Description 21
2.2.1 The scattergram 21
2.2.2 Measures of bivariate relation 23
2.3 Univariate Spatial Description 24
2.3.1 Location maps 25
2.3.2 The h-scattergram 27
2.3.3 Measures of spatial continuity and variability 28
2.3.4 Application to indicator transforms 34
2.3.5 Spatial continuity of metal concentrations 38
2.4 Bivariate Spatial Description 48
2.4.1 The cross h-scattergram 48
2.4.2 Measures of spatial cross continuity/variability 48
2.4.3 The scattergram of h-increments 51
2.4.4 Measures of joint variability 52
2.4.5 Application to indicator transforms 54
2.4.6 Spatial relations between metal concentrations 56
2.5 Main Features of the Jura Data 58
Chapter 3: The Random Function Model 61
3.1 Deterministic and Probabilistic Models 61
3.2 The Random Function Model 65
3.2.1 Random variable 65
3.2.2 Random function 70
3.2.3 Multivariate random function 74
Chapter 4: Inference and Modeling 77
4.1 Statistical Inference 77
4.1.1 Preferential sampling 78
4.1.2 Histogram declustering 79
4.1.3 Semivariogram inference 84
4.1.4 Covariance inference 88
4.2 Modeling a Regionalization 89
4.2.1 Permissible models 89
4.2.2 Anisotropic models 92
4.2.3 The linear model of regionalization 97
4.2.4 The practice of modeling 99
4.3 Modeling a Coregionalization 109
4.3.1 Permissible models 110
4.3.2 The linear model of coregionalization 110
4.3.3 The practice of modeling 118
Chapter 5: Local Estimation: Accounting for a Single Attribute 127
5.1 The Kriging Paradigm 127
5.2 Simple Kriging 129
5.3 Ordinary Kriging 134
5.4 Kriging with a Trend Model 141
5.5 Block Kriging 154
5.6 Factorial Kriging 160
5.7 Dual Kriging 171
5.8 Miscellaneous Aspects of Kriging 176
5.8.1 Kriging weights 176
5.8.2 Search neighborhood 180
5.8.3 Kriging variance 181
5.8.4 Re-estimation scores 183
Chapter 6: Local Estimation: Accounting for Secondary Information 187
6.1 Exhaustive Secondary Information 187
6.1.1 Kriging within strata 189
6.1.2 Simple kriging with varying local means 192
6.1.3 Kriging with an external drift 196
6.1.4 Performance comparison 201
6.2 The Cokriging Approach 205
6.2.1 The cokriging paradigm 205
6.2.2 Simple cokriging 207
6.2.3 Ordinary cokriging 226
6.2.4 Standardized ordinary cokriging 234
6.2.5 Principal component kriging 235
6.2.6 Colocated cokriging 237
6.2.7 Accounting for soft information 243
6.2.8 Performance comparison 250
6.2.9 Multivariate factorial kriging 253
Chapter 7: Assessment of Local Uncertainty 261
7.1 Two Models of Local Uncertainty 261
7.1.1 Local confidence interval 263
7.1.2 Local probability distributions 264
7.2 The MultiGaussian Approach 267
7.2.1 The multiGaussian model 267
7.2.2 Normal score transform 268
7.2.3 Checking the multiGaussian assumption 273
7.2.4 Estimating the Gaussian ccdf parameters 277
7.2.5 Increasing the resolution of the sample cdf 280
7.3 The Indicator Approach 286
7.3.1 Indicator coding of information 287
7.3.2 Updating into ccdf values 295
7.3.3 Accounting for secondary information 308
7.3.4 Correcting for order relation deviations 321
7.3.5 Interpolating/extrapolating ccdf values 328
7.3.6 Modeling uncertainty for categorical attributes 330
7.4 Using Local Uncertainty Models 333
7.4.1 Measures of local uncertainty 335
7.4.2 Optimal estimates 342
7.4.3 Decision making in the face of uncertainty 349
7.4.4 Simulation 353
7.4.5 Classification of categorical attributes 356
7.5 Performance Comparison 360
Chapter 8: Assessment of Spatial Uncertainty 371
8.1 Estimation versus Simulation 371
8.2 The Sequential Simulation Paradigm 378
8.3 Sequential Gaussian Simulation 382
8.3.1 Accounting for a single attribute 382
8.3.2 Accounting for secondary information 387
8.3.3 Joint simulation of multiple variables 392
8.4 Sequential Indicator Simulation 395
8.4.1 Accounting for a single attribute 397
8.4.2 Accounting for secondary information 402
8.4.3 Joint simulation of multiple variables 402
8.5 The LU Decomposition Algorithm 405
8.6 The p -field Simulation Algorithm 407
8.7 Simulated Annealing 411
8.7.1 Simulated annealing paradigm 411
8.7.2 Implementation tips 414
8.8 Simulation of Categorical Variables 422
8.9 Miscellaneous Aspects of Simulation 426
8.9.1 Reproduction of model statistics 428
8.9.2 Visualization of spatial uncertainty 433
8.9.3 Choosing a simulation algorithm 436
Chapter 9: Summary 439
Appendix 445
A Fitting an LMC 445
B List of Acronyms and Notation 449
B.1 Acronyms 449
B.2 Common notation 450
C The Jura data 459
C.1 Prediction and validation sets 459
C.2 Transect data set 466
Bibliography 467
Index 479
The book is published by Oxford University Press (ISBN: 511538-4) in a new series called "Applied Geostatistics" which also includes the book by Isaaks and Srivastava and the second edition of GSLIB. It costs $137 in the US (487 pages, hard cover) and could be ordered through amazon.com .