Geostatistical Simulation

1 Simulation versus estimation The following problem was raised by Alfaro (1979). A submarine cable has to be laid across the straits of Gibraltar. How can its length be predicted if the depth of the sea floor has been measured sparsely along its trajectory? Fig. 1. 1. Part of the actual trajectory and sample data points An exact determination of the length requires knowledge of the depth at each point of the trajectory. But these are mostly unknown. In a geostatistical set ting, they are considered as random and can be estimated by linear regression starting from the available data points. This suggests estimating the actual length as the length of the estimated trajectory. The results turn out to be disappointing. The length of the trajectory is seriously underestimated (see Figure 1. 2). Clearly, the estimated trajectory is much smoother than the actual one. Fig. 1. 2. Part of the actual trajectory and its estimate from linear regression. In this particular example, the estimated trajectory is piecewise linear because the linear regression has been carried out using an exponential covariance function 2 1. Introduction What is really questionable in this procedure is not the construction of an estimator for the length starting from the depth estimator, but the depth estimator itself. Linear regression estimation requires only the mean and the covariance function. But the covariance function does not tell us much about the length of the trajectories. Figure 1.

Clear separation of models, from methods and algorithms Coherent overview of the subject Includes supplementary material: sn.pub/extras

Klappentext
Within the geoscience community the estimation of natural resources is a challenging topic. The difficulties are threefold: Intitially, the design of appropriate models to take account of the complexity of the variables of interest and their interactions. This book discusses a wide range of spatial models, including random sets and functions, point processes and object populations. Secondly,the construction of algorithms which reproduce the variability inherent in the models. Finally, the conditioning of the simulations for the data, which can considerably reduce their variability. Besides the classical algorithm for gaussian random functions, specific algorithms based on markovian iterations are presented for conditioning a wide range of spatial models (boolean model, Voronoi tesselation, substitution random function etc.) This volume is the result of a series of courses given in the USA and Latin America to civil, mining and petroleum engineers, as well as to gradute students is statistics. It is the first book to discuss geostatistical simulation techniques in such a systematic way.

Inhalt

Introduction.- 2. Investigating stochastic models.- 3. Variographic tools.- 4. The integral range.- 5. Basic morphological concepts.- 6. Stereology: some basic notions.- 7. Basics about simulations.- 8. Iterative algorithms for simulation.- 9. Rate of convergence of iterative algorithms.- 10. Exact simulations.- 11. Point processes.- 12. Tessellations.- 13. Boolean model.- 14. Object based models.- 15. Gaussian random function.- 16. Gaussian variations.- 17. Substitution random functions.