The papers in this Geotechnical Special Publication explain and demonstrate a broad range of methods of probabilistic site characterization using soil data obtained at one or more National Geotechnical Experimentation Sites (NGES), in particular the NGES locations at Texas A&M University, Treasure Island Naval Station, University of Massachusetts at Amherst, University of Houston, and Northwestern University. Among the topics covered are statistical estimation procedures based on homogenous random field theory, with as parameters the mean, coefficient of variation and scale of fluctuation, as well as two alternate fractal representations of spatial variation in soil deposits, a neural network model, and an approach to correlating soil properties based on the concept of fuzzy subsets. The aim of this publication is to provide both a state-of-the-art assessment of statistical methods of soil profile modeling and a set of baseline statistics, representative of the well-documented NGES, intended to serve as a priori information for probabilistic site characterization.

Random variation is a fact of life that provides substance to a wide range of problems in the sciences, engineering, and economics. There is a growing need in diverse disciplines to model complex patterns of variation and interdependence using random fields, as both deterministic treatment and conventional statistics are often insufficient. An ideal random field model will capture key features of complex random phenomena in terms of a minimum number of physically meaningful and experimentally accessible parameters. This volume, a revised and expanded edition of an acclaimed book first published by the M I T Press, offers a synthesis of methods to describe and analyze and, where appropriate, predict and control random fields. There is much new material, covering both theory and applications, notably on a class of probability distributions derived from quantum mechanics, relevant to stochastic modeling in fields such as cosmology, biology and system reliability, and on discrete-unit or agent-based random processes.Random Fields is self-contained and unified in presentation. The first edition was found, in a review in EOS (American Geophysical Union) to be “both technically interesting and a pleasure to read … the presentation is clear and the book should be useful to almost anyone who uses random processes to solve problems in engineering or science … and (there is) continued emphasis on describing the mathematics in physical terms.”

The papers in this Geotechnical Special Publication explain and demonstrate a broad range of methods of probabilistic site characterization using soil data obtained at one or more National Geotechnical Experimentation Sites (NGES), in particular the NGES locations at Texas A&M University, Treasure Island Naval Station, University of Massachusetts at Amherst, University of Houston, and Northwestern University. Among the topics covered are statistical estimation procedures based on homogenous random field theory, with as parameters the mean, coefficient of variation and scale of fluctuation, as well as two alternate fractal representations of spatial variation in soil deposits, a neural network model, and an approach to correlating soil properties based on the concept of fuzzy subsets. The aim of this publication is to provide both a state-of-the-art assessment of statistical methods of soil profile modeling and a set of baseline statistics, representative of the well-documented NGES, intended to serve as a priori information for probabilistic site characterization.

Random variation is a fact of life that provides substance to a wide range of problems in the sciences, engineering, and economics. There is a growing need in diverse disciplines to model complex patterns of variation and interdependence using random fields, as both deterministic treatment and conventional statistics are often insufficient. An ideal random field model will capture key features of complex random phenomena in terms of a minimum number of physically meaningful and experimentally accessible parameters. This volume, a revised and expanded edition of an acclaimed book first published by the M I T Press, offers a synthesis of methods to describe and analyze and, where appropriate, predict and control random fields. There is much new material, covering both theory and applications, notably on a class of probability distributions derived from quantum mechanics, relevant to stochastic modeling in fields such as cosmology, biology and system reliability, and on discrete-unit or agent-based random processes.Random Fields is self-contained and unified in presentation. The first edition was found, in a review in EOS (American Geophysical Union) to be “both technically interesting and a pleasure to read … the presentation is clear and the book should be useful to almost anyone who uses random processes to solve problems in engineering or science … and (there is) continued emphasis on describing the mathematics in physical terms.”