Submitted by ASA-ENVR for competition;
Co-sponsored by the International Association for
Mathematical Geology
Session Slot: 8:30-10:20 Tuesday
Estimated Audience Size: 75-100
AudioVisual Request: xxx
Session Title: Advances in Geostatistics
Theme Session: No
Applied Session: No
Sponsored by: ASA Section on Statistics and the Environment (ASA-ENVR)
and by the International Association for Mathematical Geology (IAMG).
The primary aim of the IAMG is to promote international cooperation in the
application and use of mathematics in geological research and
technology. In keeping with this aim, IAMG is interested in fostering
increased collaboration with statistical organizations, hoping that
this will result in increased cross-disciplinary research. Two of
IAMG's journals, Mathematical Geology and Computers and Geosciences,
regularly publish state-of the-art theory and methodology pertaining to
environmental statistics, spatial statistics, and geostatistics. Many
ASA/ASA-ENVR statisticians, particularly those whose research interests
include spatial statistics, are members of IAMG or read its journals.
Session Content: The purpose of the session is to familiarize the JSM
audience with the statistical work of internationally-renown researchers
in other disciplines that routinely use and develop statistical
methodology, most notably in geostatistics, in their research. The
intent is to foster future collaboration/educational opportunities
between interested JSM participants and IAMG scientists and to encourage
ASA participation at international conferences sponsored by the IAMG or
other, similar international organizations. Most of the speakers for
this session do not usually attend the JSM, and many JSM members may not
attend the international meetings that often serve as the primary forum
for the introduction of new methods in spatial statistics and
geostatistics. Since these speakers do not normally attend the JSM,
necessary funding for these speakers will be provided by the IAMG.
Session Organizer: Gotway Crawford, Carol A. Centers for Disease Control and Prevention
Address: Centers for Disease Control and Prevention, Atlanta, GA 30333
Phone: (770) 488-7428
Fax: (770) 488-7335
Email: cdg7@cdc.gov
Session Timing: 110 minutes total (Sorry about format):
110 minutes total First Speaker - 30 minutes Second Speaker - 30 minutes Third Speaker - 30 minutes Discussant - 15 minutes Floor Discussion - 5 minutes
Session Chair: Gotway Crawford, Carol A. Centers for Disease Control and Prevention
Address: Centers for Disease Control and Prevention, Atlanta, GA 30333
Phone: (770) 488-7428
Fax: (770) 488-7335
Email: cdg7@cdc.gov
1. Geostatistical Estimation Applied to Highly Skewed Data
Clark, Isobel, Geostokos Limited
Address: Geostokos Limited, Alloa Business Centre, Whins Road, Alloa FK10 3SA Central Scotland
Phone: +44 1259 211904
Fax: +44 1259 211904
Email: isobel@stokos.demon.co.uk
Abstract: Many variables in the geological and earth sciences follow highly skewed distributions. For example: precious metal grades; toxic pollution of groundwater aquifers; grain size distributions and so on. Transformations can be applied to 'normalise' such data and geostatistical studies have been undertaken on these transforms. The major problem is to 'backtransform' predicted values to provide unbiased and reliable estimates for the original measurements. This presentation will use a case study on lognormal data to illustrate various backtransformations which have been published and/or used in practice. In particular, the problem of predicting the average value over an area or volume will be considered in detail.
2. Conditional Simulation of Random Sets
Lantuéjoul, Christian, Ecole des Mines de Paris
Address: Ecole des Mines de Paris, Centre de Géostatistique, 35 rue Saint-Honoré, 77305 Fontainebleau, France
Phone: (33) (1) 64 69 47 60
Fax: (33) (1) 64 69 47 05
Email: lantu@cg.ensmp.fr
Abstract: Predicting the geometry of a porous medium is a standard problem encountered in many domains of applications. It is often tackled by simulations (Monte Carlo procedure) which provide not only a prediction, but also its variability.Any simulation exercise requires a stochastic model describing the statistical properties of the porous medium. This model can be defined either explicitly or implicitly, and in the second case, it is not always easy to check whether it makes sense. This question will be considered first.
In order to reduce the variability of the prediction, it is preferable to consider only conditional simulations. These are simulations compatible with the available information, typically honoring data points, satisfying domain proportions or respecting connectivity constraints. The second part of the presentation will be devoted to a review of conditional simulation algorithms as a function of the model considered and of the set of conditions to be respected. Several examples will be given to illustrate these ideas.
3. Incorporating Model Uncertainty in Geostatistical Methods of Risk Analysis
Dowd, Peter A., Department of Mining and Mineral Engineering
Address: Department of Mining and Mineral Engineering, University of Leeds, Leeds LS2 9JT England, UK
Phone: (44) (113) 233-2796
Fax: (44) (113) 246-7310
Email: P.A.Dowd@leeds.ac.uk
Abstract: A major use of conditional simulation is as a means of risk assessment. In many earth science applications accurate field measurement of a single value of a variable (e.g. fractures, permeability) is difficult and measurement of all values is impossible. Thus a stochastic approach, such as conditional simulation, is the only effective way of modelling these variables. However, such approaches are based on models which can only be inferred from sparse data. Often the uncertainty on the model is greater than the uncertainty in the variable. The talk will look at ways in which model uncertainty can be incorporated into the simulation. In particular, the maximum likelihood approach will be discussed.
Discussant: Hohn, Michael Ed. West Virginia Geological and Economic Survey
Address: West Virginia Geological and Economic Survey, P.O. Box 879, Morgantown, WV 26507-0979
Phone: 304-594-2331
Fax: 304-594-2575
Email: hohn@geosrv.wvnet.edu
List of speakers who are nonmembers: Clark, Lantuejoul, Dowd -- (Hohn is ASA member 4-4)