Agenda

Denis Allard, Biostatistique et Processus Spatiaux (BioSP), INRA, Avignon

Abstract:
The need for combining different sources of information in a probabilistic framework is a frequent task in earth sciences. For example, the probability of occurrence of a certain lithofacies at a certain location for example can easily be computed conditionally on the values observed at each source of information. The problem of aggregating these different conditional probability distributions into a single conditional distribution arises as an approximation to the inaccessible genuine conditional probability given all information.

This talk makes a formal review of most aggregation methods proposed so far in the literature with a particular focus on their mathematical properties. Exact relationships relating the different methods is emphasized. The concepts of calibration, sharpness, and reliability, well known in the weather forecasting community for assessing the goodness-of-fit of the aggregation formulas, and a maximum likelihood estimation of the aggregation parameters are introduced.

These results are illustrated on simulations and on from two common stochastic models for earth science and on sand dune images. It is found that the log-linear pooling provides the best prediction while the linear pooling provides the worst.

Joint work with Philippe Renard(2), Alessandro Comunian(2;3), Dimitri D’Or(4), Roland Froidevaux(4)
2 CHYN, Université de Neuchâtel, Neuchâtel, Switzerland
3 now at Dept. of Geoscience, Universita degli Studi de Milano
4 Ephesia Consult, Geneva, Switzerland

Bio Sketch:
Denis Allard obtained his MSc and PhD in Geostatistics from the Ecole des Mines de Paris (Fontainebleau, France). He has been Assistant Professor at the Statistics Department, University of Washington (Seattle, U.S.A.) and Senior Geostatistician with BP Exploration Production, Subsurface technology (UK). Since 1996, he holds a Senior Researcher position at INRA, Avignon (France).
From 2005-2011 he was head of the INRA unit BioSP (Biostatistics and Spatial Processes) in Avignon.
His research interests focus developing models and methods spatial and spatio-temporal environmental data, with two main application domains: soil science and geology on the one hand; and climate and weather variables in view of building stochastic weather generators on the other hand.