Dates: July 7th-8th (9am – 5pm)
Leaders: Nicolas Desassis & Didier Renard (Mines ParisTech, France)
Didier Renard graduated from the Ecole des Mines de Saint-Etienne, France. He is now a senior geostatistician and has been working in the Team of Geostatistics in the Center of Geosciences of MINES ParisTech for more than 30 years. At the head of the computer group, he actively contributes to the inception, development and testing of new models. He is one of the main authors of several well-known geostatistical packages such as BLUEPACK and more recently ISATIS. He also developed the geostatistical surface modeler ISATOIL dedicated to volumetric calculations in layer-cake environments. He is the main author of RGeostats which provides a complete toolbox of geostatistical methods available on the R platform. He carries consulting activities for petroleum industry and worked for companies such as Shell, Statoil, ENI, Total, BHP. Finally he is involved in educational activities, teaching courses to students of Paris School of Mines, giving lectures during geostatistical courses and training more several hundreds of practitioners all over the world.
Nicolas Desassis is currently a research fellow in the Geostatistics team of the Centre de Géosciences of MINES ParisTech. He received a master (2003) in biostatistics and a PhD (2007) in statistics from the University of Montpellier 2, France (partnership with the Biostatistics for Spatial Process laboratory, INRA, Avignon, France). He also worked at INRIA (National Institute of Research in Automatic and Informatics) (2008) on spatio-temporal simulation of forest dynamics. Dr. Desassis’s research interests are in the area of geostatistics and spatial modelling more specifically in the inference of the spatial models (automatic variogram fitting, plurigaussian models…) and in conditional simulations. He also contributes to develop new Bayesian methods for inverse problems in geophysics.
Description: The main objective of this 2-day course is to give an overview of the main geostatistical state-of-the-art concepts that are routinely applied in environmental sciences. It will focus on practical applications through a variety of examples or real case studies and underline the common pitfalls when using integrated computer-based modeling.
- Introduction to data analysis
- Definitions: random variables, random functions
- Experimental: measures, distributions, sampling, support
- Multivariate statistics
- Variography under stationary or intrinsic hypotheses
- Experimental variograms: calculated on regular sampling, irregular sampling, variogram cloud, variogram map
- Model characteristics: sill, range, anisotropies (zonal or geometric), nested structures, fitting procedures
- Estimation for gridding and spatial interpolation
- Simple and Ordinary Kriging, Kriging with external drift
- Cokriging and Collocated cokriging
- Neighborhood search, cross-validation
- Simulations for reproducing data variability while still honoring information
- Gaussian anamorphosis
- Simulation methods (Turning Bands)
The course involves a combination of theoretical discussions, application examples and computer-based exercises. All concepts are demonstrated using RGeostats (free R package downloaded from http://rgeostats.free.fr).
Workshop material: The course material can be downloaded from the RGeostats forum.