Introduction¶

Didier RENARD, Nicolas DESASSIS, Mike PEREIRA, Xavier FREULON, Fabien ORS

In this document, you will find the list of the different files which will be browsed during this course as well as a short description.

Global presentation¶

What is gstlearn?

• Discover the web site
• Documentation (Application Programming Interface, Tutorials)
• Road Map (Stable version vs. Version under development). Github site
• Internal organization: C++, R, Python (accessor, graphics, ...)
• Some important features: . Object language . Processing through SWIG . Memory and garbage collector . Adapting documentation (built using Doxygen from C++)
• How to install it (Rstudio, Jupyter-notebook, Terminal)

Overview of Geostatistics¶

The principle is to browse the different steps of a standard geostatistical study.

Get familiar with the library¶

The contents of this paragraph is described in the Tutorial for Db.

1. Download Data Sets (using URL): we will use the 2D Scotland files.

• Pay attention on the language and platform specifics
• Other importing format (CSV, Panda Frame, Neutral File)
• Various features for accessing data
• Unique name for variables
• Discussion on locators
• Particular file organization (Grid)

1. Basic operations using a (numerical) Data Base (Db)

• Adding and suppressing one or several variable(s)
• Masking samples using selection (only keep inshore information)

1. Basic statistics

• Summarize basic statistics as a table
• Show (C++) documentation (interpretation of Doxygen information)

Basic Geostatistics¶

1. Variography

• The contents of this chapter is described in the Tutorial for Variography
• Variogram cloud
• Experimental variogram: isotropic and directional variogram
• Fitting a Model: automatic version, selecting basic structures, with constraints ,...
• Variogram maps

1. Estimation using Kriging

• The contents of this chapter is described in the Tutorial for Estimation
• Simple Kriging
• Ordinary Kriging (adding Universality condition)
• Produce maps for estimation and standard deviation of the estimation errors
• Neighborhood definition (unique, moving, possibility to introduce faults)
• Cross-validation (K-Fold option)

1. Simulations

• The contents of this chapter is described in the Tutorial for Simulations
• Using Turning Bands method
• Non-conditional or conditioning to data

1. Multivariate case

• The contents of this chapter is described in the Tutorial for Multivariate
• Define several target variables simultaneously and re-run previous steps
• Simple and cross-variograms
• Coiging

Advanced usage¶

How to turn gstlearn for personal usage?

• Object principle (heritage)
• Example: enhance one capability of Moving Neighborhood search

Quick overview of SPDE¶

• Some theory (meshing, precision matrix, sparse matrix)
• Using API for performing estimation and somulations
• Highlighting the non-stationarity: variable anisotropy
• Working in different spaces: on the sphere