# LEARNING GEOSTATISTICS THROUGH INTERACTIVE MODULES BASED ON R-SOFTWARE

A. Vervoort, A. Govaerts, P. Darius

K.U.Leuven (BELGIUM)
The concepts of geostatistics are commonly difficult to understand for students, even for those with a strong mathematical background. A main problem is certainly the link between the variation of the parameter value in space or time, and the calculated experimental semivariogram as a function of the interdistance, but over the entire area investigated. Furthermore first time users of geostatistics often do not want to struggle through different books with the whole geostatistical background before using it. Too often this leads to using geostatistics as a black box with all the risks involved. Although kriging claims to be the best linear unbiased estimator, it can only fulfil this role when the spatial variation is properly translated in a correct semivariogram model. And the latter is significantly affected by the sampling campaign and the way the semivariogram is calculated and modelled afterwards.

That is the main reason that some interactive tools based on the free public domain software R have been developed. They aim to assist students and teachers, but also individual first time users to get a feeling of the importance of the various aspects in conducting a good geostatistical study. In this way, valuable experience can be gained prior to starting real estimation projects, without being confronted with the consequences of a bad study. The geostatistical interactive modules can be accessed through: www.bwk.kuleuven.be/geostatistics

So far, three different modules are developed. The first module allows interactive experimentation with the experimental semivariogram of different datasets in one direction. Here the user can learn to understand the experimental semivariogram, but also the effect of the number of sampling points, the length of the dataset, and other parameters. The second module goes a bit further than the first, this time the datasets are two-dimensional. This module does not only focus on how to calculate the semivariogram, but it also looks at modelling the semivariograms. The third module is about kriging with two-dimensional datasets. One can choose sampling points, an unknown point and a semivariogram model, to estimate the value in the unknown point. The results of this kriging procedure are compared with the results from conventional statistics. In this module one can learn about the effect of the sampling positions, the position of the unknown point and the semivariogram model on the estimation results.

Experience when using the modules in the past 6 months in MSc courses has shown that these modules have a significant added value to get familiarized with the basic concepts of geostatistics, but that a certain supervision when using the modules is needed. A manual has been prepared where specific problems and questions are presented, which can be solved by using the different modules.