Interactive visualization applied to multivariate geochemical data: A case study

Abstract

Geochemical survey data have commonly been analysed combining methods from several disciplines - statistics, geostatistics, geographic information technology, visualization. In initial stages of analysis, tables are often used to describe the data and present statistical measures. Far too often the original data are manipulated in one or another way, for example, using mathematical transformations, or interpolation of points to a surface. It is the author's opinion that raw geochemical data should be used in initial stages of data description, thus preserving the original details. This is not a simple task, as geochemical data are commonly complex, multivariate, and collected on irregular grid. Data contain outliers, element contents vary within thousands of ppm (parts per million), and different chemical elements may be correlated. In the present study a graphical approach has been used to study distribution of 5 heavy metals in glacial till. Using interactive visualization and multiple linked views of the data, the following issues were addressed: multi-element outliers, spatial trends, multi-element correlations and patterns. Interactive graphical techniques proved to be especially suitable for studying outliers and identifying and locating samples that are redundant and may be removed from data without loss of information. Visualization using linked views gave valuable insights about metal correlations and spatial trends. As the development of appropriate tools for analysing multivariate spatial data are still in its early stages, visualization freeware seems to be a good alternative providing powerful, easy to use and intuitive techniques for exploratory data analysis.