Knowledge discovery in the environmental sciences: Visual and automatic data mining for radon problems in groundwater

Efficiently exploring a large dataset with the aim of forming a hypothesis is one of the main challenges in environmental research. The exploration of georeferenced environmental data is usually performed by established statistical methods. This paper presents an alternative approach. The aim of this study was to see if a visual data mining system and an integrated visual-automatic data mining system could be used for data exploration for a particular environmental problem: the occurrence of radon in groundwater. In order to demonstrate this, two data mining systems were built, one consisting of visualisations and the other including an automatic data mining method - a Self-Organising Map (SOM). The systems were designed for exploration of a large multidimensional dataset representing wells in Stockholm County.