When preparing data such as point observations of soil attributes for geostatistical interpolation to map characteristic patterns and to determine areas where values may exceed required thresholds, it is useful first to examine the data for spatial homogeneity, stationarity and normality. Exploratory Data Analysis (EDA) is not only useful in the pre-interpolation phase, but it is also helpful for validating the results. The quality of kriging predictions is largely determined by the choice of variogram model, which can be seriously affected by a few extreme attribute values if these are closely located in space. Dynamic interaction with the data, in which several views of the data such as histograms, scattergrams, maps and variogram clouds can be examined simultaneously, permits the relations between extreme data values or trend residuals and geographical location to be easily seen. This study was carried out using the REGARD software package developed at Trinity College Dublin, with modifications made by us. The use of EDA for detecting spatial anomalies that can affect geostatistical interpolation is demonstrated using data on polluted floodplain soils in The Netherlands.