The assessment of point-source and diffuse soil metal pollution using robust geostatistical methods: a case study in Swansea (Wales, UK)

The spatial variation of soil metal content arising from diffuse pollution in industrial regions cannot be analysed by conventional geostatistical methods because predictions are influenced by metal content from natural sources and extreme values from point-source pollution. We analyse a survey of soil arsenic, copper, lead, and tin at 372 locations around Swansea (Wales, UK). We use the approach of Hamon et al. (2004) to determine the natural metal concentrations in contaminated regions from the iron content. However, we find that this indicator is not appropriate to the area around Swansea because the iron content is elevated across the contaminated region. Therefore the natural concentration of each metal is approximated by the median concentration on nearby uncontaminated rural soils on the same parent material. We divide the remaining variation between diffuse pollution and point-source pollution by the robust winsorizing algorithm of Hawkins & Cressie (1984). This leads to a plausible log-Gaussian model with a constant mean which represents the diffuse pollution and estimates of the contribution of point-source pollution at each observation site. Point-source pollution occurs at sites historically associated with production, transport and disposal of industrial wastes. The pattern of diffuse pollution is consistent with emissions from multiple smelters located throughout urban Swansea and the effects of prevailing wind and topography are evident.