Combining marginal and spatial outliers identification to optimize the mapping of the regional geochemical baseline concentration of soil heavy metals.

The geochemical baseline concentration is used as a reference to determine the state of an area in relation to soil pollution. Various methods have been developed to determine this concentration based on filtering either the marginal or the spatial outliers. Marginal outlier identification (MOI) classifies data as belonging to the geochemical baseline or representing pollution using a globally defined single threshold value. As a result it neglects the local scale variability of the geochemical baseline level that arises from possible differences in parent material and the presence of multiple pollutants with variable degrees of influence. Hence it might lead to the identification of enrichments below the globally defined threshold but still larger than the local geochemical baseline level as belonging to the geochemical baseline. Spatial outlier identification (SOI) focuses on detecting unusual values in a local neighbourhood. As SOI is strongly dependent on data configuration, clusters of high values might wrongly be accepted as being geochemical baseline data that can inflate geochemical baseline level in pollution risk areas. The limitations of MOI and SOI can be severe when applied for a large scale study. To avoid these limitations and maximize the benefit of the two methods we proposed a combined methodology: integrated outliers identification (IOI) using fuzzy and robust means to determine the geochemical baseline measurements of Cr for Flanders, Belgium. Through the use of IOI it was possible to identify both scattered and clustered outliers resulting in determination of Cr geochemical baseline level that does not deny the local as well as the regional scale variability and display a higher degree of spatial structure as expected for the geochemical baseline data.