Geostatistical filtering to map a 3D anthropogenic pedo-geochemical background for excavated soil reuse

International audienceThe safe reuse of excavated soils in urban areas requires the guarantee of geochemical compatibility between this scarce resource and the local pedo-geochemical background. Previous works have shown that the usual statistical tools (high order quantiles, upper whiskers) are not always adequate for separating anomalies from background contents, especially when there are many anomalies.This paper presents an application of factorial co-kriging (geostatistical filtering) used to separate anomalies from the pedo-geochemical background, according to their different scales of spatial variability. The hypothesis is that the background corresponds to long distance variations, while the anomalies, which can be either positive or negative,- represent short distance variations.The variable of interest is the lead content (Pb) in the subsoils of the “Ile de Nantes” urban area, where some areas were previously dedicated to industrial activities. The data were collected from different surveys, not specifically aimed at determining the pedo-geochemical background, but rather in a context of pollution diagnostics. The lengths of the samples were strongly variable. In a first step, only the samples having approximately the same length were retained for the estimation, which significantly reduced the amount of data. In a second step, the geostatistical estimation was performed using all the data. The resulting maps are significantly more accurate for the total Pb and its background component. The distinction between anomalies and background is discussed based on the calculated maps.Factorial co-kriging appears to be relevant to map the anthropogenic pedo-geochemical background, using data acquired for other purposes. It allows separating anomalies and background for any parameter presenting significant contrasts