Variability in contamination by heavy metals: Sampling implications

The choice of a sampling design for assessing soil contamination is a key issue. The level and spatial pattern of contaminants needs to be taken into account by those buying, selling or developing such land. This paper evaluates sampling design and size in order to provide satisfactory and efficient estimates of contamination levels. Geostatistical methods are used to quantify the spatial dependence of heavy metals. Semivariance is calculated for log-transformed data to minimise the effects of extreme outliers. Several models at different sampling stages are presented for characterising spatial pattern of heavy metals at the investigated site. Variogram models are calculated for a range of sample sizes. The results are assessed by a jack-knifing technique and then used for block kriging. The overall conclusion is that a sample size of 50 to 60 is optimal for characterising variability in contamination levels at an old industrial site of 6.25 ha.