Report No DWI0660


Final Report to the Department of the Environment


Jun 1986


The extent to which current soil sampling techniques provide reliable estimates of soil metal concentrations has been examined. Three main aspects were investigated:

    1. The magnitude of spatial variations in soil metal concentrations;
    2. The effect of the number of sub-samples on the bias and precision of soil metal concentration estimates; and
    3. The extent to which the choice of sampling device influences the bias or error in estimates of soil metal concentrations.

Two field sites, one arable and one pasture, were sampled by both WRc and a Water Authority over a 10 m grid pattern. Soil samples were analysed for cadmium, copper, nickel, zinc, lead, molybdenum and chromium. In addition, archive data from a further arable site was made available by the Authority.

In all three cases, statistically significant differences in soil metal concentration were found across both the length and breadth of the grid, although their extent was different for each of the fields investigated.

A computer program was developed which enabled the effects of a range of conventional 'W' sampling patterns to be simulated using the data obtained from the field experiments. The results showed that:

    1. For conventional 'W' sampling patterns based on 25 subsamples, mean soil metal concentrations could vary by up to 10%.
    2. In addition, the choice of orientation of the 'W' could lead to biases of up to 5% of the 'true' mean. Thus a positive bias of this magnitude, coupled with a precision of 10%, would result in an overall error range of -5% to +15% of the 'true' mean soil concentration.
    3. 'W'-pattern sampling was more precise than random sampling when only 10 sub-samples were used, but this effect was reduced, and even in some cases reversed, with sub-sample numbers above 30.

Differences between the performance of sampling devices made the greatest contribution to the overall error in a composite sample; in one instance - for copper in the arable field - a bias of 64% in the estimated concentration was found. Moreover, cheese and pot auger samples were more variable than those obtained using a tube auger.

A review of sewage sludge and soil sampling procedures currently in use by the water industry has been undertaken. This has been reported separately, together with recommendations for further work, in report DoE 1261-M, June 1986.

Copies of this report may be available as an Acrobat pdf download under the 'Find Completed Research' heading on the DWI website.