Using model-based geostatistics to develop design-based sampling guidelines for estimating arsenic contamination around cattle-dip sites

In the past Arsenical pesticides were used across New South Wales, Australia to control ticks in livestock at now disused cattle dip sites which resulted in the contamination of surrounding soils with elevated and extremely variable As concentrations. Estimating the spatial distribution of As in soil is imperative to delineate its level and spread of contamination. Therefore, the aims of this study were to (1) quantify spatial-variation in soil As content around a cattle dip site and map this using linear mixed models, and (2) use the information on the spatial variability pattern to give guidelines on the sample size required to estimate the mean soil As values. The soil cores (n=102) were taken in a systematic design (0.3 m×0.35 m) to a depth of 0−0.6 m and cut into three sections to give 0−0.2, 0.2−0.4 and 0.4−0.6 m depths. The results demonstrated that total (at 0–0.2 m depth) and phosphate extractable (at three depths) As concentrations varied over short (\u3c 1 m) distances at their corresponding depths, and this extreme variability was apparent in the predicted maps. Both total and extractable soil As concentrations significantly (p \u3c0.001) increased towards the cattle dip (northings-trend). Using the information of this spatial variability trend, we suggest that for this site a sample size of 15 would be sufficient to statistically prove that mean total As (826 mg/kg) with standard error of mean (sem) of 99.0 and a confidence interval of [613.7, 1038.3] in the surface soil was higher than the ecological investigation level (20 mg/kg) limit of As in soil. Based on this we propose that a generic sampling scheme should have 5–20 samples, where sampling area around the dip bath can be stratified based on the closeness to the cattle dip