Abstract Soil physicochemical properties and microbial communities are highly heterogeneous and vary widely over spatial scales, necessitating careful consideration of sampling strategies to provide representative and reproducible soil samples across field sites. To achieve this, the study aimed to establish appropriate sampling methodology and to determine links between the variability of parameters, utilising two sampling strategies. The first (design 1) involved extracting 25 cores from random locations throughout the field and pooling them into five sets of five cores. The second (design 2) involved a further 25 cores within five 1 m 2 sub-plots. Sub-samples from each sub-plot were pooled in order to determine between and within sub-plot variability. All samples were analysed independently and as pooled sub-samples. Results indicate that pooling spatially separated samples significantly reduced the variability in pH, compared to individual samples. Pooling samples from a small area resulted in lower within sub-plot variability than between sub-plots for pH and bacterial community composition assessed by terminal-restriction fragment length polymorphism analysis. Following multivariate statistical analysis, a large amount of variation in community composition was explained by soil pH, which is remarkable given the relatively small size of the sampling area and minor differences in pH. Moisture content was also important in determining bacterial communities in the random design (design 1). In the 1 m 2 sub-plot design (design 2), the spatial location of the plots explained a large degree of the variation in bacterial community composition between plots, which was due to spatial autocorrelation of pH and possible additional environmental parameters. This study emphasises the importance of sampling design for obtaining representative samples from soil.