Spatial variability presents a challenge for effective measurement of ecological processes in soil ecosystems. This study characterized the heterogeneity of seven different biotic and abiotic soil properties at three ecological scales: land-use type (vegetation type), site (scale of km to tens of km, and subplot (10-30 m). We addressed three questions about soil properties: (i) Do they differ more at any ecological scale when compared with other scales? (ii) Does physically combining samples from replicate subplots affect their measurement? (iii) What are the implications of this variation for the detection of differences caused by changing land-use? There were five land-use types: remnant rainforest, pasture, ecological restoration plantings 10-15 yrs old, 40-50 yrs regrowth dominated by the non-native tree camphor laurel (Cinnamomum camphora) and rainforest regrowth 3-6 yrs after poisoning of camphor trees within older regrowth. Sites were interspersed within a 750 km2 region of subtropical Eastern Australia. In general, variability was greatest at the site and subplot levels. Soil water content, soil organic matter and pH were less variable than total organic carbon, microbial biomass, nitrate and nitrification. Power analyses using total organic carbon and nitrate showed that adequate replication at both site and subplot levels was important for detecting effects of land-use type, although increased replication was slightly more effective for subplots than for sites. Subplot spacing was relatively unimportant. Physically combining subplot samples to yield one aggregate measurement per site was successful in reducing analytical effort without sacrificing accuracy for all soil properties except total organic carbon and nitrification.