Variability in soil properties at different spatial scales (1 m-1 km) in a deciduous forest ecosystem

The purpose of this research was to test the hypothesis that variability in 11 soil properties, related to soil texture and soil C and N, would increase from small (1 m) to large (1 km) spatial scales in a temperate, mixed-hardwood forest ecosystem in east Tennessee, USA. The results were somewhat surprising and indicated that a fundamental assumption in geospatial analysis, namely that variability increases with increasing spatial scale, did not apply for at least five of the 11 soil properties measured over a 0.5-km² area. Composite mineral soil samples (15 cm deep) were collected at 1, 5, 10, 50, 250, and 500 m distances from a center point along transects in a north, south, east, and westerly direction. A null hypothesis of equal variance at different spatial scales was rejected (P?0.05) for mineral soil C concentration, silt content, and the C-to-N ratios in particulate organic matter (POM), mineral-associated organic matter (MOM), and whole surface soil. Results from different tests of spatial variation, based on coefficients of variation or a Mantel test, led to similar conclusions about measurement variability and geographic distance for eight of the 11 variables examined. Measurements of mineral soil C and N concentrations, C concentrations in MOM, extractable soil NH₄-N, and clay contents were just as variable at smaller scales (1-10 m) as they were at larger scales (50-500 m). On the other hand, measurement variation in mineral soil C-to-N ratios, MOM C-to-N ratios, and the fraction of soil C in POM clearly increased from smaller to larger spatial scales. With the exception of extractable soil NH₄-N, measured soil properties in the forest ecosystem could be estimated (with 95% confidence) to within 15% of their true mean with a relatively modest number of sampling points (n?25). For some variables, scaling up variation from smaller to larger spatial domains within the ecosystem could be relatively easy because small-scale variation may be indicative of variation at larger scales.