Presence and prediction of fractal behavior in particle-size distributions as affected by the sample pretratment and soil properties

The fragmentation fractal dimension has been used to characterize soil particle and aggregate-size distributions. Deviations from strict selfsimilarity were often reported. In particle-size distribution determination, different dispersion and pretreatment methods can create different fragmentation. The objectives of this work were (i) to investigate fractal behavior of soil as influenced by the presence of the fragmentationenhancing pretreatment before particle-size distribution determination and (ii) to predict the presence/absence of the fractal behavior from soil chemical and physical properties by applying discriminant analysis and classification trees and comparing the efficiency of the two methods. Paired particle-size distribution determinations with and without removal of organic and inorganic binding agents were conducted for 85 samples of soils from various genetic and textural classes found in southern Italy. A total of 20 chemical and physical properties of soil were measured in the same samples. Discriminant analysis and classification trees were applied to predict the presence or absence of the fractal scaling in particle-size distribution from soil basic properties. Only 31 samples displayed strict fractal behavior after pretreatment, whereas 44 did when the aggregating agents were not removed. Classification trees rendered better prediction of the presence of the strict fractal behavior of soil particle-size distributions for each of the determination methods when compared with the discriminant analysis. The presence of the strict fractal behavior was primarily defined by fine silt content and influenced by contents of coarse silt, coarse sand, and exchangeable calcium content.