The present study aimed at investigating whether a nonlinear and scaling approach is suitable for statistically characterizing the spatial variability of soil geochemical element contents at field conditions. Spatial distributions of 20 soil geochemical elements' contents in an agricultural field soil were investigated. Two indicators of nonlinear scaling were employed, empirical probability distribution functions (pdfs) and structure functions. For the pdfs, a trend was discovered with major elements naturally occurring in the soil being hyperbolic and minor trace elements being close to Gaussian. Further, fertilizer management was found to modify the behavior of related elements. The structure functions were generally nonscaling for major elements and scaling for minor elements, however, without any noticeable effect of fertilizer application. The scaling was of both monoaffine and multiaffine type. This suggests that the random cascade models recently used within geophysics may be useful also for geochemical element variability in soil. ¿ 2001 American Geophysical Union