Abstract: Soil has significant spatial variability, and an important indicator for describing the spatial variability of soil is the autocorrelation distance. This paper proposes a method based on soil microstructure simulation to calculate the autocorrelation distance of soil. For generating the numerical model of the microstructure of a soil more reasonable, the traditional quartet structure generation set (QSGS) method is modified, by considering the particle size distribution information of the soil. The input parameters required for the modified quartet structure generation set (MQSGS) method can be obtained by combining the scanning electron microscopy test and the particle size analysis test of soils. Based on the generated numerical model of soil microstructure, the 2-point spatial autocorrelation function can be calculated and the autocorrelation distance of soil can be obtained by curve fitting for the 2-point spatial autocorrelation function. The study shows that soil microstructure generated by the MQSGS method is more similar to the real soil structure in nature than that generated by the QSGS method; the autocorrelation distances calculated from the microstructure numerical model generated by the MQSGS method are slightly smaller than those obtained by the QSGS method.

Key words: spatial variability, microstructure simulation, quartet structure generation set method, scanning electron microscope, particle size distribution, autocorrelation distance