Abstract: Abstract: Soil properties vary randomly from location to location, due mainly to different genetic types, geologic structures, depositional conditions, stress history, and weathering conditions. Serving as a key parameter to describe the spatial variability, the Scale of Fluctuation (SOF) can be defined as the minimum distance within which a signi?cant correlation exists among soil properties. Specifically, the spatial fluctuation of soil parameter decreases with the increase of the SOF. The SOFs of soil properties is also critical to the reliability analysis on geotechnical structures. The random field and geostatistical methods are commonly used to calculate the SOF of soil properties. In this study, a theoretical derivation was conducted to comparatively analyze the Auto-Correlation Function (ACF) in the random field method, the Semi-variance Function (SVF) in the geo-statistical method, and the corresponding SOFs. Two typical cases (one has many test data, and another has very limited test data) were selected to explore the influences of the trend in the test data, the initial values of curve fitting parameters, and the number of fitting data, on the calculated SOFs for the random field and geo-statistical method. The results demonstrated that the random field and geo-statistical method were essentially the same, and a clear one-to-one relationship was found in the ACF of random field method, and the SVF of M2 method in the geo-statistics method. It infers that the test datum can be detrended before they were used for calculating the SOFs. Prior to the curve fitting for the SOFs of soil properties, the estimated SVF can firstly be normalized. A unified approach was also proposed to determine the initial values of curve fitting in the random field and geo-statistical method, particularly from the experience values of SOFs for soil properties. The optimum number of fitting data was determined for the ACF and the SVF. In calculating the SOFs of soil parameters, it needed to consider only the first several estimated ACFs greater than or equal to zero, and that of SVFs less than or equal to one. In calculating the vertical SOFs of soil parameters, several laboratory tests were conducted for an unsaturated clay, where soil samples were vertically taken from three bore holes. Three physical parameters (water content, density, and specific density of soil particles), three fitting parameters (a , n and θs) of Soil Water Characteristic Curve (SWCC) in the van Genuchten model, , and then the SOFs of all these soil parameters were calculated using the random field and geo-statistical method. The values of SOF for these soil properties of unsaturated clay were 0.14-0.51 m in the random field method, and 0.19-0.62 m in the geo-statistical method, indicating that the calculated SOFs from the random field method were a relatively smaller than those from the geo-statistical method.