【Objective】Effects of sampling scale on spatial variability of soil properties are of great significance to soil digital mappings, soil survey, and agricultural production. So, in this study, a tract of the grey eel-mud field (Soil Species), a special land use—soil type complex unit(Lu-SoTy), in Dongxiang County, Jiangxi Province, a hilly red soil region in South China was selected as an object. Using the multi-level nesting sampling protocol to ensure that the total number of soil sampling points remains unchanged whatever sampling scale is used, the study is to exposit characteristics of spatial variability of soil total element concentrations with soil sampling scale and provide important implications and scientific references for rationalizing layout of sampling points based on Lu-SoTy.【Method】In the light of spatial distribution of soil sampling points, eight soil sampling scales (from the smallest to the largest is coded in turn as A1~A8) were designed with smaller scales located inside larger ones. By taking into full account vastitude and uniformity of spatial distribution of the sampling points, each soil sampling scale had 60 points distributed evenly. Using the multi-level nesting sampling protocol, 337 sampling points were laid out across the County. In December 2014 after the crop of late rice was harvested, soil samples were collected from the 0～20cm topsoil layer at all the sampling points for analysis of soil total element concentrations. Scale characterization indices were calculated using ArcGIS 10.2.2, variability characterization indices using SPSS 20.0, and regression fitting analysis of scale characterization indices and variability characterization indices, using SPSS 20.0. Based on the level of significance and R² of the regression, the best characterization indices and the best fitting functions were screened out. 【Result】Results show that the scale effects on spatial variation of soil total elements are obvious; and coefficients of variation (CV) of TN, TP and TK go up with expansion of the sampling scale, but the increasing rates display a decreasing trend; the scale effects on the spatial variability differed with soil total element: the best fitting function for charactering the scale effects on CV (%) is CV = b× D^a for TN and TP (R²>0.87, p<0.001), and CV = e^{(a/D b)} (R²>0.93, p<0.001) for TK (D is a scale index of sampling points; namely: average distance to the nearest one point or maximum distance (km) of all the sampling points); the shifting points of scale effects between rapid and slow changes of CV varied with soil total element, for example, the shifting point of TN, TP and TK is D ₀ = 6.756 km, D₀ = 9.061 km and D₀ = 3.408 km, respectively (at this point, the angle between the tangent to the curve and the horizontal axis is about 30°). All the findings demonstrate that independent sampling protocols and analysis skills need to be considered in studying spatial distribution of different soil total elements. 【Conclusion】The scale effects characteristic of spatial variability of soil total elements in soil of a special complex Land Use-Soil Type unit, could provide important implications and scientific references for laying sampling points at county scale soil surveys in hilly red soil regions.