Soil samples were collected from the depression (200 m x 100 m) in Karst area's Mulun National Nature Reserve by grid sampling method (20 m x 20 m), with the spatial heterogeneity of soil nutrients analyzed by the methods of classic statistics and geo-statistics. The soil pH showed small variation, while the soil nutrients showed moderate variation, being in the order of available phosphorus (AP) > available potassium (AK) > available nitrogen (AN) > organic matter (OM) > total potassium (TK) > total phosphorus (TP) > total nitrogen (TN). Spherical model fitted best for soil pH, exponential model fitted best for soil TK and AK, and Gaussian model fitted best for other variables. The variation range of soil pH and AK was smaller, being 58.1 m and 41.1 m, respectively, that of soil OM, TN, TP, AN, and AP was from 100 m to 150 m, and that of soil TK was the largest (463.5 m). Soil TK and TN showed moderate spatial autocorrelation, while other soil nutrients showed strong spatial autocorrelation. Soil pH and AK presented fragmented patch distribution, showing high heterogeneity, while soil OM, TP, and TK changed gently, high in the middle and low in two sides. The spatial patterns of soil AN and AP were similar, i.e., increased with landform slope. Soil TN had a distribution pattern of low in the middle and high in two sides. Vegetation, topography, and high heterogeneity of micro-habitat were the main factors caused the differences of the spatial distribution patterns of soil nutrients in the Reserve.