Abstract: Abstract: To understand the soil nutrients (C, N, P, and etc.) heterogeneity at hillslope and catchment scales has been considered to be critical to biogeochemical processes, and particularly important for vegetation conservation and restoration in the fragile regions. Karst peak-cluster depression, widely distributed in southwestern China, is one of the most vulnerable ecosystems to human disturbance due to thin soil cover, dual structure of hydrological processes, and rapid loss of surface water. The accelerated loss of soil nutrients resulted from human disturbance is one of the most important key processes which led to the degradation of ecosystems and the occurrence of rocky desertification in karst areas. However, the mechanisms caused soil nutrients loss are still unclear because little is known about the spatial variability of soil nutrients and the relative importance of influencing factors, such as soil matrix, vegetation, and topography. This study aimed to investigate the spatial variability of soil nutrients in a typical karst peak-cluster depression landscape based on classical statistics and geostatistical analysis in virgin forest in Mulun national nature reserve in southwestern China. Moreover, the relative importance of the environmental factors influencing soil nutrients variability and accumulation were studied. Using line sampling method, 145 soil samples were collected from the upper 15 cm of calcareous soil in Dec. 2008, and 64 plots of the sampling locations were selected for the investigation on litter quality and soil physical property. Soil organic carbon (SOC), total nitrogen (TN), total phosphorus (TP), and total potassium (TK) were measured, and the data were analyzed using geostatistical and multiple linear regression analysis. Sample variograms of SOC and TN were fitted well by exponential models with the nugget effect, while TP and TK were fitted well by spherical models with the nugget effect. SOC and TN had the similar spatial variability structure, which showed moderate spatial dependence (the nugget effect between 25% and 75%), while TP and TK showed strong spatial dependence (the nugget effect <25%). The spatial ranges of influence calculated for soil nutrients ranged from 167.7 to 308.7 m. Kriged maps drawn based on the selected variogram models showed a regularly distribution of SOC, TN and TP, and they decreased with the increase of altitude on hillslope. However, TK had a different spatial pattern, and it had higher contents on the middle slope and depression. The multiple linear regression analysis showed that, environment factors explained the highest variation of TK but the lowest variation of SOC, which was corresponded to the results of spatial dependence analysis. The cation of Ca2+ was the dominant influencing factor of SOC and TN variations, litter quality (litter N content, C:N ratio) also had significant effect on SOC and TN distributions, but it explained only 10.9% and 11.8% variation of SOC and TN. This indicated that the formations and accumulations of SOC and TN are mostly controlled by soil matrix and its interaction with the quantity of litter input. Litter P content was the dominant affecting factor of TP, which explained 66.2% variation of TP, indicating that the accumulation of TP is controlled by the return forest litter P. The influencing factors of TK were a little complex, and the cation of Ca2+ was the dominant factor, but litter N:P ratio, altitude, and soil clay particle contents also had significant contribution to the variation of TK.