Objective  Analyzing the spatial distribution of soil organic carbon (SOC) and its density (SOCD) at small spatial scales aims to provide theoretical support for accurate estimation of carbon storage in alpine soils at large scales.

  Method  Taking Carex alpine meadow (CAM), Rhododendron tanastylum shrub (RTS) and Rhododendron aganniphum shrub (RAS) lying in western slope of Sejila Mountain at altitudes of 4 200−4 400 meters as objects of research, and collecting soil samples at the depths of 0−10 cm with grid method, we analyzed spatial heterogeneity and influencing factors of SOC and SOCD.

  Result  (1) The content of SOC was rich with an average up to 100.97 g/kg in the study area, and content of SOC declined in the following rank order: RAS (146.45 g/kg) > CAM (95.60 g/kg) > RTS (60.43 g/kg), while SOCD declined in the following rank order: CAM (7.34 kg/m²) > RAS (6.32 kg/m²) > RTS (4.80 kg/m²). (2) SOC fitted well spherical model while SOCD fitted well exponential model in CAM, as well as SOC and SOCD of RTS both matched spherical model. SOC and SOCD of CAM and RTS both had strong spatial autocorrelation with nugget/sill ratio in 1.46%−12.51% range, suggesting that spatial variability of SOC and SOCD was mainly caused by structural factors. In RAS, SOC and SOCD conformed to linear model, spatial variation was caused by random factors. This indicated that vegetation types may be the main factors affecting the spatial distribution of SOC in topsoil. Ranges of SOC and SOCD in CAM were 30.29 and 20.31 m, respectively, which was larger than RTS (17.68 and 17.44 m), indicating that spatial variability scales of SOC and SOCD in CAM were larger than RTS. At a sampling interval of 10 m, pure nugget effect appeared in SOC and SOCD of RAS, it may be that the spatial variability scale was less than 10 m, so a smaller scale research of SOC and SOCD in RAS was needed. (3) Kriging interpolation showed that the horizontal distribution of SOC and SOCD in CAM was affected by the moisture content with high value areas in the eastern and southern edges of the sample plot, in attention, it may also be related to the existence of slope and gully. The continuity of the horizontal distribution of SOC and SOCD in the RTS area was poor, and the degree of fragmentation was high, which may be related to the mosaic distribution of ground cover (shrub, herb, bare land). (4) Redundancy analysis showed that soil water content, soil density, and pH were the key factors affecting spatial heterogeneity of SOC and SOCD in 3 vegetation types, mechanical composition and slope were the second. However, the effect of total phosphorus was not apparent in this study.

  Conclusion  SOC content of Sejila Mountain is relatively rich. The spatial heterogeneity of micro-topography, litter volume, and soil properties (water content, bulk density, etc.) at a small spatial scale significantly affect the spatial distribution and prediction of SOC and SOCD.