Abstract: Vegetation, litter, topography, and soil attributes are driving factors of soil organic carbon (SOC) variations. However, few studies have examined those multiple driving factors simultaneously. The mixed spruce-fir-broad-leaved forest was selected  with 200 groups of data, including soil data of different soil layers, vegetation data, and data of different litter decomposition layers. Topographic data were obtained with remote sensing technology. To quantify the effects of each driving factor, correlation analysis and structural equation model were used to establish the relationship between observation variables and latent variables. According to the results of correlation analysis, five latent variables including vegetation, topography, litter return characteristics, soil properties, SOC content and SOC density (SOCD) were constructed, and 15 observation variables were selected. The test coefficients of the structural equation model basically passed with a multivariate square coefficient of 0.94, indicating that the equation construction was reasonable and that the convergence validity of the model met the standard. The four predicted latent variables had significant positive effects on SOC content and SOCD. The total effects were in order of soil attribute (0.938)> topography (0.383)> vegetation (0.131)> litter return characteristics (0.099). The contributions of interpretable partial effects were 60.5%, 24.7%, 8.4%, and 6.4%, respectively, with elevation, soil total phosphorus at depth of 20-40 cm, soil total nitrogen at 20-40 cm, soil total nitrogen at 0-20 cm and soil moisture content at 20-40 cm as the top five contributors to SOC content and SOCD. Topography (elevation and aspect) and vegetation (Shannon diversity index, number of species, canopy density) had indirect positive effects on SOC content and SOCD. Topography reached a very significant level through two paths: soil attribute and litter return characteristics (P<0.001), while vegetation only reached a significant level of 0.1 through soil attribute path. The relationship among driving factors was very complex. Vegetation affects SOC by affecting soil nitrogen, phosphorus, and soil properties of deep layers, litter has a positive effect on SOC through direct input, and topography indirectly affects SOC by affecting other soil properties through water and heat allocation. The structural equation established in this study expounds the action relationships of various driving factors in mixed spruce-fir-broadleaved forest, which provides a theoretical basis for SOC protection and soil nutrient management for natural forests.

Key words: litter, environmental factor, soil organic carbon, structural equation model