Abstract <p>Indicators of variations in soil carbon stock of coniferous–broadleaf forest are of high predictive value. This article provides an assessment of the contribution of environmental factors to the variation of carbon stock in forest soils. The study was performed in the Bryansk Forest Reserve on 45 test plots located in different landscapes. Five groups of factors, characterizing vegetation, macrofauna, landscape, topography, and history of management, were analyzed. Statistical relationships between the parameters of carbon stock and environmental factors were assessed, using machine learning methods. The quality of litter formed by dominants of tree, herbaceous, and of the moss layers is the main factor, determining the carbon stock in it. Position in the landscape and the topographic characteristics are less informative. The carbon stock of the OL litter subhorizon is the highest in forests with a great proportion of pine. The mosses cover is the most informative indicator of variation in carbon stock in the OFH litter subhorizon. The results of regression analysis of carbon stock in the A horizon and in the upper 30-cm-thick layer reflect a significant contribution of parameters of increased hydromorphism, as well as of characteristics of functional organization of forest ecosystems, the ecological-cenotic structure of plant communities (the proportion of nemoral species in the undergrowth and shrub layers), and the functional diversity of earthworms.</p>