Abstract: This study aims to explore the spatial variability of soil organic carbon density (SOCD) and influencing factors in the carbon sink of farmland soils. Taking Beijing as the study area, the SOCD data was measured for each layer from 0 to 60 cm sampled in the field from June to October 2022. A conceptual model was established using 3D GMS. Then the SOCD was simulated to determine the spatial variability in three dimensions. According to the dummy variable models, the mantel test, the geographical weighted regression, and the GeoDetector models were used to explore the importance of different factors on the SOCD and the interactive contribution among the factors on both quantitative and spatial scales. The results showed that: 1) The SOCD shared a spatial trend of decreasing from the surface to the deeper layers, with the SOCD significantly higher (P < 0.05) in the 0-15 cm soil depth than in the 30-60 cm soil layer. The variability of SOCD was more substantial in the deeper soil layers. The soil organic carbon stock in 0-60 cm was 10.80 Tg. 2) The spatial and scale effects were found in the intensity of different factors on the SOCD. The soil subclass, soil parent material, soil texture, soil water content, and topographic site were the significant influencing factors on the variation of SOCD in the study area. Among them, the soil water content, soil subclass, and topographic site were dominated (P<0.05) SOCD at the soil depths of 0-15, 15-60, and 45-60 cm, respectively. Soil subclass, soil parent material, soil texture, and topographic site were spatially strongly associated with the SOCD. The intensity of their association gradually increased from the surface layer to the deeper layer. The disturbance by human activities gradually decreased in the soil below the surface layer. The factors that caused the variation of SOCD were shifted to the natural factors. 3) The expression of SOCD showed a two-way or non-linear enhancement after the interaction of the significant factors. The soil subclass showed the most prominent enhancement to explain SOCD after interacting with each other factor. Among them, there was the strongest explanatory power for the SOCD at 0-30 cm soil depth after interaction with soil water content, with 0.724 (0-15 cm) and 0.789 (15-30 cm), respectively. The explanatory power of SOCD was the strongest at 30-60 cm soil depth after interaction with soil parent material, with 0.649 (30-45 cm) and 0.784 (45-60 cm), respectively. The interaction of multiple factors should be considered for the spatial variability of soil organic carbon and driving factors. Among them, soil subclasses (soil types) can be incorporated as a priority reference indicator, indicating the differences and the soil-forming parent material, texture configuration, and moisture condition of the environment. The finding can also provide the scientific references to optimize the spatial structure of farmland resources, in order to realize the carbon sequestration and sink enhancement in farmland.