Lakeside soil organic carbon (SOC) serves as a critical indicator for assessing cropland quality and soil carbon sequestration potential. Lakeside farmlands, shaped by unique water-land interactions, often form high-density organic carbon reservoirs, making them pivotal zones for enhancing agricultural carbon sequestration and addressing climate change. Therefore, accurately identifying the dominant drivers of SOC in lakeside farmlands is essential for improving soil fertility and achieving sustainable agricultural development. The geographical detector model is an effective tool for identifying influencing factors and their interactions. However, the commonly used optimal parameter geographical detector (OPGD) requires traversing five discrete methods, leading to computational inefficiency and instability in obtaining optimal discretization results, which limits the explanatory power of variables. To address this, this study developed a robust geographical detector (RGD) based on variance change point detection. Taking the lakeside farmland in Pei County, Xuzhou City, as a case study, the impacts of natural factors, human activities, and their interactions on SOC were systematically investigated. The results demonstrated that: (1) RGD outperformed OPGD in computational speed, data discretization effectiveness, and explanatory power. (2) Key drivers identified by RGD included mean annual precipitation, mean annual minimum temperature, distance to Weishan Lake, elevation, and farmland land use types. They could explain 44.5%, 43.3%, 37.8%, 23.4%, and 23.2% of SOC variation, respectively. (3) SOC content in eastern paddy fields near the lake was significantly higher than that in western drylands, with mean values of 6.82 g/kg versus 6.22 g/kg. The spatial pattern indicated that SOC decreased with increasing distance from the lake and stabilized beyond 18.8 km, a critical threshold spatially aligned with the boundary between paddy and dryland fields. This highlights the dominant role of lake-regulated planting and irrigation patterns in shaping SOC distribution. (4) Nonlinear-enhancement or bi- enhancement interactions were prevalent among influencing factors, indicating that SOC dynamics are driven by complex synergies rather than simple additive effects. The interaction between mean annual precipitation and mean annual minimum temperature exhibited the highest q-value (0.536). Further analysis revealed that high mean annual precipitation combined with elevated mean annual minimum temperatures significantly promoted SOC accumulation. This study validated the superiority of the RGD model and provides a reference for analyzing dominant drivers of soil attributes in other lakeside farmlands. Future research will integrate more hydrological factors (e.g., lake water level fluctuations) and agricultural management practices (e.g., fertilization amount and cultivation method) to further unravel the spatial heterogeneity of SOC in these regions.