Northern high-latitude regions store large amounts of soil organic carbon (SOC) but are also experiencing significant climate change impacts. Increased temperature accelerates SOC decomposition while simultaneously driving vegetation shifts into previously barren areas, potentially increasing carbon storage. However, the effect of a warmer climate on SOC dynamics are not fully understood. To address this gap, more data are needed to refine Earth system models. This study quantifies SOC from 72 tundra soil samples collected near Abisko, Sweden, visualizes its spatial distribution, and assesses the effectiveness of predictive modelling approaches. SOC was quantified using loss on ignition (LOI) and three forest-based models incorporating digital elevation model (DEM) derivatives, UAV imagery, or a combination of both were tested. Model performance was assessed using mean squared error (MSE), coefficient of determination (R²), and variable importance metrics. The UAV-based model showed the highest predictive accuracy (MSE = 11.1, R² = 0.91 in validation), highlighting the value of high-resolution spectral data for SOC mapping. SOC storage varied significantly between habitats, with mesic heath, semiwetlands, and snowbed habitats containing the highest carbon stocks, while barren and dry heath habitats stored the least. This study demonstrates that UAV-based predictive modelling is a powerful tool for SOC estimation in tundra environments. However, data limitations and model uncertainties highlight the need for further refinement and increased sampling. These findings could contribute to improving carbon flux predictions and understanding ecosystem responses to climate change.