We assessed long-term trends in soil organic carbon (SOC) in volcanic soils with a process-based soil genesis model, SoilGen2.25. The relation between soil geochemistry and SOC was applied in a model context, where significant soil properties were identified and used to modify the decay rates of SOC pools. We used data from Indonesian sites with different land use (tropical primary forest, secondary pine forest, and agricultural land) and calibrated major soil processes in volcanic soils, viz. clay migration and weathering of primary minerals. The model evaluated the decay rates of SOC pools using three calibration approaches: (i) a site-specific calibration, (ii) a generic calibration, and (iii) a generic calibration modified by a geochemical proxy. The best calibration for each approach was then used to estimate the future of SOC under different climate projection scenarios, viz. representative concentration pathways (RCP) 2.6 and 8.5. The SoilGen2.25 model was generally sensitive to the change of selected soil process parameters. A four-pool SOC model (Roth-C) with site-specific decay rates best reproduced total SOC with a percentage difference between measured and simulated SOC between 1 and 10%. Application of a geochemical proxy to modify the generic rate calibration also improved, relative to a generic calibration, the simulation quality of the included humus pool and total SOC in most study sites. Agricultural soil showed higher susceptibility to global warming (i.e. RCP 8.5) than forest soils. Projective scenarios based on the three calibration scenarios highlighted the importance of the calibration method on the accuracy of SOC projection.