Soil organic carbon (SOC) modelling is a useful approach to assess the impact of nutrient management on carbon sequestration. RothC was parameterized and evaluated with two long-term experiments comparing different fertilizer treatments in north (Zhengzhou) and northeast (Gongzhuling) China. Four nutrient application treatments were used: no fertilizer (Control), mineral nitrogen–phosphorus–potassium fertilizers (NPK), NPK mineral fertilizer plus manure (NPKM), and NPK mineral fertilizer plus straw return (NPKS). The comparison between simulated and observed data showed that the model can adequately simulate SOC contents in the Control, NPK and NPKM treatments but overestimated in the NPKS treatment at both sites. By changing the value of decomposable plant material:resistant plant material (DPM:RPM) ratio from the default value to 3.35 for the NPKS treatment at the Zhengzhou site, dynamics of simulated SOC agreed with measured values. A pseudo-parameter, straw retention factor was introduced to adjust the amount of straw incorporated into soils. Using the inverse simulation method and the modified value of the ratio, the best-fitted value was 0.24 for the NPKS treatment at the Gongzhuling site. This result indicated that retaining straw on the soil surface makes less contribution to carbon sequestration than if it is incorporated. With this modification for straw, the model produced reasonable predictions for the two sites. The model was run for another 30 years with the modified parameter values and current average climatic conditions for different fertilizer treatments at both sites. The results suggested that the NPK application plus the addition of manure or straw would be better management practices for carbon sequestration.