When the site-specific soil and climate data were available to initialize simulations, both K-model and CENTURY have the capacity to represent the dynamics of SOC. Statistically, there were no significant differences between measurements and modeled results, although the RMSEs and relative errors of CENTURY were larger than those of K-model. When non site-specific soil and climate data were used to initialize simulations, considerable uncertainties are introduced because of the uncertainties in initial soil conditions and history of management, as well as errors that may exist in the climatic data, K-model can still represent the dynamics of SOC with no significant reductions in accuracy. The CENTURY, on the other hand, may produce poor results that significantly differ from measurements in the clod and dry agricultural ecosystem. This study shows that K-model, in comparison with CENTURY, improved predictions of effects of chemical fertilizer, manure application, residue management and fallow on SOC dynamics. The possible reasons for these improvements were: 1) K-model simulated the SOC dynamics within whole soil profile where soil water, nutrient and aeration conditions allow root growth, while CENTURY only simulated the SOC dynamics of the upper 20 cm soil layer; 2) in comparison with CENTURY, K-model had strong capacity to adjust carbon allocation, as well as the root growth and root distribution within soil profile according to the soil texture, water, nutrient and aeration conditions, these adaptive adjustments of crop growth modeled by K-model reduces the negative effects of water and nutrient stresses on crop growth; 3) K-model allocated correct amount of carbon to leaf and stem which allows increases in the SOC resulting from complete RR to be accurately represented, CENTURY, on the other hand, is not as flexible; and 4) K-model and CENTURY use different methods to deal with initial soil, climate and management. The single carbon pool of K-model and three carbon pools of CENTURY did not generate large differences on carbon decomposition rates. However, K-model only requires a single decomposition parameter comparing with three parameters required by CENTURY. The less and easily measurable parameters and other data required by K-model improved the utility of model. Auto-correction function of K-model was helpful for accurate predictions if measurements are available to be used as references.