Forest carbon stock is an important indicator for evaluating the ecological benefits of forest ecosystem. Thus, accurately estimating the carbon stocks for the different components (i.e. stem, branches, foliages, and roots) of individual trees is essential and necessary. In this study, a total of 44 Korean pine (Pinus koraiensis) trees were sampled from the plantations in Heilongjiang province, China. The biomass and carbon stocks data were collected for different tree components. Based on the theory of non-linear simultaneous equations, the compatible models of the biomass and carbon stocks were developed for the whole tree and each tree component (i.e. stem, branches, foliages, and roots) by using the proc model procedure of SAS9.22 software to estimate the model coefficients. The prediction accuracy of two methods (i.e., direct and indirect methods) for predicting the Korean pine carbon stocks was compared. The direct method directly predicted the carbon stocks of the tree components using the developed compatible models of carbon stocks. The indirect method predicted the carbon stocks in two steps: (1) predicting the biomass of the tree components using the developed compatible models of biomass, and (2) multiplying the estimated biomass by carbon content percentages to obtain the carbon stocks. Three kinds of carbon content percentages were used (a) the carbon content conversion factor 0.5 commonly used in the literature over the past decades; (b) the average carbon content percentage of individual trees measured from the sampled trees; and (c) the average carbon percentages of tree components measured from the sampled trees. The results indicated that the coefficients of determination (R²) of both compatible biomass and carbon stocks models ranged 0.76-0.99, and the modeling efficiency (EF) of the two models were 0.80-0.98. The prediction accuracy of carbon stocks by the direct method was 91.03% for stem, 70.24% for foliages, 80.02% for branches, 87.10% for roots, and 93.08% for total, respectively. In comparison, the prediction accuracy of the indirect method using the common carbon content conversion factor 0.5 decreased 1.39% for stem, 0.13% for foliages, 1.5% for branches, 1.09% for roots, and 2.2% for total amount, respectively. On the other hand, the indirect method using the measured carbon content percentages reduced the prediction accuracy within 0.3% for each tree component and the total. We also investigated the error sources for predicting carbon stocks and found that the prediction accuracy was mainly dependent on the factor of C_i%/C % (where C_i%represents the carbon content percentage of tree components and C % is the average carbon content percentage of individual trees). It was evident that the direct method was the best for predicting the tree carbon stocks of Korean pine. The commonly used carbon content conversion factor 0.5 was significantly different from the measured carbon content percentages. Thus, the indirect method using the factor 0.5 performed poorly for predicting the carbon stocks, while using the measured carbon content percentages of tree components as the indirect method would greatly improve the prediction accuracy. The results of this study will provide basic models and direct method to predict carbon stocks of stand or large scale forest for Korean pine plantation.