Forest is an important vegetation type of terrestrial ecosystems that maintains the dynamic balance of the biosphere and geosphere, and it plays a key role in terrestrial ecosystem carbon sequestration. With the implementation of Grain for Green Project, the Three-North Shelterbelt Project, and other major ecological projects, the total area of plantations has dramatically increased; thus, the role of plantations in absorbing and fixing CO₂ has attracted more attention. Investigating carbon storage of major forest types is important for understanding regional carbon cycles. Pinus. tabulaeformis is a common plantation species in sub-humid and semi-arid regions for the restoration of forest ecosystems. In western Loess Plateau, artificial forests of P. tabulaeformis provide a substantial proportion of the terrestrial ecosystem carbon sink. To understand the carbon storage features of these plantations in response to the precipitation gradient, three typical distribution regions on western Loess Plateau were selected for plot surveys. Tree biomass was estimated separately for stems, branches, leaves, and roots, using previously constructed allometric biomass equations based on the diameter at breast height (DBH) and tree height. Carbon contents in the different parts and layers of the tree were each measured by collecting corresponding samples. Carbon density was calculated by multiplying the biomass of each plant part (from a specific forest area) and the corresponding carbon content. The results showed that in the three sites (Lanzhou, Taizishan, Xiaolongshan) with precipitation of 372 mm, 519 mm and 632 mm, respectively, biomass carbon densities of P. tabulaeformis plantations were (49.08±2.86) t/hm², (73.90±9.36) t/hm², and (82.55±7.36) t/hm², respectively. Both the carbon density of the ecosystem and biomass carbon density were significantly different (P < 0.05) between Xiaolongshan and Lanzhou. Because the shrubs were sparse, the contribution of trees was greatest, reaching 95.21%-98.91% of the biomass carbon density across the precipitation gradient. Overall density of organic carbon in soil did not differ between the three sites. Soil organic carbon in the upper soil layers showed significant differences among the sites, but were not statistically different (P>0.05) in other soil layers. Correlative analysis showed that the biomass carbon density was positively correlated with annual precipitation, with a Pearson coefficient of 0.820. However, there were no significant positive correlations between the biomass carbon density and the age of stand. Similarly, partial correlative analysis showed that there was a significant positive correlation between the carbon density of P. tabulaeformis plantations and precipitation in this sub-humid and semi-arid region of the Loess Plateau. These results indicate that precipitation is a key factor that affects biomass production and carbon fixation in semi-arid forests. Furthermore, it implies that appropriate water conservation measures are necessary, particularly in semi-arid areas, to satisfy the hydric demands of P. tabulaeformis plantations and to maintain normal growth. Such measures may also be helpful for improving carbon sequestration potential and comprehensive services of plantation ecosystems. Therefore, this study provides valuable insight into how carbon storage characteristics of plantations respond to precipitation gradients, and provides useful information for regional forest management with respect to productivity and carbon storage.