Abstract: In the context of global climate change, enhancing forest carbon sink capacity is a crucial measure to mitigate global warming. This study proposed a tree species planning strategy for arboreal forest land based on carbon sink objectives. Taking Sanfengsi Town, Huarong County, Yueyang City in Hunan Province as a case study, a BP neural network was used to construct growth models for tree species (or species groups) suited to the local environment. This was to enable a comprehensive analysis of the carbon sink potential of different species across various forest plots. Additionally, the D-ANP method was applied to evaluate the afforestation suitability of 1862 forest plots by integrating factors such as land use, transportation accessibility, and forest quality, which helped to establish an appropriate sequence for species replacement. The results indicated that: (1) There were significant differences in the carbon sink capacities of different tree species (or species groups). In monoculture afforestation, Cedrela, Liquidambar, and Phoebe exhibited strong carbon sink potential, contributing 202242 tCO2e, 184669 tCO2e, and 177884 tCO2e, respectively. In contrast, Cupressus and Quercus contributed a lower value of 95386 tCO2e and 74527 tCO2e, respectively. (2) Optimization of tree species configurations could significantly increase the overall regional carbon sink. Through scientific tree species planning, the carbon sink potential of the study area for the period of 2021-2041 could be increased from the originally projected value of 116924 tCO2e to 259590 tCO2e, which was an increase to 222.01% of the original one. (3) A comprehensive multi-factor analysis supported the stable growth of the carbon sink. The phased planning of tree species replacement ensured a smooth transition in carbon sink capacity, especially for the case of the replacement of low-efficiency species by the high-efficiency species. These findings provided valuable reference and practical guidance for tree species planning and management in arboreal forest lands, with a potential broad applicability.