Tree-ring δ13C analysis has been widely used in reconstructing past climate and environment. However, it is difficult to correlate the results from samples taken with different sampling strategy. It is still an open question, how to take sample, in which radii to take sample, that the δ13C of the samples can represent the tree δ13C trend better and more efficient. To address the issue, we analyze the correlation and similarity of the different radii sequence and the different combination sequence with the full- ring-sample sequences, based on 642 δ13C data of cellulose fixed in tree-ring at 8 different radii of two Cryptomeria fortunei tree disks from Mt. Tianmu, a subtropical mountain of China. It suggests that 2 to 4 radii mixture sample is the better and more efficient choice for the tree-ring δ13C analysis. It had better to take samples at obtuse angle with each other when to take samples at two or three radii. When to take four samples in one tree, commonly the samples are taken in “cross” relation, but the analysis shows it is not the best choice. The more the samples are taken at obtuse angle with each other, the better the representation of the samples. If, sometimes, only one core can be taken, the core in the sunny aspect has the better representation of the tree.