The present article deals with the numerical appraisement on soil classification for silviculture use and the assessment of forest productivity on those soils by means of taking the principal component analysis as a main program associated with clustering and regression to examine the rationality of the current taxonomy of soils in the low mountain and hill area around Nanjing, and to test simultaneously the effectiveness of the above-mentioned numerical method per se. It turned out that by using five soil attributes, i. e., the total depth of soil, the texture, the stone content, the pH value and the effervescent effect in subsoil, the principal component analysis applied to the data collected from 70 sites can mainly distinguish most types of soils which have been classified into the current taxonomic system according to their parent materials and soil properties. Clustering and regression carried out on the very basis further proved that the soil types based on the first two principal components were related to a certain extent to their productivity for forest.The result of principal component analysis showed that there are 3 ecological sequences of soil in the area according to their parent materials as well as soil reaction and the basisols were also confirmed statistically. The further clustering indicated that the shallow soils of yellow-brown loam and acidic stony ground scattered on to the coordinate plane of the first two components were of crossed distribution, so that it is reasonable to put them into one group as rankers; and the shallow soil of neutral stony ground ought to be allocated to the shallow type of limestone soils. The regression of tree growth to the principal components of soil illustrated that the productivity of a certain type of soil might extend across two grades, and the critical value for the assessment of soil productivity also varied with different ecological sequences of soil.