Soil fertility quality assessment based on scientifically determined indicators has significant implications for the study of stand productivity and multi-objective forest management. We carried out plant census and community analysis, and compared soil physicochemical properties among forest communities in Kanghe Nature Reserve. The four community types, i.e., scrub, evergreen broadleaved forest, coniferous-broadleaved mixed forest, and coniferous forest, represented different vegetation restoration patterns originating from a selective-logging (at 12 cm diameter) event on the subtropical evergreen broadleaved forest. The main soil factors influencing tree composition and distribution were screened by Canonical Correspondence Analysis (CCA), which were subsequently used to build the minimum data set (MDS). Soil fertility quality was quantitatively assessed with Factor Analysis on MDS.The results showed that distribution of tree species was unevened. With the lowest species richness, the coniferous forest was exclusively dominated by Cunninghamia lanceolata, whereas in the most species rich evergreen broadleaved forest, codominant tree species existed. Average tree DBH (diameter at breast height) decreased successively among coniferous forest, coniferous-broadleaved mixed forest and evergreen broadleaved forest.Soils were highly acidic (pH < 5.0) and the content of AvK, ExMg, AvMn, AvZn and AvFe had no significant difference(P > 0.05)across the four forest communities. A lower canopy density due to less canopy trees resulted in the lower soil moisture content in the scrub community, which had a higher mineral element content than other vegetation patterns as a whole. Distribution of trees was closely related to soil environmental factors, but the results of CCA indicated that there was no significant correlation (Monte Carlo Test,P > 0.05) between composition and distribution of tree species and the seven soil factors, i.e., AN, ExCa, AvK, TP, ST, AP and NCP, which should be excluded when building the MDS. Therefore, the findings revealed that CCA was an effective tool for screening soil fertility quality indicators and establishing MDS by demonstrating the effect of soil fertility on plant patterns.Factor analysis was applied on MDS to assess the fertility quality and the cumulative variance explained by five principal components was 84.67%, indicating this evaluation method was reliable. The integrated soil fertility quality assessment scores for scrub, evergreen broadleaved forest, coniferous-broadleaved mixed forest, and coniferous forest were, in a descending order, 0.438, 0.414, -0.170, and -0.331, respectively. Vegetation restoration did not obviously change soil structural parameters, such as soil texture, capillary porosity, non-capillary porosity, and bulk porosity, however, soil fertility quality declined in the early succession stage (19 a). In the three community types without human disturbance, coniferous forest had the highest biomass accumulation, whereas evergreen broadleaved forest from natural regeneration was more conducive to the increase of biodiversity and soil nutrient accumulation. Site-specific indicators, including soil physicochemical factors and biological factors, should be selected in future assessment of soil fertility quality. The appropriate quantitative ecological methods could make the determination of evaluation index more scientific.