Plant diversity plays an important role in maintaining the stability and service functions of terrestrial ecosystems. In paleoecology, pollen records can not only provide a long-term perspective for understanding plant diversity changes and ecosystem dynamics, but also provide scientific basis for ecological conservation and management. However, the correlation between pollen diversity and plant diversity, as well as its relationship with landscape structure and altitude, remain incompletely understood. Given this, we calculated pollen and plant richness and evenness using the modern plant community survey data from 24 quadrats and the topsoil pollen records within the quadrats from the subtropical forest region of central Zhejiang Province. Besides, we computed patch diversity indices (patch richness, patch diversity and patch evenness) based on the vegetation distribution map of the study area. We subsequently analyzed numerically the relationships between modern pollen diversity and plant diversity, landscape structure, as well as elevation. To assess the impacts of pollen production, pollination strategy, and dispersal capacity of pollen on the relationships between pollen and plant diversity, we adjusted the pollen data for seven main pollen taxa using the pollen representation value (R-value), and divided the data sets into different sub-data sets according to the life form (woody or herbaceous taxa) and the mode of pollination (anemophilous and non-anemophilous) to calculate the correlations between pollen and plant diversity in each pair of (sub-) data sets. The results showed that there were significant differences in plant richness and pollen richness in subtropical evergreen broad-leaved forest, but no such differences were observed in other three vegetation types (mixed forest, coniferous forest and mountain shrub). Neither pollen richness nor evenness accurately mirrors the quadrat-scale plant diversity, but the correlation between pollen richness and plant richness can be enhanced by classifying pollen data into woody and herbaceous taxa. Correction of pollen data with R-values could not improve the significance and correlation between pollen and plant diversity. This probably because the R-values, estimated as the average values of the R-values from China's subtropical forest area, may not be able to accurately reflect the representation of the selected pollen taxa to their parent plants in the study area. Our study found that pollen richness and evenness decrease with increasing altitude, and unraveled a significant positive correlation between pollen evenness and patch evenness within 300 m. We suggested that pollen richness and evenness in subtropical mountain forest could not be used as reliable indicators for quadrat-scale plant diversity, but could reflect, to a certain extent, characteristics of the landscape structure and vertical distribution pattern of plant diversity. However, dividing pollen data into woody and herbaceous taxa can enhance the robustness of reconstructing long-term changes in plant diversity over geological periods. Furthermore, this study has demonstrated that shifts in vegetation types can result in significant overestimation or underestimation of plant diversity when relying on pollen data. Therefore, it is crucial to carefully consider changes in vegetation types over geological history when using fossil pollen records to reconstruct variations in plant diversity.