Performance of vegetation cover reconstructions using lake and soil pollen samples from the Tibetan Plateau

The relationship between modern pollen and vegetation cover is the basis for reconstructing past vegetation cover quantitatively. However, this relationship has received relatively little attention on the Tibetan Plateau so far. For this research, pollen samples were collected from lake surface sediment (n = 117) and from the soil surface (n = 117) in the same region on the central and eastern Tibetan Plateau, to explore the relationship between modern pollen data, both as percentages and concentrations, and vegetation cover, as well as to make comparisons between modern pollen assemblages from lakes and soils. The results show that pollen taxa and their abundances extracted from soils are, in general, consistent with those from lakes. Redundancy analysis (RDA) results reveal that pollen assemblages from the two different datasets, from lakes or soils, have similar relationships with climate and vegetation in the study region, although soil surface pollen assemblages are more likely to be influenced by local pollen deposition. Thus, to some extent, pollen assemblages from lake surface sediments can be replaced by soil surface assemblages in vegetation reconstructions of areas with simple pollen compositions such as the Tibetan Plateau. Random forest (RF) models show good relationships between pollen concentrations and vegetation cover, and both pollen concentrations and percentages have the potential to reconstruct vegetation cover. However, the reconstructed vegetation cover for Saiyong Co and Koucha lake (covering the last 21.3 cal ka bp and 33.8 cal ka bp , respectively) are more reliable when using pollen concentration rather than percentage data. Our study suggests that pollen concentrations obtained from lake sediments are better than percentages for the quantitative reconstruction of vegetation cover at glacial and interglacial scales.