Numerical Analysis of Modern and Fossil Pollen Data from the Tibetan Plateau

Abstract

This study applies a number of multivariate numerical techniques in the analysis of modern and fossil pollen data for the purpose of paleovegetational reconstruction in the Tibetan Plateau. Modern pollen spectra from 227 sampling sites in the Tibetan Plateau were classified into groups using cluster analysis and detrended correspondence analysis. By comparing the quantitatively derived groups with local vegetation types at sampling locations, the modern pollen spectra were assigned into five major vegetation types—shrubland, forest, meadow, steppe, and desert. These were used as a priori groups for discriminant analysis. A set of discriminant functions was derived that correctly classifies 93.3 percent of the surface samples. These functions were then applied to a fossil pollen record from Yidun Glacial Lake near the Sichuan–Tibet border to reconstruct the major pattern of vegetational changes since the last glacial maximum. The results show that steppe and meadow prevailed around the site during the Late Glacial period (17.3–11.5 ka BP). Regional vegetation changed from meadow to forest during the transitional period of 11.5–9.2 ka BP. Forests have persisted in the Yidun area since 9.2 ka BP. A comparison of our reconstructed paleovegetation with other proxy data suggests that the transition to forest around 9.2 ka BP followed the early Holocene intensification of the southwest Asian monsoon. Our study also demonstrates that discriminant analysis is a useful technique for reliable reconstruction of paleovegetation in the Tibetan Plateau.