Modern pollen-vegetation coverage relationships and its application for Holocene vegetation reconstructions in the central Tianshan Mountains

Abstract

The Tianshan Mountains are the largest mountain system in Central Asia and comprise islands of humidity within a widespread arid zone. In this study, we improve knowledge of the quantitative relationship between pollen and vegetation coverage in the region, to quantitatively reconstruct Holocene vegetation coverage. Specifically, we statistically analyzed 1058 modern pollen samples from the central Tianshan Mountains and neighbouring regions, utilizing redundancy analysis (RDA), and boosted regression trees (BRT), stepwise regression (Stepwise), the modern analogy technique (MAT), weighted average partial least squares (WA-PLS) and local weighted average (LWWA) methods, to construct a series of modern calibration sets between pollen and fractional vegetation cover (FVC). The results show that (1) FVC and modern pollen assemblages display a strong correlation (correlation coefficient of 0.66) and (2) BRT outperforms Stepwise, MAT, WA-PLS and LWWA under leave-one-out cross-validation, bootstrap, and spatial autocorrelation tests, making the BRT model the best choice. Applied to Holocene datasets, we infer that from 11.4 to 4.6 cal kyr BP, FVC changes were mainly influenced by natural climate changes. However, after 4.6 cal kyr BP, FVC changes were influenced by a combination of natural climate changes (with humidity playing a dominant role) and the intensity of human activities.