Human-activity modified vegetation composition at Hangzhou Bay, Zhejiang Province, China over the last 10,000 years

Vegetation composition has been greatly modified by human activity since Neolithic times, providing valuable insights into the evolution of human-landscape relationships. In this paper, we integrated 26 pollen records from the Zhejiang Province of China where a Neolithic culture of rice farming existed at an early date, to examine the dynamics of agricultural development and its impact on vegetation. Firstly, we synthesized the relative abundance of Oryza pollen and charcoal (Cpeak) using generalized additive models, to produce a record of agricultural development and the intensity of human activities. We found that changes in Oryza abundance and Cpeak were positively correlated, and identified three phases as follows: agricultural initiation before 8000 cal yr BP, agricultural development between 8000 and 6000 cal yr BP and the onset of intensive rice farming after 6000 cal yr BP. Pearson correlation between each pollen taxon and Oryza indicates that arboreal taxa exhibit an overall negative correlation with Oryza, but terrestrial herbs and ferns show a positive correlation, indicating that deforestation accompanied agricultural development. Rosaceae, Castanea, Fagus and Typha display positive correlation with Oryza during the earliest stage dominated by hunter-gathering, reflecting the protection and cultivation of food-producing vegetation. During the intensive farming stage, the Liangzhu ancestors reclaimed more farmland and began to pay attention on farmland management including clearing the fields of weeds. This is the main reason for the enhanced negative correlation between Artemisia, Chenopodiaceae, Brassicaceae, Asteraceae and Lamiaceae with Oryza at 6000-5000 cal yr BP. Finally, we confirm the potential of integrated Oryza abundance in pollen records as a human-impact index in the Zhejiang region, and emphasize that temporal dynamics in vegetation and human activities should be critically pre-examined when applying relevant humanimpact indices.