The four archaeological sites are located on the Ningshao Plain on the southern shore of Hangzhou Bay in eastern China (Figure 1A). The sequence of Neolithic cultures mainly consisted of the Hemudu Culture (7,000–5,000 cal a BP), which is generally subdivided into four periods at approximately 500-year intervals (Wang, 2000), and the Liangzhu Culture (5,000–4,300 cal a BP) (Figure 1B). The mean annual precipitation and temperature in this region are c.1000–1,400 mm and c.15–16°C, respectively. Regional vegetation is under the influence of the East Asian Monsoon and is characterized by subtropical mixed forests of evergreen trees, for example, Lithocarpus, Cyclobalanopsis, and Quercus, and deciduous trees, for example, Liquidambar, Castanea, and Celtis.

Jingtoushan site (30°2′N, 121°22′E) is situated in the Yaojiang Valley, about 2.3 km west of the Tianluoshan site and 7.6 km north of the Hemudu site, with elevation of 2.5 m above local mean sea level (Zhejiang Provicial Institute of Cultural Relics and Archaeology et al., 2021). The Jingtoushan site was first discovered and drilled in 2014, and the cultural layers were about 7–8 m below the ground, which was the deepest archaeological site in the coastal region of China. Three sections of No. 4 core were selected for pollen analysis, that is, 4 h (1,020–920 cm), 4 g (920–820 cm), and 4f (820–720 cm). The cultural layer (930–805 cm) was subdivided into three layers: the silt layer (930–890 cm), the charcoal layer (890–860 cm), and the shell layer (860–805 cm), which was formed upon weathered crust and overlaid by thick marine sediments (Supplementary Figure S1). A total of 35 samples were collected at 4 cm intervals, including 25 samples from the cultural layer and five samples from the natural sediment layers below and above, respectively.

The Hemudu site (29°58′N, 121°21′E) is situated beside the Yaojiang River and has been excavated twice in 1973 and 1977, covering approximately 50000 m² (Zhejiang Provincial Institute of Relics and Archaeology, 2003). A 715-cm long core HMD1602 was retrieved from the site reservation of pile dwellings southeast of the Hemudu museum in 2016. The upper 140 cm was composed of continental and artificial layers that were not analyzed. The sediment of 370–140 cm was the Hemudu cultural layer and was sampled at 10 cm intervals. The sediment of 558–370 cm containing numerous foraminifera and marine diatoms was interpreted as a neritic layer and sampled at approximately 20 cm intervals (Supplementary Figure S2) (He et al., 2020b). A total of 27 samples were collected for pollen analysis, including 17 samples from the cultural layer and 10 samples from the natural layers below.

The Yushan site (30°02′N, 121°33′E) is situated on the eastern entrance of the Yaojiang Valley, 7.3 km west to the present coastline (Ningbo Municipal Institution of Cultural Relics and Archaeology et al., 2016). The Yushan site was excavated in 2013, and a 275-cm long profile was collected from the south section of trench T0213 (Supplementary Figure S3). The upper 60 cm of the profile encompassed historic and modern sediment and was not sampled. The lower 215 cm was subdivided into three natural layers of marine transgression at 275–250 cm, 215–180 cm, and 120–60 cm, and two cultural layers of Hemudu Period II (250–215 cm) and Hemudu Period III-Liangzhu Culture (180–120 cm) (He et al., 2018). A total of 43 samples were collected at 5 cm intervals for pollen analysis.

The Xiawangdu site (29°46′N, 121°26′E) is situated beside the Fenghua River, which merges with Yaojiang River to form Yongjiang River at Ningbo City (Ningbo Municipal Institution of Cultural Relics and Archaeology et al., 2019). This site was excavated in 2017, and a 195-cm long profile sampled was in the west section of trench T0602 (Supplementary Figure S4). The upper 10 cm of the profile encompassed historic and modern sediment and was not sampled. The lower 185 cm encompassed three main units, including natural layers of marine transgression at 195–160 cm, cultural layers of Hemudu Period III—IV (160–50 cm), and Liangzhu Culture (50–10 cm) (He et al., 2020a). A total of 40 samples were collected at approximately 5 cm intervals for pollen analysis.

In total, 145 samples were collected from cores/profiles of the four archaeological sites, which were located in the core region of each archaeological site with the most complete cultural sequence and sediments, ranging from the coast to foothills in space and spanning the middle Holocene (8.2–4.2 cal ka BP) in time. Previously, pollen records of profiles from the Yushan and Xiawangdu sites have been reported (He et al., 2018; He et al., 2020a). For this study, palynological analysis of the cores at the Jingtoushan (Supplementary Figure S5) and Hemudu (Supplementary Figure S6) sites was conducted to get a full view of human–environment interaction during the whole middle Holocene. Each core/profile encompassed both natural (light gray) and cultural layers (dark gray) (Figure 1B) for further comparison of environmental change and human activities.

Thirty-four samples in total were collected for dating, which have been conducted in four radiocarbon laboratories: Beta Analytic (Beta), Peking University (BA), Paleo Labo Co. (PLD), and the University of Georgia (UGa). With regard to the cores at the Jingtoushan and Hemudu sites, plants and charcoal were collected systematically during sampling; as to the profiles of the Yushan and Xiawangdu sites, samples were collected at the boundaries of each layer and screened to retrieve short-lived seeds and charcoal for dating. The radiocarbon ages were calibrated uniformly using the IntCal20 dataset with the OxCal v4.4 program (Bronk Ramsey, 2009; Reimer et al., 2020). Dates of the Jingtoushan site are first published in this study, and details of all the dates are displayed in Table 1 and Figures 2, 4.

Two grams of each sample were weighed for pollen analysis. Pollen samples were treated according to the standard procedure developed by Moore et al. (1991), and generally, over 400 grains were counted for each sample using a Leica DM 750 microscope at ×400 magnification. Microscopic charcoal was counted on the pollen slides while identifying pollen. Identification of pollen and spores was made with reference to modern and Quaternary atlases (Institute of Botany and South China Institute of Botany, 1982; Wang et al., 1995; Tang et al., 2016). Quercus (oak) pollen included two categories based on the surface, tricolporoidate and size, that is, Quercus-deciduous and Quercus-evergreen (including Cyclobalanopsis) (Tang et al., 2018). Poaceae pollen was divided into three size categories (<35, 35–40, and >40 μm), and the large size category (≥35 μm) had been identified as domesticated rice pollen in the sediment of eastern China (Yang et al., 2012). Zones in the pollen diagram were divided according to the sediment and pollen assemblages using C2 software (Juggins, 2007). In addition, a correlation analysis between the percentage of Quercus and Poaceae ≥35 μm pollen was conducted by Origin 2021 software.

In this study, the terrestrial pollen assemblages of the four archaeological sites were all predominated by arboreal Quercus (average 20.4–46.3%) and herbaceous Poaceae (average 16.9–51.8%), and exhibited diverse patterns with time. In the natural layers of the Jingtoushan (Zones I and III) and Hemudu (Zone I) sites, the pollen diagram was dominated by the primeval vegetation of Quercus and Pinus. In contrast, the percentage of Quercus declined conspicuously and was progressively replaced by that of rice-type Poaceae ≥35 μm in the cultural layers of the Jingtoushan (Zone II, 8,100–7,800 cal a BP) and Hemudu (Zone II, 7,100–6,500 cal a BP) sites (Figures 2A,B). However, the percentages were problematic among different sedimentary facies. The decrease of Quercus in the cultural layer may not result from deforestation but be caused by the relative rise of Poaceae ≥35 μm instead (Li et al., 2012), which could be corroborated by the strong negative correlation between that of Quercus and Poaceae (Figures 2C,D).

In addition, the decline in the percentage of Quercus and the increase of charcoal in the cultural layer of the Jingtoushan (Zone II-2) and Hemudu (Zone II-1) sites may imply the management of thinning the stands of acorn using fire to harvest efficiently (Pan et al., 2017). Although large quantities of wooden remains, such as pile dwellings, had been excavated at the Jingtoushan and Tianluoshan sites (Figures 3A, B), only a few were identified as Quercus (11.4%) (Figure 3C) (Suzuki et al., 2011), which implied that Quercus may have been consciously protected for the collection and storage of acorns (Figures 3D, E) rather than deforestation.

In the profiles of the Yushan and Xiawangdu sites, the concentration of Pinus, Quercus, and Poaceae almost changed synchronously, implying that the relative abundance may not be affected by different sedimentary facies (Figures 4A, B). Moreover, the negative correlation between the percentage of Quercus and Poaceae was weak (Figures 4C, D), indicating these two pollen types change independently and the percentage could reflect the real evolution of local vegetation. In the later stages of the cultural layers of the Yushan (Zone IV, 5,600–5,000 cal a BP) and Xiawangdu (Zone II and III, 5,600–4,300 cal a BP) sites, the decline of the percentage of Quercus and Pinus coincided with the increase of that of Poaceae ≥35 μm and charcoal, suggesting deforestation of Quercus forest induced by the intensification of rice cultivation. Archaeobotanical research suggested that extensive paddy systems, such as Tianluoshan, Shi’ao (ca 4,650 m²), and Maoshan sites (ca 55000 m²) (Figures 5A–C) (Zheng et al., 2009; Zhuang et al., 2014), and new farm tools, such as stone plow and bone spade (Figures 5D–F) (Fuller et al., 2008), had been applied since late Hemudu-Songze and Liangzhu culture, which further support the hypothesis of slash-and-burn practices due to the possible shortage of cultivated land.

Acorns are the nuts of Quercus sp., which are the constructive species of the subtropical mixed forests of evergreen and deciduous trees in southern China, and constitute the main arboreal component in the palynological records since the late Quaternary (Tang et al., 2018). Ethnographic and archaeological evidence suggest that acorn is a significant resource commonly consumed by Native Americans (Anderson, 2005) and recovered from prehistoric sites in Levant Upper-Palaeolithic (Barlow and Heck, 2002), Japanese Jomon (Takahashi and Hosoya, 2002), and Chinese Early-Middle Neolithic (Yang et al., 2009; Liu et al., 2010a). The earliest evidence for the exploitation of acorn in the lower Yangtze River was starch grains extracted from grinding stones of the Shangshan culture (10,000–8,500 cal a BP) (Liu et al., 2010b; Wang and Jiang, 2021). In addition, large quantities of acorns have been excavated from the storage pits at the Kuahuqiao, Hemudu, and Tianluoshan sites (Figure 6A) (Zhejiang Provincial Institute of Relics and Archaeology, 2003; Fuller et al., 2011; Jiang, 2014).

Acorn is a starch-rich resource that was likely to be exploited as a staple food, predating the establishment of rice agriculture. In the early stages of rice cultivation, acorns accounted for approximately 41.2 and 52.2% of the whole plant remains at the Kuahuqiao and Tianluoshan sites (Figure 6B), respectively (Fuller et al., 2009; Pan, 2011). In addition, more than twenty storage pits of acorns had been recovered at the Tianluoshan site, and pieces of acorn shells had also been sieved out (Sun, 2013). Thus, in the subsistence strategy, acorn can be used as a staple food for both reserve resources and daily consumption. As a result, the Quercus may have been managed consciously by limiting burning to open habitats and increasing per-tree yield (Pan et al., 2017), and the decline of the concentration of Quercus in the cultural layer of Jingtoushan (Zone II-2, ca 8,000 cal a BP) and Hemudu (Zone II-1, ca 7,000 cal a BP) sites may have just reflected this human activity. Considering the low-level production of rice cultivation prior to 6,000 cal a BP, ancient humans may not sacrifice the resource of acorns to give way to undependable rice cultivation.

However, several factors may restrict the acorn exploitation. First, acorns are a seasonal resource, concentrated from August to October, and their annual yields vary dramatically (Pan et al., 2017). Second, acorns need special processing methods to remove tannins and an appropriate temperature and humidity during storage (Takahashi and Hosoya, 2002). Finally, a cooling trend through the middle to late Holocene may cause regional declines in oaks (Fuller and Qin, 2010). Therefore, the proportion of acorns in plant remains declined significantly to approximately 6.8 and 2.4% during the later stages of the Tianluoshan and Yushan sites, while that of rice increased progressively to 25.5 and 34.0% (Figure 6B), respectively (Fuller et al., 2009; Zheng et al., 2019). As the subsistence economy shifts toward an increasing focus on rice, acorns may degenerate from a staple to a famine food reserve (Hosoya, 2011). The labor input and land use of rice cultivation conflicted with that of acorn exploitation, resulting in the local deforestation that is recorded at the Yushan and Xiawangdu sites.

Deforestation for farming was the main type of impact on land use made by prehistoric humans on vegetation that has been detected in Europe and China since the middle Holocene (Ren, 2000; Fyfe et al., 2015; Roberts et al., 2018). The entire ecological process of human-induced deforestation could be modeled into five stages, that is, primeval vegetation, deforestation, cultivation, abandonment, and restoration, which can be reflected in the palynological records (Li et al., 2008). The pollen assemblage of anthropogenic deforestation was generally characterized by a decrease of zonal broadleaf wood (e.g., Quercus) and a rise of secondary pioneer trees (e.g., Pinus), ferns, and herbs (e.g., Poaceae), accompanied by an increase in charcoal (Zheng et al., 2004; Li et al., 2008; Xu et al., 2010). The lower Yangtze River was densely distributed with over 4,000 archaeological sites during the Neolithic and Bronze Age (Hosner et al., 2016) and was substantially transformed by humans. Based on the relative distance to the archaeological sites, palynological research works on the timing of deforestation induced by human activities can be divided into three categories: natural (>3 km), intermediate (50 m-3 km), and archaeological sites (Figure 7A), respectively.

The earliest signals of anthropogenic deforestation in the lower Yangtze River could be traced back to ca 8,000 and 7,700 cal a BP from the Jingtoushan and Kuahuqiao sites (Table 2) (Zong et al., 2007; Shu et al., 2012); however, the average and median dates of deforestation among archaeological sites were ca 7,000 cal a BP. As to the intermediate site that is situated 50–3,000 m outside the adjacent archaeological site, the timing of deforestation ranged from ca 7,000 to 3,000 cal a BP, with an average and median date of ca 5,000 cal a BP. Finally, the timing of deforestation from the natural sediments ranged from ca 4,000 to 1,500 cal a BP, which was generally later than ca 3,000 cal a BP. Significantly, the timing of deforestation decreased progressively from ca 7,000 cal a BP of the archaeological to ca 5,000 cal a BP of the intermediate site and ca 3,000 cal a BP of the natural sites at the end (Figure 7B).

Therefore, the synthesized pattern of pollen records in the lower Yangtze River suggested that possible anthropogenic deforestation in the middle Holocene was confined to local scales (Zhao et al., 2009), and no consistent vegetation change occurred at regional scales induced by human activities as reported previously (Ren, 2006). Instead, the main driver of regional vegetation change at the mid-late Holocene transition in eastern China may be attributed to climatic deterioration (Innes et al., 2014). In general, the effects of human disturbance on vegetation at local scales intensified gradually from archaeological to intermediate sites since the middle Holocene, and became an increasingly important factor in the vegetation of natural sites at regional scales until the last 3,000 years (Liu and Qiu, 1994; Zheng et al., 2021), which coincided with a noticeable increase in the number of archaeological sites after 3,500 cal a BP in southern China (Hosner et al., 2016).