A total of 46 samples were collected from four typical site profiles for OSL dating. The samples were prepared under subdued red-light conditions using standard laboratory methods. The purity of the quartz particles was determined using the infrared method. Small aliquots were used in our test, and the sample distribution diameter of the particles was approximately 2 mm.

The radioactive energy absorbed by the samples mainly emanates from the decay of ²³⁵U, ²³⁸U, ²³²Th, and their daughters, as well as ⁴⁰K of the geological environment and the sample itself (Aitken, 1998). The concentrations of U, Th, and K were measured by inductively coupled plasma mass spectrometry (ICP-MS) and atomic absorption spectrophotometry in the key Laboratory of subtropical resources and environment of Fujian Province at Fujian Normal University. The contribution of the cosmic dose rate to the total dose rate was calculated according to Prescott and Hutton (1994) and Prescott and Stephan (1982). The final dose rate was calculated using the DRAC program (Durcan et al., 2015). An automated Risø TL/OSL DA-20C/D reader installed with a ⁹⁰Y/⁹⁰Sr beta source was used for the quartz equivalent dose (D_e) test in a laboratory at Fujian Normal University. The improved single-aliquot regenerative-dose (SAR) protocol was used to estimate the single-aliquot D_e of the quartz grains (Jin et al., 2017).

A preheat-plateau test was performed to observe the electron transfer in the samples during preheating. Six different preheat temperatures between 180℃ and 280℃ in 20℃ increments were used in the D_e test with the SAR protocol, where the cut-heat was 20℃ less than the preheat temperature. Three aliquots were measured at each preheat temperature. The results showed that the preheat temperature had no significant influence on the D_e values in the range of 180-240℃ (Figure 3a). Therefore, the preheat temperature was set to 220℃ for 10 s.

A dose recovery test was performed on 16 aliquots of sample DHQ-09 at a preheat temperature of 220℃ for 10 s. First, we bleached the natural signal in the aliquots for 48 h, and then administered a known dose of 18 Gy. Figure 3b shows that the D_e values of the sample were largely consistent with the supplied dose within the error range, and that the D_e test is reliable at a preheat temperature of 220℃.

The quartz OSL signal comprises a variety of components (commonly 2-7) with different bleaching characteristics, decay rates, saturation limits, thermal stabilities, and susceptibilities to recuperation and sensitization (Bailey et al., 2001). Dating using quartz OSL involves measuring the optical signal from the so-called fast component (Jin et al., 2021). Figure 4a shows the OSL signal decay curve and the regeneration dose growth curve of sample DHQ-09. The decay curve of the natural luminescence signal shows that the luminescence signal has a rapid decay rate, and will fade to the background value in approximately 1 s. This indicates that the OSL signal is composed mainly of fast components, and that the SAR protocol is suitable to measure D_e.

On multi-grain aliquots from sample DHQ-09, linearly modulated (LM) OSL measurements were performed, and the curves were deconvoluted using the ‘fit-LMCurve’ feature in R to determine the signal components after different treatments (Kreutzer, 2020). The results show that the natural luminescence signals of the samples were mainly composed of fast components. Supplying a laboratory dose (5.7 Gy) after bleaching the luminescence signal of the sample, and simultaneous stimulation with blue light at 220℃ for 10 s, the results showed that the decay curves of the samples showed one fast, one medium, and two slow components (Figure 4d). This indicates that the ratio of medium and slow component signals may be increased during the regenerating dose measurements.

To further understand the characteristics of the natural luminescence signal of quartz samples, we calculated the fast ratio (F_R value) of all aliquots (n=16) from sample DHQ-09 (Durcan and Duller, 2011). The photoionization cross-section values were used to calculate the F_R of all aliquots using the method described by Durcan and Duller (2011). The F_R values of the 16 quartz aliquots were found to be dominated by the fast component of the sample (Figure 5). Comparisons between the supplied dose and the regenerative dose signal F_R ratios in the dose recovery test data by contrast, showed no systematic differences. This suggests that the aliquots may have been bleached and then irradiated before dose recovery and are affected by the same medium and slow components as the regenerative dose signals (Neudorf et al., 2019; Jin et al., 2021). Thus, these signal components are largely non-existent in natural OSL signals and are probably thermally unstable over geological time.

The distribution of D_e values can be used to indicate whether the sediment has been completely bleached before burial (Fuchs et al., 2012). The frequency density distribution of the D_e values in the samples that were completely bleached before burial showed obvious symmetry and narrow peaks. In contrast, the frequency density distribution of the D_e values of sample DHQ-09 showed a roughly normal distribution (Figure 4b), meaning that the sample was not completely bleached. The above luminescence signal characteristics indicate that the OSL samples of Haitan Island are suitable for dating.

The OSL ages of all the samples were calculated and are shown in Table 2.

Among the four sites selected for this study (Figure 2), two samples were taken from Donghuaqiu site (DHQ). Figures 2d-2e shows the sampling site of the first field survey (sample numbers DHQ-15 to DHQ-26). The red sand layer (also called the “old red sand”) at the bottom of the site is approximately 7.08±0.86 ka in age, and a Neolithic cultural layer appeared at 5.06±0.28 ka. According to the dating results and the artifacts excavated from the site, the other cultural layers were severely disturbed by human activities. Figures 2a-2c shows the location of the second sampling site. The median luminescence ages of the four “old red sand” samples (DHQ-11 to DHQ-14) at the bottom of the site were 6.01±0.88 ka in age. The burial stratigraphic ages of the three layers were 4.39±0.65 ka, 4.09±0.56 ka and 2.84±0.4 ka, respectively (Figure 2c). In general, the quartz burial luminescence ages of the cultural layers at the site were concentrated between 4.39±0.65 ka and 1.82±0.18 ka, corresponding roughly with the Huangguashan cultural period in the coastal area of Fujian Province.

Guishan site (GS) has undergone several formal archaeological excavations. The samples reported in this study were collected from the main excavation area in the first survey (Figures 2f-2h). Figures 2f-2g shows the main sampling profiles. The buried age of the “old red sand” at the bottom of the site is 10.51±0.6 ka, and the ages of the cultural layers are 5.15±0.4 ka and 3.25±0.25 ka, corresponding to the Tanshishan and Huangguashan cultural periods, respectively. Figure 2h shows the sampling location of the trench excavated approximately 100 m northwest of the main excavation area. The two OSL ages are 7.36±0.82 ka and 5.90±0.93 ka, which is consistent with the archaeological ages of the excavated artifacts and correspond to the Keqiutou cultural period. According to the classification of Yang et al. (2004), Guishan site should be classified as a culturally superimposed site.

Unlike the DHQ and GS sites, Chitanghou (CTH) site was formed on a low hill of weathered granite. Therefore, the OSL samples were often mixed with granite debris. At the same time, due to long-term cultivation, mineral particles close to the surface cultural layer were likely to be exposed. The three ages obtained from the main cultural layer at the site are 2.45±0.19 ka, 6.03±0.58 ka and 5.44±0.35 ka, corresponding to the Keqiutou cultural period.

Houqishan site (NH) is approximately 25 m above sea level and has not been formally excavated; only three OSL samples were collected from the cultural layer (Figure 2i). The burial age of the main cultural layer is 4.71±0.35 ka, which is classified as a single site from the Tanshishan cultural period.

Figure 6a integrates the age data from the four typical sites and statistically analyzes the frequency density, relative frequency, and cumulative frequency to understand the spatial and temporal distribution of sites in the region. The results show that the dune/shell mound site of Haitan Island was mainly formed at 8 ka, which is consistent with the results of regional sea-level change and archaeochronology. Human activities on Haitan Island were mainly concentrated in two high sea level stages about 7-5 ka and 4-2 ka, which roughly correspond to the Keqiutou cultural period (6.5-5.5 ka) and the Huangguashan cultural period (4.3-3.5 ka) in the lower Minjiang River (Lin, 2011).

Sea level change is one factor that affected prehistoric human activities in coastal areas. Due to the influence of climate, geological structure, and paleotopography, there are many Holocene sea-level patterns in different regions (Chen et al., 2015). Sea level during the early Holocene was lower than modern sea level and in the middle Holocene (7-4 ka), due to hydro-isostacy caused by the continuous influx of Arctic meltwater to the Atlantic Ocean. Sea water was also affected by topographical features in the process of upwelling to the estuary. This was followed by a relatively high sea level stage through a phenomenon called the backwater effect (Baker et al., 2003). For example, the relative sea level near the equator, South America, Southeast Asia along the Mekong River, the Red River, Singapore, and the delta areas of northern Australia were 1-5 m above modern mean sea level (Hesp et al., 1998; Tanabe, 2003). In contrast, the sea level near the Yangtze River Delta did not rise above modern sea levels in the middle Holocene (7-4 ka) (Wang et al., 2018) or during the high sea level phase of the Holocene that occurred before 7 ka (Zhu et al., 2002); this was verified by a continuous Neolithic cultural sequence in the Ningshao Plain. The Neolithic cultural period in the lower reaches of Minjiang River was not found to be continuous at a small temporal scale. Human activities from 7-5.5 ka were concentrated on the coastal islands. In contrast, human activities from 5.5-4 ka were mainly concentrated on the low hills or tableland near the inland margin of the Fuzhou Basin. Humans from 4-3 ka mainly lived near the coastal zones. The temporal and spatial distribution characteristics indicate that there was a secondary fluctuation trend of “low-high-low” sea levels in the lower reaches of the Minjiang River during the period of high sea level between 7 and 4 ka. Table 1 shows that the Neolithic sites dated to 7-5 ka on Haitan Island all accumulated on the platform formed by the old red sand or by the granite weathering crust at an elevation of 10 m above sea level during the late Pleistocene. This may indicate that there was a high sea level stand caused by “backwater” in the middle Holocene. Such distribution characteristics are consistent with sea level changes in the coastal areas of Fujian Province (Zeng, 1991), indicating that the period of prehistoric human activities on Haitan Island corresponds to the Holocene high sea level stage, and that the landscape differed significantly from that of the present day.

Coastal aeolian activities under the background of sea level change is another important factor affecting prehistoric human activities on Haitan Island. The stratigraphic structure of the Neolithic sites on Haitan Island indicates that most sites were distributed in the late Pleistocene to early Holocene aeolian sand (old red sand), which was associated with Holocene aeolian sand. This suggests that the development of the Neolithic culture on Haitan Island was accompanied by aeolian activities in the region. The sedimentary processes of coastal dunes in Fujian Province are mainly affected by the sediment source, prevailing wind, and topography (Wu et al., 1995b). Yanshanian granites are widely distributed along the coast of South China, and it is easy to form a huge, thick weathering crust under the humid monsoon climate conditions of tropical and subtropical regions, and produce a large amount of sand through erosion by flowing water, coastal currents, and waves, providing rich sand sources for the development of coastal dunes in South China (Wu et al., 1995b). The average annual runoff and sediment transport of the Miniang River can reach 5.9×10¹² m³ and 1.02×10⁷ t, respectively. In the middle and late Holocene, sea level fluctuations were small, the sand source was abundant, and the geomorphology was relatively stable; therefore, the development of coastal dunes was mainly affected by wind conditions.

Coastal dune deposits can reflect variations in the East Asian monsoon on the millennium-centennial scale (Jin et al., 2015; Xu et al., 2017; Jin et al., 2019b). Coastal dunes on the Fujian coast mainly accumulated during the middle and late Holocene, and can be roughly divided into three stages: 7-5, 4-2 and 1-0 ka (Figure 6b). This division is closely related to the intensity of regional aeolian sand activities and regional coastline changes (Wu et al., 1995a; Jin et al., 2015). The first stage from 7-5 ka is an important active aeolian period, which is consistent with the intensity of the East Asian winter monsoon reflected in Huguangyan Maar Lake deposition (Yancheva et al., 2007) and sea surface temperature (SST) variability in the Okinawa Trough (Jian et al., 2000). There are also several obvious millennial-centennial scale depositional gaps during this period. Cai et al. (1992) analyzed the aeolian sand sequence from the Luyangpu Plain on Haitan Island through borehole deposition and determined that the aeolian sand in Luyangpu was deposited at 2.4 ka. Therefore, aeolian sand deposition in ancient Changjiang bay, where the site group is located, should be the product of the third active aeolian period, which is related to the backwater effect in the middle Holocene in the low-latitude area (Chen et al., 2015). The rising water stripped away previously formed dunes. The periods of human activity recorded in the cultural stratigraphy correspond well with the periods of regional coastal aeolian sand activity (Figure 6b). There are several centennial-scale depositional discontinuities in most dunes over 2 ka of age (Jin et al., 2015). Due to numerous high tides and storm surges in the region, the elevation of most coastal dunes older than 1 ka is above 10 m. Below 10 m, coastal dunes have been deposited mainly over the last thousand years, especially since the Little Ice Age. In future studies, the altitude of the dunes can be used as an index to indirectly determine the age of regional coastal dunes.

In addition, this study also found that most of the old red sand was late Pleistocene deposits. Old red sand, developed over the past 20 ka, is rarely found and reported on; it is only found on leeward slopes and at higher elevations. Globally, red aeolian sand during this period did not exist in isolation (Jin et al., 2018b; Jin et al., 2019b; Reuter et al., 2020), and its distribution and paleogeographic significance need further study.

Coastal dunes are products of land-sea-air interactions (Dong, 2010) and also reflect abrupt climate events over short time scales (Dong et al., 2016; Yang et al., 2017). Haitan Island and its adjacent islands are located south of the Minjiang River estuary, and the highest elevation is approximately 400 m. Strong winds and high waves in the Taiwan Strait and the continuous supply of sediment from outside the Minjiang River estuary provide abundant sand sources and energy for the development of aeolian geomorphology on Haitan Island (Cai et al., 1992; Shi et al., 2009). In the Neolithic period, when productivity was reduced, the spatial scope of prehistoric human activities was not only controlled by slow, natural, environmental processes, but it was also affected and restricted by abrupt events such as typhoons and storm surges. Chen et al. (2015) proposed the concept of “lowest habitable surface” in view of the relationship between sea level and human activities in the estuarine area. They pointed out that the height of safe human habitation in estuaries or coastal zones is correlated with the maximum tidal range and the reachability of storm surges. Modern observations show that the maximum spring tide height in the northern part of Haitan Island can reach more than 6 m (Tian et al., 2011). Storm surges occur frequently in the coastal areas of central and southern Fujian Province from July to September. In September, the coastal areas of Fujian Province experience an astronomical high tide period (Yuan et al., 2018), and the increase in water height may exceed 7 m due to the superposition effect of storm surges and high astronomical tides. In other words, the lowest habitable surface for human activity in the region should be 7 m above sea level. Influenced by prevailing winds, topography, and the ocean, prehistoric people chose to inhabit the low hills and high terraces near the leeward slope of the ancient Changjiang’ao bay.

In a study on the environmental background of archaeological remains in Guanting Basin, Qinghai Province, Yang et al. (2004) divided the accumulation forms of archaeological sites into superposed sites and unitary sites, representing stable and unstable lifestyles, respectively. High terraces near the river seem to have been the site of long-term human settlement, representing the continuity of culture. Affected by floods, humans on the lower terraces would have had to migrate to higher terraces periodically, indicating discontinuous cultural inheritance (Yang et al., 2004; Yang et al., 2005). Although the spatial distribution patterns of archaeological sites in coastal zones are not affected by rivers and floods, they are affected by sea level fluctuations and should have similar distribution characteristics (Figure 7). According to the spatial distribution characteristics of the Neolithic sites on Haitan Island, archaeological sites are mainly concentrated in the northeast region of Haitan Island. In terms of geomorphic features, most of the sites are located at the edge of isolated islets or aeolian plains close to the northwest edge of the ancient bay (Figure 1); that is, they are mainly distributed on the slope foothills of the two rows of the NNE-SSW trending high hills. With an increase in altitude, the sites features are distributed as “single site (I)-superposed site-single site (II)” and appear successively. Single type sites (I) mainly appeared at low sea level, while single type sites (II) mainly appeared at high sea level. Superimposed sites are not subject to sea level changes. Dong et al. (2021) proposed the “Fulcrum Cognitive Model” to explore the mechanism of past human-land coevolution based on a case study of past man-land relationships. They pointed out that the balance between natural ecosystems and human social systems is broken by changes in the climatic environment or human activities at certain points in time (Dong et al., 2021). The relationship between Haitan Island’s natural ecosystem and the human social system can be explained by a “quantitative balance model.” Within a certain threshold, the regional human social system is in balance with the natural ecological environment through human migration. The regional environment at 7-5 ka is characterized by high sea level (Zeng, 1991; Chen et al., 2015), strong winter winds (Jian et al., 2000), and strong sandstorm activity; this breaks the balance between the natural and human social systems, shrinks the environment that is suitable for human survival, and increases the pressure for human survival. Faced with this situation, human societies may achieve a new balance between the weakened carrying capacity of the environment by migrating to higher altitudes and to more inland regions. The period between 5-4 ka is the most important period of human activity in the lower reaches of the Minjiang River, and is related to weaker winter monsoon intensity. From 4 to 3 ka, sea level decreased gradually, approaching the modern sea level. Although the intensity of winter winds gradually increased after 5-4 ka and aeolian sand activity increased, the rapid development of rice farming during this period reduced the living pressure of human beings and increased the available living resources and living space. Human beings were able to achieve a new balance between the development of human society and the enhanced carrying capacity of the land by improving the level of productivity (Figure 7).