The published results of three national systematic archaeological field surveys in China, which were performed in 1956, 1981 and 2007, provide the original archaeological information used in this study. In the surveys, the archaeological sites of all prehistoric and historical periods were registered. The State Administration of Cultural Heritage compiled the basic information such as location, culture type and size into the series Atlas of Chinese Cultural Relics by provinces, with coarse chronologies of the sites mainly based on ceramic typology. The data of archaeological sites examined in this study were extracted from the following Atlas of Chinese Cultural Relics: Inner Mongolia Autonomous Region Volume (2003), Beijing Volume (2008), Liaoning Volume (2009) and Hebei Volume (2013). The chronology of and quantitative changes in the archaeological sites of the prehistoric culture sequence studied in this paper are summarized and presented in Figure 2. Although there are still some deficiencies, such as controversial periods of certain culture, rough site chronology, omission of undiscovered sites and inaccurate location of certain site, the long time span and large number of sites guarantee that most of the results and conclusions in this study are reliable.

3.2.1 Primary coverage and mapping of sites

The primary topographic coverage in this study is a 30 m digital elevation model of the research area that was extracted from an Aster satellite image and slightly adjusted based on known ground control points. The major rivers and lakes were extracted from the China Historical Geographic Information System project (CHGIS, 2007). After preparation work, the archaeological and geographical data of all sites were imported into GIS. With the longitude and latitude data, the sites were plotted as points on the map in GIS. The coordinate system of each layer was adjusted according to the World Geodetic System 1984 (WGS 1984).

3.2.2 Spatial density

Since the purpose of this paper is to study the evolution and spatial-temporal distribution patterns of archaeological sites of different cultures, and the sites are small compared to the entire study region, the sites can be regarded and calculated as points scattered in space. An archaeological site represents a settlement where human lived and can be regarded as a basic unit of archaeological analysis (Chang, 1967). Thus, the density of sites indicates population density to an extent and can be used to estimate population size (Han et al., 2008; Dong et al., 2012b). The Point Density tool in GIS calculates a magnitude per unit area from point features that fall within an area around each cell. The spatial density features of sites were obtained with this tool.

3.2.3 Nearest-neighbour analysis

Spatial density analysis provides the average value of point density in a certain area but cannot show cluster features. Therefore, nearest-neighbour analysis is required. The value of the average nearest-neighbour (ANN) ratio indicates whether points are clustered (< 1), random (= 1) or dispersed (> 1) in space. It is calculated based on the average distance from each point to its nearest neighbouring point. The calculation formulas are as follows:

$ANN=\frac{{{{\bar{D}}}_{O}}}{{{{\bar{D}}}_{E}}}$

${{\bar{D}}_{O}}\frac{\sum\limits_{i=1}^{n}{{{d}_{i}}}}{n}$

${{\bar{D}}_{E}}=\frac{0.5}{\sqrt{{n}/{A}\;}}$

where ${{\bar{D}}_{O}}$ is the observed mean distance between each point and its nearest neighbour, and ${{\bar{D}}_{E}}$ is the expected mean distance for the point given in a random pattern. In the preceding equations, d_i is the distance between point i and its nearest neighbouring point, A is the area of a minimum enclosure and n is the total number of points.

The ANN tool in GIS, calculates not only the ANN ratio of each culture, but also the p-value - representing probability - and the z-score - representing standard deviation. The p-value and z-score are measures of statistical significance used to ascertain whether to reject the hypothesis of complete spatial randomness. The smaller that the p-value is and the larger the absolute value of the z-score is, the less likely the observed spatial pattern is the result of a random process. If the p-value is less than 0.01 and the absolute value of the z-score is larger than 2.58 (i.e., < -2.58 or > +2.58), the confidence level would be over 99% (ArcGIS, 2016).

The density analysis results reveal that the Lower Xiajiadian culture has the largest maximum density value, followed by the Hongshan culture (Table 1). The Xiaohexi, Xinglongwa, Zhaobaogou and Fuhe cultures have similar maximum density values (approximately 0.037 site/km²). Additionally, spatial density maps (Figure 4) were drawn for each culture. The spatial density values were classified into five classes under a unified standard with geometric intervals, which ensures that each class range has approximately the same number of sites and that the change between intervals is relatively consistent. In Figure 4, three types of spatial distribution pattern can be observed. The first type consists of the Xiaohexi, Xinglongwa, Zhaobaogou, Fuhe and Xiaoheyan cultures. The Hongshan and Lower Xiajiadian cultures represent the second type while the Upper Xiajiadian culture represents the third type. The characteristics and dynamic mechanisms of the revealed spatial distribution patterns are further investigated in the following.

Using the ANN tool in GIS, the ANN value, p-value and z-score of each culture were calculated (Table 2). According to their critical values, the archaeological sites of all cultures exhibit clustering with the confidence levels over 99% except the Xiaohexi culture, whose sites tend to be randomly distributed. Among all cultures, the Lower Xiajiadian culture has the highest level of clustering with the highest confidence level. The Hongshan and Upper Xiajiadian cultures come second.

The prehistoric culture sequence studied here includes the entire Holocene Megathermal period. Because of the sensitivity of the environment to climate in this area and the drastic fluctuations in precipitation and temperature during the Holocene Megathermal period in North China (Shi et al., 1992; Yancheva et al., 2007; Avni et al., 2010), the development of local prehistoric cultures and human activities were extensively influenced by climatic and environmental changes (Xia et al., 2000; Mu et al., 2014). The significant impact of such changes on the evolution of ancient cultures in other areas has been demonstrated (Hoelzmann et al., 2001; Gao et al., 2007; Wu et al., 2010; Dong et al., 2012a). Thus, outlining the Holocene climate fluctuations in North China and environmental changes in the Yan-Liao region contributes to a better understanding and further discussion. Pollen, sedimentary and geochemical archives for this region provide background information.

Studies have demonstrated that the Holocene Megathermal in northern China began at approximately 8.5 ka BP, before which the climate was cold and dry (Feng et al., 2006). Temperature and rainfall increased dramatically since 8.5 ka BP and then gradually stabilized to a more easily habitable level (Cui and Kong, 1992; Shi et al., 1992). The variation in δ¹⁸O values (Figure 5a) collected from the Dunde Ice Core (Site A in Figure 1) records from the Qilianshan Mountains represents credible evidence of climate changes in northern China and in the Northern Hemisphere (Thompson et al., 1989; Thompson et al., 1990). The records indicate strong summer monsoons prior to 5.0 ka BP, succeeded by weak monsoon events with a decrease in temperature and precipitation in northern China. The concentrations of total organic carbon (TOC, %) and total inorganic carbon (TIC, %) of a sediment core recovered in the depocenter of Dali Lake (Site B in Figure 1) denote the balance between the water input to the lake and the evaporation of the lake water, indicating fluctuations of precipitation and humidity in the Dali Lake area (Xiao et al., 2008). Drought events, depicted as shaded bars in Figure 5b (6.6-5.8, 5.1-4.85, 4.45-3.75, 3.15-2.65 ka BP), coincide with the weak summer monsoon period in Figure 5a. Pollen data from Hulun Lake (Site C in Figure 1), northeast to the study region, suggest that drought events also occurred in relatively cold intervals (Wen et al., 2010). Pollen records from the Anguli-nuur Lake (Site D in Figure 1) in northern Hebei Province illustrate changes in the ratios of arboreal plants (AP) pollen to non-arboreal plants (NAP) pollen (Figure 5c). From approximately 6.8 ka BP to 2.0 ka BP, the pollen of conifers, primarily Pinus, and herbs was generally abundant, whilst pollen of Chenopodiaceae and Artemisia gradually increased, indicating a drop of water level. Correspondingly, the pollen of deciduous broad-leaved trees was sparse (Wang et al., 2010). Flora remains in the Chahai Site (Site E in Figure 1) prove that many deciduous broadleaved forests were present in this region before 7.0 ka BP (IALP, 2012a), indicating an increase in the number of trees during warm, humid, strong summer monsoon periods (before 6.5 ka BP and approximately 5.8-5.1 ka BP) and the expansion of steppe during cool, arid, weak summer monsoon periods (after 5.1 ka BP) in this region. The ¹⁴C dating data and geological analysis of northern China also demonstrated that the environment became colder and drier in 6.0-4.0 ka BP (Mo et al., 2003). A reconstruction of the deviation from the present mean aridity and temperature of southern Liaoning (Figure 5d), derived from sediment and sporo-pollen assemblages (39°30′N, 122°E, Site F in Figure 1), demonstrates that the humidity and temperature increased dramatically in 8.0 ka BP and gradually decreased after 6.5 ka BP, with an extreme low temperature event after 3.0 ka BP (GIG, 1977). Although these palaeoenvironmental records are from different sites in the study region or the neighbouring area, the environmental conditions they reflect represent the condition of a fairly large area. Furthermore, the coherence of these records implies that their integration characterizes the environmental conditions of the Yan-Liao region.

Line graphs of the total number, maximum density and ANN value of archaeological sites are appended here to show their correspondences with environmental changes. Total number and maximum density have increased slightly since 8.5 ka BP, with slight decreases in the ANN value (Figure 5e). This trend is in line with the changes in environmental conditions, which were humid and warm when the Xinglongwa, Zhaobaogou and Fuhe cultures rose and developed. However, the peak in the number and maximum density of archaeological sites and the valley in the ANN value occur in the Lower Xiajiadian culture period between 4.0-3.0 ka BP, when it was drier and colder than the previous period. Additionally, the second peak and valley in Figure 5e between 6.5-5.0 ka BP also occurred in a relatively dry and cool period, which corresponds to the Hongshan culture. To understand the potential causal factors, the subsistence strategies, spatial patterns and evolution process of this culture sequence are discussed with the previously described calculations.

The Xiaohexi culture was the first Neolithic culture in the Yan-Liao region. Because of climate fluctuation and primitive level of technology, Xiaohexi has the smallest number of sites (33 sites). The subsequent Xinglongwa, Zhaobaogou and Fuhe cultures developed during a period of relatively comfortable climate conditions. However, the numbers of these sites only increased slightly. Flora and fauna remains from archaeological studies on the Zhaobaogou (IACASS, 1997), Xinglongwa (Liu, 2001), Xinglonggou (Zhao, 2004), Baiyinchanghan (IAIMAR, 2004) and Chahai (IALP, 2012a) sites demonstrated that hunting-gathering was the primary survival strategy of these four cultures, albeit primitive agriculture emerged in the Xinglongwa culture period. The turning point appeared in the Hongshan culture. Archaeological surveys and studies on a number of large sites from the Hongshan culture demonstrate that hunting-gathering remained the primary subsistence strategy in the early stage. However, in the late Hongshan stage, when the climate deteriorated, primitive agriculture was predominant while hunting-gathering was a supplementary activity (IALP, 2012b). Environmental archaeology demonstrates that the temperature decrease in the late Hongshan period was not dramatic. Because of the presence of rivers and their branch streams in the area, the temperature and water supply could support the adoption and development of primitive agriculture, despite the environmental deterioration (IACASS, 1996; IALP, 2012b; Sun and Zhao, 2013). Cohen (1977) demonstrated that only when hunting-gathering cannot provide enough food supplementation, do people adopt agriculture as their primary subsistence strategy to resolve the crisis of food shortages and starvation through cultivation and domestication of plants and animals. Figures 2 and 5 show that the Hongshan culture developed in a relatively dry and cool period with the expansion of steppe. The climate and environment were more favourable for cultivators than hunter-gatherers with respect to obtaining sufficient food. Since the main distribution area of the Hongshan culture is the same as that of the Xinglongwa and Zhaobaogou cultures (Figures 4b, 4c and 4e), the climate change that occurred in this period was likely a primary contributor to the adoption of agriculture. Hunting-gathering was no longer able to provide enough food for the population, which increased during the habitable period. The additional food provided by agriculture resulted in further population growth, as indicated by the remarkable increase in the number and density of Hongshan sites. The lower ANN value is also consistent with the higher clustering level of agriculture. Additionally, the environment in this area during the Xinglongwa and Zhaobaogou cultures was suitable for human habitation, and the food provided by hunting-gathering was abundant such that people were not sufficiently motivated to turn to agriculture, which requires more labour.

Agriculture declined in the Xiaoheyan culture (Han, 2010) when there was another drought event (Figure 5b). The subsequent Lower Xiajiadian culture developed in a relatively humid period between two drought events, with a significant increase in site number and density that was interpreted as the zenith of the human occupation of this region (Shelach et al., 2011). Fieldwork has revealed that primitive agriculture was more widespread in the Lower Xiajiadian culture than in the Hongshan culture (IACASS, 1996), which is also verified by ANN values. However, because of the vulnerability of primitive agriculture to aridity and low temperatures, the Lower Xiajiadian culture was replaced by the Upper Xiajiadian culture with a major shift in subsistence strategies from primitive agriculture to pasturing (IAIMAR and LMNC, 2009) as a result of the large amplitude of aridity and cooling after 3.0 ka BP (Figures 5b-d). Pasturing requires greater living space but can support a smaller population. Therefore, the number, density and clustering level of sites decrease noticeably in the Upper Xiajiadian culture.

The spatial distribution patterns of archaeological sites can reveal important information on human cultures and the environment. The ANN values (Table 3) reveal that all cultures appear to cluster in space except the Xiaohexi culture, which tends to be a random distribution. The density map of the Xiaohexi culture (Figure 4a) also indicates that its archaeological sites are scattered evenly across space. Although the Xiaohexi, Xinglongwa, Zhaobaogou and Fuhe cultures have approximate maximum density values and their spatial density maps exhibit similar patterns, their ANN values vary. Among these cultures, the Zhaobaogou culture has the smallest ANN value, which indicates the highest clustering level. Based on different catchments and human activities divided between hunting-gathering and agriculture, the sites where these two survival methods were practised present different spatial features. Consequently, the similar spatial density characteristics of the Xiaohexi, Xinglongwa, Zhaobaogou, and Fuhe cultures indicate that their survival strategies were similar, while primitive agriculture occupies a relatively high proportion of food-based activity in the Zhaobaogou culture. This inference is supported by archaeological evidence (IACASS, 1997; IAIMAR, 2004; Liu, 2006; IALP, 2012a).

The Fuhe culture is exceptional because of its larger maximum density value and ANN value. The Fuhe sites are primarily distributed along the Olji Moron River, where the terrain is dominated by hills. Archaeological research has demonstrated that the Xinglongwa culture might be an important origin of the Fuhe culture, and connections have been drawn between the Zhaobaogou culture and Fuhe culture (IALP, 2012a). However, evidence of agriculture was scarcely found in the Fuhe sites. The larger maximum density and ANN value indicate that Fuhe sites are distributed more evenly and randomly. One can conclude that hunting-gathering was dominant in Fuhe culture and that the environmental conditions of the Olji Moron River might not have been suitable for primitive agriculture to develop during that period. The Xiaoheyan culture has a similar ANN value to that of the Fuhe culture but a larger maximum density. This outcome indicates that more sites of the Xiaoheyan culture were distributed evenly in a smaller area (Figure 4f). A colder climate and the improvement of survival skills (Chen, 2009) may be the main contributors to this development.

The areas with the highest density of the studied cultures are centralized within a limited area and surrounded by areas of low density (Figure 4). The decrease in the ANN value and the increase in density imply that more sites were distributed in high-density areas with fewer sites in surrounding areas. This outcome means increasingly more people inhabited the highly populated areas, while marginal areas had increasingly fewer residents. This trend is apparent in the Hongshan culture with a smaller ANN value of 0.405 and in the Lower Xiajiadian culture, with the smallest ANN value of 0.159. Adoption of primitive agriculture is a crucial causal factor of this trend. The high-density areas on their density maps are obvious (and coloured orange and red, respectively, in Figures 4e and 4g). Patterns of concentric rings appear. However, there are differences between them. In the Hongshan culture, the marginal area with lower site density primarily expands towards the west and north, particularly along the Xar Moron River. In contrast, the marginal area expands more towards the south in the Lower Xiajiadian culture, with more sites along the Dalinghe River and fewer sites along the Xar Moron River. A trend for people to move south in response to colder and drier climate conditions is evident.

A different pattern than that of previous cultures occurs for the Upper Xiajiadian culture due to the adoption of pasturing. Compared to agriculture, pasturing is more active pursuit and more resilient to difficult environments, and requires a larger space. However, its capacity to support a human population is smaller than that of agriculture (Han, 2010). This different pattern elucidates why the Upper Xiajiadian culture has more sites, a larger ANN value and a smaller density value than the Hongshan culture. The Upper Xiajiadian sites are distributed more evenly in space, with several core areas of relatively high density (Figure 4h). These core areas are distributed extensively along river basins, including the Laohahe River, the Jiaolaihe River, the Dalinghe River, the Luanhe River and branches of the Xiliaohe River. This development coincides with characteristics of nomadic production, in which herdsmen prefer areas with adequate water and grass. This development also implies pasturing’s noticeable position among subsistence strategies in the Upper Xiajiadian culture and the spread of steppe in this region.

The most densely populated areas of these cultures, except the Fuhe culture, all overlap along the Laohahe and Jiaolaihe river basins. Because of the predominance of hunting-gathering before the Hongshan culture and the preference of hunter-gatherers for geographical environment (Yang et al., 2015), the available flora and fauna resources in this area were more sufficient than in other areas in the Yan-Liao region at 9.5-6.5 ka BP. Primitive agriculture progressed and the relative population index reached a high level in the Hongshan and Lower Xiajiadian cultures (CICARP, 2003). Thus, the proportion of primitive agriculture practised at sites distributed along the Laohahe and Jiaolaihe river basins was also larger than that in other areas. This inference is supported by archaeological evidence of stone implements and remains (Yang et al., 2000; IAIMAR, 2004; Suo et al., 2005; IALP, 2012a; Sun and Zhao, 2013). Considering the thousands of years of continual occupation by human beings and the high density of the Hongshan and Lower Xiajiadian archaeological sites in this area, the population capacity of this basin must have been much higher than that of other areas, which also means a habitable environment in the prehistoric period.