As the Asian Water Tower and the Third Pole of the earth, the Qinghai-Tibet Plateau (QTP), with its typical alpine environment, is the most unique natural region in the world (Qiu, 2008; Lu, 2020). Since 1949, the population and footprint of human activities in the QTP have been increasing. From 1951 to 2019, the population of Qinghai and Tibet increased from 2 703 200 to 9 583 800, an increase of approximately 2.55 times in population size, or an average annual increase of 1.88%. Such population growth continuously reshapes the urban scale system of the QTP, accelerates urban expansion and social-economic development, and has an important impact on the sustainable development of the plateau (Cheng and Shen, 2000; Qi et al., 2020). The QTP, a region that has the most vulnerable ecological environment on the earth and is most strongly affected by human activities, is an important ecological security barrier of Asia (Yao et al., 2017), and a natural laboratory for the research of human-land relations (Yao et al., 2015).

The studies on population distribution and change in the QTP have focused on, for example, the spatial distribution of county-level population on the basis of population census data (Liao and Sun, 2003a; Zhang et al., 2004; 2005), sustainable development of a population, resources and the environment (Cheng and Shen, 2000; Liao and Sun, 2003b), intensity of the impact of human activities (Li et al., 2018b), urbanization (Fan and Wang, 2005), village clusters (Li et al., 2018a), and ethnic distribution (Paik and Shawa, 2013). Studies have also discussed the stability of the QTP as the sparse population area northwest of the ‘Hu Line’, from the perspective of population distribution patterns and evolution in China (Bai et al., 2015; Yang et al., 2016; Liu et al., 2019a) and discussed the regional differentiation regularity of population in the QTP on the basis of the idea of the ‘Hu Line’ (Qi et al., 2020). The QTP is a vast region with prominent differences between its natural and socio-economic environments. Thus, what should be studied are the law of population distribution and its variation characteristics on a precise scale, to describe a more accurate spatial heterogeneity, a more accurate demographic and geographical conditions, and more abundant laws on human-land relations in the QTP. Notably, in the Development of Western China, the QTP has witnessed substantial development in its transportation infrastructure and social economy (Gao et al., 2019a), which has led to population growth and mobility and promoted the remodeling of its population spatial pattern. Additionally, population, as a primary carrier of economic and social activities and a core element in the regional system of human-land relations, has a prominent spatial response to the sustainable development and ecological protection in the QTP and the Third Pole.

Using the population of the QTP at the township level as the research object, this study attempts to assess the population distribution regularity, population change, and influencing factors in the QTP from the perspective of population. Furthermore, this study uses the QTP as the research area, based on the 1982, 1990, 2000, and 2010 national census data and 2017 statistical data, by using mathematical statistics, a concentration index, cold- and hot-spot analysis, the Geodetector method, and geographic information system (GIS) spatial analysis to analyze the spatial distribution and change characteristics of the township-level population of the QTP. The aforementioned population perspective includes two major aspects: first, spatial pattern of population concentration and dispersion, population concentration index and population center of gravity, and inter-annual change in the population; second, the influencing mechanism of natural and socio-economic factors on the population change at the township level is discussed by using human-land relations as a reference and the Geodetector method.

The QTP is located on the southwestern border of China, between 73°29′E–104°41′E and 25°59′N–39°50′N. The total area of the QTP is approximately 2.54 million km², in which there are six provinces (Qinghai, Tibet, Xinjiang, Gansu, Sichuan, and Yunnan), two capital cities (Xining and Lhasa), five prefecture-level cities (Haidong, Changdu, Linzhi, Shannan, and Shigatse), and seven county-level cities (Delingha, Golmud, Yushu, Shangrila, Maerkang, Kangding, and Hezuo) (Fig. 1). To maintain the continuity of analysis and the comparability between years, the administrative division of 2017 is taken as standard in data processing and analysis (Ministry of Civil Affairs, 2018). The spatial scope and population of towns that have undergone administrative division adjustment are adjusted, and the smaller streets are merged. This paper considers 1972 township-level units in the QTP.

Figure 1.  Location of the Qinghai-Tibet Plateau

The data of the permanent resident population at the township level in the QTP are mainly drawn from previous national population censuses, including the third, fourth, fifth, and sixth national censuses of the six provinces and autonomous regions. The data of permanent resident population at the township level in 1982, 1990, 2000, and 2010 are obtained. The data of permanent resident population at the township level in 2017 is from China County Statistics Yearbook 2018 (Department of Rural Social Economic Investigation, National Bureau of Statistics, 2019). The missing permanent resident population data are improved by county annals, yearbooks, or statistical data of corresponding or similar years. The economic data at the township level are obtained from China’s GDP spatial distribution kilometer grid dataset from the Resource and Environmental Science Data Center of the Chinese Academy of Science (Xu, 2017). The years of the selected data are 1995, 2000, 2005, 2010, and 2015. Based on the grid data, the regional statistics module of ArcGIS is used to obtain the economic data of each township in the region. Additionally, the gray prediction model GM (1, 1) (Lai et al., 2004) has been used to obtain the township economic data of 1982 and 1990. The division of township-level administrative regions, GDP of township, average annual relative humidity, annual average temperature, and DEM are derived from the Resource and Environmental Science Data Center of the Chinese Academy of Science (http://www.resdc.cn).

The spatial scope of the QTP is derived from the National Earth System Science Data Center of China (http://www.geodata.cn).The economic data of administrative divisions at and above the county level are derived from the China County Statistical Yearbook (Department of Rural Social Economic Investigation, National Bureau of Statistics, 2019), and statistical yearbooks of each province and city. The road network is derived from the national basic geographic database (the scale is 1 : 250 000) and the corresponding road network is obtained through atlases of different years (Gao et al., 2019b). The total urban economic volume is derived from the statistical yearbook, and by referring to the literature (Gutiérrez, 2001; Murthy et al., 2016; Gao et al., 2019b), the model is established to calculate the location conditions of towns over the years (Gutiérrez, 2001; Murthy et al., 2016; Jacobs-Crisioni and Koomen, 2017; Gao et al., 2019a; b).

The population concentration index is a quantitative index to assess the trend of concentration or dispersion of regional population distribution (Pan et al., 2013). The smaller the population concentration index is, the more dispersed the population distribution is, and vice versa. The population center of gravity is an important index representing population distribution, and its change trajectory can reflect the basic trend and situation of population distribution and its evolution (Ding et al., 2014).

The inter-annual change in population is calculated by the annual average rate of change (Qi et al., 2015; 2020). According to the inter-annual change in population, the population change in the study area is divided into five types of areas according to the annual average change rate r_i and one standard deviation δ of the study area: rapid growth area, r_i ≥ r_ave + δ; stable growth area, r_ave ≤ r_i < r_ave + δ; slow growth area, 0 < r_i < r_ave; slow decline area, –r_ave ≤ r_i <0; rapid decline area, r_i < –r_ave. Among them, r_i is the annual average change rate of population of town i, r_ave is the annual average change rate of population in the area, and δ is the standard deviation.

The G^* index can effectively reveal the local correlation characteristics of each spatial element attribute (Liu et al., 2020). The map of the G^* index directly reflects the spatial distribution of the population and makes the spatial location of the population clear. The formula used is as follow Liu et al. (2020). A positive value of G^* indicates a local concentration of high values, and a negative value of G^* indicates a local concentration of low values. According to the size of G_i^*, the research area is divided into seven types: first-level hot spots, second-level hot spots, third-level hot spots, not significant, third-level cold spots, second-level cold spots, and first-level cold spots.

The Geodetector method was proposed by Wang’s team and can be applied to the study of influencing factors and mechanisms (Wang and Xu, 2017). The Geodetector method is suitable for the analysis of the type of variable we use and compensates for the limitation of the use of variables in a general regression model. Factor detection can detect the interpretation degree of different influencing factors on population spatial differences, and interaction detector can analyze the interaction among influencing factors.

We refer to the literature (Chi, 2012; Kotavaara et al., 2012; Liu et al., 2019a; Wang et al., 2019) and combine the population distribution characteristics of the QTP (Liao and Sun, 2003a; b; Qi et al., 2020), and based on the accessibility and integrity of the data, The adaptability index of living environment (HESI), GDP per capita (GDP_p), location conditions (Loc), land use degree index (L_aj), and urbanization level (U_level) were selected as independent variables, and the annual rate of change of township population (r_i) was taken as the dependent variable. HESI (Liu et al., 2019a; b) indicates the comprehensive factors of natural environment such as climate and terrain. GDP per capita represents the level of economic development (Wang et al., 2019). U_level is the proportion of urban construction land area, that is, land urbanization. Location conditions combine transportation and economic development, reflecting the economic interaction ability between towns and cities (Chi, 2012; Kotavaara et al., 2012). The land use degree index (L_aj) refers to the land use of production and living. Independent variables all passed the collinearity test.

In terms of population density, the QTP is far lower than the average population density of the whole of China and western China (Bai et al., 2015), with the characteristics of a dense population in few regions and a sparse population in most regions. The population size and population density in the QTP are increasing (Table 1). The total number of permanent residents increased from 9 546 600 in 1982 to 14 897 800 in 2017. The population density correspondingly increased from 3.52 people/km² in 1982 to 5.49 people/km² in 2017.

The population distribution presents the characteristics of ‘dense in southeast and sparse in northwest’ on the whole. The distribution pattern of the population density from 1982 to 2017 (Figs. 2a–e) is drawn according to the cumulative percentage, which has typical regional differentiation characteristics: towns with high population density are concentrated in the northeast of the plateau, the Valley Area of ‘One River and Two Tributaries’ in Tibet (the Valley Area), the southeast edge of the plateau, and seats of prefecture-county government. The towns with the second-highest population density are mainly distributed in the southeastern QTP, such as Hainan Prefecture, Gannan Prefecture, and Changdu, namely, the southeast side of the Qilian-Jilong Line (‘Qiji Line’) (Qi et al., 2020). Towns with low population density are concentrated on the northwest of the plateau, for example, Ali, the Qiangtang Plateau, Hoh Xil, Qaidam Basin, and Kunlun Mountain. From the evolution of the spatial pattern of population distribution, we find that the population distribution pattern in the QTP is relatively stable.

Figure 2.  Spatial distribution of population density and population gravity center in QTP

The population center of gravity is gradually approaching Yushu from east to west (Fig. 2f), which indicates that the population center of gravity is still located in the east of the Qiji Line. The population growth of towns in the hinterland of the plateau is greater than marginal areas in the east, prompting the population center of gravity to move westward. Among them, the movement distance of the population center of gravity from 2011 to 2017 is the longest (approximately 26 km), followed by that from 1982 to 1990 (approximately 21 km), from 1991 to 2000, and from 2001 to 2010 (approximately 6 km and 15 km, respectively). From 1982 to 1990, under the influence of the macro situation of China’s reform and opening up (Yang et al., 2016), as well as the combined effect of the social culture, education level, and ethnic distribution (Paik and Shawa, 2013; Li et al., 2018a), the population in the eastern edge of the plateau was more likely to migrate to the southeast coast of China than to the hinterland of the plateau, which promoted the substantial movement of the population center of gravity. Since 1990, the movement distance of the population center of gravity has gradually increased, indicating that the proportion of the population in the hinterland of the plateau is accelerating.

From 1982 to 2017, the concentration index of population distribution in the QTP decreased annually (Table 2), and the distribution pattern of the population cold and hot spots is shown in Fig. 3. The distribution of cold and hot spots has relatively stable regional distribution regularities: cold spots are mainly distributed in the areas near Shigatse-Lhasa-Linzhi-Kangding-Maerkang, and hot spots are concentrated in the Three Parallel Rivers Areas, northeastern QTP, and the Qaidam Basin. The area of population hot-spot concentration in the QTP is increasing, and the number of towns is decreasing. The area of population hot spots and the number of towns in the QTP are both decreasing, indicating that the population distribution is mainly distributed dispersedly.

Figure 3.  Distribution pattern of cold and hot spots of township-level population in QTP

In 1990, the surrounding areas of Golmud were transformed into hot spots, and the area of population hot spots has increased from three to four (i.e., northeastern QTP, the Three Parallel Rivers Areas, Hotan of Xinjiang, and the surrounding area of Golmud). Guolemude Town, Zongjia Town, and other towns have a large area of land, increasing the hot concentration area. However, the number of hot-spot towns, especially the number of first-level hot-spot towns declined significantly.

The area of cold spot concentration and the number of towns have been shrinking. In 1982, four zones were formed in the cold spot concentration area: western Sichuan (i.e., Ganzi, Aba; and southern Guoluo), eastern Tibet (i.e., the surrounding area of Linzhi, which is in the eastern Lhasa and western Changdu), western Tibet (i.e., western Lhasa, mainly in Shigatse, southern Nagqu and other areas), and the surrounding area of Ali (i.e., Gaer and Zada). In the past 35 yr, the cold spots decreased continuously, and the cold spots around Ali disappeared. The cold spot concentration area in eastern and western Tibet shrank sharply. By 2017, the first-level cold spots were distributed only in western Sichuan.

In addition to obvious cold and hot spots, the population distribution of most towns in the QTP is random, and the number of towns belonging to insignificant cold and hot spots has increased by 328 over the 35 years. Over time, the population hot spots in the QTP will continue to shrink and divide into Xining, Haidong, Golmud, Lijiang, and other major plateau cities and nearby areas. The cold spots will gradually shrink to Aba and Ganzi, and the spatial distribution of the population of the QTP will become more dispersed on the whole.

Table 3 and Fig. 4 show the growth of permanent resident population at the township level in the QTP from 1982 to 2017. From 1982 to 1990, the rapid growth areas and stable growth areas are mainly located in Tibet, southern Qinghai, western Xinjiang, and the surrounding areas of Golmud. The slow growth areas are mainly distributed in western Sichuan, the Three Parallel Rivers Areas, the eastern edge of the QTP, the southeastern Changdu, and other areas. Areas with decreased population are distributed mainly in Shiqu, the Qaidam Basin, and the northern edge of the plateau. From 1991 to 2000, despite the decreased permanent population of towns in northern QTP and border areas, the permanent population of remaining towns increased rapidly or stably. The towns with slow growth are mainly distributed in western Sichuan, the Three Parallel Rivers Areas, the Hehuang Valley, the Qilian Mountains, southern Qinghai, and other areas.

Figure 4.  Population growth types at the township level in QTP

From 2001 to 2010, most towns in the Qaidam Basin, western and central Tibet, Linzhi, southeastern Qinghai, and Ganzi increased rapidly or in a stable fashion. The slow growth areas are concentrated in the Valley Area, southeastern Tibet, the Three Parallel Rivers Areas, and Hotan. The towns with decreasing populations are distributed mainly in Hehuang Valley, Hoh Xil, Qilian Mountains, Aba, Shannan, and other areas. From 2010 to 2017, different types of population growth in towns are mosaic distribution. Among these towns, those with rapid, stable population growth are distributed mainly in southern Qinghai and Tibet. Others with reduced population are concentrated in the Qaidam Basin, Jiuquan, Shannan, Lijiang, and other areas.

Overall, the annual change in population has relatively stable regional differentiation regularity. The population growth rate of towns in the plateau hinterland is higher than the average rate. The population growth rate in the northern and eastern edges of the plateau and towns in the southeast of the plateau is lower than the average rate. It promotes the moving of the population center of gravity westward. In conclusion, the population growth rate of the majority of towns in Tibet is higher than the average growth rate with the rapid and stable growth as the main types. The growth rate of permanent resident population in most towns in western Sichuan and the Three Parallel Rivers Areas is lower than the average growth rate of the plateau and mainly the type of slow growth. The change of permanent resident population in the Qaidam Basin is the most drastic, and the permanent resident population there is most affected by resource development and economic situation; additionally, the rapid growth towns coexist with those of rapid decline.

The population distribution is influenced by the natural environment, economic development, and social and historical conditions (Hu, 1990). The effect intensity of social and historical factors is far lower than that of the natural environment and economic development in the QTP . In order to reveal the influencing factors and socio-economic implications of township-level population change in the QTP, this paper selects natural environment and socio-economic indicators, and uses the Geodetector method to make quantitative analysis.

Based on the Geodetector method, the effect intensity (q value) of each factor on population change is calculated (Table 4). The results show that the influence of natural factors on population change in the QTP is extremely weak, and socio-economic development is the leading factor affecting population change. The influence degree of different factors on population change has significant differences in different periods. GDP_p (GDP per capita) is the decisive factor of population change in the QTP, followed by Loc (location conditions) or U_level (urbanization level).

Getting more economic income is the main driving factor of population mobility and change. GDP_p is the leading factor of population change in township units of the QTP in each period. From 1982 to 1990, the main influencing factor of population change is GDP_p (with the q value of 0.3020), followed by Loc (0.0386) and U_level (0.0066). From 1991 to 2000, the influence of various factors on population change in the QTP decreased obviously, with the leading factor GDP_p (0.0538), followed by U_level (0.0096) and Loc (0.0035). From 2001 to 2010, the effect intensity of GDP_p is 0.3330, and the influence intensity of U_level is close to that of Loc. They are 0.0116 and 0.0162 respectively. From 2011 to 2017, the dominant factor is per capita income (0.3507), and the effect intensity of U_level is 0.0264.

Over a longer time scale, the effect intensity of GDP_p on population change from 1982 to 2000 is 0.0980, far lower than that from 2001 to 2017 (0.5688), however, the effects of factors such as HESI (the adaptability index of living environment) and U_level are more intensive than those after 2000. This indicates that with the implementation of the Great Western Development Strategy of China in 2000, and the policy of returning grazing land to grassland program, and the ecological protection of the Three River Headwaters Region, the QTP has gradually changed from the extensive economic development mode to the man-land harmonious development mode. The change of economic development mode not only improves the diversity of herdsmen’s livelihood, but also reduces the dependence of local residents on grassland and cultivated land. L_aj on population change decreases from 0.0142 to 0.0036. Altitude, climate and other factors are the natural geographical background conditions of population distribution in the QTP (Liao and Sun, 2003a; Paik and Shawa, 2013; Qi et al., 2020), and they are also important influencing factors of population distribution. However, the influence of natural environment factors on population change in the QTP is weak. This shows that with the continuous progress of science and technology and the deepening of reform and opening up, people’s ability to adapt to the natural environment of the QTP has been significantly enhanced, and socio-economic factors have become the leading factors affecting population distribution and mobility, and have reshaped the spatial distribution of township population in the QTP.

The results of interaction between different factors on population change show that the interaction between GDP_p and other factors in different periods has passed the significance test (Table 5), which is a nonlinear enhancement relationship, especially after 2000, and the explanatory power of the interaction measurement is obviously higher than that before 2000. The interaction among GDP_p, HESI, U_level, Loc and L_aj will enhance the explanatory power of population change, which also reveals the leading role of per capita income level in population change in the QTP.

The effect intensity of different factors on population change in the QTP is quite different in different periods. In this paper, the relative effect intensity is used to compare the influence differences of different factors in different periods, so as to reveal the effect and mechanism of different factors on population change in the QTP. It can be seen from Table 6 that the relative effect of GDP_p on population change in four short periods is greater than 0.70. The relative effect intensity of GDP_p, Loc, U_level and other factors has changed obviously in a long period (1982 to 2000, 2001 to 2017).

The relative effect intensity of HESI and L_aj on population change is low, and it first increases and then decreases in four short periods. The relative effect intensity of U_level increases from 0.0188 during 1982–1990 to 0.1405 during 1991–2000, then decreases to 0.0315 during 2001–2010, and changes to 0.0690 after 2010. The relative effect intensity of Loc on population change shows a gradual decline in four short periods, from 0.1097 to 0.0443 during 2001–2010, especially only 0.0079 during 2011–2017.

Over a longer time scale, the per capita income level can promote population change, and the relative effect intensity of GDP_p on population change is significantly enhanced. Especially since 2000, China has implemented the great western development strategy, and the relative effect intensity of GDP_p on population change has increased from 0.5160 to 0.9188. The relative effect intensity of other factors on population change shows a downward trend around 2000. The location condition and U_level decrease from 0.2172 and 0.1751 to 0.0289 and 0.0455 respectively, and L_aj and HESI also decrease significantly.

The above analysis finds that before 2000, the relative effect intensity of U_level and Loc on population change of towns in the QTP is strong, and it decreases significantly after 2000. In the 1980s, the household registry reform, especially the change from ‘agricultural to non-agricultural’ status, released the original demand and motive force for farmers and herdsmen to move to cities and towns (Wang and Tong, 2016). After 2003, the urban-rural integration system was gradually promoted, which led to the obvious weakening of the relative effect intensity of U_level on population change around 2000. By 2000, the QTP had initially formed a transportation network dominated by highways (Zhao, 2002; Miao et al., 2020). After 2000, railways and airports in the QTP developed rapidly (Gao et al., 2019a). However, highways deep into the hinterland of the plateau are still the main mode of transportation of the QTP at present. The construction of major transportation infrastructure, such as railways and expressways, has greatly improved the location conditions and opportunities for socio-economic development of towns in the QTP. For the vast QTP, most towns have farming and animal husbandry economic structures, which are too far away from cities, too high in transportation costs, and weak in economic interaction ability, prompting the relative effect intensity of location conditions on population changes to gradually decrease.

The above analysis reveals that compared with the natural geographical background conditions of population distribution, GDP_p is the most influence factor of the changing intensity of permanent population in towns of the QTP. Especially after 2000, China’s Great Western Development Strategy promoted urban economic development in the QTP. Meanwhile environmental protection policies (e.g., eco-migration and national parks moved herders from the Three River Headwaters Region and other areas to urbanization areas with higher per capita income such as Yushu and Xining), which further strengthened the impact of GDP_p on population changes in the QTP, followed by Loc and U_level. The improvement of traffic conditions reduces the traffic time between towns and cities and improves the location conditions and economic potential of towns, but it does not bring direct economic benefits. Coupled with the characteristics of sparsely populated plateau areas and land ownership in China, the development of transportation can only attract people to gather in a few areas such as cities. Therefore, the relative effect of location conditions on population in the QTP is declining. HESI and L_aj have a certain influence on population change, but the relative effect intensity is weak.

Based on five terms of permanent resident population data at the township level in the QTP from 1982 to 2017, our discussions have focused on the spatial distribution and regional differentiation characteristics of population at the township level in the QTP from the aspects of, for example, population density and size, center of gravity trajectory, and spatial pattern. By establishing natural and socio-economic indicators, and using geographical detectors, this research has analyzed the influencing factors and socio-economic connotations of township-level population change in the QTP from 1982 to 2017 for the first time. It draws the following conclusions:

(1) The population distribution of towns in the QTP is unbalanced, showing the spatial distribution characteristic of ‘dense in southeast and sparse in northwest’, and the population center of gravity moves westward to the hinterland of the plateau. From 1982 to 2017, the permanent resident population of towns in the QTP showed an increasing trend. The densely populated towns are concentrated in the Hehuang Valley, the Valley Area of ‘One River and Two Tributaries’ in Tibet, the southeast edge of the plateau, and seats of prefectural government. The sparsely populated towns are distributed in the northwest of the Qiji Line, with typical regional differentiation characteristics. Affected by population outflow from the eastern edge of the plateau and population growth in the hinterland of the plateau, the population center of gravity of towns in the QTP gradually approaches Yushu from the east to the west and shows a trend of acceleration.

(2) The decentralized distribution trend of township-level population in the QTP has improved, and the distribution of cold and hot spots shows regional distribution regularities of obvious spatial correlation and relative stability. From 1982 to 2017, the concentration index of population distribution in the QTP decreased annually: from 0.6763 in 1982 to 0.6654 in 2000 and 0.6573 in 2017. Both the cold and hot spots of population show spatially aggregated distribution. The number of towns with the type of cold and hot spots that are aggregating is decreasing. The hot-spot aggregating area has increased from the northeastern QTP, the Three Parallel Rivers Areas, and Hotan of Xinjiang to four zones, with the surrounding areas of Golmud being added. The cold spots are mainly distributed in areas near Shigatse-Lhasa-Linzhi-Kangding-Maerkang. The number and scope of towns where cold spots are aggregated are sharply reduced. Especially since the implementation of the Development of West China, the first-level cold spots are distributed only in western Sichuan.

(3) The annual change of township-level population has relatively stable regional distribution regularities. From 1982 to1990, 1991 to 2000, 2001 to 2010, and 2011 to 2017, the annual change of permanent resident population in the QTP was respectively 1.74%, 1.14%, 1.21%, and 1.05%, and the annual change of permanent resident population aged over 35 yr was 1.28%. The towns whose population presents rapid, stable growth are mainly distributed in Ali, the Qiangtang Plateau, Shigatse, southern Qinghai, and other plateau hinterlands. Most towns in the northern edge, eastern edge, and southeastern part of the plateau are areas of slow population growth, and the towns with a rapidly decreasing population are mainly Wutumeiren Town, Lenghu Town, and other resource-exhausted towns.

(4) The analysis of the influencing factors of population change shows that the per capita income level (GDP_p) is the leading factor of population change in towns of the QTP, and it is gradually increasing. Location conditions (Loc) and urbanization level (U_level) also have a certain impact on population changes. Natural conditions, land use and other factors have little influence on population change. Comparing the relative effect intensity of influencing factors on population change in different periods, we find that the relative effect intensity of GDP_p has increased significantly after 2000, while the Loc and U_level have decreased significantly. The change of explanatory power of various factors on population change is closely related to the stages of China’s macro-economic and social development, such as China’s Great Western Development Strategy, household registry reform, and environmental protection.

A township is the most basic administrative and regional division, and the basic unit of national population census in China. Based on the permanent population data of the QTP from 1982 to 2017, this paper explores the evolution and regional differentiation characteristics of population distribution patterns on township scale for the first time and analyzes the influencing factors of population change of towns. With the influence of environmental protection policies and eco-migration, population will gradually concentrate in the towns with higher per capita income level, which are the main cities of the QTP. In view of the environmental characteristics of the QTP, this process may break the balance of resources and environmental carrying capacity in population concentrated areas, and even cause a wider range of ecological problems. Thus, future research should pay attention to issues relating to population change and environmental carrying capacity in urban areas of the QTP and the influence of ecological environmental protection policies such as returning grazing land to grassland policy and national parks on population change and their ecological effects, which will provide new perspectives for sustainable development of man-land relationship in the QTP.