1. Introduction
The southeastern Tibetan Plateau is one of the key birthplaces of prehistoric agricultural civilization on the plateau. Historically, it served as a vital node for trade and ethnic exchange along the ancient Tea-Horse Road, while today it remains a critical transportation hub linking Tibet with Sichuan, Qinghai, and Yunnan. This region has been recognized as the “gateway to Tibet” since antiquity [
1,
2]. Prehistoric sites in this region, including Karuo, Lapo, Xiaoenda, and Jiangqin, document the emergence and spread of Neolithic agricultural practices. Archaeobotanical evidence reveals cultural linkages between prehistoric populations and Yellow River agricultural traditions, marking the inception of farming on the Tibetan Plateau. The establishment and development of the ancient Tea-Horse Road further support the pivotal historical role and gateway function of southeastern Tibet in the socioeconomic development of the Tibetan Plateau [
3]. Since 1951, the development of transportation infrastructure, including the Sichuan-Tibet Highway and Railway, has further enhanced the geographical significance of southeastern Tibet. These developments have revitalized academic investigations into historical transformations of settlements, population dynamics, and agricultural land use patterns in the region.
As a fundamental component of historical Land Use and Cover Change (LUCC) research, cropland reconstruction provides both a direct measure of anthropogenic impacts on terrestrial ecosystems and a crucial interface between socioeconomic systems and ecological processes [
4]. Since the mid-1990s, many international scientific programs (IGBP, IHDP, LUCC, GCTE, and GLP) have advanced our understanding of global and regional land use/cover change through datasets like HYDE [
5,
6], KK10 [
7,
8], PJ [
9], and SAGE [
10]. However, the coarse spatial resolution of these global-scale datasets limits their utility for regional studies [
11,
12,
13]. Consequently, researchers worldwide have extensively utilized historical archives to conduct regional-scale cropland reconstructions that address these limitations [
14,
15,
16]. While significant progress has been made in reconstructing historical cropland patterns in China, notable gaps remain in both temporal coverage and geographic representation. Temporally, most studies focus on the past 300 years [
17,
18], with few extending to millennial scales [
19,
20]. Geographically, research has predominantly concentrated on well-documented eastern agricultural regions [
21,
22,
23,
24]. In contrast, the Qinghai–Tibet Plateau region presents particular challenges due to limited historical documentation. Existing studies have focused on relatively well-documented areas like the Huangshui River Basin [
25,
26,
27,
28] and the Three-River Region (Yarlung Zangbo River, Lhasa River, and Nianchu River) [
29,
30,
31], while in southeastern Tibet historical agricultural patterns remain understudied because of its limited arable land, geographic isolation, and paucity of historical records.
For this purpose, this study draws on the Second Comprehensive Scientific Expedition to the Qinghai–Tibet Plateau, conducting in-depth field research and historical data analysis. Leveraging the technical expertise and experience of our team in reconstructing the Qinghai–Tibet Plateau, we employ a grid-based reconstruction model. Specifically tailored to the complex topography, small and fragmented farmland areas, and other characteristics of the southeastern Tibet region, the model estimates the farmland areas of the Tubo, Yuan, Ming, and Qing dynasties and reconstructs the historical patterns of farmland distribution in complex mountainous regions at a fine-scale level. This study explores the historical trajectory of agricultural development in mountainous regions under complex topographical conditions, providing a reference for historical farmland reconstruction in other similar regions.
2. Study Area
2.1. Physical Geographical Setting of the Study Area
The study focuses on the southeastern Tibet Autonomous Region, specifically encompassing the Changdu and Linzhi prefectures, which span latitudes 26°52′–32°28′ N and longitudes 90°05′–98°47′ E (
Figure 1). Encompassing approximately 224,700 km
2. This area represents a critical transition zone between the Qinghai–Tibetan Plateau and Hengduan Mountain ranges. It is characterized by a significant elevational gradient that decreases from north to south. The highly dissected topography features steep mountain ranges and profound river gorges, with well-developed fluvial landforms. The area contains abundant hydrological resources, including glacial meltwater, extensive wetland systems, numerous lakes, and high levels of precipitation, supporting a dense river network. Major river systems comprise the Jinsha, Lancang, Nu, and Yarlung Zangbo Rivers. The predominant north–south topographic gradient directs all major rivers southward through deep gorges, which have historically been the primary focus of human settlement and activity.
2.2. Historical Trajectory of Agricultural Development in the Region
Archaeological evidence dates agricultural activities in southeastern Tibet to the Neolithic period (5300–4300 BP), as shown by sites including Changdu’s Karuo, Chayu’s Jiangqin, and Linzhi’s Yunxing and Jumu [
32]. Notably, the Karuo site revealed foxtail millet remains and lithic agricultural tools, suggesting cultural diffusion from the Yellow River’s Majiayao Culture. This established a millet-based agricultural system representing a key node in the westward expansion of millet agriculture [
33,
34]. During this phase, farming was limited to river terraces due to primitive stone tools and slash-and-burn methods. Throughout the Qin and Han dynasties, agricultural techniques remained rudimentary without major technological breakthroughs, advances, resulting in limited farmland development. The Tubo dynasty expanded agriculture expansion through plateau unification. According to
the Old Book of Tang [
35], at that time, the Changdu region was characterized by all dwellings being multi-story structures, with the king living in a nine-story building and the common people in six-story buildings. The king wore a blue silk skirt, a collarless shirt, and a blue robe with sleeves reaching the ground. In winter, he wore a lambskin coat decorated with embroidered brocade, which indirectly reflects the prosperous agricultural and pastoral economy of southeastern Tibet at the time. In the late 13th century, after the Yuan dynasty unified Tibet, the central government established the ‘Xuanzheng Yuan’ (Bureau of Buddhist and Tibetan Affairs) and implemented the ‘Wan hu’ system, conducting population surveys and tax collection, which marked Tibet’s formal incorporation into the central administrative system [
36]. Although southeastern Tibet was not fully incorporated into the thirteen ‘Wan hu’ system, the region maintained a lenient tributary relationship with the central Yuan authority. The introduction of the Yuan post station system significantly improved transportation and connectivity between this area and inland China, facilitating the movement of people and goods and strengthening political and economic ties. Following the establishment of feudalism in the late 13th century, the manorial economy gradually replaced self-cultivating peasant agriculture across much of Tibet. However, in the southeastern region, the manorial system remained underdeveloped. Only in a few isolated areas were manors organized on the scale of thousands or tens of thousands of households established. As a result, self-sufficient peasant agriculture remained widespread in southeastern Tibet, and the feudal manor economy had relatively limited impact on local agricultural production. The Ming dynasty continued the governance strategies of the Yuan, establishing the ‘Wusi Zang’ and ‘Duo gan’ Regional Military Commissions and maintaining relations with local religious powers through a system of conferment and tribute. Southeastern Tibet fell under the jurisdiction of Wusi Zang (administering Linzhi) and Duogan (administering Changdu). By the Ming and Qing dynasties, with the development of the plateau courier network, southeastern Tibet became a vital passageway into Tibet from Shanxi, Gansu, and Sichuan. By the mid-Qing dynasty, the central government’s governance over Tibet had further deepened through the conferment of the Dalai Lama, the appointment of resident commissioners (ambans), and multiple surveys of households and land to strengthen administration. With the establishment of military and administrative officials and an increase in garrisoned troops, many of Han Chinese immigrants moved in. Some retired soldiers or garrisoned troops cultivated wasteland locally and gradually settled down, opening farmland cultivation of grain and vegetables. This led to the saying, “Producing vegetables, therefore, many Han Chinese settled near the city and cultivated land as gardens, with dozens of melon trellises and bean frames, resembling the inland regions [
37]”. Historical records indicate that the main crops grown in southeastern Tibet during historical periods included cold-resistant varieties such as barley, wheat, and buckwheat, with some rice cultivation also present in the Linzhi region [
38]. Additionally, small-scale vegetable cultivation, such as radishes and cabbage, likely existed in river valleys. This type of garden agriculture developed further during the Qing dynasty with the introduction of Han Chinese migrants. In terms of livelihood strategies, the agricultural system in the region exhibited diversified characteristics, including agro-pastoral combination, terrace farming, and utilization of forest resources (e.g., gathering and hunting), adapting to the variable and unstable climatic conditions of the high-mountain environment.
3. Data and Methods
3.1. Data Sources and Description
Historical settlement data were compiled from multiple authoritative sources, including the Third National Cultural Relics Survey,
the Historical Atlas of China [
39],
Tibetan Buddhist Monasteries [
40], and
Tibetan History during the Qing Dynasty [
41]. Settlement records were georeferenced by cross-validating historical toponyms with modern gazetteers and local historical archives, employing both toponymic analysis and spatial verification techniques. Our analysis identified 457 significant historical settlements distributed as follows:88 from the Tubo Dynasty, 90 from the Yuan Dynasty, 130 from the Ming Dynasty, and 149 from the Qing Dynasty. Demographic data were obtained from authoritative historical compilations, including
China Population: Tibet Volume [
42] and
Studies on the History of Tibet During the Yuan Dynasty [
43]. Historical cropland areas were reconstructed through a multi-proxy approach integrating population records, historical seed yields, and documented cultivation extents, ensuring chronological consistency across different historical periods.
3.2. Estimation of Cropland Area
Population size serves as a key factor influencing arable land dynamics and is commonly employed as a proxy variable for cropland area reconstruction. However, extant Tibetan and Chinese historical records contain only fragmentary and imprecise demographic records for this region. While the Changdu region has some dispersed records of settlements, population, and cropland, such documentation is particularly scarce for the Linzhi region. This paucity of primary historical sources necessitates estimation of historical population parameters and agricultural characteristics through indirect proxy methods. Based on this, this study utilizes modern population distribution data from the Tibet region, applying certain rules to estimate the historical population sizes at different times. Using these population estimates, we reconstructed arable land areas in southeastern Tibet during the Tubo dynasty, as well as during the Yuan, Ming, and Qing periods (
Table 1). This methodology is predicated on a core assumption: the relative proportions of regional populations within Tibet’s total population have remained relatively stable over historical time periods. The validity of this assumption derives from the enduring constraints imposed by the natural geography of the Qinghai–Tibet Plateau on population distribution over millennia, and is further supported by empirical evidence. For instance, Tibet’s total population in 1952 was recorded as 1.15 million [
42]. Applying the modern proportion of Linzhi’s population relative to Tibet’s total (approximately 6%) yields an estimated population of around 69,000 for the Linzhi region during that period. This estimate closely matches the population figure documented in the Annals of
Linzhi Prefecture for the same time (approximately 70,000) [
44], with a relative error of only 1.43%. This correlation demonstrates that, in the absence of direct historical demographic records, the use of the stability hypothesis combined with modern proportional data can offer a reasonably reliable approximation of historical population size and cultivated land extent.
3.2.1. Tubo Dynasty (633–842 AD)
Following the unification of the Qinghai–Tibet Plateau by Songtsen Gampo, the territory was administratively organized into five ‘Ru’: Weiru (also known as Wuru, corresponding to the present-day Lhasa River Basin), Yueru (encompassing contemporary Shannan, Gongbu, Jiangda, Milin, Linzhi and Lang counties), Yeru (including present-day Nangmulin, Renbu, Nimu, Xietongmen, and Angren counties), Rula (covering parts of modern Rikaze Prefecture such as Sangzhuzi District, Jiangzi, Bailang, Lazi, Saga, Gangba, Dingjie, Dingri, Nielamu, Kangma, Yadong, Jilong, Saga, Zhongba, and other counties), and Supiru (also called Sunboru, which included eastern Naqu, northwestern Changdu, and much of the Yushu Tibetan Autonomous Prefecture in Qinghai Province) [
45]. Historical sources indicate that the population of Weiru was approximately 740,000, Yueru 700,000, Yeru 700,000, Rula 720,000, and Supiru 120,000, yielding a total estimated population of 2.98 million across the Tibetan Plateau during this period. For the Tubo Dynasty, historical documents specifically note a population of approximately 200,000 in the Changdu region [
42], but no explicit records survive for the Linzhi region. Therefore, this study estimates the population of Linzhi by applying the modern population proportion of Linzhi relative to the total population of Tibet to the historical population total of the Yueru region, to which Linzhi historically belonged. According to the fifth and seventh national population censuses, the contemporary population of Linzhi accounts for approximately 6% of Tibet’s total population. This proportion has remained relatively stable over the past four decades, suggesting that the historical population ratio may have been similar. Accordingly, the population of Linzhi during the Tubo period is estimated to have been approximately 178,800. Combined with the documented population of 200,000 in Changdu, the total population of southeastern Tibet during this era is estimated to have been approximately 378,800. Historical demographic data from 1990 and 2000 indicate that the agricultural population accounted for approximately 70% of the total population in Changdu and 83% in Linzhi [
44,
46]. Assuming that the agricultural population proportion in historical periods was broadly consistent with these modern figures, the agricultural population in southeastern Tibet during the Tubo period is estimated to have been approximately 290,000.
During the Tubo Dynasty, agricultural development reached its peak thanks to the implementation of a series of agricultural policies and advancements in agricultural technology. During this period, arable land on the Qinghai–Tibet Plateau was primarily concentrated in the Three-River region, as well as the Nu River and Lancang River basins. The per capita arable land area was 1.46 mu [
47], and based on population estimates, the arable land area in the southeastern Tibet region during the same period was approximately 28,085 hectares.
3.2.2. Yuan Dynasty (1271–1368 AD)
In 1271, the Tibetan region was formally integrated into the territory of the Yuan Dynasty. To facilitate tax collection and administrative governance, Kublai Khan, the founding emperor of the Yuan Dynasty, dispatched officials to Tibet on three occasions—1260, 1268, and 1287—to conduct population censuses. The unit of enumeration adopted was the ‘huoerdu [
43]’. It should be noted that these censuses excluded monastic populations and the Menluo region, as well as areas such as Changdu, Linzhi, and Ali. According to the
Dictionary of Chinese Ethnic Minority Cultures: Southwest China Volume [
48], which draws upon both Tibetan and Chinese historical sources along with the three Yuan-era censuses, the Yuan administration established thirteen ‘wanhu’ in Tibet, with an average household size of six individuals. This indicates a population of approximately 780,000 in the Weizang region, 20,400 in Ali, 50,000 in Changdu, and 150,000 in Menluo. An additional 14,000 individuals in Weizang were not affiliated with any wanhu’ unit. The monastic population in Tibet was estimated at around 70,000, resulting in a total population of approximately 1.23 million for the region during this period [
46]. Based on data from the fifth and seventh national population censuses, which indicate that the population of Linzhi accounts for about 6% of Tibet’s total population, the estimated population of Linzhi during the Yuan Dynasty is calculated as 73,800. Combining this with the documented populations of Changdu (50,000) and Menluo (150,000), the total population of southeastern Tibet during the Yuan Dynasty is estimated to be approximately 273,800. After accounting for the proportion of the agricultural population in the Changdu and Linzhi regions, the agricultural population in southeastern Tibet during the Yuan period is inferred to be around 220,000.
The arable land data for the Yuan Dynasty were derived from conversions based on the ‘huoerdu’ unit. One ‘huoerdu’ was defined as a household consisting of six people, 24 livestock, a house supported by six pillars, and a plot of land capable of sowing 12 ‘Mongoliagrams’ [
43]. The term ‘Mongoliagrams’ is hypothesized to have been synonymous with ‘Zangke’ [
49], a historical land measurement unit in Tibet. During this period, one ‘Zangke’ of arable land was equivalent to the area sown with approximately 14 kg of seeds, which corresponds roughly to one mu in the contemporary Chinese measurement system. Based on this conversion, the average per capita arable land during the Yuan Dynasty is estimated to have been approximately 2 mu. Using the previously established population figure for southeastern Tibet during the Yuan Dynasty, the total arable land area in this region is thus estimated to have been approximately 29,449 hectares.
3.2.3. Ming Dynasty (1368–1644 AD)
In 1368, the Ming Dynasty was established, and it administratively divided the Tibetan region of the Qinghai–Tibet Plateau into three major areas: Duogan, Weizang, and Ali [
50]. The Changdu region fell under the jurisdiction of the Duogan Military Command, while the Linzhi region was governed by the Weizang Military Command. According to demographic research presented in A History of China’s Population, the total population of Tibet during the mid-Ming period was approximately one million [
51]. However, historical records from this era do not provide specific population figures for the Linzhi and Changdu regions. Consequently, this study estimates the populations of these two areas based on proportional data derived from the fifth and seventh national population censuses of China, which indicate that the contemporary Linzhi and Changdu regions account for approximately 6% and 20% of Tibet’s total population, respectively. Applying these proportions to the mid-Ming demographic context yields an estimated population of approximately 60,000 for the Linzhi region and 200,000 for the Changdu region during the Ming Dynasty. The total population of southeastern Tibet in this period is therefore estimated to have been around 260,000. Furthermore, based on the agricultural population proportion in the Changdu and Linzhi regions, the farming population in southeastern Tibet during the Ming Dynasty is inferred to have been approximately 189,800.
Based on the reconstructed population data for southeastern Tibet during the Ming Dynasty and considering the continuity of agricultural management practices inherited from the Yuan Dynasty into the Ming and Qing periods, the per capita arable land area was estimated using the Yuan Dynasty benchmark of 2 mu per person. This calculation yields an estimated total arable land area of approximately 25,319 hectares for the southeastern Tibet region during the Ming Dynasty.
3.2.4. Qing Dynasty (1644–1911 AD)
During the reigns of the Yongzheng and Qianlong Emperors, the Qing government conducted detailed censuses in Tibet. Historical records indicate that the total population of Tibet in the second year of the Qianlong reign (1737) was approximately 940,000, with the Changdu region accounting for an estimated 200,000 inhabitants [
42]. According to data from the fifth and seventh national population censuses, the contemporary population of the Linzhi region constitutes approximately 6% of Tibet’s total population. Applying this proportion to the Qing Dynasty demographic context gives an estimated population of 56,400 for Linzhi during that period. Combining this figure with the documented population of Changdu (200,000) yields a total estimated population of approximately 256,400 for southeastern Tibet during the Qing Dynasty. Furthermore, based on the agricultural population proportion within the Changdu and Nyingchi regions, the farming population in southeastern Tibet during the Qing Dynasty is estimated to have been approximately 186,800.
The cultivated land data for southeastern Tibet during the Qing Dynasty were mainly derived from
the Inventory of the Year of the Iron Tiger [
52]. As this register primarily documents tax obligations of temples, nobles, and manors across Tibet, a conversion process was necessary to estimate actual cropland area [
53]. In Tibet, the unit ‘Gang’ was applied predominantly to government-owned farmland, whereas ‘Dun’ was used for lands owned by nobles and temples [
52]. The standard conversion ratio between these units was approximately 1 Dun to 2 Gang. Actual farmland measurement employed units such as ‘Tu’ (referring to the area cultivated in one day) and ‘Zangke’ (a unit based on seed quantity) [
54]. Specifically, 1 ‘Zangke’ of land denotes the area required to sow 1 ‘Zangke’ (approximately 14 kg) of seeds, which is equivalent to about 1 standard mu [
55]. Although Gang and Dun cannot be directly converted into arable land area, the Qing government explicitly stipulated in the 16th year of the Guangxu reign (1891) that “every 40 ‘Zangke’ of land shall incur 1 Gang of tax liability [
55]. Based on this regulation, the conversion relationship for arable land area adopted in this study is as follows:
Subsequently, a statistical compilation was conducted according to the various types of taxes levied on individual manors. Items that could not be converted into arable land area were excluded, and duplicate accounting of obligations paid by serfs across nobles, temples, and governmental manors was carefully avoided. The tax records
the Inventory of the Year of the Iron Tiger include both farmland and fallow land where serfs could not fulfill corvée labor obligations or pay taxes. Since the extent of tax evasion was not explicitly recorded, this study adjusts the data by incorporating an estimated proportion of concealed land. Surveys from certain regions during the Qing Dynasty suggest that the concealment rate often exceeded 20% [
56]. Accordingly, this paper applies a 20% concealment rate to correct the underreported cultivated land area in the Linzhi region, resulting in a revised estimate of 5694.95 hectares of cultivated land for Linzhi in 1830. Since the Inventory of the Year of the Iron Tiger did not cover the Changdu region, the arable land area for this area was estimated using population data from the Qing Dynasty period and assuming continuity of the agricultural management model established since the Yuan Dynasty. Applying the same per capita arable land standard of 2 mu adopted for the Yuan Dynasty yields an estimated arable land area of 18,676 hectares for Changdu during the Qing period. Therefore, by combining the reconstructed cultivated land areas of the Linzhi and Changdu regions, the total arable land area in southeastern Tibet during the Qing Dynasty is estimated to have been 24,371 hectares.
3.3. Gridding Reconstruction Model
Historical farmland reconstruction commonly utilizes documented data on population, taxation, seed usage, and registered land area, integrated with spatial allocation models to simulate the distribution of cultivated land across suitable areas. These results are often expressed in terms of grid-based cultivation rates to represent land reclamation intensity. The present study adopts the grid-based model developed by Luo et al. [
25] for the river valleys of the Qinghai–Tibet Plateau. The methodology involves first reconstructing theoretically cultivable zones, followed by combining historical records of settlement locations and population size to estimate the distribution of farmland near inhabited areas. Given the topographical constraints of deeply incised rivers and steep gorges in the region, arable land is typically limited to narrow, fragmented patches. To enhance the accuracy and reliability of the spatial reconstruction, a grid resolution of 500 m × 500 m was employed in this study.
The spatial distribution of arable land in southeastern Tibet is predominantly constrained by topographic and edaphic factors, including elevation, slope, and soil type [
57]. The upper elevational limit of cultivation is a key determining factor, leading to a distinct distribution pattern compared to plains and low-altitude mountainous areas [
58,
59]. Elevation is a key factor constraining the distribution of cultivated land in southeastern Tibet, significantly influencing its spatial pattern. According to regional chronicles and field survey data [
44,
46], the upper elevation limit for cultivated land in the study area is approximately 4200 m. This value is therefore adopted as the maximum altitude threshold for cultivated land distribution in this research. Slope gradient is another critical factor affecting soil and water conservation as well as land reclamation potential. Statistical analysis of modern cultivated land data reveals that over 95% of the region’s farmland is located on slopes with gradients below 25°. Consequently, a maximum slope limit of 25° is applied for arable land distribution in this study. Furthermore, soil provides the essential loose surface layer required for plant growth and constitutes a fundamental condition for crop cultivation and development. Variations in soil types lead to differences in soil fertility and tillage intensity, thereby influencing land use conversion. The Changdu region is characterized mainly by alpine meadow soils and brown soils, which generally have relatively shallow profiles, whereas the Linzhi region is dominated by brown earths and dark brown earths with deeper soil layers. In this study, land suitable for cultivation is assigned a value of 1, while unsuitable land—such as permafrost areas—is assigned a value of 0. Therefore, elevation, slope, and soil type were selected as limiting factors in this study. Each factor was converted into a binary constraint layer to first estimate the potential arable area before excluding regions unsuitable for farming. The calculation is expressed as follows:
In the given formula, i denotes the grid cell index. The terms NE(i) and NS(i) correspond to the normalized elevation and normalized slope of grid cell i, respectively, each constrained to the interval [0, 1]. The variables E(i) and S(i) indicate the actual elevation and slope values of grid cell i. Furthermore, max(E(i)) and max(S(i)) refer to the maximum elevation and maximum slope values observed across the entire grid domain.
Non-restrictive factors primarily modulate the proportional distribution of arable land. In highland river valley regions, slope aspect serves as an indicator of spatial variations in hydrothermal conditions, significantly affecting the allocation of cultivable area. Aspect values were categorized and weighted as follows: 90–180° (assigned a value of 4), 0–90° (value of 3), 180–270° (value of 2), and 270–360° (value of 1). Soil organic matter content, a critical indicator of soil fertility, exerts a direct influence on both crop distribution and productivity. Thermal conditions represent another key constraint on cultivation, with ≥0 °C being the minimum temperature threshold required for seed germination. Settlements, which function as fixed points of population distribution, are typically situated within suitable agricultural areas. Consequently, farmland tends to concentrate around these settlements, with cultivation intensity diminishing as distance increases. This study utilized reconstructed historical settlement locations in southeastern Tibet as centers to generate multi-ring buffers at distances of 500 m, 1000 m, and >1000 m. These zones were assigned values of 3, 2, and 1, respectively, reflecting the principle that farmland allocation decreases with increasing distance from settlement centers. Organic matter content and climatic production potential were standardized using the following formula:
In the formula,
NO(
i) and
NC(
i) denote the standardized soil organic matter content and climatic production potential for grid cell
i, respectively, with both indices scaled to the range [0, 1]. The terms
O(
i) and
C(
i) refer to the original values of soil organic matter content and climatic production potential in grid cell
i, while
max(
O(
i)) and
max(
C(
i)) indicate the maximum observed values of these variables across the entire study area. All standardized influencing factors were subsequently integrated into a grid-based spatial allocation model, expressed as follows:
In the formula,
M (
i,
vn) indicates the cultivation suitability score of grid cell i in year
vn;
Q represents the total number of limiting factors considered;
Rk refers to the constraint value of the
k-th limiting factor; and
P indicates the number of non-limiting factors. The weight of the
y-th non-limiting factor is denoted by
βγ, with weight values referenced from Luo et al. [
25];
Aγ represents the effect value of the
y-th non-restrictive factor. Since arable land in the study area is predominantly located in river valleys characterized by a significant presence of water bodies, the variable
Ø is introduced in the 500 m × 500 m grid-based reconstruction to account for the percentage of land area within each grid cell, excluding water surfaces.
The proportion of arable land area, total arable land area, and cultivation rate for each grid cell
i within the study area were calculated according to the following methods:
In the formulation, Z (i, vn) and C (i, vn) denote the proportion of arable land area (%) and the total arable land area (km2) within grid cell i located in county Wj for the year vn, respectively. The term E(Wj, vn) represents the total arable land area (km2) across the entire county Wj in year vn, while areai corresponds to the land area (km2) of grid cell i. During the reconstruction of historical arable land patterns, situations may arise where the grid-level cultivation rate, FR(i, vn), exceeds the theoretically valid range of [0, 1] due to an over-allocation of cultivated area. In such cases, a correction procedure is applied: grid cells with values exceeding 1 are truncated to 1, and the surplus cultivated area is systematically reallocated to grids whose cultivation rates remain below 1, following the iterative adjustment process defined by Formulas (6)–(8). This redistribution continues iteratively until all grid-cell cultivation rates are constrained to values less than or equal to 1.
4. Results
4.1. Reconstruction of Settlement Patterns
We compiled and spatially processed historical data for various settlement types—including post stations, residential areas, temples, cemeteries, and watchtowers—dating from the Tubo, Yuan, Ming, and Qing dynasties in southeastern Tibet (
Figure 2). Owing to the profound socio-political influence of Tibetan Buddhism and the region’s theocratic system, temples were particularly numerous and constituted a dominant category within the historical settlement record.
The evolution of settlement types reveals distinct temporal patterns. During the Tubo Dynasty, 88 settlement sites were identified in southeastern Tibet, mostly consisting of burial sites constituting the dominant category. These cemeteries not only reflect burial practices but also provide direct archaeological evidence for the presence of a settled agricultural population in the region during this period, indicating a sedentary lifestyle and the long-term use of specific land resources. By the Yuan Dynasty, the number of sites increased to 90, with temples emerging as the predominant type, accompanied by the initial appearance of postal stations and estates. This shift related to 13th-century political reforms that established a dual governance system integrating monastic and aristocratic authority, along with the Yuan court’s expansion of the postal station network, which revitalized transportation infrastructure. This improvement enhanced both intra-regional connectivity and integration with the Central Plains. During The Ming Dynasty, the number of sites increased to 130. Although monastic sites remained predominant, there was a significant rise in postal stations and manors, reflecting enhancements in both frontier transportation systems and agricultural productivity. During the Qing Dynasty, the number of sites remained elevated at 149. While temples maintained a significant presence, other settlement types—including manors, Zong, and postal stations—underwent concurrent expansion, signifying a diversification of socio-economic structures. This complexity in settlement patterns resulted from the interaction of religious institutions, administrative mechanisms, economic modalities, and military-logistical networks.
The spatial distribution of settlements is predominantly constrained by the region’s high topographic fragmentation and pronounced vertical zonation, exhibiting strong altitudinal dependence. Settlement frequency decreases stepwise with increasing elevation. The majority (64.34%) are concentrated within the mid-altitude zone (3000 to 4000 m), whereas the low-altitude zone (below 2000 m) contains a mere 2.9%. Above 4000 m, settlement occurrence declines precipitously, comprising less than 12% of the total. Regarding terrain preference, settlements are overwhelmingly located on gentle slopes proximal to water sources. Specifically, 65.7% occupy areas with slopes less than 10 degrees, in stark contrast to the 15.16% found on steep slopes greater than 15 degrees. River valleys, characterized by lower elevations, gentle slopes, fertile soils, and accessible water for irrigation, form the core zones for both dense population and agricultural activity. In contrast, high-mountain canyon areas, impeded by extreme terrain, exhibit a scattered and isolated distribution of small, point-like settlements. Development is largely confined to river terraces, which support limited, clustered settlement patterns.
Significant spatial variation is evident among the various settlement types. Post stations exhibit the highest average elevation (3809 m), a pattern attributable to their dependence on major transportation corridors. To maintain the continuity of long-distance routes and control key junctions, these stations were typically situated at high mountain passes linking river valleys or on other elevated landforms, rather than in low-lying valley bottoms. This distribution reflects a reliance on macro-scale transport networks over micro-scale topographic convenience. Furthermore, it suggests that earlier post stations located at lower altitudes may have been more susceptible to destruction or burial resulting from intensive human activity in later periods—such as modern urban development—thereby introducing a potential altitude-related bias in the surviving archaeological record. Temples are characterized by both high elevation and notably steep slopes (average slope of 9°), reflecting the dual imperatives of religious symbolism and defensive needs in their siting. In contrast, manors occupy the lowest average elevation (2807 m) and the gentlest slopes, underscoring the prioritization of favorable topographical conditions for agricultural production (
Table 2). In terms of topographic preference, settlements are predominantly located on flat, resource-rich alluvial plains at mid-to-high elevations. Areas of moderate and high relief constitute the secondary concentration, whereas extremely rugged mountainous regions, with their steep terrain and inhospitable environment, are largely devoid of settlements. Those situated in gentle slope zones within mountainous areas display a fragmented, patchy distribution. In the highest altitude zones, settlements are exceedingly scarce and exist only in isolated pockets due to extreme environmental constraints. Conversely, settlements within mid-to-low elevation river valleys adopt a distinct linear pattern, extending along drainage systems and forming core areas of population aggregation.
4.2. Reconstruction of Historical Cropland Distribution
Based on the reconstruction framework outlined above, this study reconstructed the total arable land area in southeastern Tibet for the Tubo, Yuan, Ming, and Qing dynasties. These totals were then spatialized and allocated to 500 m × 500 m grid cells, resulting in the generation of spatially explicit land use patterns for each respective period (
Figure 3).
Regional analysis indicates that arable land is predominantly distributed across river valley plains, alluvial fans, and terraces, occurring at elevations between 2500 and 4000 m with slopes of 0–10 degrees. These areas, which include the valleys of the Yarlung Tsangpo, Nu, and Lancang rivers and their tributaries, possess gentle topography, deep soils, and sufficient water resources, forming the core agricultural zones of the region. Within them, cropland exhibits linear or patchy patterns that extend along the fluvial networks. Beyond these favorable zones, some cultivation occurs in localized micro-topographic environments such as slopes, hillsides, and intermontane basins. Nevertheless, limited by natural constraints, farmland in these areas is generally scattered and fragmented, often appearing as narrow strips—a typical spatial manifestation of agricultural systems in high-mountain gorge landscapes.
An analysis of the spatiotemporal distribution of farmland reveals significant changes in southeastern Tibet. During the Tubo Dynasty, cultivation was concentrated mainly in the valleys of the Yarlung Tsangpo, Jinsha, Nu, and Lancang Rivers. This development was supported by the warmer climate of the period and the social stability afforded by strong centralized governance. In the Yuan Dynasty, farmland remained concentrated in these major river valleys but expanded further into tributary valleys. This expansion was driven by the unification of Tibet under Yuan rule, which ended prolonged regional fragmentation and created a stable environment that supported agricultural development. Additionally, improvements in irrigation infrastructure further supported the extension of cultivated land. During the Ming and Qing Dynasties, however, the total farmland area decreased. This shift was primarily driven by two factors. First, the period overlapped with the Little Ice Age, a significant climatic cooling phenomenon that shortened the growing season in high-altitude and high-latitude regions. This led to a retreat of the cultivation margin and the abandonment of some formerly farmed lands due to inadequate thermal conditions and increased frost risk. Second, the rapid expansion of monastic economies drew a large labor force toward religious activities, further undermining the human capacity for agricultural production. Under the combined pressure of these dual factors, the total area of cultivated land gradually contracted during this time. Although the Little Ice Age generally suppressed agricultural productivity in mid- to high-latitude regions, the colder climate also conferred certain relative advantages in some areas. For example, the drier and colder environment reduced the incidence of pests and diseases in certain regions. At the same time, in some river valleys and low-lying areas, lower temperatures actually promoted grassland productivity or facilitated the cultivation of certain cold-tolerant crops [
60,
61]. Furthermore, climatic changes during this period had significant impacts on agriculture in other parts of the world. In China, for instance, climate fluctuations during the Ming and Qing dynasties triggered variability in agricultural and pastoral yields and prompted adjustments in farming systems, illustrating the complex responses and adaptation mechanisms of agricultural societies under climate-induced stress [
62,
63].
Historically, the level of land cultivation intensity in southeastern Tibet was generally low, a phenomenon attributable to the region’s harsh natural environment and underdeveloped agricultural technologies and tools. These constraints collectively resulted in persistently low overall cultivation rates throughout the studied periods. During the Tubo period, the overall cultivation rate was 0.12%, with cropland grids comprising 0.42% of the total grids. The average cultivation rate within these cropland grids was 29.49%, with a maximum value of 73.32%. Low-cultivation zones (cultivation rate less than 10%) accounted for 41.05% of the total cropland grids, medium-cultivation zones (cultivation rate between 10% and 30%) represented 10.37%, and high-cultivation zones (cultivation rate greater than 30%) constituted 48.58%. In the Yuan Dynasty, the overall cultivation rate slightly increased to 0.13%, and cropland grids accounted for 0.46% of all grids. The average grid-based cultivation rate was 28.42%, peaking at 76.70%. The proportions of low, medium, and high cultivation zones were 39.20%, 11.03%, and 49.77%, respectively. By the Ming Dynasty, the overall cultivation rate declined to 0.11%, with cropland grids making up 0.29% of total grids. The average cultivation rate per cropland grid rose to 38.10%, and the maximum reached 88.23%. The share of low-, medium-, and high-intensity cultivation zones was 20.02%, 16.48%, and 63.51%, respectively. During the Qing Dynasty, the overall cultivation rate remained at 0.11%, and cropland grids again constituted 0.29% of all grids. The average grid cultivation rate was 37.57%, with a peak of 89.51%. Low, medium, and high-cultivation zones accounted for 19.15%, 15.03%, and 65.82% of the total cropland grids, respectively.
5. Discussion
Existing research has predominantly focused on the major agricultural regions of the Qinghai–Tibet Plateau, while southeastern Tibet—an area characterized by scarce historical documentation and highly complex topography—has received comparatively little scholarly attention. Moreover, most previous reconstructions of historical cultivated land have concentrated on the past century, with a limited temporal depth that fails to capture long-term dynamics. To address these gaps, this study extends the temporal scope to cover the millennium from the Tubo Kingdom to the Qing Dynasty, systematically reconstructing the spatiotemporal patterns of cultivated land in southeastern Tibet over this long-term sequence. This work fills a critical void in the historical land use research of this region. On this basis, we constructed spatial distribution databases of settlements across four historical periods and employed correlation analysis to quantitatively reveal the dynamic influences of human factors on cultivated land distribution. This approach significantly improves the alignment between reconstruction outcomes and actual historical conditions. Our findings not only contribute to the supplementation and optimization of global cultivated land datasets (e.g., HYDE) by enhancing their regional applicability and spatial resolution, but also provide an important case study for understanding universal patterns of human adaptation and land use in global high-altitude mountain regions. As a transitional zone between the Qinghai–Tibet Plateau and the Hengduan Mountains, southeastern Tibet exhibits distinctive characteristics in its historical farmland distribution. While sharing certain commonalities with other high-mountain regions—such as elevation-dependent zonation, landscape fragmentation, and spatial coupling between settlements and farmland—it also demonstrates unique local features. These include the profound influence of monastic economies on land development practices. These findings not only enrich empirical case studies on human–land interactions in high-altitude regions but also offer important insights into the driving mechanisms of cultivated land distribution under extreme topographic conditions.
This study reconstructed the historical distribution of cultivated land in southeastern Tibet primarily by extrapolating from historical population data in conjunction with modern population structure ratios, owing to the severe scarcity of direct historical sources. While this approach is feasible under existing data constraints, it possesses inherent limitations. Moreover, different methodologies were applied to estimate cultivated land in the Linzhi and Changdu regions during the Qing Dynasty: Linzhi utilized direct conversion based on the Inventory of the Year of the Iron Tiger, whereas Changdu, which lacks systematic historical archives, relied on a per capita cultivated land indicator (2 mu/person) for estimation. This methodological divergence stems from variations in the administrative precision of the Qing Dynasty across southeastern Tibet and differences in the preservation integrity of historical records. We respected the heterogeneity of these data sources and avoided imposing a uniform methodology. Although different approaches were applied, the per capita arable land metric is widely adopted in historical reconstructions across the Qinghai–Tibet Plateau and demonstrates considerable temporal stability. Furthermore, both regions exhibit a consistent trend of agricultural land reduction during the Ming and Qing dynasties, with macro-scale evidence supports the reliability of this study’s conclusions. Nonetheless, this reconstruction method inevitably introduces a “homogenization” of spatiotemporal heterogeneity in historical cultivated land. By relying on relatively stable population parameters and modern proportional relationships, the model smooths out the impacts of social institutions, economic structures, and climatic fluctuations on agricultural practices over time. Future research should seek to systematically validate and calibrate population projections through micro-level case studies incorporating archaeological evidence, paleoenvironmental proxies, and Tibetan historical documents. Particular attention should be paid to the phased influences of major historical events and climatic transitions on cultivated land use, thereby improving the dynamic and spatiotemporal accuracy of reconstruction outcomes.
A significant methodological challenge arises from the substantial discordance between historical and modern administrative boundaries, which complicates the spatial reconstruction of historical cropland. For instance, the spatial extent of the ‘Five Ru’ of the Tubo dynasty exhibits only partial overlap with contemporary prefectural and county borders, introducing potential spatial mismatches in cropland allocation. Compounding this issue, southeastern Tibet’s historical status as a peripherally administered territory, marked by frequent political transitions, has resulted in complex and often poorly documented jurisdictional shifts, further obscuring historical territorial extents. Although our study employed a high-resolution (500 m × 500 m) grid-based allocation model, the inherent uncertainty in historical administrative delineations remains a potential source of spatial error in our results. Future work should therefore integrate critical analysis of historical cartography and textual sources to achieve more precise period-specific boundary delineations, which would significantly enhance the spatial fidelity of land-use reconstructions.
In reconstructing the historical distribution of arable land in southeastern Tibet, modern environmental constraints—such as an upper altitude limit of 4200 m and a maximum slope of 25°—were applied as suitability parameters. This approach was adopted primarily due to the extreme scarcity of high-resolution paleoenvironmental data from historical periods, which precluded direct support for grid-based model reconstruction. Moreover, in relatively stable river valley regions, the relationship between modern cultivated land distribution and environmental factors such as topography and soil may partially reflect historical suitability patterns. However, this method has inherent limitations. Significant fluctuations occurred in historical climate conditions; for example, the Ming–Qing Little Ice Age likely resulted in lower temperatures and shorter growing seasons, potentially depressing the actual upper cultivation limit below modern reference values. On the other hand, although topography and soil properties have remained relatively stable over millennia, evolving agricultural techniques, crop varieties, and social organizational structures—such as the constraints of primitive farming methods during the Tubo dynasty on steep slopes and the expansion of terrace farming technology in the Qing dynasty—could have altered the practical usability of land resources. This study integrated historical documents and field survey data when establishing environmental constraints to improve the plausibility of the reconstruction. Future research should incorporate higher-resolution paleoclimate reconstructions, archaeological findings, and ethnographic data to refine environmental parameters and land-use strategies across different periods, thereby enhancing the spatiotemporal accuracy of cultivated land reconstruction. Additionally, the complex geomorphological processes in southeastern Tibet—including river erosion, slope deposition, landslides, and debris flows—may significantly affect the preservation and detectability of historical settlements and cultivated land remnants. Alluvial deposits in valleys may bury early cultural layers, while soil erosion on steep slopes can remove traces past cultivation. These natural processes not only complicate archaeological exploration but may also lead to an underestimation of historical arable land extent. Future studies should integrate geomorphological archaeology and sedimentological analyses to more accurately assess the historical patterns of land use.
6. Conclusions
This study integrated multi-source historical data with a grid-based reconstruction model to spatially reconstruct the distribution of cultivated land across four historical periods—the Tubo, Yuan, Ming, and Qing dynasties—in southeastern Tibet. From the Tubo to the Qing Dynasty, the number of settlements increased steadily, accompanied by gradual transformation in their typological structure, reflecting historical processes such as the establishment of Tibet’s theocratic system, the expansion of transportation networks, and the intensification of the agricultural economy. Spatially, settlements consistently clustered in river valleys, highlighting human adaptation to favorable hydrothermal conditions and agricultural potential in these areas. The overall cultivation intensity remained low throughout the four dynasties, with reclamation rates of 0.12%, 0.13%, 0.11%, and 0.11% during the Tubo, Yuan, Ming, and Qing periods, respectively. Cultivated land was primarily distributed in the river valleys of the main and tributary streams of the Yarlung Tsangpo, Jinsha, Lancang, and Nu rivers, within elevation zones of 2500–4000 m and slope gradients of 0–10 degrees—areas characterized by relatively favorable hydrological and thermal conditions for agricultural activity.
This study not only enhances understanding of the human–land interaction mechanisms in southeastern Tibet, revealing the topographical constraints and historical dynamics of cultivated land distribution, but also provides methodological and empirical support for historical land use reconstruction in high-altitude regions. The scientific contributions extend beyond a regional case study. The reconstruction framework and empirical results offer valuable references for global research on historical land use in high-altitude areas, particularly in filling gaps in high-resolution historical cropland data, validating theories of human adaptation in mountainous regions, and elucidating complex feedback mechanisms within human–environment systems. Future research should further incorporate multidisciplinary evidence to improve reconstruction accuracy and explore in greater depth the dynamic mechanisms of human–land interactions throughout history.