Introduction

Climate is considered a key component in human evolution and history1,2,3,4,5,6,7,8. Environmental conditions along the ancient Silk Road system contributed to the economic and cultural development in Eurasia since 1 CE6,9,10,11,12,13,14,15,16,17,18,19,20,21. As the gateway from the Mongolian Plateau to Western Asia and Europe (Fig. 1), the Tien Shan Mountains and its surrounding oases played a critical role in nomadic migration into the inner Eurasian steppes22,23. The Tien Shan Mountains not only supplied water and pastures for the steppe empires but also enabled trade caravans from East Asia to cross the vast Gobi Desert along the Silk Road system.

Fig. 1: Map of the study area.
figure 1

a Route of Mongol conquests (arrows), capital cities of the Mongolian Empire (Karakorum) and Timurid Empire (Samarkand), and silk road. The river basins are also shown (Rectangle). b Map showing the locations of the four sampling sites and hydrological stations with main rivers in Central Asia. The Pre-Balkhash plague focus area of southern Kazakhstan is also shown (Rectangle).

Full size image

The hydroclimatic history of the Silk Road system over past centuries has been studied using tree-ring chronologies at high temporal resolution6,8,24,25,26,27. Tree ring-based hydroclimatic reconstructions were also used to provide environmental insights into periods of historical transformation in Inner Asia, such as the invasion of the Huns in the 4–5th centuries, Turks in the 6–8th centuries and the Mongols in the first half of the 13th century6,13,19,21. However, high-resolution streamflow reconstructions, as an important indicator of hydrological variation are still insufficient in Central Asia for interpreting the relationship between the supply of fresh water and historical societal changes in a long-term perspective. Here, we present an annually resolved streamflow reconstruction from the Tien Shan for the period 1225–2009 CE that covers the rapid expansion of the Mongolian Empire in the early 13th century. Our reconstruction is based on long-lived spruce trees from the different elevation area where tree growth is predominantly controlled by soil moisture availability.

Results and discussion

Streamflow reconstruction from the Tien Shan Mountains

The composite chronology was significantly positively correlated (p < 0.05) with temperature in the prior September and November and current February–March and June, August, September, and it was also significantly positively correlated (p < 0.05) with precipitation in the prior July and November and current July (Fig. 2a). Correlations were stronger with streamflow as well as with Normalized Difference Vegetation Index (NDVI) (Fig. 2a). Overall, correlations were strongest with July–April precipitation (r = 0.64, p < 0.01), August–April composite streamflow (r = 0.77, p < 0.01), November–October temperature (r = 0.49, p < 0.01) and August–October NDVI (r = 0.72, p < 0.01). Meanwhile, high correlation (r = 0.69, p < 0.01) between August–July composite streamflow and the composite chronology was also found. Warm temperatures with relatively abundant precipitation can prolong the growth period and lead to an increase in radial growth of spruce trees in high altitude area. We found that the seasonal response to the NDVI and streamflow tended to occur later than the responses to precipitation, suggesting that hydroclimate variability is the root cause of streamflow and vegetation change.

Fig. 2: Response of tree-ring width to to climate, NDVI and streamflow and streamflow reconstruction.
figure 2

a Responses of the composite chronology to monthly mean temperatures, precipitation sum, composite streamflow, and NDVI. Correlations are computed from the previous year in July to the current year in September over the common period. The horizontal dashed lines indicate the 95% significance levels. b Actual and estimated total streamflow of the previous August to the current July for 1962–2009. c Comparison between tree-ring index and total July–April precipitation for 1962–2009. d Reconstructed August–July composite streamflow of the Syr Darya, Ili, Chu, and Jing rivers (blue curve) shown together with the two-tailed 95% bootstrap confidence limits (green) and ±1 root mean square error (RMSE) uncertainty estimates (gray) since 1225 CE. Red curve is a 21-year low pass filter. Reconstructed streamflow from 1225 to 1239, 1252 to 1260, 1361 to 1400, 1437 to 1467, 1494 to 1511, 1678 to 1692, 1693 to 1705/1731 to 1756, 1844 to 1905 and 1961 to 1996, corresponding to conquests of Batu Minien, Hulagu Khan’s conquests, rise of Timurid Empire, rise of Esen Choros, fall of Timurid Empire, defeat of Galdan Choros, recovery of Dzungar Khanate, Russian conquest and drying up of the Aral Sea, respectively. Blue and red boxes represent high and low streamflow conditions, respectively. Central horizontal line shows the average of streamflow estimates, and dotted lines indicate one standard deviation.

Full size image

Based on the above analysis, to cover more months and develop a water-year composite streamflow reconstruction, we selected August–July composite streamflow as the target for our reconstruction. The model (Y = 96.051 + 266.905X) has significant skill based on the positive reduction of error (RE) and coefficient of efficiency (CE) statistics (Supplementary Table 3). The first differences of the tree-ring series with instrumental streamflow data were correlated with r = 0.50 (p < 0.01). The composite chronology showed good consistency with the regional streamflow and precipitation variations (Fig. 2b, c). We also found strong positive correlations up to r = 0.72 (p < 0.01; 1981–2009) with the August–October NDVI, and some significant positive correlations were found in the Central and Eastern European plains as well as in Scandinavia. Moreover, significant negative correlations were also found in the Mediterranean Basin, Middle East, and western Asia (Fig. 3a). The PDSI reconstruction of the Mediterranean Basin16 also had a good correlation with NDVI, and the spatial distribution pattern was different from that of our streamflow reconstruction (Fig. 3b and Supplementary Fig. 4).

Fig. 3: Spatial correlations between the dendrohydrological reconstructions and the gridded NDVI dataset62.
figure 3

a Correlation patterns of the streamflow reconstruction with August-October NDVI, a satellite-derived measure of grassland productivity, over their period of overlap (1981–2009). Insignificant correlations (P> 0.05) are masked out. The mouse icons indicate that the rodent population density increased under the good vegetation and water conditions in Central Asia. The inset figure shows the comparison between regional NDVI (averaged over 39–45°N, 77.5–86.5°E) and reconstructed streamflow (this study). b Correlation patterns of the PDSI reconstruction of the Mediterranean Basin with July–September NDVI. The mouse icons and arrows show that poor vegetation conditions in Central Asian steppe (blue shape) forced the rodent to migrate to the Mediterranean basin and Western Asia (red circle) with the good vegetation and water conditions. The inset figure shows the comparison between regional NDVI (averaged over 30–45°N, −5–37°E) and reconstructed PDSI16. The numbers in parentheses in the maps indicating the recorded time of plague outbreaks per city.

Full size image

The average annual (August–July) reconstructed streamflow (1225–2009 CE) was 358.9 × 108 m3, with a standard deviation of 51.1 × 108 m3. The streamflow reconstruction is the longest streamflow reconstruction current in Central Asia, and covers the two Mongol expansions in 1235–1242 CE and 1252–1260 CE9,14,28,29, the onset and establishment of the Black Death in 1346–1348 CE (Supplementary Table 2)15, the rise of the Timurid Empire in 1361–1400 CE, the rise and fall of the Zunghar Khanate in 1678–1756 CE29,30, and the recent drying up of the Aral Sea from the 1960s to present31 (Table 1).

Table 1 Some high and low periods in the streamflow reconstruction of Tien Shan Mountains and comparisons with important historical events.
Full size table

Streamflow levels were mainly above average from 1225 to 1260 CE during the westward conquests of the Mongols9, including pluvials from 1225 to 1239 CE during the conquests of Batu Minien, and from 1252 to 1260 CE during Hulagu Khan’s conquest. These pluvial periods were interrupted by a brief period of low streamflow (1240–1251 CE) followed by a multidecadal dry period in the late 13th century.

Another pluvial occurred in the second half of the 14th century during the rise of the Timurid Empire (1361–1400 CE), when streamflow was above average from 1364 to 1371, 1381 to 1387, and 1394 to 1400 CE. The decline of the Timurid Empire in 1507 CE, on the other hand, fell within an 18-year period of low streamflow conditions (1494–1511 CE).

During a pluvial from 1437 to 1467 CE, the king of the Oirat Mongols, Esen Choros, reunited the Mongolian tribes and captured the Zhengtong emperor of the Ming Empire after the Battle of Tumu in 1449 CE9. This series of events allowed the Oirat Mongols to control the Tien Shan from the mid-15th century until the attack of the Qing Empire (1755–1757 CE)30 and found the Dzungar Khanate in 1634 CE. The army of Galdan Choros, a descendent of Esen Choros, was defeated by the Qing Empire during a relatively low streamflow period in 1678–1692 CE, although the power of the Dzungar Khanate was restored during two subsequent high-streamflow periods in 1693–1705 and 1731–1759 CE. This latter period was characterized by frequent wars with the Eastern Mongols, Kazak Khanate, and Qing Empire. In the warm-wet climatic condition32, epidemics resulted from wars were expanded in the Dzungar Khanate, and 40% of the Dzungar Khanate population died of smallpox33. Due to the prevalence of smallpox and the civil strife in the Dzungar Khanate, the Qing Empire conquered the Dzungar Khanate30, and the subsequent low streamflows from 1758 to 1790 CE coincided with weak hostile forces in Xinjiang.

In the late 19th century (1844–1905 CE), the period during which Russia gradually completed its conquest of Central Asia34, the streamflow was generally above average. The mean streamflow value (332.0 × 108 m3) during the 20th century (1900–1999 CE) was lower than in any other century since 1225 CE (Supplementary Table 4). The low streamflow (−6%) during the period 1961−1996 contributed to the continuous reduction in upstream water supplies and marked the unprecedented drying of the Aral Sea and its surroundings31.

We find significantly high streamflows 15 years prior to the 18 plague events of Europe and the Mediterranean Basin that linked with Asian plague outbreak (Supplementary Table 2). After these abnormal streamflow increases, the long-term downward trends of streamflow or some drought events occurred until the outbreak of the plague in Europe and the Mediterranean Basin, and the drying process may be last for up to a decade, and on the contrary, the Mediterranean Basin and Near East is in the relatively humid state when the plagues occurred (Fig. 4). Results from GCA also support the claim that streamflow Granger-cause plague outbreaks (P < 0.05) (Supplementary Table 5).

Fig. 4: The response of Central Asian streamflow to the plague outbreaks.
figure 4

Results of the superposed epoch analysis (SEA) testing the response of Central Asian streamflow (a) and PDSI of the Mediterranean Basin (b) to the outbreak of the plague events (Supplementary Table 2). The solid lines indicate the 95% significance level. c Comparison between the reconstructed August–July total streamflow (red curve) and PDSI reconstruction of the Mediterranean Basin (blue curve)16. Thick curves are trend lines of the reconstructions along with the circulars representing the outbreak years of the plague in the Mediterranean Basin and Europe.

Full size image

Impact of river streamflow on the rise and fall of Central Asian empires

The Mongol Empire, the Timurid Empire, the Oirat Mongols, and the Dzungar Khanate had some important influences on the history of Eurasia, such as the Mongol conquests9,28,29, the Timur conquests29,35, the Turko-Mongol fusion35, the Mongolia unification by Esen Choros, the battle of Tumu36 and the Dzungar–Qing Wars30,37,38. Such outstanding military and political accomplishments likely required a strong material base that relied on grassland and oasis productivity, favorable climatic conditions, and abundant water. Because rainfall in Central Asia is scarce, the rivers originating from Tien Shan glaciers became the most valuable water resource, greatly affecting oasis agriculture and grassland animal husbandry, as well as the power balance among different empires in Central Asia39,40. In contrast to oasis settlers, ancient nomadic tribes mainly depended on grazing, hunting, and gathering for food41. Moreover, the survival and reproduction of livestock and wild animals are closely linked to the net primary productivity (NPP) of the steppe6, which in turn depends on rainfall and water42. As a good indicator of NPP, the significant positive correlation of NDVI data with our streamflow reconstruction for the Central Asian steppe supports this connection. The high streamflow periods of 1225–1239, 1252–1260, 1361–1400, 1437–1467, 1693–1705, and 1731–1759 CE were likely associated with humid conditions and relatively high NDVI values (+4.1%) (Fig. 2d); therefore, the grassland provided more food for the armies of Central Asian empires (Table 1). During that time, more water drained from the Tien Shan into the arid plains, lakes, and wetlands, including the Aral Sea, Balkhash lake, and Ebinur lake. This increased water might have led to rising NPP in the middle and lower reaches of nearby rivers, which was conducive to the armies of Central Asian empires advancing westward towards Europe and Western Asia. Adequate water resources also promoted the restoration and development of oasis agriculture, and the armies had access to pasture and food supplies from the conquered oases. The most typical examples are Samarkand and Ili. As the capital of the Timurid Empire, oasis city Samarkand grew to have a population of about 150,000 during the second half of the 14th century, and a large number of artists, craftsmen, and scholars gathered from all over the country. By forced Uighurs to move to Ili and other parts of Dzungaria to engage in agricultural activities on farms, the Dzungar Khanate established a progressive oasis agriculture in Ili and its surrounding regions38,43. The transfer of the political center of the Dzungar Khanate to Ili supports the hypothesis of abundant wealth from highly productive farms and steppes supporting the concentrations of nomadic and semi-nomadic tribes. Access to water enabled the Timurid Empire and Dzungar Khanate leadership to establish military and political power. Of particular interest is the similar response of NDVI of Central Asia and the Central and Eastern European plains to our streamflow reconstruction (Fig. 3a), which may be linked to the middle-latitude westerly circulations, with contributions from middle and high latitudes44,45,46,47. Wet conditions in the Central and Eastern European plains also supplied more fodder and clean water for the Mongol invasions in the early-mid-13th century.

Prolonged bad climate conditions and the resulting poor economies also drove the outward expansion of Central Asian invaders, leading to events such as the fall of the Timurid Empire and the rise of the Mughal Empire13,19,48. Although the Timurid Empire was destroyed during the low-streamflow phase of 1494–1511 CE with a relatively low NDVI (–5%), the influence of the empire did not completely disappear. The remnants of the Timurid Empire entered the Indian subcontinent with the relatively good climate and good economic conditions in 1510s48,49, and established the Mughal Empire in India.

Although some scientists consider climate change to have been an important driver of the rise of the Mongol Empire and the migration of nomadic tribes in Eurasia6,14, the role of river streamflow in arid to semiarid Central Asia during the rise of the empire has not been addressed. Our streamflow reconstruction suggests that the rise of the Central Asian empires during the last 750+ years was related to favorable water conditions, which were conducive to increased pasture yields and powerful nomadic tribes. Perhaps even more importantly, sufficient water resources met the need for oasis irrigation and provided resources for the rise of the empire and the foundation of a persistent civilization. The large-scale expansions and conquests based on sufficient water resources not only enabled the empires to obtain greater territories and resources but also eventually affected Eurasian civilizations. At the same time, it needs to be emphasized that the rise and fall of the Central Asian empires did not depend entirely on the hydroclimate-driven NDVI change, which simply provided a substantial material base. Although hydroclimate changes are the far-reaching process affecting human society, the rise and fall of empires are also influenced by military power, the political system, the ability of leaders, and other factors, especially in multi-ethnic and multicultural Central Asia9,29,35.

Role of river streamflow in the spread of plague

The correlation between the streamflow estimates and NDVI values indicates that the grassland productivity in mountainous terrain benefits from abundant water, providing favourable conditions for the pastures and livestock of nomadic tribes. However, increased precipitation that leads to high streamflow also supports rodents in grasslands and mountains, thus leading to an increased risk of plague10,50,51,52,53. We detected significantly high streamflows 15 years prior to the 18 plague events of Europe and the Mediterranean Basin that were linked with previous plague outbreaks in Asia (Fig. 4a and Supplementary Table 2). This phenomenon resembles other finding in surrounding area15, and suggests the influence of solar activity and its related atmospheric circulations on the hydroclimate variation54,55 (Supplementary Fig. 5). The high streamflow period of 1332–1337 CE preceded the onset and establishment of Europe’s Black Death from 1346 to 1348 CE by approximately a decade. More seriously, with the low streamflows, drought events, and decrease in the NDVI (−2.9%) before the outbreak of the plague, significant decreases in the amount of food available to rodents were observed (Fig. 3). This change would have led to changes in the population structure of rodents in Central Asia and caused rodent migration51. Meanwhile, some studies have shown that the Mediterranean Basin, the Near East, and Western Asia were relatively humid during the Black Death periods12,16, which may have provided rodents with sufficient food thereby increasing the risk of plague outbreak. The NDVI and precipitation patterns also support such a connection (Fig. 3 and Supplementary Fig. 4). As shown in Figs. 3 and 4a, the abnormal decrease in the NDVI persisted for 5–10 more years before the outbreak of plague not only occurred in the Tien Shan and surrounding grasslands but also may have existed in the Kipchak steppe stretching from the northern shores of the Black Sea as far east as the Aral Sea and Caspian Sea. After the unification of Inner Asia by the Mongol Empire, the trade routes of the northern steppe began to prosper. With the increase in logistics and human movement, the risk of plague transmission in the Mediterranean Basin, the Middle East, and Western Asia was further increased. The most typical example was observed in 1346 CE. When the Golden Horde army attacked the trading city Caffa at the Black Sea coast56, it also brought plague from the natural foci north and east of the Caspian Sea and Black Sea in the upper reaches of the rivers (with low NDVI) to the Mediterranean Basin and Near East (with relatively high NDVI) (Fig. 4 and Supplementary Fig. 4). Therefore, the spread of plague may not have been simply caused by the hydroclimate-driven NDVI change and its associated rodent migration but also by anthropogenic activities, such as trade activities or military activities. Nevertheless, large-scale comparisons between tree-ring records still provide evidence of the possible role of hydroclimate variation in the outbreak of plague.

Link between 20th century low streamflows and the Aral Sea crisis

The 785-year streamflow reconstruction presented here provides a long-term perspective for the recent period of rapid economic, social and cultural transformation in Central Asia. During the 20th century (with the unprecedented warming), regional water resources were less abundant than during any other period of the past seven centuries, including the ten decades when they were below the long-term average. Based on the range of streamflows reconstructed from 1901 to 1999, there would be a 70% chance that streamflow would not meet or exceed 358.9 × 108 m3. Thus, in almost seven out of every 10 years the surface water supply would not be sufficient to meet the water right allocations and river flow targets. The persistent low runoff in Tien Shan with climate warming also raises doubts over the sustained water resources supply to meet the rapidly growing water consumption in Central Asia, and this is a vital information that can not be obtained from short instrumental records. Even worse, these conditions also occurred during a period of rapid and widespread agricultural development in the former Soviet Republics of Central Asia. After World War II, the former Soviet Union built reservoirs and canals and developed large regions of Central Asia into granary and cotton production bases, including Uzbekistan, which is still the second-largest cotton exporter in the world57,58. The combination of continuous natural streamflow reduction (−6%) and dramatically increased water irrigation successively reduced river streamflow into many lakes of Central Asia59, and Aral Sea has lost 90% of its volume over the past 50 years and caused Central Asia’s largest ecological disaster31. Thus, although the Aral Sea crisis is mainly due to unreasonable water diversion, our tree-ring-based reconstruction implying that unprecedented 20th century low streamflow has exacerbated the crisis, and this valuable paleo-hydrologic information is in the process of being incorporated into reasonable water resources management policy for Central Asia.

Based on tree-ring data from the Tien Shan, China, and Kyrgyzstan, we presented a streamflow reconstruction for the Tien Shan that revealed Central Asian streamflow variation over the past 785 years. The atypical hydrological conditions of the Tien Shan, such as those occurring during 1361–1400 CE and the 20th century, had far-reaching consequences, including the rise of empires, civilization formation, and ecological changes in Central Asia. The pluvial periods in our 785-year composite streamflow reconstruction associated with the rise of Central Asian steppe empires, and plague spread were marked by sequential years (≥ 4 years) of above/below average streamflow rather than individual years of extremes. Although water resources played an important role in the historical and societal changes of Central Asia, it must be emphasized that water resources represent only one of the factors that influenced these complex historical processes. Due to the warm-wet climate in Central Asia, streamflow has also shown significant recovery since 198760. Due to the accelerated melting of glaciers in the warming 21st century, the rivers originating in the Tien Shan will benefit from additional meltwaters within a relatively short time39. However, in the future, as temperatures continue to rise and glaciers disappear, evaporation will exceed precipitation; as a result, ecological disasters similar to the disappearance of the Aral Sea will likely increase/continue in Central Asia. Therefore, it is necessary for Central Asian countries to establish a rational water distribution mechanism to ensure sustainable development and maintain ecological balance.

Methods

Streamflow and climate data

The study focused on the Tien Shan, China, and Kyrgyzstan, the wettest region in Central Asia. Tien Shan, referred to as Central Asian water tower, are the source of some of Central Asian major rivers, include Syr-Darya and Ili river, and also help to regulate regional climate31,39 (Fig. 1). As the most densely populated area in Central Asia, this region is an important part of the Silk Road in Central Asia and the meeting point of eastern and western civilizations9. Many imperial capitals were built in the study area, include the Timurid Empire (Samarkand) and the Dzungar Khanate (Ili), and developed splendid civilizations. The glacier retreats in Tien Shan with climate warming raises doubts over the sustained fresh water supply to meet the rapidly growing water consumption39. At the same time, the conflict over water resources in the region has become increasingly acute, and the irrational use of water resources of transboundary rivers caused the Aral Sea crisis31,59. Monthly streamflow data are obtained from the five downstream hydrologic hydrometric stations, include Akdjar, Ushjarma, Chapaeva, and Jinghe for the period 1962–2009 (Supplementary Table 1 and Supplementary Fig. 1). The streamflow data were used to produce time series of monthly composite streamflow and correlation analysis was used to evaluate relationships to the tree-ring chronology. We also used Climatic Research Unit (CRU) gridded monthly climate data (averaged over 77.5–86.5°N, 39–45°E, 1962–2009)61 and Normalized Difference Vegetation Index (NDVI, averaged over 77.5–86.5°N, 39–45°E, 1981–2009)62 data to better understand the climatic controls on the vegetation growth.

Tree-ring data and streamflow reconstruction trials

The composite chronology is based on raw measurements of tree-ring widths derived from 396 individual series from 210 drought-sensitive spruce trees that were growing at four sampling sites between 42–45°N and 79–84°E (Supplementary Table 1 and Supplementary Fig. 2). The RBAR (mean correlation between ring-width series) and expressed population signal (EPS) used 50-year windows with a lag of 25 years to evaluate the adequacy of replication in the early years of the chronology63 (Supplementary Fig. 3). We truncated the composite chronology at the year 1225 CE based on the EPS value of 0.85, when sample depth dropped below six series from three trees. A simple linear regression model was used to develop a composite streamflow reconstruction for the Tien Shan, and model skill was verified with standard methods64. In order to indicate the response of the NDVI patterns in Eurasia to the regional reconstructions, we also used the KNMI Climate Explorer65 to calculate spatial correlation maps of our streamflow reconstruction and the PDSI reconstruction for the Mediterranean Basin16 with the NDVI dataset. We investigated hydrology-driven plague outbreak signals in the Tien Shan streamflow reconstruction using a Superposed Epoch Analysis (SEA, https://rdrr.io/cran/dplR/man/sea.html)66 and Granger Causality Analysis (GCA)67. Details on the research area, data, methods, and analyses are provided in the supporting information.