There is no academic consensus on the causes of historic cycles of war and peace. Historians cite government mismanagement and concomitant social turmoil; military scientists and politicians emphasize power imbalances among competing groups or entities; psychologists and biologists relate warfare to innate human aggressiveness; while the Marxists hold that warfare is the unavoidable consequence of class struggle. However, none of these explanations adequately accounts for war–peace cycles from a macro-historical perspective.

Webster (1975), using Darwin’s concept of natural selection to examine a number of prehistoric and early historic societies, concludes that warfare is an ‘adaptive ecological choice’ under circumstances of population growth and resource limitation. Recent studies also support the argument that resource limitation and environmental degradation are both significant factors in generating armed conflicts throughout human history (Westing 1988; Stranks 1997; Suhrke 1997).

The Panel on Climate Variability on Decade-to-Century Scales of the National Research Council (1998) emphasizes that human societies depend on, or are controlled by, climate. This is especially so for historic agrarian societies. Temperature is probably the most important climatic variable influencing human societies, which are particularly vulnerable to long-term temperature changes. Temperature fluctuations directly impact agriculture and horticulture, exacerbate natural disasters, and can adversely affect plant, animal, and human rates of disease. There is a long-standing scholarly tradition that in the historic past, organized armed conflicts and climate change are correlated. Some scholars (e.g., Huntington 1907; Hinsch 1988; Hsu 1998; Atwell 2001; 2002) suggest that in the technologically simple societies of pre-industrial Europe and China cooling temperatures could hugely impact the availability of critical crops and herds. In situations of ecological stress, warfare could become the ultimate means of redistributing shrinking resources. However, this hypothesized relationship has never been substantiated with quantitative data. In this study, we attempt to quantitatively test this hypothesis.

Study Area

In China, there is voluminous archival documentation dating back to 880 bc recording major events that seemed significant to different dynasties. This valuable and relatively comprehensive repository of data is the basis of our study. Eastern China, which includes the regions east of 105°E and within the latitude zones of 25°N and 40°N (Fig. 1), is the historic, political, economic, and cultural center of China. Since the establishment of Qin dynasty (221–206 bc), the first united Chinese empire, nearly 80–90% of the population lives in this area, and its agricultural production supports the majority of the Chinese population. The documentary records for this area are more complete than those of other regions, leading us to make it our primary study area.

Fig. 1
figure 1

Physical regionalization of eastern China.

According to the basic principles of physical regionalization (Ren et al. 1985; Zhao 1986), we divided the study area into two macro regions: (1) the northern area that is characterized by continental humid, semi-humid, and semi-arid temperate climate influenced by both the monsoons and the westerlies (i.e., prevailing winds from the west that occur in both hemispheres between latitudes of about 35° and 60°). Average annual temperatures range from 8 to 14°C and average annual precipitation ranges from 250 to 750 mm. Major agricultural products are wheat and maize, with some pasturage in the northwest. Frost days per annum exceed 100. (2) The southern area where the climate is dominated by the monsoons, with average annual temperatures ranging from 14 to 18°C, and between 10–20 frost days. This region serves as China’s major rice-growing area. In both regions, temperature and precipitation decrease from the southeast to the northwest.

Climate Change and Social Stability

We assume climate affected historic societies by altering agricultural productivity and consequently social stability (see Zhang et al. 2005; 2006). Warm but temperate climate may augment agricultural production, while significant cooling can directly impede agricultural production or even lead to crop failure. Temperature change influences agricultural production by affecting the length of growing seasons, intensity of average summer warmth, and reliability of rainfall, which can adversely affect food production, especially at high and middle latitudes (Bryson and Murray 1977; Galloway 1986). In addition, a long period of cooling will lower the elevation at which crops can be grown, thus decreasing the amount of land available for cultivation and leading to a decline in total output or more intensive cultivation but with lower yields. Lower yields may also result from the biological inability of certain grains to effectively withstand cooler temperatures and associated variability in short-term weather patterns.

According to Gong et al. (1996), agricultural yields in China in ad 1840–1890 (cold phase) were reduced by 10–25% in comparison with ad 1730–1770 (warm phase), despite the fact that the total arable area was at least 10% more than that in the earlier warm period (Liang 1980). Agricultural contraction caused by climatic cooling is also evidenced in the history of rice cultivation in the middle and lower reaches of the Yangtze River (i.e., the southern portion of eastern China). Double cropping of rice started here in the Tang dynasty (ad 618–906) and was further developed in the late fifteenth century during the Ming dynasty (ad 1368–1643). However, between ad 1620 and 1720, the double cropping system collapsed due to cooler temperatures (Yin et al. 2003). Nevertheless, coinciding with the warmer phase that started c. ad 1720 double cropping rice cultivation began to dominate the region again, although by the beginning of the nineteenth century (demarcated by a cold climate) it once more proved unsustainable despite government promotion (Yin et al. 2003). Today, double cropping rice cultivation is again functioning successfully. Nowadays, even in countries with industrialized agriculture where technology is utilized to control as many environmental variables as possible, agricultural production is still limited by climate (Bryson and Murray 1977; Bryson and Ross 1977; Parker and Smith 1978). Even though agricultural production may well increase over time, its upward trend is tempered by oscillations caused by long-term climate change. We believe that such oscillations drove the historic war–peace cycles we examine here.

Chinese history has been shaped by nomadic invasions originating outside the Great Wall, which is also related to climate change. The territory outside the fortification that skirts around the heartland of Manchuria from the northeast, runs parallel with the present Great Wall in the middle, and curves southwestward to separate Kokonor and Tibet from China proper is ecologically fragile (Huang 1988; Lattimore 1988). It is estimated that a 2°C drop of air temperature can shorten the grass growing season there by 40 days. Cooling also changes the vegetative composition of the grasslands and the resulting forage shortage can cause the death of nearly 90% of domestic animals annually (Fang and Liu 1992). It is not unreasonable to suppose that in the past cooling temperatures could lead to economic distress for the nomads in this area, triggering them to migrate. Historic nomadic migrations are in two directions: the eastward movement of nomads in western China and the adjacent former Soviet territory; and the southward movement of the nomads in Manchuria and the Inner Mongol Autonomous Regions of China and in the south of the Mongolian People’s Republic. These nomadic migrations threatened the Han Chinese who lived in the south. Occasional nomad raids to pillage Han farmers could escalate to large-scale invasions and reprisal attacks launched by the defenders (Huang 1988; Lattimore 1988; Zhao and Xie 1988; Fang and Shi 1990; Wang 1996).

Cold periods are often associated with great climatic variability, including extremes of drought, flood, and even occasional extreme summer heat, which further disturb agriculture practices already handicapped by a short growing season (Gribbin 1978), potentially causing severe economic disruption and demographic stress. In the “Little Ice Age” (ad 1500–1900), many areas of the world experienced famine (Ponte 1976; Bryson and Murray 1977; Lamb 1995) and some witnessed large-scale population migrations in search of food (Hsu 1998). Some researchers suggest that such severe climatic disasters were also responsible for civil wars and especially for peasants’ uprisings in Chinese history (Fu and Wang 1982).

Methodology and Data

Warfare Data

The Tabulation of Wars in Ancient China (Editorial Committee of China’s Military History 1985) is a multivolume compendium scrupulously recording the wars that took place in China from 800 bc to ad 1911. It includes an appendix detailing each war, including year of inception, type of participants, location, causes, and in most cases, the numbers of soldiers or combatants, causalities, progress, and results. The 899 wars in eastern China listed as occurring between ad 1000 and 1911 were used as the database for this study. To avoid bias associated with different sources of information, we used only those that included comparable data on inception year, type of participants, and location. Geographically, wars were categorized as either “north” or “south” (please refer to Study Area). We also categorized wars on the basis of types of participants, particularly the leaders of the two sides in the armed conflicts, as either “rebellion” or “others” (state and tribal wars). Further warfare categorization is impossible due to data limitations.

Temperature Data

Briffa and Osborn (2002) have analyzed the five most influential Northern Hemispheric climate series of the last millennium, including data on China, in order to resolve the differences among the results of various independent studies (Fig. 2a). Despite the diverse data sources, all five high-resolution climate series register a close matching of warm and cold phases, suggesting a high degree of accuracy with reference to both temperature and timing, and thus provide us with a reliable basis to investigate the relationships between climate changes and warfare in eastern China. We adopted the records from ad 1000 to 1980 as the standard climate variations in this study. These records were reconstructed by using multi-proxy data, including tree-ring, coral, ice-core, borehole, and historic document studies. The data were recalibrated by Briffa and Osborn (2002) with linear regression against the ad 1881–1960 mean annual temperature observations averaged over the land area north of 20°N, and the results were smoothed with a 50-year low-pass filter. We then averaged these recalibrated records to quantitatively define the boundaries of the cold and warm phases. A cold or warm phase would be determined if the average temperature change (bold black line, Fig. 2a) has an amplitude exceeding 0.14°C, in order to get an equal aggregate duration of cold and warm periods. We identified six major cycles of ‘warm’ and ‘cold’ phases from ad 1000–1911 based on the average reconstruction. These phases were also reflected in many other climatic reconstructions of China and Northern Hemisphere (e.g., Zhang 1981; Seabury and Codevilla 1989; Wang 1990; Ge et al. 2002; Mann et al. 2003). The boundaries between warm and cold phases were delineated at the mean temperature point between the minimum and maximum values of two contiguous phases on the average reconstruction. The aggregate duration of the cold phases was 459 years and that of the warm phases was 453 years. The cold phases spanned ad 1110–1152, 1194–1302, 1334–1359, 1448–1487, 1583–1717 and 1806–1911, while the warm phases spanned ad 1000–1109, 1153–1193, 1303–1333, 1360–1447, 1488–1582 and 1718–1805 (Fig. 2a). The winter-half-year temperature reconstruction of eastern China (30-year resolution) was also included (Fig. 2b). This reconstruction is based on historical documentation, and then recalibrated with mathematical equations and models (Ge et al. 2002). While it also approximately agrees with Fig. 2a, it is used only for visual comparison because its time-resolution did not reach the annual scale that we achieved in this research.

Fig. 2
figure 2

Climate changes and war frequencies in eastern China during the last millennium: a Normalized temperature change records of the last millennium over the land areas north of 20°N of the Northern Hemisphere: Mann and Bradley (1999, pink line), Briffa (2000, yellow line). Jones et al. (1998, blue line), Cowley and Lowery (2000, light blue line), Esper et al. (2002, purple line) and the averages of variation of the above five normalized series (bold black line); Cold phases (gray shade); b Normalized temperature changes of winter-half-year in eastern China since ad 1000 (Ge et al. 2002); c Total number of wars (dark blue line), north wars (black line), and south wars (red line); d Total number of rebellions (light blue line), north rebellions (pink line), and south rebellions (dark green line).

Population Data

Whereas warfare and temperature change are comparatively well documented, it is very difficult to obtain records of population changes and agricultural production for the study’s entire time frame because early dynasties had no registry systems for the collection of such data. Fortunately, records were compiled under the “Baojia” system introduced in China in ad 1741, which gathered annual population, harvest and land data for taxation and administrative purposes. It continued until ad 1851, when the Taiping Rebellion (ad 1851–1864) disrupted the entire system. Population counts in the Baojia period are generally valid in terms of the magnitude of China’s population size changes over a long period (Durand 1960). As there are no major changes in territory and frequency of warfare during the Baojia period, which also covers both warm and cold climatic phases, we deem the associated data to be appropriate in identifying the effect of climate change, if any, on population growth and spatial distribution. We also included the first modern Chinese census conducted in ad 1953 for comparison.


Temperature and War Cycles

Warfare frequency in eastern China demonstrates a cyclical pattern (Fig. 2c). War frequencies are summed by decades and grouped into three classes: very high (>30 wars/10a), high (15–30 wars/10a), and low (<15 wars/10a). All four very high peaks and eight out of the 11 peaks above the very high and high groups coincide with cold phases. Three high peaks stand well above the others, two of which occur in the coldest phases. All cold phases have high warfare frequency. Warfare generally lags 10–30 years behind the start of a cold phase.

The geographic distribution of these wars (Fig. 2c) reveals an interesting and significant pattern. Most warfare occurred in the south. Nevertheless, the high warfare frequencies were generally initiated by wars in the north, except in the fourteenth and nineteenth centuries when China was ruled by northern nomadic tribes (respectively, Mongol and Manchu) and outbreaks of war began in the south. Warfare peaks in the north closely matched cold phases––five out of the six peaks (>10 wars/10a) occurred in cold phases. The frequencies of warfare in the north are relatively constant. At the same time, six of the seven war peaks in the south occurred during cold phases. The maximum war frequencies in the south came 20–50 years after the start of cooling, except in the fourteenth century when the Mongols ruled China. The aberrant peaks in the sixteenth century (Fig. 2c) were mainly caused by northern nomadic invasions led by the Mongol leader Altan Khan and wars with “wokou” (Japanese pirates) along China’s eastern seaboard. In general, rebellion was the dominant category of war. The variation in the frequency of rebellion was highly correlated with climate change (Fig. 2d), as all three outstanding peaks were in the cold phases.

Eastern China’s winter-half-year temperature reconstruction (Fig. 2b) shows that all four very high war peaks and nine out of the 11 peaks above the very high and high groups also occurred in the coldest times of a cold spell (<−0.3°C). The notable difference between the two temperature reconstructions as shown in the Fig. 2a and b was the period during the thirteenth century when eastern China had favorable temperatures while the Northern Hemisphere experienced the opposite. However, the war peaks in this century were generated by the tribal invasions originating from western China (Mongols and Jin), where the climate was rather cold and dry (Shi et al. 1999; Tan et al. 2003).

War Ratios in Different Climate Phases and Correlations Between Temperature and War Frequency

To refine our analysis, we calculated the number of wars in each climate phase and compared the ratios of warfare in warm and cold phases (Table I). Results show the warfare ratios in the cold phases were higher than in the warm phases, especially for the “south wars” and “rebellions.”

Table I Number of Wars and Ratio of Wars in Cold and Warm Phases of Various War Categories in Eastern China

Pearson’s correlation coefficients between warfare frequency and temperature anomalies were calculated at three different time scales: phase, decade, and annual. The phase scale calculation could determine whether outbreaks of warfare were related to the lowest or average temperature anomalies. In each phase, the highest war frequency, lowest temperature anomalies, and average temperature anomalies were included in the calculation. Results show that the highest frequencies of “total number of wars,” “south wars,” and “rebellions” were significantly correlated with the lowest temperature anomalies of the phases (Table II). Only the highest frequencies of “south wars” and “rebellions” were significantly correlated with the average temperature anomalies. The rebellions were predominantly peasant uprisings induced by famine and heavy taxation, since farmers were always the first to suffer from declining agricultural production. The three outstanding peaks of warfare were dominated by peasant uprisings. Wherever they occurred rebellions are always significantly correlated with temperature change (Table III).

Table II Pearson’s Correlation Coefficients Between War Numbers, Average Temperature Anomalies, and Minimum Temperature Anomalies at the Phase Scale
Table III Pearson’s Correlation Coefficients Between Rebellion Frequencies, Average Temperature Anomalies, and Minimum Temperature Anomalies at the Phase Scale

At the decade scale, we computed the correlations between the number of wars in each decade, average temperatures, and lowest temperatures with 0–30 year time lags (Table IV). “South wars” and “rebellions” are significantly correlated with the anomalies in average temperature and the lowest temperature anomalies at the 0-, 10-, and 20-year time lags. The “total number of wars” is significantly correlated only with the two anomalies at the 10- and 20-year time lags. Such a pattern also appears in the annual scale correlation analysis (Table V). Annual warfare counts are significantly correlated with temperature anomalies in the categories of “total number of wars,” “south wars,” and “rebellions” for 0–30 years time lags. In addition, the highest correlation coefficients for different time lags vary with warfare categories. For the “total number of wars” and “south wars,” the most correlated time lag is 10-years, compared to “rebellions” at 15-years.

Table IV Pearson’s Correlation Coefficients Between War Numbers and Temperature Anomalies at the Decade Scale
Table V Pearson’s Correlation Coefficients Between Average Temperature Anomalies and Different War Categories at the Annual Scale

Nevertheless, correlation analysis based on decadal and annual scales cannot reveal a precise exact correlation between the local short-term temperature changes and occurrences of warfare. The temperature records we use are averaged and smoothed, so that any local temperature extremes were removed prior to statistical analysis. However, correlation analysis was still appropriate in assessing the correlation between the frequency of warfare and ambient temperature anomaly. The time lag between outbreaks of warfare and temperature changes (Fig. 2c), revealed from the analysis, is most likely accounted for by the buffering capacity of stored food resources which could sustain individuals for some time as well as maintain the power of the state.

There is discrepancy in the timing of temperature changes between eastern China and the Northern Hemisphere (Fig. 2a and b). Nevertheless, when we compute the correlations between the temperature anomalies of eastern China and frequency of occurrences of warfare at the 30-year scale, we reach the same conclusion: “south wars” and “rebellions” were both significantly correlated with temperature changes (Table VI).

Table VI Pearson’s Correlation Coefficients Between Eastern China Temperature Anomalies and Different Categories of Wars at the 30-year Scale

Northern China is colder and drier and its ecology is more vulnerable to temperature change than the warmer and wetter south. However, in our correlation analysis, the frequencies of warfare in the north are not significantly correlated with the temperature anomalies. In addition, the cold–warm ratio of occurrences of warfare in the north is lower than that in the south (Table I), highlighting the unique history and geography of northern China. During the periods when China was ruled or dominated by northern nomadic tribes (the South Song, Yuan, and Qing dynasties—more than half of the study period), people in the north could freely move to the south, either temporarily or permanently, to pursue their livelihoods. Thus, the frequency of warfare in northern China was lowered in the cold phases in the thirteenth, fourteenth, seventeenth and nineteenth centuries (Fig. 2c). At the annual scale correlation analysis, if we exclude the periods of ‘conquest/nomadic dynasties,’ the temperature-frequency of warfare correlations in the north is higher than in other regions (Table VII). Another view is that during the periods when China was politically divided between the south and north, the disruption of food supplies in times of cooling led to war, and opposing armies conducted cross-border raids on each other’s crops. During the periods when China was unified, which was often under the northern Jurchen, Mongol, and Manchu nomadic tribes, inter-provincial southward migrations were driven by cooling and led to repeated conflicts between migrants and the indigenous/minority groups in southern China. This seems to explain why the southern wars are more closely correlated with temperature variation than elsewhere.

Table VII Pearson’s Correlation Coefficients Between Annual Temperature Anomalies and Annual War Records of the North Part of Eastern China with the Period of ‘Conquest/Nomadic Dynasties’ Excluded

Population Dynamics in Face of Climate Changes

In the mid-Qing dynasty, the Chinese “Malthus,” Hong Liang-chi (ad 1744–1809) opined that excessive population growth threatened Chinese society (Silberman 1960). We used official population census data from ad 1741 to 1850 (Ho 1959) to compare population growth rates between warm and cold climatic phases and found no occurrences of large-scale warfare and out-migration that might affect population growth. For China as a whole, the average annual growth rate from ad 1741–1805 (warm phase) was 1.27%, while from ad 1806–1850 (cold phase) the rate was just 0.60%. Such a discrepancy might reflect the effect of cooling in reducing agricultural yields, which subsequently lowered natural population growth. The problem may have been aggravated by population growth during the previous warm period.

The patterns of population growth in the north of eastern China (Henan, Hebei, Shanxi, Sha’anxi, Gansu, and Shandong provinces) and in the south (Jiangsu, Zhejiang, Anhui, Jiangxi, Anhui, Guizhou, Shchuan, Hubei and Hunan provinces) are shown in Table VIII and are based on the notably accurate population censuses of ad 1787, ad 1850, and ad 1953 (Liang 1980). From ad 1787 (warm period) to ad 1850 (cold period), population increased rapidly in the south but decreased in the north. Population growth rate in the north was even negative. Conversely, from ad 1850 (cold period) to ad 1953 (warm period), population increased very slowly in the south but very rapidly in the north. At the same time, the north–south ratio of population size varied in different climate phases. In the warm period, the population percentage increased in the north but decreased in the south, and vice versa. Although the ratio difference was only about 10%, the size of the population in question was over 30 million. It is instructive to tentatively explain the above changes even in the face of incomplete information.

Table VIII Population Distribution in Eastern China Based on the Official Population Census

Despite the estimated ten million deaths caused by various natural calamities in Jiangsu, Zejiang, and Jiangxi during the cold peri2d (Deng 1998), population in the south grew very fast between ad 1787–1850, at an annual rate of 1.3%. This figure is higher than the annual growth rate of any other country in the world (highest = 1.2%) (Durand 1960). Besides, the negative growth rate in the north during the same period is puzzling. Except for large-scale warfare or the shrinkage of the area controlled by the state, depopulation only occurred in the periods of frequent pandemics and natural calamities. Between ad 1787 and 1850, several epidemics and natural disasters occurred in Gansu, Hebei, and Shandong (in ad 1810, 1811, and 1846) (Deng 1998), but the reduction in population did not occur in those provinces but elsewhere in the north. The most likely explanation is internal migration. We suggest that during the cold phase people in the north migrated to the south to take advantage of the relatively favorable climate, and during the warm period, many people in the south went north to avail themselves of the increase in arable land in formerly marginal areas. Subject to warmer temperatures, some pastoral steppe land or wasteland can be converted to farmland (Fang and Liu 1992). There were two large-scale waves of southward migrations over the past millennium: one over the period ad 1000–1300, and the other from ad 1630 to 1900 (Fang and Liu 1992). In fact, all of the three long cold phases (ad 1194–1302, ad 1583–1717, and ad 1806–1911) overlap with the two mass migration periods. In the early ad 1900s, the Northern Hemisphere became warmer again, and people from Shandong and Henan moved in large numbers to northeastern China (i.e., Manchuria).

Between ad 1787 and 1953, China’s total population increased 77%, while the two northwestern provinces of Sha’anxi and Gansu, which are located in the transitional zone between wheat-farming and pastoral regions and underwent severe desertification, experienced population declines of −5.4 and −16.2% respectively. This may reflect the ecological fragility of dry and cold areas in periods of cooling that not only reduce local bio-productivity, but also trigger long-term desertification and land degradation. Thus, unlike other provinces, the population in these two provinces did not recover to their previous levels.

The Fall of the Ming Dynasty

There are two periods of climatic cooling during Ming dynasty (ad 1368–1644), which are also two of the coldest since ad 1000. The first period (ad 1448–1487) overlaps with a major low point in Ming history. In ad 1448, there were massive famines in the central Yangtze region and on the eastern reaches of the Yellow River coinciding with severe drought in the southeast and northwest. Conditions in China overall did not improve much over the next several decades. There were major famines in Jiangxi and Huaidong in ad 1452; the Southern Metropolitan Region in ad 1455; Shandong and the Northern Metropolitan Region in ad 1457; Huguang in ad 1458–1459; Huguang and Guangxi in ad 1460; Shaanxi in ad 1462; Henan in ad 1463; Hebei, Henan, and the lower Yangtze in ad 1465; Shanxi and the Southern Metropolitan Region in ad 1466; Huguang and Jiangxi in ad 1467; northern and central China in ad 1468; and Shanxi in ad 1469. Compounding these problems were a series of destructive locust attacks in the mid-ad 1450s and two unusually cold winters in ad 1449–1450 and 1453–1454. During the latter, large numbers of people and animals in southeastern China died of cold and starvation; Lake Tai in the heart of the Yangtze delta froze for the first time in perhaps a century (Atwell 2001; 2002). Adverse climatic conditions and food shortages during these years also caused serious difficulties in Mongol-controlled territory north of the Great Wall, possibly contributing to the intensification of Mongol pressure on the Ming Empire that culminated in their destructive invasion of northern China in ad 1448–1449 (Atwell 2001; 2002).

In China, anti-government uprisings began during the mid-ad 1440s in the border areas of Zhejiang, Fujian, and Jiangxi provinces and were not brought under control until the early ad 1450s. That same period also saw Ming armies mount a series of costly campaigns in southwestern China and northern Burma. In July ad 1449, after tensions on the northern and northwestern frontiers had been escalating for some time, due in part it appears, to deteriorating climatic conditions, Esen, the chieftain of the Oirat Mongols, began a large-scale invasion of China. This led to a disastrous campaign during which the Ming Emperor was captured by the invaders and Esen’s forces advanced to the walls of Beijing. Although the Ming government managed to establish relatively peaceful relations with the Oirats over the next few years, the dynasty’s security problem continued. In ad 1450–1456, Ming forces were forced to put down anti-government uprisings in Guizhou, Huguang, Guangdong, Fujian, Zhejiang, and Sichuan. Moreover, cooling during the late ad 1450s and early 1460s appears to have caused serious problems for the Chinese and the Mongols living in the area, with increasing tensions between the two groups contributing to Mongol participation in the abortive coup of ad 1461 in which disaffected members of the Ming government attempted to remove the emperor from power (Atwell 2001; 2002). Fortunately, this cooling period was comparatively short, and had been preceded by a long warm phase that might have allowed sufficient granary storage to preserve the state power. Within a few decades a warming trend set in ushering in what is termed the High Ming period. In the mid-sixteenth century, warfare was mainly efforts to rebuff northern nomadic invasions led by the Mongol leader Altan Khan and wars with “wokou”––Japanese pirates who marauded along China’s eastern seaboard. Nevertheless, neither Altan nor the “wokou” mounted a challenge serious enough to force a major reorganization of the Ming Empire (Huang 1988).

The fate of the Ming regime is seemingly virtually sealed in the second cooling period (ad 1583–1717). There was drought in the Beijing region in the autumn and winter of ad 1584, famine in Huguang during the late summer of ad 1585, severe flooding in Jiangnan, Zhejiang, Huguang, Fujian, Yunnan, and Liaodong in the summer of ad 1586, and heavy rains and serious flooding in both northern and southern China during the summer of ad 1587. At various points during ad 1587, drought and famine were reported in areas north of the Yellow River. By the early months of ad 1588, the areas experiencing severe food shortages had widened to include parts of Shaanxi, Shanxi, Shandong, Henan, Zhejiang, and Jiangnan. In ad 1589, an extended period of severe drought began in southeastern China, with parts of Zhejiang, Huguang, Jiangxi, and Jiangnan being particularly hard hit. After a brief respite in ad 1590, the next few years saw heavy rains and floods damage agricultural production in many regions of the country. By ad 1594, people in some parts of China were reduced to eating the bark of trees, the seeds of grass, or even seeds from the excrement of wild geese (Atwell 2001; 2002). Although the epicenter of the ad 1594 famine is usually considered to have been north-central China, there were also very poor harvests that year in the southeastern provinces of Fujian and Guangdong. During the autumn of ad 1596, there were persistent rains and severe floods in northern Zhejiang, one of China’s key agricultural regions. In ad 1601, there was extended drought in northern China, followed by summer floods in both the north and the southeast of the country. In ad 1609, the Ming Empire was hit by an extended drought in Huguang, Sichuan, Henan, Shaanxi, and Shanxi and by heavy rains and floods in Fujian, Zhejiang, and Jiangxi. By the late spring of ad 1610, famine conditions were reported in parts of Bei Zhili, Shandong, Shanxi, Henan, Shaanxi, Fujian, and Sichuan (Atwell 2001; 2002).

There is abundant evidence of both increased drought and cold in China during this period, when the growing season in north China was 2 weeks shorter than it is now (Jiang 1993). Both the lakes of the middle Yangtze and the Huai River froze over in winter (Wakeman 1986). As conditions in China deteriorated during the late sixteenth and early seventeenth centuries, the Ming government faced a number of significant military challenges to its rule. Among the more notable of those were a major rebellion in Shandong in ad 1587, repeated attacks in the northwest by the so-called Ordos Mongols, raids by the Eastern Mongols in Manchuria, hostilities between Chinese and Burmese forces on the southwestern frontier, Vietnamese incursions along the southern border, and a lengthy rebellion in central and western China.

The Manchus originated from Mount Changbai in northwestern China, and were descended from the Jurchens of the Jin dynasty that ruled between ad 1127 and 1234 (Wang 1991). During the early seventeenth century the Manchu leader Nurhaci began to organize his people into a powerful political and military force. In April 1619, he routed a force of 100,000 Ming soldiers (Huang 1988). To what extent Nurhaci was spurred in his organizational efforts by climatic and economic problems in the Manchu homeland remains to be carefully studied, but the rise of the Manchus, like the rise of the Jurchens in the late eleventh and early twelfth centuries and that of the Mongols in the late twelfth and early thirteenth centuries, occurred in a cold climatic phase (Atwell 2001; 2002).

Drought and famine continued to plague northern China the following years. At the same time, the Ming Empire had to fight a protracted two-front war against peasant guerrillas on the one hand and the Manchu cavalrymen on the other. Unusually severe weather struck in ad 1620–1640, when the earth’s climate fell to its lowest temperatures since ad 1000. Extreme droughts were followed by major floods. Frequent famines, accompanied by plagues of locusts and smallpox, resulted in starvation and mass death. The five worst years of consecutive drought in China as a whole occurred in ad 1637–1641 (Wakeman 1986). Peasant rebellions spread to the northern and southern parts of eastern China (Fig. 2d) and drastically reduced the populations of many localities in northern China, nearly exterminating the population of the Red Basin of Sichuan (Ho 1959). From the Huai Valley to the Northern Metropolitan Region, all the bark had been stripped from the trees and people even dug up corpses for food. Following a rebellion in Shandong during which a large number of people were killed, the dead were quickly cut up for food (Atwell 2001; 2002). The result was extraordinary depopulation during the late Ming era.

Unusually heavy snowfalls and food shortages caused serious problems in southeastern China during the early months of ad 1641, and it remained very dry in much of the north with infestations of locusts reported in many areas. As grain prices rose to extraordinary levels, local people organized bandit gangs and reportedly even resorted to cannibalism (Jiang 1993). Nor were China’s problems limited to the lower Yangtze region. There was severe flooding in Henan during the summer of ad 1642 as well as famine in portions of Henan, Shandong, Huguang, Zhejiang, and the Northern Metropolitan Region. By this time, nation-wide rebellions had already destroyed the major military power of Ming (Atwell 2001; 2002). Ominously for Chinese authorities, ad 1642 also saw the final collapse of Ming military power north of the Great Wall and an extended campaign by Manchu forces in eastern China. The extent to which the Manchus were experiencing climatic-induced stress of their own remains to be thoroughly studied but, as was the case during the late sixteenth century, there is considerable evidence to suggest that their attacks on China during the ad 1630s were inspired, at least in part, by serious economic problems in the Manchu homeland (Hsu 1998; Atwell 2001; 2002). In early ad 1644, Li Zicheng, the leader of peasant rebellions, declared his new dynasty in Xi’an, and then crossed the Yellow River, marched across the entire length of Shanxi Province, seized the fortress on the Great Wall, overpowered the capital garrison and occupied the capital. Finally, the Ming General Wu Sangui opened the gates of the Great Wall for the Manchus, who defeated both the rebels and the remaining forces of the Ming regime (Huang 1988).

There is no easy way for historians to explain how the Manchus, with a population of about one million, could by ad 1644 seize the throne of China with such ease. They had devised a writing system for their spoken language only in ad 1599. The “banner system,” which gave their tribal organization a bureaucratic touch by regulating the mobilization procedure and its agricultural support, was introduced no earlier than ad 1601. By ad 1635 they began to call themselves Manchu. Another year elapsed before the Qing dynasty was formally proclaimed. It took less than a half-century for this loosely constructed confederation of tribes to swallow an enormous empire with a profound cultural heritage (Huang 1988). We believe that the force driving the Manchu invasions was the extreme climatic conditions in northeastern China during the coldest period of the Little Ice Age, and the consequent serious economic problems (Hsu 1998; Atwell 2001). This explains why nearly all of the Manchus left their homeland and moved southward after the conquest (Ge 1997).

Discussion and Conclusion

Over the past millennium, eastern China suffered from periodic ecological stress and a significantly reduced anthropocentric carrying capacity during climatic cooling. Since the expanded population encouraged by the previous warm phase could not be sustained, famine and nationwide uprisings, predominantly mobilized by peasants, were seemingly fueled during cold phases. Such domestic chaos weakened state power, which in turn invited northern nomadic invasions. Tensions among different states, ethnic groups, and tribes were further intensified during the cooling periods, accounting for higher frequencies of warfare in cold phases. The fact that warfare occurred more frequently in the south than in the north can be attributed to its unique local cultural and geographic settings, which are also reflected in the differing time lags between onset of cooling and outbreaks of warfare in the north and south, as well as the temperature–war correlation in the north (with the periods of ‘conquest/nomadic dynasties’ removed). Although cold phases reduced agricultural yields, the outbreaks of warfare generally lagged behind the onset of cooling because of social buffers (i.e., granary storage). We also found that dynastic collapses also followed the oscillating temperature cycles over the past millennium (Fig. 2c). Almost all of the dynastic changes occurred in cold phases, with the exception of the of Yuan dynasty, which collapsed 8 years after the end of a cold phase (ad 1368), although it had lost most of its territory in the “Late Yuan” peasant uprisings during the cold phase. The delayed collapse was largely a result of power struggles among different rebel groups.

Past research, using the paradigm of “historical particularism,” has tended to reduce the fundamental causes of warfare to various economic, political, or ethnic sources. None of these deal specifically with the probable contribution of climate change as a causal triggering factor. For example, historians of China speak of a war–peace cycle characterized by prosperity in the upswing when a new line of emperors was established, and by misery in the downswing when the dynasty became old and feeble. The fall of dynasties and the related social turmoil (i.e., internecine warfare and population collapse) are attributed to the growing degeneracy of the court (Zhao 1994). Although such generalization can account for instances of war–peace cycles, it cannot explain their timing or periodicity.

Thomas Malthus (1798) pointed out that population can increase geometrically, while means of subsistence can increase, at best, arithmetically. Thus, if population growth is not controlled, human societies will proceed to the limits of carrying capacity, which will be followed by wars, famines, and epidemics, and ultimately population collapse. Following Malthus, studies of China’s historic war–peace cycles usually conclude that excessive population growth is the ultimate cause of state instability (Usher 1989; Chu and Lee 1994; Turchin 2005). However, our study suggests that from a macro-historical perspective (not relying on particular cases, but on all known occurrences of warfare to reach any conclusion), it was the oscillations of agricultural production brought by long-term climate change that drove the China’s historical war–peace cycles. Our methodological framework reveals a near perfect match between high war frequencies and the cold phases, doubled war ratios in cold phases, and significant correlations between frequency of occurrences of warfare and temperature variation in phase, decadal and annual scales in the last millennium. These are unlikely to be fortuitous. This study also finds that during cold phases, reduced critical resources could not support the size of population achieved during the previous warm period. People of different social groupings (class, tribe, state) in agrarian society had to compete for shrinking resources (land and food in particular) during cold phases, and consequently, war–peace cycles closely followed temperature variations during the past millennium. Given that fact, instead looking only to human factors, it might be well for scholars to keep climate change in mind as they consider the anthropology of warfare in the historic past.

Recently global warming has attracted much attention. It must be pointed out that, unlike the “warm phases” we discuss here, current global warming is unprecedented in the last two millennia (Mann et al. 2003; Moberg et al. 2005). Global temperature is expected to rise faster and faster in the foreseeable future. Its impacts will be fairly unpredictable because both natural and anthropogenic forces are involved. In spite of technological changes, most of the world’s populations will continue to rely on small-scale agriculture, which is vulnerable to climatic fluctuations as were the historical societies examined here. Furthermore, in an increasingly densely populated world, habitat-tracking as an adaptive response will no longer be an option (Weiss and Bradley 2001). In the last century, with rapid advances in industry, people often overestimated their power to control the natural environment. As a related phenomenon, anthropocentrism came to dominate the fields of war studies, social sciences, and history in the last few decades. However, history suggests that climate change has played a crucial role in shaping agrarian societies and may be the major underlying mechanism of cycles of warfare. Now that the human population has reached its current level, even if food supply does not become a problem in highly developed societies, shortages of other essential resources, such as fresh water, agricultural land, energy sources, and minerals, may very likely trigger more armed conflicts among human societies.