The causality analysis of climate change and large-scale human crisis

Responses to Climate Change in Human Society.

Our study period covers both mild and cold phases of the Little Ice Age in the NH. Based on the NH (Fig. 1A, red line) and European (Fig. 1A, black line) temperature anomaly series, we divided our study period into Mild Phase 1 (A.D. 1500–1559; average temperature = 0.43σ), Cold Phase (A.D. 1560–1660; average temperature = −0.59σ), and Mild Phase 2 (A.D. 1661–1800; average temperature = 0.24σ). The Cold Phase coincided with the GCSC. In Mild Phase 2, there was brief cooling in A.D. 1700 and A.D. 1750. To elicit the real association between climate change and the cyclic pattern of various variables, the variables with obvious long-term trends (agricultural production index, grain price, real wages, body height, and population size) were linearly detrended (23, 24).

Fluctuations of all agro-ecological, socioeconomic, human ecological, and demographic variables corresponded very well with temperature change and were in successive order. The variables of the bio-productivity, agricultural production, and food supply per capita (FSPC) sectors responded to temperature change immediately, whereas the social disturbance, war, migration, nutritional status, epidemics, famine, and population sectors responded to the drop in FSPC with a 5- to 30-y time lag. The adverse effect of the two short-term, minor cooling episodes in Mild Phase 2 (Fig. 1, blue dotted lines) also was reflected by the variables’ fluctuations in annual and decadal units, such as NH tree-ring width, grain yield, grain price, and agricultural production index.

Cooling in the Cold Phase dampened agro-ecosystem output by shortening plant growing seasons and shrinking the cultivated land area (23). The ratio of grain yield to seed decreased along with temperature decline (Fig. 1B, red line). Tree-ring width (a variable of bio-productivity) also varied in response to temperature change, decreasing rapidly in A.D. 1560–1650 (Fig. 1B, black line). Grain yield links directly to agricultural production, which is represented by the agricultural production index. Although in the long term the agricultural production index moved upwards with population size, it decreased or stagnated in a cold climate and increased rapidly in a mild climate at the multidecadal time-scale (Fig. 1C, black line).

Although agricultural production decreased or stagnated in a cold climate, population size still grew. Hence, two variables of FSPC—grain price and real wages of labor—changed considerably, and economic crisis followed. Grain price is determined by both demand and supply and is an important indicator of the boom-and-bust cycle in an agrarian economy. The detrended grain price (Fig. 1C, red line) was inversely correlated with every fluctuation of the agricultural production index and temperature. Real wages of labor (Fig. 1D, red line) varied inversely with grain price and followed agricultural production and temperature change closely. Given the low FSCP, famine became more frequent (Fig. 1D, black line), resulting in deteriorating nutritional status and ultimately in reduced human body height (25). The average height of Europeans followed temperature closely (Fig. 1F red line) and declined 2 cm in the late 16th century. It increased slowly with rising temperatures only after A.D. 1650.

Inflating grain prices and declining real wages bred unbearable hardship in all walks of life, triggering many social problems and intensifying existing social conflicts. Peaks of social disturbance such as rebellions, revolutions, and political reforms followed every decline of temperature, with a 1- to 15-y time lag (Fig. 1E, black line). Many disturbances eventually developed into armed conflicts. The number of wars increased by 41% in the Cold Phase (Fig. 1E, red line). Although the number of wars decreased in the interval A.D. 1620–1650, these wars were comparatively more lethal and longer lasting (e.g., the Thirty Years War) (26). Annual war fatalities from 1620–1650 were >12 times those in the period A.D. 1500–1619 (Fig. 1G, red line).

More frequent and severe economic chaos, famine, social disturbance, and war pushed people to emigrate. In Europe, migration (Fig. 1G, black line) peaked during A.D. 1580–1650, overlapping exactly with the peak of social disturbance. This correlation indicates that social conditions are imperative in driving migration (27). Migration, coupled with individuals’ deteriorating health caused by poor nutrition, facilitated the spread of epidemics (7). The number of plagues peaked during A.D. 1550–1670 (Fig. 1, black line), reaching the highest level throughout the study period. Population growth rate, which is codetermined by famine, epidemics, and war, fluctuated complexly. When peaks in war fatalities and famine occurred during A.D.1620–1650, the annual population growth rate (Fig. 1H, black line) dropped dramatically from 0.4 to −0.3%. Population collapse occurred (Fig. 1H, red line), and the European population dropped to its lowest point (105 million people) in A.D. 1650.

In general, variables in European societies (except population) reacted linearly to temperature change at the multidecadal time scale (Fig. 1 and SI Appendix, Table S1). Some variables, however, responded exponentially to cooling in A.D. 1620–1650. We further examined the time-series of those variables and found that, after cooling, population pressure rose after A.D. 1560 (the agricultural production index declined, and annual population growth was ∼0.4%) to the point that a significant reduction of population size was necessary to ease food strain in Europe. The triggers of population collapse were war and famine. After A.D. 1618, many large-scale wars and famines occurred in Europe. The war fatality index was 20 times higher than the A.D. 1500–1617 average and persisted at a similar level for the next 32 y (Fig. 1G, red line). During A.D. 1618–1649, ∼10 million people perished in wars (26).

Humans have served as both producers and consumers in Earth's ecosystem since the Agricultural Revolution. During great wars and famines, death rates exceeded birth rates, causing substantial reduction of the agricultural production workforce. Also, collapse of agricultural production infrastructure caused by wars left behind massive damage to carrying capacity and sustainability (8). Consequently the role of the human population as a producer became less significant. Although the temperature and grain yield in A.D. 1600–1620 and A.D. 1620–1650 were similar, in 1621 the feedback effect of population collapse brought about a 13% reduction in agricultural production, which had stagnated for ∼50 y. Such a huge decrease caused an exponential increase in grain price (+200%), famine (+250%), war fatality (+1,350%), social disturbance (+100%), migration (+250%), and other population checks. On the other hand, real wages, body height, and epidemics remained at the same level, and the number of wars dropped slightly (Fig. 1). This complex relationship between agricultural production and population size continued until A.D. 1650, when temperature and thus agricultural production increased.

At the end of the Cold Phase there was an augmentation of agricultural production that, together with a population slump, led to a rise in FSPC and the recuperation of most of European society after A.D. 1660. This date marks the end of the GCSC and the start of the Enlightenment era. The mild climate in the 18th century created human ecological harmony, leading to a speedy economic and population recovery in Europe. Although the short cooling in A.D. 1700 and A.D. 1750 caused minor fluctuations in grain yield, real wages, grain price, famine, war, social disturbance, and migration, its impact was not strong enough to cause general crisis and population collapse (Fig. 1).

Statistical Verification of Causal Linkages Between Climate Change and Large-Scale Human Crisis.

Cooling triggered a chain of responses in variables pertaining to European physical and human systems. All 16 of the variables we identified are categorized into 11 sectors according to the response time of variables to cooling, together with natural laws and social theories related to different variables. Five of these sectors contain two variables with the same properties (e.g., the variables “NH temperature” and “European temperature” belong to the climate change sector). We then identified a set of causal linkages among the 11 sectors, demonstrating how climate change brings about general human crisis (Fig. 2 and SI Appendix, Text section 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Set of causal linkages from climate change to large-scale human crisis in preindustrial Europe. The terms in bold black type are sectors, and terms in red type within parentheses are variables that represent the sector. The thickness of the arrow indicates the degree of average correlation, which is calculated from SI Appendix, Table S2.

In the set of causal linkages, climate change and associated bio-productivity fluctuation are revealed as the ultimate cause of economic, social, human ecological, and demographic problems. If the climate change and bio-productivity sectors are disregarded, various linkages within the human system seem to be driven endogenously by population growth. The concept of a population-driven human system is prevalent among social scientists, demographers, and economists (28, 29), but ignoring the impact of climate forces on human systems may lead to false conclusions. Although the causal linkages in Fig. 2 are theoretically reasonable, the strength, consistency, predictability, and time sequence of the linkages should be verified statistically before any definite conclusions are drawn.

We cross-correlated the 16 variables (Materials and Methods) to validate the strength of the set of causal linkages in Fig. 2. All 120 cross-correlations were statistically significant (P < 0.05), and 116 of them were highly significant (P < 0.001) (SI Appendix, Table S2). Patterns of the correlations reveal the following:

• i) Correlation between temperature data and a variable in another sector became weaker as the sector's distance from climate change increased. For instance, the sector's distance for climate change → bio-productivity vs. climate change → population showed a distance decay effect indicating that the impact of climatic forcing was partially offset by human adaptation or natural factors.

• ii) Compared with the NH temperature variable, the European temperature variable was better correlated with other variables, because, aside from NH temperature and NH tree ring variables, all other variables are for Europe only.

• iii) The causal linkage from the climate change sector to the bio-productivity sector (e.g., European temperature and grain yield variables) was comparatively stronger.

• iv) The variables Population size and Population growth rate had weaker correlations with other variables because they were determined by multiple variables. The strength of association among different sectors is shown in Fig. 2.

We also used multiple regression analysis to validate the consistency and predictability of the causal linkages shown in Fig. 2 (Materials and Methods). In regression models, the independent variables were time and causal variables, and the dependent variable was the “effect” variable. For example, in the relation European temperature → grain yield, European temperature is a causal variable, and grain yield is the effect variable. Time (t) presumably represents technology and/or capital accumulation. An attempt was made to eliminate the trend from the dependent variable (effect variable) using parabolic (t and t²), squared (t²), and cubic (t³) terms (23). The various detrending procedures did not affect the regression results significantly, and all the elasticity of the effect variable in response to the change in the causal variable was statistically significant (SI Appendix, Table S3). The causal relationship between variables/sectors was statistically valid when the effect of societal development was controlled.

We further validated the time sequence and predictability of the causal linkages in Fig. 2 by using GCA (SI Appendix, Text section 2). Via GCA, the causal relationship between variables is confirmed only if the cause precedes the effect in time and the causal series contains special information that could better explain and forecast the series being caused (30). The causal linkages in Fig. 2 boiled down to these relationships: Climate change → bio-productivity → agricultural production → FSPC; FSPC → social disturbance → war; FSPC → famine → nutritional status; FSPC, social disturbance, war, and famine → migration; nutritional status and migration → epidemics; war, famine, and epidemics → population; population → agricultural production; and population → FSPC. Our GCA results show that all null hypotheses of these linkages were rejected (13 linkages with P < 0.01 and 4 linkages with P < 0.05), implying that causal relationships between climate change and human crisis are statistically valid (Table 1 and SI Appendix, Text section 2.1).

Because the alternation of periods of harmony and crisis in Europe followed variations in FSPC (Figs. 1 and 2), we suggest that FSPC is a key sector bridging climate change and human systems. Because FSPC is codetermined by agricultural production (supply) and population size (demand), it can be epitomized by grain price (the ratio of supply to demand). We used GCA to test whether grain price is the direct cause of all social and human ecological crises. Grain price was the Granger-cause of social disturbance, war, migration, epidemics, famine, and nutritional status (five linkages with P < 0.01 and one linkage with P < 0.05) (Table 2 and SI Appendix, Text section 2.2). Hence, grain price could be taken as an indicator and direct cause of conditions of harmony or crisis in preindustrial Europe.

Simulation of Periods of Harmony and Crisis in Europe and the NH.

Grain price.

Based on the above findings, we used a longer grain price series to simulate the alternation of conditions of harmony and crisis in Europe further back in time. To eliminate the effect of long-term inflation upon the market price of grains, real grain price was used (SI Appendix, Text section 3). We found that real grain price followed temperature change inversely (Fig. 3A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Real grain prices and the alternation of periods of harmony and crisis in Europe, A.D. 1200–1800. (A) European temperature anomaly (σ, orange line), real grain price (Ag/L, bold black line), and the threshold of general crisis (real grain price = 0.2, pink dotted line). (B) Agricultural production index (orange line) and population size (in millions, green line). European temperature, real grain prices, and agricultural production index were smoothed by 40-y Butterworth low-pass filter. The light gray stripe represents a period of general crisis (real grain price >0.2); the dark gray stripe represents a period of demographic collapse.

When the GCSC began in A.D. 1560, real grain price was 0.2. Therefore, we set a real grain price = 0.2 as the general crisis threshold. The periods in which real grain price was >0.2 or <0.2 represent periods of crisis or harmony, respectively. With that threshold, our simulated crisis periods were A.D. 1264–1359 and A.D. 1559–1652, consistent with the time spans of the Crisis of the Late Middle Ages and the GCSC, as delimited by historians. In both crisis periods, real grain price was driven up by population pressure (i.e., steady population growth over a long period), bringing about demographic collapses at later stages. Each demographic collapse lasted for ∼30 y. The collapse of the 14th century started in A.D. 1315 when the Great Famine began. Our simulated periods of harmony (A.D. 1360–1558 and A.D. 1653–1800) coincided with the prosperous Renaissance and Enlightenment eras (SI Appendix, Text section 4.2) (21, 31, 32).

The complex relationship among temperature, real grain price, agricultural production, population size, and social conditions during the period is illustrated clearly in Fig. 3B. The finding echoes the key notion of Malthusian theory (33): When population size overshoots agricultural production, human misery follows. Malthus (33) argued that rapid population growth was the cause of human misery. Our findings, however, indicate that the misery in fact was triggered by climate-induced agricultural decline. Malthusian theory emphasizes increasing demand for food as the cause, whereas we found the cause to be shrinking food supply.

Temperature.

Although the alternation of harmony and crisis tracked fluctuations in the real grain price in preindustrial Europe, GCA results show that temperature change was the Granger-cause of real grain price (SI Appendix, Text section 2.3), because agricultural production was climate dependent at the time. Indeed, temperature change is the ultimate cause of human catastrophes, in that it affects first agro-economy and then people's livelihood.

We used a European temperature series as another indicator of conditions of harmony or crisis to simulate the “golden” and “dark” ages in Europe over the past millennium. We set temperature = −0.1σ (according to the 100-y smoothed European temperature series) as the general crisis threshold. The periods in which temperature was lower than −0.1σ or greater than −0.1σ represent dark ages and golden ages, respectively. With that threshold, the dark ages we calculated were A.D. 1212–1381 (the Crisis of Late Middle Ages) and A.D.1568–1665 (the GCSC), whereas the golden ages were the 10th to 12th centuries (the High Middle Ages), the late-14th to early 16th centuries (the Renaissance), and the late-17th to 18th centuries (the Enlightenment) (SI Appendix, Text section 4.1), largely in agreement with time intervals delimited by historians (SI Appendix, Text section 4.2). The mild cooling in Europe in the late 18th and 19th centuries brought about an upsurge in prices, social disturbance, war, and migration, but not demographic crisis, because of social buffers such as cross-continental migration, trade, and industrialization. Hence, the crisis that occurred in the early 19th century (the Age of Revolution) was not a general one.

Periods of harmony and crisis in the NH are reflected by fluctuations in population growth and the frequency of famine, epidemics, and war (SI Appendix, Text section 5). The NH temperature and European temperature was highly correlated (Fig. 4 A and B). The troughs of population growth (Fig. 4C) and the peaks of various mortality factors (Fig. 4 D–F) in the NH coincided with a cold climate. In fact, the alternation of periods of harmony and crisis in the NH corresponded to the alternation of such periods in Europe. In addition, regression results indicate that all the aforementioned variables were determined significantly by temperature change (SI Appendix, Table S4). Just as in Europe, temperature could be taken as the indicator of conditions of harmony or crisis in the NH in historical time. However, in the NH warming also caused widespread famine between the 11th and 12th centuries (the Medieval Warm Period), because high temperature caused drought in North Africa and Western Asia (34, 35). However, the warmth was not severe enough to engender global crises.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Temperature change and the alternation of periods of harmony and crisis in the NH during the past millennium. (A) European temperature anomaly (σ). (B) NH temperature anomaly (°C). (C) NH annual population growth rate (%). (D) Famine years in the NH (number of famine years per decade). (E) Number of deadly epidemic events (malaria, plague, typhus, measles, smallpox, and dysentery) per decade in the NH. (F) Number of wars per year in the NH. All data were smoothed by a 100-y Butterworth low-pass filter. Gray stripes represent periods of crisis in Europe as delimitated by historians (SI Appendix, Text section 4.2).