Komaga-take (; 42∘ N, 140∘ E; 1131 m a.s.l.), also known as Hokkaidō Komaga-take (), Oshima Komaga-take (), or Oshima Fuji (), is an andesitic stratovolcano located on Hokkaidō, Japan. According to the Global Volcanism Program (GVP, 2013a), much of the truncated volcano is of Pleistocene age and the breached crater (resulting from an edifice collapse in 1640) remains visible (Yoshimoto and Ui, 1998). The collapse was presumably caused by a phreatic eruption (Fig. 3a).

The yearly records of Matsumae, or Matsumae Nennen-Ki, report that on 31 July 1640, part of Komaga-take's summit collapsed to send pyroclastic flows and a huge debris avalanche (estimated at 0.25 km3) down its eastern slopes to produce a tsunami in Uchiura Bay (Katsui et al., 1975; Katsui and Chimoto, 1985; Yoshimoto et al., 2003), drowning at least 700 people in the present-day Toya and Usu region (Tsuji et al., 1994; Nishimura and Miyaji, 1995). Furious eruptive activity continued on 1–2 August 1640, so that contemporary sources report that “the day became dark like nighttime with little sunshine” (Katsui et al., 1975). While ashfall continued, the color of the cloud turned purple on 3 August in Matsumae. Activity weakened substantially thereafter and stopped after about 70 d (Katsui and Ishikawa, 1981). Ashfall deposits of the Komaga-take's 1640 eruption are 2 m thick next to the volcano and still reach 10 cm in Tsugaru region (northern Honshū), some 130 km south-southwest of the volcano (Yamada, 1958) for an estimated ejected volume of 3.5 km3 (Sasaki et al., 1970; Katsui et al., 1986). The 1640 Komaga-take eruption is thus one of the largest in Japan during historical times and deposited ash even in central Honshū, located >800 km to the south.

Mount Parker (6∘ N, 125∘ E; 1824 m a.s.l.), locally known as Mélébingóy, is a densely vegetated, andesitic–dacitic stratovolcano near the southern tip of Mindanao Island (Philippines; GVP, 2013b). Its English name originates from 1934 when American general Frank Parker spotted the mountain and claimed its discovery during a flight he piloted (Davis, 1998). Mount Parker has a base diameter of 40 km; its summit is truncated by a 2.9 km wide caldera with steep walls rising 200–500 m above Lake Maughan (Lake Hólón locally; Fig. 3b).

The 1641 eruption was first described in western sources by the French naturalist Guillaume Le Gentil, who visited the region some 140 years after the cataclysm. In his two-volume work, A Voyage in the Indian Ocean (published in 1779 and 1781), he referred to a big eruption (erroneously dated to January 1640) that would have devastated large parts of the Philippines. Relying on contemporary Hispanic sources and geological investigations, Delfin et al. (1997) provided more insights into the sequence of the eruption. Mount Parker started erupting on 26 December 1640, and produced a major explosive event on 4 January 1641, resulting in the formation of the caldera as well as voluminous pyroclastic flows and lahars on its northern and eastern flanks (Delfin et al., 1997). The eruption produced pumice comparable in type to that described for the 1991 Pinatubo eruption (David et al., 1996).

While this eruption is known from contemporary Hispanic documents, an eyewitness report of Jesuit monk Raymundo Magisa written in 1641 has long been misattributed to Gunung Awu volcano in Indonesia (3∘ N, 125∘ E), ca. 300 km south of Mount Parker. The dating of nine charcoal pieces and oral traditions of local indigenous people finally allowed the attribution of the 1641 eruption described by Magisa to Mount Parker (Delfin et al., 1997). The same authors also translated Magisa's text testifying to the eruption:

“On … 4 January by 9 AM, the (supposed) artillery fire increased to such an extent that it was feared that the squadron (sent to investigate) might have run into some Dutch galleons. It lasted for about half an hour. People soon became convinced, however, that the noise originated from a volcano which had opened up, because by noon we saw a great darkness approaching from the south … By 1 PM we found ourselves in total night and at 2 PM in such profound darkness that we could not see our hands before our eyes. This darkness … lasted until 2 AM [5 January 1641] when a little moonlight appeared, to the great relief of Spaniards and Indios who had feared that they would be buried beneath the mass of ash which had started to fall on them at 2 PM. … At 10 AM the ships found themselves in such deep darkness and terrible blackness that they believed the Judgement Day to be at hand. It started raining so much rock, soil, and ash that the ships considered themselves in grave danger and had to light lanterns and quickly lighten the heavy load of soil and ash that was accumulating” (Delfin et al., 1997, p. 35).

To quantify NH summer temperature (JJA) and identify possible volcanic cooling induced by mid-17th-century eruptions, we employed the NVOLC v2 reconstruction of Guillet et al. (2017, 2020). NVOLC v2 comprises 25 proxy records (13 tree-ring width (TRW) and 12 maximum latewood density (MXD) chronologies). A principal component regression (PCR) approach was applied to reconstruct NH JJA temperature anomalies (compared to the 1961–1990 mean). A bootstrap method was then combined with the PCR to estimate the skill of the reconstruction and to compute confidence intervals of reconstructed JJA temperatures. To gradually adjust to the changing number of records available back in time, the PCR was combined with a nested approach (Guillet et al., 2017). The NVOLC v2 reconstruction is consequently based on 23 subsets of tree-ring chronologies or nests, with the most replicated nest containing 25 chronologies spanning the period 1230–1972. For each nest, a principal component analysis was calculated on the proxy predictors, and the first n principal components (PCs) with eigenvalues >1 were retained as predictors to develop a multiple linear regression model. The regression models were all calibrated on JJA temperature obtained from the Berkeley Earth Surface Temperature (BEST; Cowtan and National Center for Atmospheric Research Staff, 2019) dataset over the period 1805–1972. For each nested regression model, a split calibration–verification procedure was repeated 1000 times using the bootstrap method to assess the robustness of the transfer function, and 1000 reconstructions were generated. The final reconstruction of each nest is the median of the 1000 realizations and is given with its 95 % bootstrap confidence interval. To quantify the reconstructed cooling within a context of climate variability at the time of the major eruptions under study, we filtered the NVOLC v2 series with a 31-year running mean (NVOLC_filt v2). For example, in the case of the year 1640, a background was calculated by averaging the window 1625–1639 and 1641–1655. The magnitude of the anomaly is then determined by subtracting this background from the temperature reconstructed in 1640.

Figure 4 illustrates the cooling induced by major volcanic eruptions of the 17th century. According to the NH JJA NVOLC v2 temperature reconstruction, the largest cooling induced by any eruption of the past 1500 years occurred after 1600 Huaynaputina in Peru (White et al., 2022) with a NH temperature reduction of -1.56 ∘C in 1601 relative to the 1961–1990 reference period, followed by a -1.48 ∘C reduction in 1816 (rank 2, Tambora), -1.42 ∘C in 1699 (rank 3, absent apparent volcanic activity), -1.40 ∘C in 536 (rank 4, unidentified source), -1.35 ∘C in 1453 (rank 5, unidentified source; Abbott et al., 2021), and -1.33 ∘C in 1109 (rank 6, the notable eruption cluster between 1108 and 1110, presumably Mount Asama, among others; Guillet et al., 2020). The assumedly largest volcanic eruption of the Common Era occurred in 1257 (Samalas; Lavigne et al., 2013; Guillet et al., 2017) and induced a cooling of -1.04 ∘C in 1259, ranking 16th in the NVOLC v2 reconstruction.

The years 1641 and 1643 stand out as the 13th and 17th coldest years of the last 1500 years with -1.13 and -1.02 ∘C anomalies, respectively. By contrast, cooling was less marked after the assumed 1636 Hekla (-0.15 ∘C; rank 628) and the 1646 Shiveluch (-0.36 ∘C; rank 348) eruptions.

To estimate the regional variability of summer cooling induced by the mid-17th-century eruptions, we used the climate field reconstruction of extratropical NH JJA temperatures developed by Guillet et al. (2017, 2020) spanning the last 1500 years. The target field (predictand) used for the reconstruction is the BEST JJA gridded temperature dataset (1∘×1∘ lat–long grid). The NH was divided into 11 regions based on the spatial distribution of the 25 chronologies and their correlations. Performance and temporal stability of reconstruction models was assessed for each grid point with a split-sample calibration and verification approach repeated 1000 times using the bootstrap method.

Figure 5 shows JJA-gridded reconstructions for some of the largest CE eruptions (Fig. 5a) and all years between 1638 and 1643 (Fig. 5b). As with other eruptions, we observe widespread yet heterogeneous NH summer cooling, above all in 1641. The most affected regions are the western US (up to -2 ∘C), Scandinavia (-1.5 ∘C), Siberia (–1 ∘C), Central Asia (-1 ∘C), and, to a lesser extent, western Europe (-0.4 ∘C). Whereas Central Asia, Scandinavia, and the western US experienced cooling until 1643, summer temperatures recovered more rapidly in the other regions.

The European summer temperature reconstruction uses 12 multi-centennial TRW (i.e., Forfjorddalen, French Alps, Tatra Mountains), MXD (Jämtland, Lauenen, Lötschental, Nscan, Pyrenees, Torneträsk, Tyrol), and stable oxygen isotope (Angoulême, Fontainebleau) chronologies covering western Europe and Scandinavia.

A PCR approach (Stoffel et al., 2015; Guillet et al., 2017) was applied to reconstruct summer temperature anomalies from these chronologies. Regression models were calibrated on JJA temperatures obtained from the E-OBS pre-1950 and v22.0e gridded datasets (0.25∘×0.25∘ lat–long) over the period 1920–2010. For each regression model, a split calibration–verification procedure was repeated 1000 times using a bootstrap method to assess the robustness of the transfer function and to generate 1000 reconstructions. The final reconstructions for each grid point comprised the median of 1000 realizations and covered the period 1500–2010. For each reconstruction, we calculated coefficient of determination (R2 for the calibration period) and squared correlation coefficient (r2 for the verification period), RE (reduction of error), and CE (coefficient of efficiency) statistics. Anomalies are given with respect to a moving 31-year average calculated around the years given in Fig. 6. We compare our reconstruction using 12 TRW, MXD, and isotope chronologies (Fig. 6a) with the Luterbacher et al. (2004) dataset in Fig. 6b. For the domain and period of interest, the latter is based on ordinal-scale climate indices from Germany, Switzerland, and the Low Countries as well as TRW records from the Lofoten Islands for JJA temperatures, at a spatial resolution of 0.5∘×0.5∘ lat–long.

Our gridded summer temperature reconstruction allows analyses of warm and cold spells that have affected Europe. Although not comparable to the cooling observed in Europe after the 1600 Huaynaputina (White et al., 2022; not shown) eruption (with -1.87 ∘C in 1601; rank 1), 1641 stands among the coldest summer seasons of the last 500 years (-1.19 ∘C; rank 16), with a cooling comparable to that observed after the Tambora eruption (-1.23 ∘C in 1816; rank 14). Notably, cooling is most obvious in Scandinavia where summer temperatures were more than -2 ∘C below average (compared to the 31-year moving average around the event) but much less marked in the Alpine and Mediterranean domains. The cooling becomes much more generalized across most of Europe in 1642 (-0.47 ∘C; rank 105) and in 1643 (-1.01 ∘C; rank 27). While it is tempting to attribute the strong, yet localized, cooling observed in Scandinavia in 1641 as well as the more widespread, but less marked, cooling in 1642 and 1643 to the eruptions of Komaga-take and Mount Parker in July 1640 and January 1641, respectively, at least some cooling may be the result of natural climate variability. Indeed, cold spells like those observed in the early 1640s and affecting most of Europe were frequent in the 17th century and occasionally occurred absent any known major volcanic activity, for instance in 1608, 1609, or 1633 (-1.57 ∘C; rank 4 for the latter).

Whereas both spatial reconstructions presented in Fig. 6 point to similarities in the overall tendency of temperatures in different years, they also point to regional departures and – on some occasions – even larger differences. Part of these differences can be explained by the much larger dataset used in Fig. 6a, but some differences may also be related to the choice of proxies, their spatial representativeness during particular years, the method applied with their the underlying assumptions, the length of the calibration period, and different choices in terms of instrumental data to derive the statistical connections.

Western Europe of the early to mid-17th century was dominated by the Holy Roman Empire, which roughly included the territory of modern-day Austria, Belgium, Czechia, Denmark, Germany, northern Italy, Luxembourg, westernmost Poland, the Netherlands, Slovenia, and Switzerland. The Empire was both religiously and politically fragmented and was riven by the Thirty Years' War (1618–1648), occurring predominantly within its own territory. The war involved multiple European powers with the combatants often divided between the different religious denominations – Catholicism and Protestantism – though the centrality of the role of religion can be questioned (e.g., Wilson, 2008) and related exceptions existed, not least Catholic France, which fought on the “Protestant” side for strategic reasons. What is certain is that the spread of diseases, the rise of poverty, lack of resources, quartering of soldiers, general violence, and looting also cost many lives and created immeasurable hardships for populations within the Holy Roman Empire over these 3 decades. In 1620, the Holy Roman Empire had about 16 million inhabitants; around 1650 it had only 10 million (Gotthard, 2016), though there are considerable difficulties in accurately estimating population loss for this period (see the review by Wilson (2009), who states that total deaths in the empire may have reached 8 million). The war was in its final phase in the 1640s and, having already been suffering from the prolonged conflict, disease and hunger were likely enhanced by the severe weather of the early 1640s.

Contemporary chroniclers detail harsh weather in the early 1640s and numerous eyewitness reports underline the cold and wet conditions prevailing across the region. Both spring and summer 1640 are reported as being wet and cold, with many rivers flooding. For the territory of modern-day Switzerland, March temperatures sank to such a low that smaller water bodies were covered with ice and snow on the Swiss Plateau into April (HS10). Grain harvest on the Swiss Plateau was also poor (Pfister, 1984). The following winter 1640/41 was long and severe, followed by a rather wet and cool spring. In the area of present Switzerland, the winter 1640/41 saw heavy snow. On the Swiss plateau, snowfall occurred in late March and April (HS11), followed by a late frost in May (HS12). Summer 1641 was cold across Europe, ice formed, for instance, several times in southern Germany (Middle Franconia), and especially on 25 July 1641 (4 August 1641) (HS13, 14), and summer was reported as anything but sultry in Europe even during the dog days. In Krupka (today's Czechia), frost was noted on 10 August 1641 – and cucumbers and cabbages froze (Brázdil et al., 2004).

At Frankfurt an der Oder, the harvest was affected by incessantly rainy, cold weather, and thunderstorms (HS15), and frost in mid-September 1641 destroyed wine in southern Germany (HS16): “The wine of 1641 was sour …, grapes did not grow at all during the dog days, [and] before harvest they froze on the vines” (HS17). Various independent sources confirm the presence of strong hoarfrost around mid-September 1641 in Germany. In the area of present Switzerland, the harvests were poor and late as well (Pfister, 1984).

In May 1642, much of present-day Germany saw a cold spell: in Bavaria, “the coldest winds and hoarfrost” continued until mid-May, damaging grain and fruit trees greatly (HS18-19). Late snowfall and hard frosts are reported for May 1642 in Krupka, causing damage to the vineyards (Brázdil et al., 2004). hoarfrost and snow are again reported for several days in southwestern Germany (HS20) and Krupka (Brázdil et al., 2004) in June 1642. By contrast, a report from Diez (Lahn) in today's Rhineland-Palatinate noted that the summer 1642 was calm and the people in Diez harvested exceptionally good crops (HS21).

The Baden wine chronicles allow a comparison of the quality of the wines produced each year during the 17th century. The 1630s and 1640s clearly appear to be very harmful for wines. Of the 1630s, only the years 1631 and 1638 produced “good” wine, while the others produced “bad” or “sour” wine. Between 1640 and 1650, just 4 years (i.e., 1644–1647) provided “good” wine. For 1641, the Baden chroniclers indicate that the vine suffered late frost in May and then early frost in September (Garnier, 2010).

In neighboring France, the grape harvest between 1640 and 1643 began a full month later than usual and wheat prices surged from 12 to 20 pounds (per setier, that is about 7 L) between 1641 and 1642 (Baulant, 1968), indicating poor cereal harvests, especially also in eastern France. A source from Dijon described summer 1640 as chilly, resulting in a grape harvest that was comparably late. Poor temperatures also resulted in high grain prices on the Île de France. The year of 1641 saw considerable precipitation during winter and rather low summer temperatures. In 1642, the grape harvest was again late in Dijon and very late around Paris, while grain prices increased rapidly in the Île de France, probably the result of cold conditions prevailing in 1642 (Parker, 2013).

Grape harvest dates in Besançon (France) are intermittent in the 1630s and 1640s, only available for 1631, 1633, 1635, 1638–1642, 1646, and 1649. In these years, harvests occurred late between 1 and 15 October (Garnier et al., 2011). Years without harvest dates are not necessarily explained by meteorological factors. Most correspond with the arrival of the Swedish and French armies in Franche-Comté (1632, 1641, 1644, 1639, 1644) or can be related to the quartering of Swiss mercenaries (1634), damage to vines by “German soldiers” of the Duke of Lorraine (1637, 1647), and the plague (1630 and 1635). Only 3 years without harvest (1636, 1648, and 1650) are attributed directly to meteorological conditions by the city clerks: more exactly to frequent rainfall episodes between June and October. Abnormally rainy conditions were obviously sufficient to justify pro serenitate processions during which people prayed for cessation of the rain (Garnier, 2010).

Paris has excellent climate observations between 1630 and 1640 in municipal acts. Parisian aldermen (échevins) met almost weekly to deliberate on city affairs. Among them, climatic conditions, more particularly extreme events, were systematically discussed. This is easy to understand since droughts, floods, or even freezing of the Seine had an almost immediate impact (in 1 or 2 d) on prices in Parisian markets. Indeed, until the mid-19th century and the arrival of the railroad, bulk goods such as wheat and wood used by city dwellers for heating and cooking were mainly transported by waterways (Garnier, 2015). The municipal acts also provide other indicators for droughts, mentioning the drying up of the water sources and public fountains supplying the main hospital and conditions that prevented hydraulic mills from making flour. Between 1630 and 1650, Paris thus suffered five very severe droughts (some long like those of 1631–1632 and 1646–1647), but the drought of 1639 was undoubtedly the most dramatic of the 17th century (Garnier, 2019). Between 1630 and 1650, freezing of the Seine affected Paris six times (1635, 1636, 1637, 1644, 1645, and 1646). In the latter case, it is likely that the freezing was favored by the drought which preceded it in summer and the autumn, so that when the cold arrived, low water levels facilitated ice formation that lasted nearly 3 weeks.

In the 17th century, most of modern-day Sweden, Finland, Estonia, and northern Latvia belonged to the Swedish Realm, a “great power” of its time. Like in the Holy Roman Empire, contemporary documents point to severe weather and resulting agricultural hardship in the early 1640s. Estonian and Livonian sources report unusual heat and drought in summer 1640 to a degree that grain crops were damaged (Tarand et al., 2013), whereas the following two summers were cold and rainy (Soom, 1940). Grain ripening was badly delayed and the first autumn frosts in early September 1641 destroyed a great part of the harvest. The winters 1640–1641 and 1641–1642 were notably cold (Tarand and Nordli, 2001). Crop failures with consequent food shortage troubled Estonia in 1641 and 1642 but did not likely escalate into an acute subsistence crisis (Seppel, 2008). In the territory of present-day Sweden, bad crop failure was reported in 1641, especially in northern Sweden (Leijonhufvud, 2001). In Stockholm, at the heart of the realm, grain prices were 13 % to 20 % higher in 1641–1643 than the 1630s average (Edvinsson and Söderberg, 2010). Southern Sweden suffered a “violent fever or burning disease epidemic [likely typhoid fever and typhus, which are hunger-related diseases], which killed both rich and poor” especially in late spring 1642 (HS22).

Within the Swedish Realm, the area of modern-day Finland was likely the most affected by the weather anomalies of the 1640s. Already in 1632–1635 the region suffered from local poor harvests, and the year 1633 witnessed a country-wide crop failure (Johanson, 1924). In August 1638, a week-long frost killed the crops and allegedly “hand-deep” snow covered the ground (Johanson, 1924). Like in Estonia, cold conditions in summer 1641 delayed the ripening of grains, and frosts in early September killed the grain (Mårtensson, 1952). Grain tithes from western Finland indicate that the 1641 harvest was half that of normal (Huhtamaa and Helama, 2017); harvests remained poor in the following year, with yields only reaching approximately two-thirds of the mean (Huhtamaa and Helama, 2017). Unlike in Sweden and Estonia, the crop failure in 1641 triggered severe hunger in Finland. The economic consequences of these harvest losses can be seen in many regions, with every fifth or fourth farmstead marked as “deserted” in the administrative records (Huhtamaa et al., 2021). The high percentage and spatial extent of abandoned and acutely impoverished farmsteads in the mid-1640s point to a severe agricultural crisis, similar to the early 1600s when the eruption of Huaynaputina in 1600 (White et al., 2022) triggered harvest failures, famine and farm desertion (Huhtamaa et al., 2021). Whereas in the Swedish Realm the year 1640 was hot and dry, unusual cold and constant rain ruined the hay in Iceland in 1640, such that farmers resorted to dried fish as cattle fodder (Ogilvie, 1984).

In Ireland, winter 1641–1642 was particularly severe. Numerous contemporary documents describe the severity of cold. Particularly important are the approximately 8000 surviving statements of mainly Protestant interviewees (or “deponents”) recounting their experiences during the 1641 Rebellion, as attested to British government authorities, beginning with the work of the Commission for Distressed Subjects in December 1641. In these documents we hear survivors (often widows from Ulster and Leinster) describing “cold snowy weather” and “the extremity of the winter” (HS23). This unusual cold exacerbated the consequences of the rebelling Gaelic and “Old English” Catholics, who killed or turned Protestant residents out of their residences, in addition to their stripping and forced expulsion (Fig. 7). Thus, Katherin Cooke of Armagh recalled “such frost [and] snow that the deponent[']s children and divers [many] other children … present at that battle would[,] where they saw the warm blood of any fall on the ground[,] thread therein with their bare feet to keep them from freezing and starving such was the extremity of the weather” (HS24). Another deponent recalled: “Whereupon the deponent and his wife and 5 small children going away were stripped of all their clothes … one poor daughter of his seeing him … grieve for their general miseries in way of comforting said she was not cold nor would [she] cry although presently after … she died by that cold and want: and the first night this deponent and his wife creeping for shelter into a poor crate were glad to lie upon their children to keep in them heat and save them alive” (HS25). Other sources also provide visual depictions (often with propagandistic intent), such as that from James Cranford's, The Teares of Ireland (1642; Fig. 7), the caption from which reads: “English Protestantes striped naked and turned into the mountaines in the frost, and snowe, whereof many hundreds are perished to death, and many liynge dead in diches and savages [mainly Irish Catholics] upbraided them saynge now are ye wilde Irish as well as wee”. These conditions have, to date, been associated with explosive volcanism by Ludlow et al. (2013) and more specifically the eruption cluster starting in 1636 by Ludlow and Crampsie (2018, 2019).

The Compendium of Chinese Meteorological Records of the Last 3000 Years (hereafter the Compendium; Zhang, 2004) is the key reference for past weather and climate anomalies for the territory of present-day China. The Compendium is compiled from 7930 volumes of the state annals and local chronicles, supplemented by representative personal diaries. For the Ming and Qing Dynasties (1368–1911), records can be considered precise at the county level and reported place names have been linked to the corresponding modern county. Here, we adapt a simple sorting and counting approach to digitize documentary records (Gao et al., 2017) and to quantify the occurrence of droughts or other disastrous events in each county Gao et al., 2021a).

At the time of the 1640s eruptions, China was suffering one of the most disastrous droughts in its known history, called the Chongzhen Great Drought (Chen et al., 2020). Extraordinarily dry conditions (Fig. 8) resulted in severe societal and economic consequences during the final years of the Ming Dynasty (1638–1644), perhaps ultimately contributing to its collapse. Drought started to emerge in the provinces of Shandong, Henan, and Shaanxi along the Yellow River Basin in 1638, with 91 counties reporting drought. Reports rose to 109 counties in 1639 and 303 in 1640, almost 6 times the average annual drought count during the 17th century. Indeed, 1640 may have been the single driest year in the region since at least 1470 (Zhang and Lin, 1992). The drought expanded from the Yellow and Huai River valleys to the middle and low reaches of the Yangtze River in 1641 (Fig. 9). Local sources report that 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”. In the early months of 1641, unusually heavy snowfall was then recorded in southeastern China, while conditions remained very dry with infestations of locusts in northern China from where they propagated rapidly to the Yangtze River valley to form a large-scale locust plague (Song, 1992; Liu et al., 2018). The drought and locust plague resulted in an estimated reduction of 20 %–50 % in per capita grain production (Zheng et al., 2014), triggering a sharp increase in rice prices that intensified widespread famine (Fig. 8).

The severity of the drought was such that part of the Grand Canal in Shandong Province dried out completely, an incident without documented parallels over the past 3000 years. Water flow of the Yellow and Yangtze Rivers as well as lake levels in Zhejiang, Shaanxi, Shandong, Henan, Shanxi, and Hebei provinces dropped and eventually dried up in 1640–1641 (Zhang, 2004). Shen et al. (2007) have suggested that the intensity of the Chongzhen Great Drought (1638–16.44) was comparable to that of 1997, in which about 700 km of the lower Yellow River dried up, stopping outflow to the Yellow Sea for more than 300 d. The Chongzhen Great Drought was likely the outcome of natural climate variability potentially intensified by the Komaga-take and Mount Parker eruptions. Chen et al. (2020) ascribe this intensification to a volcanically induced decrease in land–sea thermal contrast and a negative soil moisture response combined with a weakening and eastward retreat of the West Pacific Subtropical High.

In many provinces, purported incidents of cannibalism occurred, and a widespread epidemic was likely triggered or at least compounded by malnutrition and weakened immune systems. By 1641, 29 counties within Shandong Province reported this pestilence. Conditions are described in the Gazettes of Tangye County: “No rain again in spring, a bucket [a unit used to measure grain, about 7.5 kg in the Ming Dynasty] of rice worth more than a thousand QIAN [a unit used to measure money, about 3.125 g in copper], cannibalism among family members, corpses buried for weeks were dug out for food. Pestilence raged in summer, no household escaped; … eight to nine out of ten people died of pestilence, hunger and robbery … it was indeed a situation that has never been seen or heard of” (Zhang, 2004, vol. 2, p. 1612).

Between 1641 and 1646, rebellion and warfare reached historical highs (Zhang et al., 2006; Zheng et al., 2014). The year 1645 ranks first in war frequency for the period 850–1911 (Fig. 8), with 16 wars in central China and 6 rebellion wars elsewhere in the country. With some parallels to Ireland (Ludlow and Crampsie, 2019), the counties in which uprisings occurred were most frequently those where extreme weather was most severe, suggesting that regional drought severity played a critical role in fueling or worsening peasant uprisings (Zheng et al., 2014).

Japan was likewise affected by a famine already underway at the time of the Komaga-take and Mount Parker eruptions in 1640 and 1641. The so-called Great Kan'ei Famine ( or Kan'ei kikin) started in 1640 and lasted until 1643, with up to 100 000 people and innumerous livestock dying from starvation and disease (Yamaguchi and Sasaki, 1971; Asao, 1975; Nagakura, 1982). The Kan'ei kikin was the first major famine to occur in Japan after the rapid urban and demographic growth of the late-16th and early-17th centuries that made many more Japanese dependent upon others for their food supply (Atwell, 1986). The eruption of Komaga-take on 31 July 1640 must thus be ascribed a compounding role to an already imminent crisis: heavy ashfall resulted in plant poisoning near the volcano and triggered local crop failure over 2 years. In 1641, the situation worsened as incessant rainfall and unusual cold propelled another poor harvest in the northeast (Yamamoto, 1989; Atwell, 2001). Winter 1641–1642 was cold as well and summer 1642 was cool, diminishing harvests once again in northeastern Japan, whereas floods and droughts destroyed harvests elsewhere in the country (Atwell, 2001).

In 1642, “the corpses of those who had starved to death filled the streets while the peasants, artisans, and merchants who begged for food were numerous. None of these people had any clothes [and could only] wrap straw and other matting around their bodies for protection” (Kanei nikki zōho in Asao, 1975). By June 1642, starving peasants started to abandon their lots to the degree that the governor ordered the abolition of tobacco plantations and strongly restricted alcohol production to favor food crops. Regardless, death tolls increased in winter 1642–1643 such that farmers performed infanticide of children below 7 years (Drixler, 2013) and lent out older children, often to pimps, in Fukushima Prefecture. The famine ceased only later in 1643 when crop yields rebounded.

The early decades of the 17th century mark the start of prolonged limited sunspot activity known as the Maunder Minimum (1645–1715). This period also marks a phase during which climatic conditions were worsening, with colder conditions also translating into advancing glaciers. In the Alps, contemporary sources report the disappearance of fields, farmsteads and entire villages as various glaciers reached their furthest extent in historical times (Nussbaumer and Zumbühl, 2012): the Mer de Glace (Mont Blanc, France), for instance, had one of its Little Ice Age maxima in 1644 after advancing by more than 200 m between 1641 and 1644. This points to a sustained, positive mass balance of the glacier, implying a multi-year succession of cold and wet conditions. By contrast, the Lower Grindelwald Glacier (Switzerland) reached its Little Ice Age maximum at some time between 1600–1606 and 1636–1645, with a second peak in 1855/6. The Upper Grindelwald Glacier had its largest extension between 1593 and 1630. Indeed, various glaciers across the European Alps reached valley floors at the beginning or during the first decades of the 17th century (Sigl et al., 2018).

In 1642, officials of the village of Chamonix (France) spoke of the daily progression of the glacier and evoked the avalanches of snow and ice that continued to carry away houses and livestock. The same was true for the Argentière glacier (France), which threatened to consume the church of the same name. The Glacier des Bois (France) exhibited the most impressive advance and in 1644 threatened to block and transform the Chamonix valley itself into a lake. Alpine mountain communities coped as best they could. Believing glaciers to be evil, they invoked divine protection, at first resorting to rogations to implore God's help. Faced with the ineffectiveness of this strategy, they then requested the intervention of the Bishop of Geneva, who in 1644 led a procession there. Thereafter, the ice slowly receded (Le Roy Ladurie, 2004).

The early decades of the 17th century were characterized by major warfare and widespread religious and dynastic disputes. The most widely known is the above-described Thirty Years' War (1618–1648), consisting of a series of battles centered in and around present-day Germany, resulting in one of the deadliest wars of Europe. While at first restricted to battles among Catholic and Protestant states of the Holy Roman Empire, fighting became part of a wider European struggle after 1635, with the Swedish Realm, Dutch Republic, and France on the one side and Spain and the Austrian Habsburgs (including large parts of the Holy Roman Empire) on the other. Estimates of military and civilian deaths range from 4.5 to 12 million, of which only 450 000 died in the fighting (Outram, 2002; Parker, 2008; Wilson, 2009). The vast majority, however, died from disease, starvation, or violence against civilians. Some sources suggest a mortality of up to 60 % of the population in certain areas of present-day Germany (Outram, 2002), while in the Palatinate and Württemberg, losses are estimated at between 70 % and 90 %, and in Alsace they average 60 % (Outram, 2002). These deaths must be considered the result of combined severe weather conditions (induced or at least often compounded by volcanism), epidemics, military violence, and famines. It is difficult to determine which was the most decisive factor, particularly given their likely interdependencies, but the scale of the combined impact is less in doubt. Evidence suggests that mortal sickness in military camps and peasant communities far exceeded death in action, perhaps in the proportion of 5 to 1 (Outram, 2002).

The entry of France into “open war” in 1635 brought a considerable increase in taxation, contributing to revolts almost every year for 10 years, mainly in the south and southwest of France. The movements of troops served as a vector for the plague which spread widely. Famous for his hostility to d'Artagnan and his musketeers in the novel by Alexandre Dumas, Cardinal Richelieu is better known to historians for having created in 1638 the heaviest regular tax, in the form of the “wintering place” (quartiers d'hiver) for troops. This obliged subjects of the King of France to provide food and shelter for soldiers during winter. Requisitions for food and equipment such as mills, ovens, and means of transport were also often carried out in a context of pillage and brutality. A further cause of discontent was the tailles, comprising all direct taxes. These, moreover, were paid almost entirely by the peasants, and the Thirty Years' War was thereby waged heavily at the expense of peasant income in France. In 1648, the share of tailles in the revenue of the royal state approached 62 %. Popular riots among peasants and modest city dwellers thus appear to be a response to fiscal and military aggression of absolutist authorities, and the socio-economic “efficacy” of shocks to agricultural production arising from inclement weather in this period and region must be understood as operating against this background.

The multi-faceted associations among climate, warfare and famine are exemplified by the case of Finland, being part of the Swedish Realm in the early 17th century. The Swedish Realm intervened in the Thirty Years' War in 1630 and from the late 1630s its army consisted increasingly of Swedish and Finnish conscripts as infantry was raised by conscripting every 10th man among the adult male peasant population (Jespersen, 2016). The Realm's constant need for soldiers increasingly impacted agricultural production, as call-ups were annual during wartime. In the case of Finland, conscriptions were particularly heavy with at least 13 000 men recruited between 1638 and 1645 (Lappalainen, 1987). This number is considerable as the area held only an estimated 27 500 peasant households at this time. Consequently, nearly every second farm was missing an adult male from its agricultural workforce. Furthermore, the Swedish recruitment system created an additional financial burden, because to avoid being shipped to the European battlefields, peasants started to hire men to take their place in the conscriptions (Villstrand, 2000). Yet even if hired men, adult sons, or farmhands were recruited preferentially, heads of farmsteads were increasingly conscripted with time and as the number of available men decreased. By 1643, nearly 60 % of all recruited men from southern and central Finland were the masters of their farmsteads (Lång, 2003). The resulting labor shortfall undoubtedly affected grain production quantity and quality. Even if battles were fought far from Finland, livelihoods were rendered more vulnerable to environmental hazards, like the extreme cold of 1641. Thus, the synergistic interaction of warfare and extreme cold was key to the early 1640s agricultural crisis in Finland (Huhtamaa et al., 2021), and it is arguably a matter of semantics (and/or disciplinary norms) as to whether both are considered “causal” or “contextual” in this sense. The poor weather also diminished yields in Estonia and the main agricultural areas in Sweden, but being located further south than Finland, they were less affected by the unseasonal summer weather (Edvinsson et al., 2009), while Estonia (and men from other newly gained territories) also escaped the onerous conscription experienced by Finnish peasants.

In Ireland, one of the most infamous rebellions against English rule erupted in October 1641 with Irish Catholics trying (unsuccessfully) to seize control of the English administration in Dublin but killing and displacing many Scots and English Protestant settlers in what evolved into the Irish Confederate Wars, 1641–1653 (also known as the Eleven Years' War, in Irish Cogadh na hAon-déag mBliana) (Ó Siochrú, 1999). This wider conflict was perhaps the most devastating in Irish history (Fig. 10), with direct deaths in conflict (including by infighting) considerably augmented by hunger, disease (including bubonic plague) and weather-related exposure (Dorney, 2014). Neither the total population of Ireland nor exact numbers of deaths between 1641 and 1653 are precisely known, but a finding of 15 %–20 % mortality countrywide has been credibly offered (Lenihan, 1997). The initial outbreak of rebellion in October 1641 was preceded by poor weather (plausibly compounded by the sequence of eruptions starting with Hekla in 1636), with consecutive years of poor harvests from 1639 to 1641 (Ohlmeyer, 1998; Gillespie, 1989). These climatic–agricultural pressures can, with care, be posited as one trigger for the rebellion's onset in 1641 (Ludlow and Crampsie, 2018), though a full understanding of the role of climate must ultimately grapple with its interdependencies with other economic, political, class, cultural, and religious factors.

Any hypotheses offered must also pay careful attention to the respective timing of events. As an example, consider the impacts of the 1641 Rebellion as described in great detail in documents known as the Depositions (https://1641.tcd.ie, last access: 10 October 2015), comprising more than 8000 (mainly Protestant) witness statements that make multiple dramatic references to extreme cold and severe weather in the 1641/42 winter (Sect. 6.2). With the rebellion beginning in October, these specific attestations will, however, refer to the months after the initial onset of rebellion and cannot therefore be strictly credited with a role as an initial trigger (though they may represent an evolution of inclement weather responsible for the aforementioned poor harvests). Nonetheless, comparing the geography of reported killings in the Depositions (Smyth, 2013) with these references to (post-rebellion-onset) extreme weather per Irish county, whilst (crucially) controlling for variable witness numbers per county, reveals a statistically significant positive correlation (Spearman ρ=0.596, p=0.003, n=32: Ludlow and Crampsie, 2019). One credible part-explanation is that regional variance in the severity of weather also influenced the scale and severity of violence once it was underway and particularly as the rebellion spread among Catholic underclasses and those disposed of lands that had become increasingly marginalized and vulnerable under the preceding decades of British rule. Such participants may have been more likely to engage in violence (and in violence of greater severity) if they were more severely affected by the weather conditions attested to in the Depositions.

Even if true, this observation tells only part of the story, though it is one largely absent from understandings of the rebellion until recently. The geography of violence during the rebellion must certainly reflect many additional factors, some more immediate to the time of the rebellion and others longer evolving (e.g., Edwards, 2013). Of the latter, mounting indebtedness among Catholic landholders in the decades preceding the rebellion is important. To default on a loan frequently meant the forfeiture of one's lands, and loans were thereby made by wealthy creditors seeking to increase their land holdings in Ireland during this pivotal era of expanding colonization (“plantation”) by English and Scottish settlers (Ohlmeyer, 1998). Unsurprisingly, therefore, accounts from the Depositions outline how some Catholics sought to destroy loan agreements (bonds) amid the chaos of 1641/42 (Canny, 1995, 2001). But it is a mistake to consider such factors unrelated to the climate (and environment). While the environment has often been oversimplistically (e.g., mono-causally) attributed a role in human affairs, its effective omission from many (perhaps most) histories similarly promotes more simplistic (or incomplete) understandings. Human societies cannot exist (and their histories did not occur) independent of the environment and its variability. Indeed, this inseparability or entanglement can be illustrated by returning to the 1641 Rebellion. Ludlow and Crampsie (2019) have shown a repeated association between harvest crises (of which there were many; Gillespie, 1989, 2018) and poor weather in the decades preceding the rebellion. They have also shown an association between these crises and borrowing, needed to cover harvest-related income shortfalls in the decades running up to 1641 (Ohlmeyer, 1998). Extreme weather was thus also entangled in the longer-term processes of mounting indebtedness, consequent loss of hereditary lands, and growing grievance of Catholics towards Protestant settlers and elites that contributed to the rebellion.

Adverse weather had already affected Chinese agricultural production in the years leading up to the eruptions of the early 1640s, with some of the wealthiest regions suffering food shortage and disease by the late 1630s. This was also at least partly the result of the Chongzhen Great Drought, featuring the 5 worst years of consecutive drought (1637–1641) in known Chinese history (Lamb, 1970; Atwell, 2001; Zheng et al., 2014; Chen et al., 2020). Food costs soared and lawlessness increased substantially. In April 1644, Beijing fell to a rebel army invading the capital from the economically devastated northwest (Liu et al., 2018). Only 6 weeks later the city fell again, this time to Manchu invaders, marking the start of a new era in Chinese political history (Atwell, 1986).

Likewise, a succession of very cool summers in northern Japan and floods and droughts elsewhere in the country drastically reduced grain yields in the late 1630s and early 1640s, resulting in the Great Kan'ei Famine (Endō, 1982; Atwell, 1986). Notably, the famine occurred in a period when the center of Japanese rice cultivation was in the process of shifting from the south and west to the north and east, driven by the desire of lords in these regions to specialize in food production to exploit growing urban populations but thereby placing the crop at greater risk of cold damage (Arakawa, 1974; Tani, 1978) and at times propelling grain prices to unprecedented levels. The famine was enhanced by the large-scale internal migration after the Shimabara Rebellion (, Shimabara no ran, 1637–1638) as well as the extreme impoverishment of low-ranking members of the samurai class. Worse, a rinderpest epizootic broke out in Kyushu in 1638 and could not be contained so that mass cattle mortality occurred across western Japan in 1640, with the consequent scarcity of working animals further reducing yields (Atwell, 1986).

Extreme weather and poor harvests thus certainly exacerbated the disastrous situations in China and Japan. As Atwell (1986) and Zheng et al. (2014) have argued, however, a deeper understanding of the terrible “famines” of the late 1630s and early 1640s and the role of weather therein must appreciate the larger economic and social problems that had been festering in the two countries for some time, including excessive public expenditure, rapid urban growth, and intense economic competition. These factors proved notably socially and politically disruptive once agricultural productivity could not be maintained.