Climatic and societal impacts of a volcanic double event at the dawn of the Middle Ages
Volcanic activity in and around the year 536 CE led to severe cold and famine, and has been speculatively linked to large-scale societal crises around the globe. Using a coupled aerosol-climate model, with eruption parameters constrained by recently re-dated ice core records and historical observations of the aerosol cloud, we reconstruct the radiative forcing resulting from a sequence of two major volcanic eruptions in 536 and 540 CE. We estimate that the decadal-scale Northern Hemisphere (NH) extra-tropical radiative forcing from this volcanic “double event” was larger than that of any period in existing reconstructions of the last 1200 years. Earth system model simulations including the volcanic forcing show peak NH mean temperature anomalies reaching more than −2 °C, and show agreement with the limited number of available maximum latewood density temperature reconstructions. The simulations also produce decadal-scale anomalies of Arctic sea ice. The simulated cooling is interpreted in terms of probable impacts on agricultural production in Europe, and implies a high likelihood of multiple years of significant decreases in crop production across Scandinavia, supporting the theory of a connection between the 536 and 540 eruptions and evidence of societal crisis dated to the mid-6th century.
In 536 CE, observers documented a mysterious cloud which dimmed the light of the sun for at least a year (Stothers and Rampino 1983; Stothers 1984). Tree rings suggest a sudden onset of a decadal-scale exceptional cooling in 536 CE in Northern Europe (Grudd 2008; Larsen et al. 2008; Esper et al. 2012), Mongolia (D’Arrigo et al. 2001), and Western North America (Salzer and Hughes 2007; Salzer et al. 2013), and in the Northern Hemisphere (NH) average (Stoffel et al. 2015). The 536 mystery cloud was linked to crop failures and famines by ancient scholars (Stothers 1999), and has been speculatively linked to a number of major societal crises throughout the NH, including the European outbreak of the plague of Justinian in 541 CE (Baillie 1999; Stothers 1999; Keys 2000).
The documented descriptions of the 536 CE mystery cloud are consistent with the optical characteristics of stratospheric sulfate aerosol resulting from volcanic eruptions (Robock 2000). A volcanic origin of the 536 mystery cloud was only relatively recently confirmed by ice core records. Larsen et al. (2008) reported the presence of two sulfate signals in Greenland ice cores separated by approximately four years around the year 536 CE, and linked the second sulfate peak—the signature of a likely tropical eruption—to the 536 mystery cloud. Updated ice core timescales, proposed initially based on matching of ice core and tree ring volcanic signals (Baillie 1994; Baillie 2008; Baillie and McAneney 2015) and recently confirmed through matching of signatures of cosmogenic isotopes in ice cores and tree rings in the 8th century (Sigl et al. 2015), place the two signals at 536 and 540 CE. This double peak structure of the ice core records is in qualitative agreement with the temporal character of many tree ring-based temperature reconstructions for this time period (Baillie 1994), with cooling maxima in 536 and again 4 or 5 years later. While more complex eruption sequences are possible, the simplest plausible scenario, assumed hereafter, is one of a volcanic “double event”, with two major eruptions from unknown locations in or around the years 536 and 540 CE.
While the updated ice core timescales clarify the timing of the 536 and 540 volcanic events, estimates of the magnitudes of climatic impacts stemming from the eruptions differ depending on the source of evidence. Multiple NH tree ring reconstructions suggest that the year 536 was the coldest single year of at least the past 2000 years, and that 536–545 was the coldest decade of the same period (e.g., Larsen et al. 2008; Sigl et al. 2015). However, current estimates from ice cores of the global volcanic radiative forcing over the same interval ranks the 536 and 540 eruptions as only the 18th and 5th strongest events respectively (Sigl et al. 2015). At face value, the magnitudes of the 536 and 540 eruptions derived directly from ice core data appear inconsistent with the exceptional cooling implied by tree ring reconstructions, and therefore do not necessarily support the theory of widespread societal crisis popularly connected to volcanic activity at the time.
To address these apparent inconsistencies in the climatic and societal impacts of the 536 and 540 eruptions implied by different records, we have reconstructed the radiative forcing of the two events with a coupled aerosol-climate model, constrained by ice cores and historical accounts. The reconstructed radiative forcing is then used in 15-year-long simulations with an Earth system model, to estimate global and regional climate anomalies resulting from the eruptions. Finally, the climate model results are used to assess the likely societal impacts in Europe of the volcanic double event.
2 Materials and methods
2.1 Volcanic radiative forcing construction
Available total sulfate flux (kg/km2) derived for the 536 and 540 events from Antarctic and Greenland ice cores (Traufetter et al. 2004; Larsen et al. 2008; Plummer et al. 2012; Sigl et al. 2013) are listed in Table S1. To account for potential sampling bias from the small number of available ice cores available for this time period, and the spatial variability of sulfate deposition to the surface (e.g., Sigl et al. 2014), Antarctic and Greenland averages are corrected using scaling factors derived from MAECHAM5-HAM simulations (Supplementary Methods, (Niemeier et al. 2009; Toohey et al. 2011; Toohey et al. 2013)). The model-based scaling factors for both Greenland and Antarctic based on the four ice core locations in Table S1 are close to unity, with values of 1.01 and 0.93 for Greenland and Antarctica respectively. For the single Antarctic ice core measurement of the 536 event, the scaling factor is 0.73.
The ratios of Greenland-to-Antarctic sulfate flux for the 536 and 540 events were compared to those of measured historical eruptions (Table S2) and MAECHAM5-HAM simulations (Table S3) to constrain the latitudes of the eruptive sources. Maximum stratospheric sulfate aerosol loading was deduced based on the method of Gao et al. (2007), using scaling factors of 1*109 km2 for tropical and 0.57 *109 km2 for high latitude eruptions. Total SO2 injection by the eruptions is based on scaling the estimated global sulfate aerosol load by the mass ratios of sulfate aerosol to sulfate (MSO4/MAer = 0.75, assuming a 25 % water mass content of the sulfate aerosol) and the molecular weight ratio of SO2 to SO4 (MWSO2/MWSO4 = 0.66).
MAECHAM5-HAM simulations were used to construct volcanic aerosol forcing timeseries based on the estimated eruption magnitudes and locations deduced from ice core data. Small ensembles of “candidate” simulations were performed (Table S3) and individual members with closest agreement to ice core-derived hemispheric sulfate deposition were selected and concatenated into a single volcanic radiative forcing timeseries composed of zonal mean aerosol optical depth (AOD) and aerosol effective radius.
2.2 Climate simulations
The climate impacts to be expected from the 536 and 540 eruptions were estimated through ensemble simulations with the Max Planck Institute Earth System Model (MPI-ESM, (Giorgetta et al. 2013)) using the reconstructed radiative forcing timeseries as prescribed forcing (as in Timmreck et al. 2010). Initial conditions for 12 ensemble members were defined by the climate state at unique points of time in a 1000-year-long pre-industrial control run performed as part of the 5th phase of the Climate Model Intercomparison project (CMIP5), and were selected so as to span a wide range of climate states in regards to NH extratropical (30–90°N) mean temperature (Fig. S2). In addition, care was taken to ensure no bias in the El Niño Southern Oscillation (ENSO) state of the ensemble mean of the initial conditions, with equal numbers of El Niño (Niño 3.4 ≈ 1) and La Niña (Niño 3.4 ≈ −1) states selected (Fig. S2).
3.1 Radiative forcing
Categorization of eruptions as tropical or extratropical is necessary to estimate their stratospheric SO2 injection and radiative forcing from ice core sulfate records, and is based on the presence or lack, respectively, of sulfate signals in ice cores from both polar regions. Historical tropical eruptions typically lead to sulfate transport to both hemispheres, and produce Greenland-to-Antarctic sulfate flux ratios of between 1:2 and 2:1 (Gao et al. 2007). In contrast, sulfate from extratropical NH eruptions is typically found in Greenland but not Antarctica (Fig. S1). Based on available sulfate flux records (Tables S1, S4), the 540 CE event can be placed in the tropical eruption category, with a 2:1 Greenland-to-Antarctic sulfate flux ratio, similar to that of the eruptions of Huaynaputina (1600, 16°S) and Cosiguina (1835, 13°N). The 536 event has a strong signal in Greenland ice cores, and while a corresponding signal is undetectable in most Antarctic ice cores, a small signal was reported in the high resolution West Antarctic Ice Sheet (WAIS) Divide ice core record (Sigl et al. 2013), suggesting some degree of cross-equator stratospheric transport. The resulting Greenland-to-Antarctic sulfate flux ratio of more than 10:1 can be safely assumed to be representative of a mid or high latitude NH eruption, which is consistent with detection of tephra in a Greenland ice core consistent in chemical composition to NH volcanoes (Sigl et al. 2015).
Using published volcanic sulfate deposition data (Table S1), and taking the 536 and 540 events as extratropical and tropical eruptions, respectively, and applying established techniques (Gao et al. 2007) for converting sulfate flux to estimates of volcanic stratospheric SO2 injection (Table S4), we estimate global stratospheric SO2 injections of approximately 30 Tg and 50 Tg for the 536 and 540 CE events, respectively. Applying the same procedure and sample of ice cores to estimate the SO2 injection by the 1815 eruption of Tambora results in an estimated SO2 injection of 50 Tg, which agrees well with other estimates (Self et al. 1984; Gao et al. 2007).
Simulations with the coupled aerosol-general circulation model MAECHAM5-HAM were performed to construct a timeseries of radiative forcing and aerosol properties. First, a series of sensitivity studies were performed to constrain further eruption parameters such as latitude, season and injection height, in order to produce best agreement between simulations and ice core and historical records of the eruptions.
To select eruption latitudes producing hemispheric aerosol partitioning consistent with the Greenland-to-Antarctic sulfate flux ratios derived from the ice cores, MAECHAM5-HAM simulations were performed of eruptions at a set of latitudes spanning 6°S to 56°N, with eruptions in January and July. Simulated Greenland-to-Antarctic sulfate flux ratios produced largest overlap with the ice core-derived 2:1 deposition ratio of the 540 event for simulated eruptions at 15°N (Fig. S1). The roughly 10:1 Greenland-to-Antarctic sulfate flux ratio of the 536 event is similar to MAECHAM5-HAM simulations of mid and high latitude eruptions. MAECHAM5-HAM simulations produce overlap with the observed Greenland-to-Antarctic sulfate deposition ratio for simulated eruptions at both 46° and 56°N, and show very little difference in the ensemble mean radiative forcing for eruptions at these two latitudes.
In terms of global mean, annual mean AOD, the reconstructed 536 and 540 events are comparable to the strongest volcanic eruptions in a reconstruction of volcanic forcing over the past 1200 years (Crowley and Unterman 2013), with magnitudes that would rank 7th and 3rd within this list, respectively. Following the Greenland-to-Antarctica sulfate flux ratios, the simulated AODs for both eruptions are stronger in the NH than in the Southern Hemisphere (SH). The aerosol load for the simulated 536 event is largely constrained to the NH, with the largest AOD found north of 30°N, similar to previous model simulations (Oman et al. 2006) and satellite observations (Bourassa et al. 2010) of other high-latitude eruptions. It is also consistent with the distribution of historical observations of diminished solar intensity in 536 CE, with the most reliably located accounts originating from Rome (42°N) and Constantinople (41°N), but a noted lack of direct observations documented by scholars residing at lower latitudes (Arjava 2005). The structure of the reconstructed forcing for the 540 eruption is qualitatively similar to that of the tropical eruption of Huaynaputina in 1600 (Fig. S2). The exceptional property of the volcanic forcing for the 536/540 double event is therefore not the magnitude or latitudinal structure of either eruption individually, but rather the temporal proximity of two events with strong forcing in the NH mid and high latitudes. In the decadal cumulative global mean AOD, the 536–545 decade would rank 3rd in the reconstruction of Crowley and Unterman (2013), while in the NH extratropical (30–90°N) mean, the reconstructed 536–545 decadal AOD would rank 1st, with a magnitude approximately 1.5 times larger than the combined impact of the unknown eruption of 1809 and Tambora in 1815 (Fig. 1d), the strongest decadal AOD of the reconstruction of Crowley and Unterman (2013).
3.2 Climate response
3.3 Societal impacts
The 536–545 cold phase has been linked to evidence of societal changes around the globe. Well-dated evidence includes documentation of food shortages or famines in the Mediterranean (Stothers 1984; Rampino et al. 1988; Stothers 1999), Ireland (Baillie 1994), and China (Weisburd 1985), and changes in building frequency in Germany and Ireland (Baillie 1991). Wide ranging societal changes inferred from archaeological and palynological data, and by their nature only roughly constrained to the 6th century, are often speculatively linked to the 536–545 cold phase. For example, the event has been connected to an apparent collapse of Scandinavian societies, evidenced by (Fig. 4a): abandoned settlements (Solberg 2000; Gräslund and Price 2012; Löwenborg 2012), findings of sacrificial gold offerings (Axboe 2001) and evidence of sudden decreases in agriculture (Tvauri 2014; Pedersen and Widgren 2011).
Using this metric, the simulated temperature anomalies for 536 CE imply greatest absolute agricultural impact in Europe in high altitude regions (e.g., over the Alps) and at the northernmost margin of nominal cultivation (Fig. 5b). The decrease in the CSIT is especially pronounced in the Baltic sea region, with absolute decreases of 0.3–0.4. In many Scandinavian locations, these decreases amount to a complete diminishment of CSIT, implying severe crop failure. Decreases in simulated CSIT after the 540 eruption (not shown) is roughly similar to that of 536, and due to the temporal proximity of the two strong volcanic events, and the persistence of each event through two summers, such crop failure is likely to have occurred for multiple years within 536–545: in the ensemble average, CSIT anomalies exceed −0.2 for 3–5 years in the Baltic sea region during the 536–545 period (Fig. 5c).
4 Discussion and conclusions
Ice core data combined with historical evidence indicates that the mid-6th century was marked by multiple major volcanic eruptions. Using eruption parameters constrained by ice core and documentary evidence, and the MAECHAM5-HAM aerosol-climate model, we have reconstructed the volcanic radiative forcing for a plausible scenario of major eruptions in 536 and 540 CE. Consistency between the ice core records, MXD tree ring temperature reconstructions, and historical observations of the 536 event can be achieved under the scenario of a high latitude NH eruption producing a high altitude sulfur injection, consistent with contemporary observations of major tropical eruptions, but as yet not observed for extratropical eruptions. This result suggests that the climate impact of extratropical eruptions may not always be as minor compared to tropical eruptions as deduced from prior studies (e.g., Schneider et al. 2009). Best agreement with ice core records of the ca. 540 CE eruption was achieved from simulation of a eruption at 15°N, consistent with the location of Ilopango, one suggested source of a major eruption at this time (Dull et al. 2010). Based on the ice core data, our simulations suggest very strong radiative forcing in the NH high latitudes resulting from both eruptions, such that the decadal average radiative forcing in the NH extratropics is 50 % larger than the largest such forcing of the past 1200 years. These conclusions are to some degree dependent on the assumption of two single eruptions: if the 536 sulfate peak was due to multiple NH eruptions (a possibility suggested by tephra analysis of a Greenland ice core (Sigl et al. 2015)), the resulting radiative forcing could potentially be weaker than that estimated here. Constraining the eruption sequence at this time, and generally the uncertainties related to estimating sulfur injection and radiative forcing from ice core sulfate signals, are topics in need of further study.
Earth system model simulations using the reconstructed forcing show good agreement with tree ring reconstructions from Northern Scandinavia, especially in regards to the cooling after the 536 CE event. While persistent decadal-scale cooling after 540 CE—apparent in some tree ring width records (e.g., D’Arrigo et al. 2001; Larsen et al. 2008)—is not reproduced by the model at the locations of the Scandinavian tree ring samples, decadal-scale anomalies of Arctic sea ice are produced, suggesting a possible mechanism for longer term climate response. If sea ice growth is an important mechanism in the prolongation of short-term volcanic radiative forcing into decadal scale climate responses (e.g., Schneider et al. 2009; Schleussner and Feulner 2013; Lehner et al. 2013), it may be that the characteristics of the 536/540 double event, which produced strong radiative forcing at NH high latitudes focused over a single decade, may have been especially effective at creating climate anomalies persisting well after the eruptions.
Finally, the simulated temperature anomalies are interpreted in terms of impact on agriculture, quantified through changes in growing degree days (GDD) and an index of cultivation suitability. We find that while temperature anomalies were likely similar across most of Europe, the direct impact of the eruptions on agricultural production in southern regions was likely minimal, consistent with documentary evidence from the time (Arjava 2005). In contrast, the simulations imply that marginal agricultural societies of Northern Europe most likely faced multiple years of crop failure within a single decade as a result of the two eruptions. It is clear that the widespread societal changes in the 6th century which mark the end of Antiquity and the beginning of the Middle Ages, deduced from documentary and archaeological evidence, are due to a complex set of causes, many of which unrelated to (or potentially indirect impacts of) volcanic activity. However, the modeling results shown here, incorporating estimates of volcanic radiative forcing derived directly from ice core records, support the theory of a direct role of the 536 and 540 eruptions on agricultural and societal changes in Northern Europe and Scandinavia.
This work was supported by the Research Council of Norway through its Centres of Excellence funding scheme to Centre of Earth Evolution and Dynamics (CEED) project number 223272, and by the Federal Ministry for Education and Research in Germany (BMBF) through the research program ‘MiKlip’ (FKZ:01LP130B). Computations were performed at the German Climate Computer Center (DKRZ). The authors thank Jan. Esper for providing the N-Scan data set, and two anonymous reviewers for their insightful comments.
- Axboe M (2001) Amulet pendants and a darkened sun. In: Magnus B (ed) Roman gold and the development of the early Germanic kingdoms, history and antiquities. Royal Academy of Letters, Stockholm, pp 119–136Google Scholar
- Baillie MGL (1999) Exodus to Arthur: catastrophic encounters with comets. BT Batsford Ltd, LondonGoogle Scholar
- D’Arrigo R, Frank D, Jacoby G, Pederson N (2001) Spatial response to Major Volcanic events in or about AD 536, 934 and 1258: frost rings and other dendrochronological evidence from Mongolia and Northern Siberia: comment on R. B. Stothers, ‘Volcanic dry fogs, climate cooling, and plague pandemics in Europe and the Middle East’. Clim. Change 49:239–246. doi:10.1023/A:1010727122905 Google Scholar
- Dull R, Southon JR, Kutterolf S, et al (2010) Did the TBJ Ilopango eruption cause the AD 536 event? AGU Fall Meet Abstr −1:2370.Google Scholar
- Keys D (2000) Catastrophe: An investigation into the origins of the modern world. Ballantine, New YorkGoogle Scholar
- Löwenborg D (2012) An Iron Age shock doctrine: did the AD 536-7 event trigger large-scale social changes in the Mälaren valley area? J Archaeol Anc Hist 4:1–29Google Scholar
- Niemeier U, Timmreck C, Graf H-F, et al. (2009) Initial fate of fine ash and sulfur from large volcanic eruptions. Atmos Chem Phys 9:9043–9057. doi:10.5194/acp-9-9043-2009
- Pedersen EA, Widgren M (2011) Agriculture in Sweden 800 BC–AD 1000. In: Myrdal J, Morell M (eds) The Agrarian History of Sweden: From 4000 BC to AD 2000. Academic Press, Nordic, pp. 46–71Google Scholar
- Self S, Rampino MR, Newton MS, Wolff JA (1984) Volcanological study of the great Tambora eruption of 1815. Geology 12:659–663. doi:10.1130/0091-7613(1984)12%3C659:VSOTGT%3E2.0.CO;2
- Solberg B (2000) Jernalderen i Norge: ca. 500 f. Kr.-1030 e. Kr. Oslo: Cappelen Akademisk ForlagGoogle Scholar
- Traufetter F, Oerter H, Fischer H, et al (2004) Spatio-temporal variability in volcanic sulphate deposition over the past 2 kyr in snow pits and firn cores from Amundsenisen, Antarctica. J Glaciol 50(168):137–146. doi:10.3189/172756504781830222
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.