The 852/3 CE eruption of Mount Churchill, Alaska, was one of the largest
first-millennium volcanic events, with a magnitude of 6.7 (VEI 6) and a
tephra volume of 39.4–61.9 km
The reconstructed climate forcing potential of the 852/3 CE Churchill eruption
is moderate compared with the eruption magnitude, but tree-ring-inferred
temperatures report a significant atmospheric cooling of 0.8
Precise comparisons of palaeoenvironmental records from peatlands across North America and Europe, facilitated by the presence of the WRAe isochron, reveal no consistent MCA signal. These findings contribute to the growing body of evidence that characterises the MCA hydroclimate as time-transgressive and heterogeneous rather than a well-defined climatic period. The presence of the WRAe isochron also demonstrates that no long-term (multidecadal) climatic or societal impacts from the 852/3 CE Churchill eruption were identified beyond areas proximal to the eruption. Historical evidence in Europe for subsistence crises demonstrate a degree of temporal correspondence on interannual timescales, but similar events were reported outside of the eruption period and were common in the 9th century. The 852/3 CE Churchill eruption exemplifies the difficulties of identifying and confirming volcanic impacts for a single eruption, even when the eruption has a small age uncertainty.
The 852/3 CE eruption of Mount Churchill in the Wrangell volcanic field, southeast Alaska, was one of the largest first-millennium volcanic events, with a roughly estimated eruptive volume of 47 km
Site and White River Ash east distribution map with thickness data for volume estimate.
The ash produced from this eruption caused considerable and long-lasting environmental disturbances in regions proximal to Mount Churchill. For example, the eruption has been linked with changes in vegetation that persisted for ca. 50–150 years in Yukon (Rainville, 2016), multi-centennial changes in peatland ecology in southeast Alaska (Payne and Blackford, 2008) and decreases in aquatic productivity lasting ca. 100 years in southwest Yukon (Bunbury and Gajewski, 2013). These spatial patterns in proximal environmental responses to the 852/3 CE Churchill eruption are diverse. The eruption and its environmental impacts are also suggested to have driven societal changes in the region (Kristensen et al., 2020), notably a decline in indigenous occupancy in the southern Yukon (Hare et al., 2004). In addition, the event may have triggered the southward migration of people, who brought their culture and Athapaskan language to the US Great Basin and the American southwest (Mullen, 2012). Several studies have therefore characterised the proximal impacts of this 852/3 CE Churchill eruption, but less is known about the wide-scale Northern Hemisphere (NH) or global impacts of this large eruption.
Several lines of evidence suggest that the 852/3 CE Churchill eruption
occurred in winter, including the stratigraphic context of the tephra in
proximal locations (West and Donaldson, 2000), the ash cloud trajectory
(Muhs and Budahn, 2006) and the timing of ash deposition in Greenland.
Cryptotephra from the eruption was identified in the NGRIP and NEEM-2011-S1
ice cores from northern Greenland in ice then dating to 847 CE based on the
Greenland Ice Core 2005 (GICC05 chronology; Coulter et al., 2012; Jensen et
al., 2014). Based on the revised NS1-2011 chronology (Sigl et al., 2015),
the event is now dated to the winter of 852/3 CE (Fig. 2) and is likely to
have occurred between September 852 CE and January 853 CE, with sulfate
deposition peaking in early 853 CE (Fig. 2e, f). The eruption also produced
large quantities of ash and chlorine, the peak deposits of which are
detected a few months prior to the sulfate peak in Greenland (Fig. 2). The
NS1-2011 chronology is precise to the calendar year in 775 and 939 CE
(Sigl et al., 2015), and it is therefore well-constrained over the time
period of interest for this Churchill eruption. The resultant conservative
age uncertainty associated with the 852/853 CE Churchill eruption is winter
852/853 CE
Geochemical characteristics of the 852/3 CE Churchill eruption based on concentrations of
Large volcanic eruptions have been implicated in global to hemispheric
climate change and societal impacts (e.g. Sigl et al., 2015; Stoffel et al., 2015; Büntgen et al., 2016, 2020; Oppenheimer et al., 2018; McConnell et al., 2020) and raise the question of whether the Churchill eruption – amongst the largest-magnitude eruptions of the Common Era – also had a
far-reaching impact. While extratropical eruptions are often thought to have
less impact on climate than tropical eruptions, recent modelling experiments
have shown that large extratropical eruptions with injection heights above
Given the extent of the Churchill WRAe isochron in glacial and terrestrial environments spanning North America and western Eurasia, our study serves dual purposes. Our first aim is to examine potential NH impacts of the 852/3 CE Churchill eruption on climate, terrestrial environments and societies, using modelled forcing data, climate simulations, palaeoenvironmental reconstructions and historical records. Our second aim is to use the WRAe tephra isochron as a pinning point between intercontinental palaeoenvironmental records to characterise and compare regional expressions of climate change near the outset of the Medieval Climate Anomaly (MCA), a period of increased temperatures ca. 950–1250 CE (Mann et al., 2008, 2009). The WRAe isochron from the 852/3 CE Churchill eruption is therefore aptly placed to identify leads and lags in MCA climate responses and improve characterisations of the spatial and temporal extent of this warm period. We similarly use the tephra isochron to critique the timing of land-use practices, inferred from pollen records, during a period of known societal reorganisation, to determine the extent to which climate change played a role in socio-economic transformation.
Despite the considerable magnitude of the eruption that deposited WRAe,
there has not been a spatially consistent estimate of its volume or
magnitude using established methods (e.g. two-piece exponential function,
Pyle, 1989; Weibull function, Bonadonna and Costa, 2012). The most recent
volume estimate for WRAe (Lerbekmo, 2008) used disparate isopach maps for
the proximal and distal regions of the deposit and the uncertainty
assessment was limited. Here we construct an updated isopach map for WRAe
using a GIS-based synthesis of Lerbekmo's distal and proximal isopachs
We develop a primary forcing reconstruction for the 852/3 CE Churchill
eruption using the EVA(eVolv2k) 550 nm stratospheric aerosol optical depth
(SAOD) reconstruction (Toohey and Sigl, 2017). Detailed explanations of the
model selection and set-up are provided in Appendix A. We also generate a
second SAOD reconstruction using the EVA_H model, which is an extension of the Easy Volcanic Aerosol Model (EVA; Toohey et al., 2016), which accounts for the SO
Climate conditions were simulated using the Community Earth System Model
version 1.2.2 (CESM; Hurrell et al., 2013). The ensemble simulation consists
of 20 ensemble members performed to study the impacts of the 852/853 CE
Churchill eruption on climate. To generate the ensemble members, initially a
seamless transient simulation is run from 1501 BCE (Kim et al., 2021a) with
time-varying orbital parameters (Berger, 1978), total solar irradiance
(Vieira et al., 2011; Usoskin et al., 2014, 2016), greenhouse gases (Joos
and Spahni, 2008; Bereiter et al., 2015) and volcanic forcing from the
HolVol v.1.0 (Sigl et al., 2021) and eVolv2k (Toohey and Sigl, 2017)
databases. The necessary prescribed spatio-temporal distribution of volcanic
sulfate aerosol for the simulations is generated using the EVA model (Toohey
et al., 2016) and follows the same procedure employed by Zhong et al. (2018)
and Kim et al. (2021a). Each ensemble member is branched off at 845 CE with a
small perturbation in the atmosphere introduced at the first time step.
Then, the simulations are seamlessly run until 859 CE. The simulations used
for the analysis have the spatial resolutions of approximately
The anomalies of temperature and precipitation are calculated by subtracting the 845–859 CE multi-year monthly means from the values at each grid point for the initial condition ensemble simulation. From these anomalies, the seasonal means of each individual ensemble simulation are computed as well the ensemble means of 20 member simulations. NH summer conditions reported here refer to climate conditions of June–July–August (JJA), and winter conditions refer to December (of the previous year)–January–February (of the reported calendar year) (DJF).
To test the statistical significance of changes in temperature and
precipitation after the 852/3 CE Churchill eruption, we use the Mann–Whitney
NH summer (JJA) temperatures in the 850s were reconstructed using 13 NH tree-ring width and 12 maximum latewood density chronologies from the NVOLC v2 dataset (Guillet et al., 2017, 2020). Full details of the nested principal component regression (PCR) used to reconstruct NH JJA temperature anomalies (with respect to 1961–1990) and the model calibration are provided in Appendix C. To place the summer temperature anomalies within the context of climate variability at the time of major volcanic eruptions, we removed longer timescale variations by filtering the final reconstruction, which involved calculating the difference between the raw time series and the 31-year running mean. Further investigation of volcanic-forced cooling was facilitated by filtering the original reconstructions using a 3-year running mean to filter out high-frequency noise. To estimate the spatial variability of summer cooling induced by the winter 852/3 CE eruption, we also developed a 500–2000 CE gridded reconstruction of extratropical NH summer temperatures (more details are provided in Appendix C).
Estimated soil moisture anomalies for the 9th century are extracted from tree-ring reconstructions of the gridded summer (JJA) Palmer Drought Severity Index (PDSI) over North America, Europe and the Mediterranean (Cook et al., 2010, 2015). The PDSI metric integrates the influence of both precipitation, evapotranspiration and storage on soil moisture balance throughout the year and is normalised so that values can be compared across regions with a range of hydroclimate conditions. Positive values indicate anomalously wet conditions, while negative values are anomalously dry for that location, and normal conditions are set to zero. Tree-ring PDSI reconstructions in North American and Euro-Mediterranean drought atlases are developed using the point-by-point regression approach described by Cook et al. (1999).
Testate amoebae are a well-established palaeoenvironmental proxy used to
reconstruct past hydroclimatic variability in ombrotrophic (rain-fed)
peatlands because species assemblages predominantly respond to changes in
peatland surface moisture during summer months, and tests are preserved in
the anoxic peat strata (e.g. Woodland et al., 1998; Mitchell et al., 2008).
For this study, testate amoeba analysis was completed on cores obtained from
11 ombrotrophic peatlands located in Maine (
The 9th century in Ireland was a time of significant socio-economic reorganisation and possibly population decline (Kerr et al., 2009;
McLaughlin et al., 2018; McLaughlin, 2020). To investigate the extent to
which these events may have been driven by the effects of either the 852/3 CE
eruption or the transition to the MCA, we compiled land-use proxy data from
five pollen records (Fig. 2e) that included the Churchill (AD860B)
tephra as a chronological tie-point (Hall, 2005; Coyle McClung, 2012;
Plunkett, unpublished data). Raw data were re-categorised by biotope, with a
specific focus on the ratio of arboreal pollen (AP) to non-arboreal pollen
(NAP) and the representation (percentage of total dryland pollen) of taxa
associated with pastoral or arable farming. Age models were constructed for
each site based on tephrochronological and
A wide range of written sources were examined to collate the extant historical record of climate, weather and societal stresses for the period 850–856 CE. This survey focused on Europe – northwestern insular Europe (Irish and Anglo-Saxon annals) and continental Europe (annals and histories covering Byzantine, Carolingian and Umayyad lands) – southwest Asia, North Africa (Abbasid and Byzantine texts), and Tang-era eastern China. To place the 852/3 CE eruption in a wider context where effects of the eruption are apparent, we also employ evidence for large subsistence crises (“famines”) and seemingly more circumscribed crises (“lesser food shortages”) spanning 800–900 CE reported in Carolingian sources, which comprise one of the densest records of subsistence crises extant for the 9th century anywhere (Newfield, 2013; Devroey, 2019).
WRAe deposit bulk tephra volume was modelled as a mean value of 49.3 km
The EVA(eVolv2k) reconstructed SAOD at
550 nm for the 852/3 CE eruption is relatively moderate, with a peak aerosol optical depth perturbation of 0.049 (95 % confidence interval
0.021–0.085) in terms of global monthly mean and 0.078 in terms of NH
monthly mean (Fig. 3a, b). In comparison, the global mean SAOD following the
Pinatubo 1991 eruption was 2–3 times larger (Thomason et al., 2018) and the
reconstructed global mean SAOD for the largest eruptions of the Common Era
(Fig. 3a) reaches 0.3–0.6 (e.g. 0.56 for the Samalas 1257 CE eruption).
For the 9th century alone, four volcanic events have a peak global mean
SAOD exceeding that of the 852/3CE Churchill eruption. The EVA_H reconstruction (Fig. 3b), which accounts for the SO
Stratospheric aerosol optical depth (SAOD, 550 nm) reconstructed for the Churchill 852/3 eruption.
NH summer temperature reconstructions based on tree-ring records reveal
long-term decadal-scale temperature fluctuations between 500–2000 CE (Fig. 4a). All tree-ring-based NH JJA reconstructions contain a short-lived
decreasing temperature trend from 851 CE that peaks in 853 CE, with
temperature anomalies (relative to 1961–1990) reaching
Spatial patterns of the hemisphere-wide JJA cooling in the early 850s are
complex (Fig. 5a): generally cold conditions prevailed over western and
central Europe as well as Scandinavia (anomalies exceeding
Reconstructed and simulated NH spatial patterns of temperature and precipitation anomalies.
Summer PDSI reconstructions based on tree-ring records reveal a shift from wet to drier conditions in parts of western Europe in 854 CE, which persists into 855 CE (Fig. 5d). Wetter conditions in 853 CE in northern Europe and dry anomalies in North Africa and parts of the Mediterranean are potentially indicative of a positive phase of the North Atlantic Oscillation. By 855 CE, dry conditions in northern and western Europe and in 855 CE in the eastern United States are more similar to the pattern expected during a negative phase of the North Atlantic Oscillation (Anchukaitis et al., 2019). Eastern North American tree-ring moisture reconstructions however are also consistently dry from 852 through 855 CE. Tree-ring records in the western half of the continent reveal a mixed PDSI anomaly, generally indicating wetter conditions to the southwest and drier in the northwest, reminiscent of the moisture anomalies during an El Niño event in the tropical Pacific (Fig. 5d).
Simulated summer (JJA) temperature anomalies derived from the CESM reveal a
widespread cooling in the NH extratropical regions in 853 CE that reaches an
ensemble mean value of approximately
The modelled summer (JJA) and winter (DJF) precipitation anomalies vary spatially and temporally between 851–855 CE (Fig. 5e, f), although the post-eruption variability of precipitation is statistically indistinguishable from that of the pre-eruption period. Parts of western Europe show slightly drier conditions in winter 853 CE, with wetter conditions prevalent in western Scandinavia. The summer of 853 CE is characterised by slightly wetter conditions in parts of western Europe (Fig. 5e). The spatially averaged ensemble mean of precipitation indicates that all variation occurs within 1 standard deviation of the pre-eruption period means (Appendix H); there is therefore no obvious statistical differences between modelled summer and winter precipitation patterns associated with the 852/853 CE Churchill eruption in the NH.
The compilation of WTD data in peatlands indicates no consistent response at the time of the WRAe deposition (Fig. 6). Both Irish peatlands record wet conditions relative to the preceding decades at the time of WRAe deposition, but the Dead Island record indicates a subsequent long-term drying whilst Cloonoolish records a temporary drying before a shift to wetter conditions. Two of the three peatlands in eastern Newfoundland record wetter conditions following the WRAe deposition. Jeffrey's Bog in southwestern Newfoundland and the peatlands in Nova Scotia become drier following the eruption but the duration and magnitude of the water table lowering vary between peatlands. For example, the longer-term drying trends in Jeffrey's Bog, Framboise Bog and Villagedale Bog persist over approximately 200 years, whilst the drying in Petite Bog is less pronounced and shorter-lived (ca. 50 years). The peatlands in Maine register a temporary shift to wet conditions following the WRAe deposition.
Although most of the sites reflect centennial-scale trends in WTD, the higher temporal resolution of Petite and Cloonoolish bogs (11 and 12.5 years, respectively) allow decadal-scale responses of the peatlands following the eruption to be considered. Each bog experienced a short-term change towards drier conditions before returning to the prior trend to wetter conditions, but the scale of each hydrological shift lies within the levels of variability of the WTD records.
We find no consistent MCA signal registered in the peatland WTD reconstructions (Fig. 6). Our peatland WTD records indicate that the medieval period was characterised by variable hydrological conditions. The onset of changes towards drier conditions, which may signal the warm Medieval Climate Anomaly, varies temporally and spatially. The earliest dry shift starts ca. 900 CE in northern Nova Scotia (Framboise Bog) and some records from Newfoundland (Jeffrey's Bog and Nordan's Pond bog), whilst this hydroclimatic change is registered ca. 100 years later in records from southern and central Nova Scotia (Villagedale Bog, Petite Bog), Maine (Sidney Bog) and Ireland (Cloonoolish). All records in this study register temporary wet shifts at approximately 850 and 1050–1150 CE, although the extent and durations of the wet shifts vary. The presence of the WRAe isochron conclusively demonstrates that the onset of the wet shift ca. 850 CE is not synchronous. There is also a high degree of spatial variability between records from sites proximal to one another, with some recording contradictory hydrological conditions, such as Saco Heath and Sidney Bog in Maine and Nordan's Pond bog, Pound Cove bog and Southwest Pond bog in Newfoundland.
Available moisture reconstructions from terrestrial and glacial archives containing the WRAe from the North Atlantic region. Records have been developed using
Pollen records from Ireland show considerable variability in the intensity and extent of farming (Fig. 7). The WRAe deposition from the 852/3 CE eruption coincides with the pinnacle of land clearance (reduced arboreal pollen) in central Ireland (Clonfert and Cloonoolish bogs), after which pastoral and arable indicators start to decline as woodland expands. Sites in the northeast of Ireland show less coherent trends than those in central Ireland. At Garry Bog, arable weeds temporarily dip at the time of the eruption, although cereals are still evident. In contrast, evidence for farming is very limited at nearby Frosses Bog, highlighting the localised nature of land use in the vicinity of Garry. At Lake View, moderate levels of farming are recorded, and these increase slightly following the eruption before a decline in activity begins later in the century. The spatial diversity in the pollen records (even within a single region) demonstrates that changes in land use in the 9th century cannot be attributed to any one environmental trigger and very likely reflect differences in local-to-regional economic organisation and demographic pressures.
Summary pollen records from five sites in Ireland, showing the ratio of arboreal to non-arboreal (dryland) taxa and indicators of arable and pastoral environments. Cereal and arable weed curves are shown with a 10-fold exaggeration. The red dashed line indicates WRAe.
Historical records from Europe characterise the 850s CE as time of apparent
climate instability, at least as indicated by the multiple documented
weather extremes and related hazards (Table 1). Carolingian sources observe
severe winter flooding in western Germany in 849–850 CE, and severe summer
heat, drought and a
Climate and climate-related events recorded in Irish,
Carolingian, Anglo-Saxon, Byzantine, Italian, Iberian, Abbasid and Egyptian
sources between 850 and 858 CE. Locations given reflect where the
texts were likely compiled at the time, though the phenomena recorded could
have been more widespread. Cases were the phenomena locations are instead
given are denoted by an asterisk (
A wider chronological consideration of the Carolingian evidence reveals that
food shortages occurred in several other decades of the 9th century in
Carolingian Europe (Fig. 8). This observation reinforces the point that a
correspondence (or near-correspondence) between the dating of the Churchill
eruption and the documented events of the 850s CE does not confirm a causal
linkage. Some food crises of the 9th century were also apparently more
vast and longer lasting than those (reliably) documented here for the 850s,
with the Carolingian sources also observing widespread crises associated
with climate anomalies in the 820s, 860s and 870s (Newfield, 2013; Haldon et
al., 2018; Devroey, 2019). One mid-10th-century source does observe a
Ninth-century reports of large subsistence crises (“famines”, black) and seemingly more circumscribed crises (“lesser food shortages”, orange) recorded in Carolingian sources (Newfield, 2013). Note again that the record of food shortages is imperfect: some crises may not have been recorded and the extent and severity of several recorded crises are difficult to determine.
Elsewhere, in Iberia, we read only of significant flooding along the Rio
Guadalquivir in 849 and 850 CE (Meklach et al., 2021), while there are no
known reports in Anglo-Saxon, Byzantine, Italian or Iberian sources of
anomalous weather
A suppression of the East African Monsoon has been noted after some large extratropical NH eruptions, with the consequence being a reduction of the agriculturally critical Nile summer flood that is primarily driven by summer monsoon rainfall in the Ethiopian highlands, including the Blue Nile and Atbara river watersheds (e.g. Oman et al., 2006; Melesse et al., 2011; Iles and Hegerl, 2015; Manning et al., 2017; Atwood et al., 2020; Singh et al., 2022). The extant historical sources do not, however, allow us to identify such a happening around the time of the Churchill eruption. Egyptian historical records appear silent, with no known incidences of food crises or other societal instability that might follow unusually low Nile summer flooding (e.g. Hassan, 2007; Ludlow and Manning, 2016, 2021). We are also unaware of sources from the Nubian Nile that might suggest hydroclimatic anomalies or related societal reactions for our period of interest (Adam Laitar and Giovanni Ruffini, personal communication, 2021).
The Islamic Nilometer record, measured for our period on Roda Island near
Cairo, provides the maximum height reached by each year's summer flood
(Hassan, 1981; Hassan and Stucki, 1987; Said, 1993; Kondrashov et al., 2005;
Hassan, 2007; Manning et al., 2017). This is an important source, but one
that presents a complex story for our years of interest. A notably poor
summer flood (the fifth lowest maximum of the 9th century, using the
Nilometer data as processed by Kondrashov et al., 2005) is recorded for 851 CE, following on from the 15th lowest value for the year previous.
Given that this extreme for 851 CE would have largely been the product of
rainfall in the summer of that year over the Ethiopian highlands, it is too
early to be credibly linked to the Churchill eruption, even when allowing
the
Chinese historical sources register local and regional weather anomalies and
impacts in (primarily) eastern China in the years following the Churchill
eruption. Of particular note is a drought in the summer of 852 CE, affecting
the Huainan Circuit, comprising some 12 prefectures and 53 counties and
situated between the Huai and the Yangzi rivers. Famine is associated with
the drought-induced migration, with people resorting to wild foods (Zhang,
2004; as per the
The VEI 6 eruption of Churchill in the winter of 852/3
Despite the moderate climate forcing potential of the 852/3 CE Churchill
eruption estimated from ice core sulfate records, there is evidence for a
strong NH cooling around the time of the eruption. Tree-ring temperature
reconstructions show temperature declines centred on summer 853 CE with a
peak magnitude of around
In some respects, however, the spatial patterns differ between the climate model simulations and the tree-ring reconstructions. In particular, the persistent cool conditions in central Asia and Siberia in 855 CE are only found in the tree-ring-based reconstructions. These deviations (changes in temperature amplitudes and in spatial patterns) are expected as the ensemble means of the simulations focus on the signal of the volcanic eruption by reducing internal climate system variability. In contrast the tree-ring-based reconstruction contains both internal variability and the potential forcing signal of the eruption and/or other external drivers. Therefore, the reconstructed cooling in Asia and Siberia in 855 CE is potentially related to internal variability of the climate, such as changes in the large-scale atmospheric circulation rather than being externally forced by the Churchill eruption.
The reconstructed climatic cooling peak in 853 CE aligns with the eruption
date of the winter 852/3 CE Churchill eruption but the timing of the start
of this tree-ring-inferred cooling trend begins in summer of 851 CE, thereby predating the eruption (and its associated age uncertainty). However, the magnitude of the temperature decline in summer 851 CE is within the range of
natural temperature variability, and it is not until the summers of 852 and
853 CE when temperatures exceed the range of natural variability. The
modelled climate scenario cooling occurs later in 853 CE, with widespread
cooling present in summer 854 CE and winter 854 CE. The results from the
tree-ring-based temperature reconstructions preclude the attribution of the
climatic cooling solely to the Churchill eruption, but the eruption timing
clearly corresponds to cooling as registered in both reconstructed and
simulated approaches. These findings therefore suggest that the winter 852/3 CE Churchill eruption exacerbated a naturally occurring cold period. This is supported by the decadal-scale step changes in temperatures recorded in the
tree-ring-based reconstructions (Fig. 4a) and NGRIP1
Hydroclimate changes driven by volcanic eruptions are less clearly defined
than those of temperature, partly due to the higher degree of variability in precipitation and the small changes in atmospheric moisture associated with
the magnitude of temperature change often associated with volcanic cooling.
In principle, two possible processes might lead to precipitation changes
after an eruption: thermodynamic or dynamic affects. The direct
thermodynamics effect is related to the Clausius–Clapeyron relationship,
which predicts that the water-holding capacity of the atmosphere decreases
by approximately 7 % for every 1
The climate forcing of the 852/3 CE Churchill eruption derived from existing
ice core records and used in the climate model simulations is the current
best estimate. Uncertainty in the stratospheric aerosol forcing (as shown in Fig. 3b) is not incorporated into the model simulations as, e.g., was
undertaken by Timmreck et al. (2021). Furthermore, additional forcing
factors have not been explicitly taken into account. In particular, this
explosive eruption is characterised by high chlorine concentrations in the
ice cores (Fig. 2) and a very extensive ash cloud across the NH midlatitudes to high
latitudes, suggesting large atmospheric loadings. Emissions of halogens and
ash have the potential to influence climate, but their climate forcing
potential is poorly constrained and so they remain unaccounted for in the
EVA and EVA_H forcing time series as well as in the CESM
simulations. The injection of a large quantity of halogens along with sulfur by the 852/3 CE eruption may have modulated the impact on surface
temperatures: some model studies suggest that the co-emission of halogens
may intensify or prolong the volcanic cooling (Wade et al., 2020;
Staunton-Sykes et al., 2021), although contrasting model results suggest the effect may be model or event dependent (Brenna et al., 2020). The influence
of ash on radiative forcing is currently unclear. For example, recent
observations for the Kelud 2014 eruption suggest that ash exerted a
radiative forcing of
The White River Ash east (WRAe) deposit from the 852/3 CE Churchill eruption has reported thicknesses of 50–80 m proximal to Mount Churchill and visibly extends in an easterly direction
Historical records gleaned from a wide range of sources across Europe,
Africa and Asia provide an opportunity to (i) assess the extent to which the 852/3 CE Churchill eruption had distal societal consequences, (ii) corroborate or critique results from the modelled and tree-ring-based
climate scenarios around the time of the Churchill eruption and (iii) identify any evidence of extreme weather conditions that is not registered in the palaeoenvironmental reconstructions based on natural archives, such as severe winters. European historical records spanning the 850s document some anomalous conditions, albeit fewer extreme weather events and associated crises than in other decades of the 9th century (Fig. 8). Food shortages and extreme weather were reported shortly before and after the 852/3 CE eruption in western Germany; a severe subsistence crisis may also have occurred in nearby northern France and Belgium that set in during the eruption year or shortly thereafter. Tree-ring reconstructions show that the growing season in 852 CE (falling within the current
The pollen records are insufficiently resolved to identify sub-decadal anomalies or extreme weather events, but they provide a useful longer-term perspective on societal adaptation to climate variability. Precise comparisons of the pollen assemblages between sites are facilitated by the presence of WRAe, which dispels any chronological uncertainty with respect to the timing of changes in land use. The pollen records clearly show spatially complex patterns in the extent and intensity of land use, implying that changes in human activity around this time were not driven merely by responses to changing environmental conditions. Rather, it would seem that any observed cultural shifts around this time reflect an interplay of social, economic and political factors.
The MCA is commonly characterised as a warm period ca. 950–1250 CE (Mann et al., 2009), with dry conditions in Europe and North America (e.g.
Büntgen and Tegel, 2011; Ladd et al., 2018; Marlon et al., 2017). There
is, however, considerable spatial variability in the timings of the MCA
onset and peak warmth (e.g. Neukom et al., 2019) as well as hydroclimatic
expressions (e.g. Shuman et al., 2018). In order to assess regional
variability in terrestrial MCA hydroclimate across northeastern North
America and western Europe, this study provides chronologically precise
hydroclimatic comparisons facilitated by the detection of the WRAe isochron
in our peatland archives as well as the NGRIP1 ice core, which acts as a
chronological tie point between the palaeoenvironmental reconstructions.
Comparisons of our 11 peatland records show that there is no consistent
multidecadal-scale hydrological response associated with the MCA; rather
hydrological conditions are variable both within and between records. There
are also no clear temperature trends associated with the MCA detected in the NGRIP1
The environmental reconstructions presented in this study highlight the
heterogeneous and time-transgressive nature of the reconstructed MCA
hydroclimate change. For example, a dry shift, which may be typical of an MCA
climate response, began ca. 900 CE in northern Nova Scotia and some records
in Newfoundland and corresponds to a change to warmer conditions in
central Greenland. However, the onset of drier conditions is delayed by ca.
100 years in more southwesterly sites in Nova Scotia and Maine as well as
on the east coast of the North Atlantic in Ireland. In addition, all
peatland records contain temporary wet shifts that occur prior to the MCA
ca. 700–850 CE, which corresponds to a period of generally colder
temperatures in central Greenland as reconstructed from the NGRIP1
Here we have reported the dominant peatland hydroclimatic patterns that are supported by multiple regional peatland records; however, some differences exist between proximal reconstructions, such as the clusters of three peatland records developed within ca. 10 km in eastern Newfoundland and two records within ca. 110 km in Maine. The differences in hydroclimate at such local levels in Newfoundland may reflect the degree of spatial hydroclimate variability during this period but may also be exacerbated by autogenically driven peatland responses such as enhanced peat accumulation under warmer MCA climates that would drive an apparent lowering of the water table (e.g. Swindles et al., 2012). The divergence between the hydroclimate reconstructions obtained from the Maine peatlands is likely influenced by a fire disturbance event at one of the sites, Saco Heath, which created a substantial hiatus in peat accumulation in some areas of the site (Clifford and Booth, 2013). Whilst the Saco Heath record presented here appears less impacted by the fire, there is a high degree of uncertainty in the hydroclimate reconstruction between ca. 1000–1250 CE when the accumulation rate slows, which may reflect a temporary hiatus (Fig. 6; Appendix D). The development of more palaeoenvironmental reconstructions from sites containing the WRAe, particularly in locations such as Maine and western Europe, will be useful to investigate further MCA trends further.
The winter 852/3
Areas proximal to Mount Churchill experienced widespread and prolonged ecological, environmental and societal changes attributed to the eruption emissions, but there is no evidence of multidecadal-scale climatic response preserved in distal palaeohydrological records from the North Atlantic region that are precisely temporally linked by the 853 CE Churchill WRAe isochron. Pollen records of vegetation change and human activity from Ireland linked by the WRAe isochron also provide no evidence to support long-lasting societal responses in Ireland associated with the eruption. Evidence of short-term societal impacts in Europe from the 852/3 CE Churchill eruption remains equivocal: some historical records from Ireland and Germany, and possibly northern France and Belgium, report harsh winter conditions and food shortages within the age uncertainties of the eruption, but similar events were reported outside of the eruption period and were not unknown in the 9th century. The 852/3 CE Churchill eruption therefore exemplifies the difficulties of identifying and confirming volcanic impacts on society even when only a small eruption age uncertainty exists.
The presence of the WRAe isochron in peatlands in northeastern North America and western Europe assists with comparisons of hydroclimatic reconstructions during the Medieval Climatic Anomaly, often defined as a period of globally increased temperatures between 950 and 1250 CE (Mann et al., 2009). Reconstructed hydroclimate conditions in 853 CE vary, highlighting leads and lags in the terrestrial responses to environmental change that may otherwise be considered contemporaneous without the temporal precision provided by the WRAe. This study shows a lack of a consistent terrestrial response to MCA climate change in the North Atlantic region; rather the MCA time period is characterised by time-transgressive and heterogeneous hydroclimatic conditions. These findings contribute to a growing body of research that cautions against the application of the globally defined MCA characteristics when interpreting individual records of palaeoenvironmental change and ultimately questions the detectability of a coherent MCA climate signature.
The EVA (eVolv2k) reconstruction (Toohey and Sigl, 2017) is the recommended
volcanic forcing dataset for climate model simulations of Phase 4 of the
Paleoclimate Modeling Intercomparison Project (PMIP, Jungclaus et al., 2017; Kageyama et al., 2018). The EVA reconstruction uses volcanic stratospheric sulfur injection estimates derived from sulfate deposition from an extensive bipolar array of ice cores (Sigl et al., 2015), which are then converted into
an SAOD time series using the idealised, scaling-based aerosol model Easy
Volcanic Aerosol (EVA, Toohey et al., 2016). The global mean radiative
forcing (RF) time series is estimated from the SAOD using the following
relationship from Marshall et al. (2020):
An initial condition ensemble was created using the Community Earth System Model version 1.2.2 (CESM), consisting of 20 ensemble members. CESM is a state-of-the-art fully coupled Earth system model composed of atmosphere, land, ocean and sea ice components. To generate the ensemble members, initially a seamless transient simulation was run from 1501 BCE (Kim et al., 2021a) with time-varying orbital parameters (Berger, 1978), total solar irradiance (TSI) (Vieira et al., 2011; Usoskin et al., 2014, 2016), greenhouse gas (GHG) (Joos and Spahni, 2008; Bereiter et al., 2015), and volcanic forcing from the HolVol v.1.0 (Sigl et al., 2021) and eVolv2k (Toohey and Sigl, 2017) databases. CESM1.2.2 uses a prescribed monthly mean sulfate aerosol mass on a predefined latitudinal and vertical grid as an input volcanic forcing. Optical properties are estimated within the model assuming that the aerosol mass is comprised of 75 % sulfuric acid and 25 % water and has a constant log-normal size distribution with a constant effective radius and following Neely et al. (2016). The necessary prescribed spatial–temporal distribution of volcanic sulfate aerosol for the simulation is generated using the Easy Volcanic Aerosol Model (Toohey et al., 2016) and following the same procedure employed by Zhong et al. (2018) and Kim et al. (2021a). In the procedure, the EVA-generated spatio-temporal distribution of sulfate was first converted to volcanic aerosol mass to be readable by CESM. This distribution of volcanic aerosol mass in CESM was scaled up by 1.49 to reconcile CESM and EVA atmospheric responses to the 1991 Pinatubo eruption. The scaling value is derived based on some sensitivity experiments for the Pinatubo eruption using the EVA-generated forcing and the available CESM forcing (Ammann et al., 2003) and after comparing the atmospheric responses related to the vertical and surface mean temperatures and radiative balances. Then, the transient simulation was branched off at 845 CE and a small perturbation was introduced at the first time step in the atmosphere. The 20 ensemble members were run from this point until 859 CE. During this 14-year period, no other volcanic eruptions were included except the 852/853 CE Churchill eruption. An eruption occurring in the Southern Hemisphere in 853 CE is excluded in the simulations.
Mann–Whitney
We employed a PCR to reconstruct NH JJA
temperature anomalies (with respect to 1961–1990) from tree-ring
records. We coupled this PCR with a bootstrap random sampling approach to quantify the robustness of our reconstruction and to estimate confidence intervals of reconstructed JJA temperatures. To account for the decreasing number of records available back in time, we combined the PCR with a nested approach. In total, our reconstruction is based on 23 subsets of tree-ring chronologies or nests. The earliest and most recent nests span the periods 500–551 and 1992–2000 CE, respectively. The most replicated nest (1230–1972 CE) includes 25 chronologies. For each nest, we reduced the proxy predictors matrix to principal components (PCs) using a principal component analysis (PCA). PCs with eigenvalues
The target field (predictand) used for the reconstruction is the BEST JJA
gridded temperature dataset (
Composite water table depth (WTD) record for Sidney Bog, Maine, USA, based on testate amoebae assemblage data obtained from two cores obtained from different coring locations on the peatland (Clifford and Booth, 2013; Mackay et al., 2021). Testate amoebae WTD reconstructions were obtained using the tolerance-downweighted weighted averaging model with inverse deshrinking (WA-Tol inv) from the North American transfer function of Amesbury et al. (2018). To produce the composite record, the chronological resolution of the Clifford and Booth (2013) WTD record has been reduced to the same resolution as the Mackay et al. (2021) record using linear interpolation between chronologically adjacent WTD values. The composite record then presents the average WTD of the interpolated Clifford and Booth (2013) and the Mackay et al. (2021) reconstructions.
Distributions of JJA temperature anomalies in the unfiltered (red) and filtered (blue, 31-year moving average filter) 500–2000 CE NH reconstructions. Blue and red vertical dotted bars indicate the 1st percentile of the filtered and the 2nd percentile of the unfiltered reconstructions, respectively. Blue and red bold vertical lines show the cooling observed in 853 CE in the filtered and unfiltered reconstructions, respectively.
Top 30 coldest years during the period of 500–2000 CE based on tree-ring temperature anomalies filtered using a 3-year mean and corresponding eruption information for proximal calendar years. Eruption dates and volcanic stratospheric sulfate injection (VSSI) estimates taken from the eVolv2k reconstruction (Toohey and Sigl, 2017), representing the most immediate preceding eruption in the dataset. Roman font: eruption occurred within 2 years of the coldest reported year; italic font: eruption occurred more than 2 years before the coldest reported year. Unidentified eruption: UE.
The spatially averaged NH extratropical (15–90
NGRIP1
The EVA code is publicly available on GitHub:
The volcanic forcing data are accessible online:
HM, GP and BJ were responsible for the conceptualisation and design of the project. HM, MA, AM, AB and GS conducted the testate amoebae analyses as well as the associated data analysis and interpretation. HM and MA conducted the testate amoebae analyses as part of projects supervised by PDMH, PGL and DC. RB and HM created the Sidney Bog composite testate amoebae record. GP and LCM designed and conducted the Irish pollen and tephra analysis for the Irish sites (exception of Dead Island record, tephrochronology by GS). TA and MT designed the forcing potential analyses, which were conducted by TA. MiS analysed the ice core chronologies and associated data. BJ and MB designed the eruption volume estimate and magnitude analyses, which were conducted by MB. WK and CR designed the climate model simulation analyses, which were conducted by WK. CC and MaS designed the tree-ring temperature reconstruction analyses, which were conducted by CC. KJA designed and analysed the tree-ring drought reconstructions. JM, TPN, NDC, FL, CK and ZY analysed the historical records. HM, KLD, TA, WK, CC, AM, KA and MB designed and produced the visualisations. HM prepared the original draft of the paper and all co-authors were involved in the writing review and editing process.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “Interdisciplinary studies of volcanic impacts on climate and society”. It is not associated with a conference.
This paper benefitted from discussion at events of the Past Global Changes (PAGES) working group “Volcanic Impacts on Climate and Society” (VICS) as well as with Angus M. Duncan and Richard J. Payne. PAGES is supported by the Chinese Academy of Sciences (CAS) and the Swiss Academy of Sciences (SCNAT). Helen Mackay and Matthew Amesbury were supported by the UK Natural Environment Research Council (PRECIP project grants NE/G019851/1, NE/G020272/1, NE/G019673/1 and NE/G02006X/1 and MILLIPEAT project grant NE/1012915/1). A Quaternary Research Association New Research Workers award was granted to Helen Mackay and the NERC Radiocarbon Facility NRCF010001 (allocation numbers 1744.1013 and 1789.0414). Christophe Corona and Markus Stoffel were supported by the Swiss National Science Foundation Sinergia project CALDERA (grant agreement no. CRSII5_183571). Woon Mi Kim and Christoph Raible are supported by the Swiss National Science Foundation (SNSF, grant nos. 200020_172745 and 200020_200492). The climate mode simulations were performed at the Swiss National Super Computing Centre (CSCS). Michael Sigl acknowledges funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement 820047). Francis Ludlow and Conor Kostick were supported by an Irish Research Council Laureate Award (CLICAB, IRCLA/2017/303). Joseph G. Manning and Francis Ludlow acknowledge support from US National Science Foundation Award no. 1824770. Francis Ludlow and Zhen Yang acknowledge additional support from an ERC Synergy Grant (4-OCEANS; grant agreement 951649). Thomas J. Aubry acknowledges support from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 835939 and from Sidney Sussex College through a Junior Research Fellowship. We are grateful to all reviewers for their constructive comments and valuable suggestions.
This research has been supported by the Natural Environment Research Council (grant nos. NE/G019851/1, NE/G020272/1, NE/G019673/1, NE/G02006X/1 and NE/1012915/1), the Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (grant nos. CR-SII5_183571, 200020_172745 and 200020_200492), the National Science Foundation (grant no. 1824770), the Irish Research Council (grant no. IRCLA/2017/303) and the H2020 European Research Council (grant nos. 820047, 835939 and 951649).
This paper was edited by Allegra N. LeGrande and reviewed by Mukund Palat Rao and one anonymous referee.