A 2000 Year Saharan Dust Event Proxy Record from an Ice Core in the European Alps

Dust events originating from the Saharan desert have far‐reaching environmental impacts, but the causal mechanism of magnitude and occurrence of Saharan dust events (SDEs) during the preinstrumental era requires further research, particularly as a potential analog for future climate. Using an ultrahigh resolution glacio‐chemical record from the 2013 Colle Gnifetti ice core drilled in the Swiss‐Italian Alps, we reconstructed a 2000‐year‐long summer Saharan dust record. We analyzed both modern (1780–2006) and premodern Common Era (CE) major and trace element records to determine air mass source regions to the Colle Gnifetti glacier and assess similarities to modern and reconstructed climate trends in the Northern Hemisphere. This new proxy SDE reconstruction, produced using measurements from a novel, continuous ultrahigh‐resolution (120‐μm) ice core analysis method (laser ablation‐inductively coupled plasma‐mass spectrometer) is comprised of 316,000 data points per element covering the period 1–1820 CE. We found that the Colle Gnifetti ice core captures an anomalous increase in Saharan dust transport during the onset of the Medieval Climate Anomaly (870–1000 CE) and records other prominent shorter events (CE, 140–170, 370–450, 1320–1370, and 1910–2000), offering a framework for new insights into the implications of Saharan dust variability.


Introduction
Dust transport from the Saharan region to Europe is linked to the state of the climate system ), yet variability in the intensity of these events is poorly constrained for the past two millennia during which both natural and anthropogenic climate change has occurred. Saharan dust influences terrestrial and ocean biogeochemical ecosystems, human health, and radiative properties of t2001). Due to the importance of Saharan dust events (SDEs) to natural systems and human environments, and given the uncertainty of future occurrences of these events in a changing climate, several studies have looked to the past as an analog to understand how SDEs are connected to climate (e.g., Angelis & Gaudichet, 1991;Bohleber et al., 2018, Maupetit & Delmas, 1994, Schwikowski et al., 1995Schwikowski et al., 1999;Thevenon et al., 2009;Wagenbach et al., 1996;Wagenbach & Geis, 1989). To further investigate SDE transport to Europe and their response to changing climates, we present in this paper the longest, continuous, ultrahigh-resolution ice core record yet produced reflecting Saharan dust transported to the European Alps. This record was developed using an innovative technique of compiling continuous measurements from ultrahigh resolution (120-μm) laser ablation-inductively coupled plasma-mass spectrometer (LA-ICP-MS) technology, applied to the Colle Gnifetti ice core below the firn-ice transition.
Europe and surrounding regions in the Northern Hemisphere (NH) experienced several anomalous global climate shifts during the late Holocene including the Medieval Climate Anomaly (MCA;~900-1300 CE) (e.g., Berner et al., 2011;Bradley et al., 2003;Lamb, 1965;Mann et al., 2009;Mann & Jones, 2003;Xoplaki et al., 2011), and the following colder era, the Little Ice Age (LIA;~1350-1900 CE) (e.g., Bradley & Jonest, 1993;Grove, 1988;Mann et al., 2008;Moberg et al., 2005). Hypothesized causal mechanisms of these long-term climate events include changes in total solar irradiance (Renssen et al., 2006;Shindell et al., 2001), volcanic aerosol loading into the atmosphere (Crowley, 2000;Shindell et al., 2003;Otterå et al., 2010), and ocean circulation (Broecker, 1991;Crowley, 2000). Atmospheric aerosol composition and air mass transport changes provide both a mechanism for and an indicator of transitions in and out of anomalous climate periods and can be measured through chemical fingerprinting of air masses captured in ice cores (Meeker & Mayewski, 2002). Saharan dust deposition reconstructions from 240 ka marine sediment records show low latitude dust deposition does not parallel the variability of glacial-interglacial changes seen in mid-latitude and high latitude dust emission. Rather, Skonieczny et al. (2019) identified a high correlation to summer insolation on millennial timescales, indicating a significant African monsoon influence on low latitude dust transport. However, relatively little is known about Saharan dust transport during significant climate events of the Common Era such as the LIA and MCA (e.g., Thevenon et al., 2009Thevenon et al., , 2011Bohleber et al., 2018), and understanding the variability and causality of past climate anomalies can offer essential insights for future climate change (e.g., Antoine & Nobileau, 2006). Saharan dust transported across the Mediterranean to Central Europe originates from North African sources, predominantly during the spring and summer months, and is typically carried at least 5,000 m above sea level (m a.s.l.; Prospero, 1996;Prospero et al., 2005). Major modern source regions of Saharan dust to Europe include three main areas: Western Sahara, Moroccan Atlas, and northern/central Algeria, confirmed through trajectories analyses from northeastern Spain to their area of origin (Avila et al., 1997), also noted in Scheuvens et al. (2013). The transport and deposition of Saharan dust over the Mediterranean in the summer depends on favorable conditions for entrainment and transport of dust above the boundary layer (Gaetani & Pasqui, 2014). Past research has shown dust emissions are highest when peak surface heating moves from the Sahel to the central Sahara during the summer months during the northward extension of the Intertropical Convergence Zone (ITCZ), a heavy precipitation band caused by converging northeasterly and southeasterly winds (Engelstaedter et al., 2006). The ITCZ has a southernmost position in the winter months and northernmost in the summer, causing dry convection and convergence, thereby enhancing near surface turbulence and facilitating dust transport (Engelstaedter et al., 2006). Favorable conditions for SDE incursions into Europe during the summer months are influenced by a steepening of the pressure gradient along a strong subtropical high (Azores High), coupled with a northeastward shifting of the Saharan High, located on the northern boundary of the Sahara, into the Mediterranean and a southeast shift in the Icelandic Low (Barkan et al., 2005). Longer transport pathways are also documented from back trajectories reaching out over the North Atlantic prior to transport to Europe (e.g., Schwikowski et al., 1995;Collaud Coen et al., 2004;Sodemann et al., 2006;Thevenon et al., 2011). Although infrequent, Saharan dust has the potential of reaching as far north as the British Isles following an anticyclone over western Europe (Wheeler, 1986;Coudé-Gaussen, 1989).
The variability of summer SDEs on an interannual to decadal time scale during the modern era has been explained by anthropogenic forcings and ocean-atmospheric teleconnections such as the North Atlantic Oscillation (NAO), El Niño Southern Oscillation (ENSO), the ITCZ, and the Atlantic multidecadal oscillation (AMO). Human-induced soil degradation and increased droughts in the Sahel are linked to increases in dust emissions across the Atlantic (Moulin & Chiapello, 2006). However, anthropogenic influences (e.g., desertification, land-use changes, and climate change) on Saharan dust emissions are difficult to resolve precisely due to large natural variability in the arid Saharan region (Moulin et al., 1997;. For example, we do not yet have the capability to estimate the percentage of anthropogenic dust (Engelstaedter et al., 2006). According to Moulin et al. (1997), enhanced dust transport across the Mediterranean is linked with an increased pressure gradient (r = 0.66, p = 0.027) associated with the winter NAO index, the difference in normalized winter sea level pressures between Lisbon, Portugal, and Stykkisholmur, Iceland (Hurrell, 1995), along with eastward transport over the North Atlantic (r = 0.50, p = 0.097). Years of high (low) NAO indices are identified by a stronger (weaker) Azores high coupled with a below (above) normal Icelandic Low-pressure system, causing drier (wetter) conditions over southern Europe, the Mediterranean Sea and northern Africa, therefore increasing (decreasing) dust mobilization and transport across the Mediterranean and North Atlantic. Shifts in the ENSO are coupled with changes in lower tropospheric atmospheric circulation over North Africa, causing stronger winds towards the Atlantic Ocean, and therefore enabling SDEs (DeFlorio et al., 2016;Prospero & Nees, 1986;Prospero & Lamb, 2003;Rodríguez et al., 2015). A southward displacement in the ITCZ, along with a colder North Atlantic Ocean (negative phase AMO; Wang et al., 2012), prompts a decrease in precipitation and increased surface winds over dust-producing regions in the Southern Sahara leading to amplified SDEs over the North Atlantic, therefore also contributing to Mediterranean transport (Doherty et al., 2012;Doherty et al., 2014). Periods of increased aridity in the Sahara (Middleton, 1985;Littmann, 1991) and decreased rainfall in the Sahel (Prospero & Lamb, 2003) caused by ocean-atmospheric teleconnections can also play a major role in dust production.
In agreement with Evan et al. (2016), we speculate that any phenomenon that facilitates the transport of dust can lead to an increase in SDEs. Saharan dust, along with other natural and anthropogenic aerosols transported across the Mediterranean, are deposited at high elevation glaciers in the European Alps (e.g., Angelis & Gaudichet, 1991;Maupetit & Delmas, 1994;Schwikowski et al., 1995;Schwikowski et al., 1999;Wagenbach & Geis, 1989;Wagenbach et al., 1996). Because of the close redundant proximity of the alpine glaciers to population centers, ice core records from these areas capture and preserve unique records of natural and anthropogenic aerosol transport not available through polar ice cores. Ice cores from this region contain the southernmost such records available for the North Atlantic. These records of past dust and sea-salt aerosols offer a history of atmospheric circulation, including strength of wind and pressure systems and changes in air mass sources. Previous ice core studies of aerosols throughout the European Alps include: Col du Dôme, Mont Blanc (Preunkert et al., 2000); Fiescherhorn, Bernese Alps (Schwerzmann et al., 2006); Ortles, Eastern Alps (Gabrielli et al., 2016); and Colle Gnifetti and Colle del Lys in the Monte Rosa region (e.g., Wagenbach et al., 2012, and references therein).
Situated in the heart of the European Alps, Colle Gnifetti (CG) glacier (4,500 m a.s.l) stands out as the only nontemperate site where net snow accumulation is low enough to record environmental signals over at least the last two millennia . Previous studies of CG demonstrate it is an ideal site to examine past changes in natural and anthropogenic source aerosols to the European Alps region (e.g., Bohleber et al., 2018;Loveluck et al., 2018;Lugauer et al., 1998;Meola et al., 2015;Schotterer et al., 1985;Schwikowski et al., 1995;Thevenon et al., 2009;Wagenbach & Geis, 1989). Due to transport and deposition in relation to meteorological conditions, the primary and most consistently deposited aerosol transported to CG is Saharan dust (Figure 1; Schotterer et al., 1985). Precise source locations of dust to the CG region have not been well identified, however back trajectories in concert with chemical tracer measurements by Thevenon et al. (2011) specify regions of Algeria and north-central to north-western part of the Saharan desert (i.e., Morocco, Tunisia, Libya, and Mali), following similar conclusions by Schwikowski et al. (1995). Lugauer et al. (1998) noted the most pronounced aerosol variability captured in modern snow at CG is accumulated during the summer months with some variability recorded during fall and spring and minimal amounts in the winter months.
We investigate the recent portion (1780-2006 CE) of the 2013 CG ice core record using annually resolved discretely sampled inductively coupled plasma-mass spectrometer (ICP-MS) measurements for signatures of potential air mass sources, then calibrate to modern climate reanalysis data and known SDEs. Using this recent portion of the record as an analog for SDEs, we extend our record back to 1 CE using subannual (0.02 year) and annually averaged LA-ICP-MS 56 Fe, followed by the identification of major dust episodes at 121-μm resolution. The Climate Change Institute's W. M. Keck Laser Ice Facility LA-ICP-MS sampling system was used to collect a total of 316,000 data points for dust element 56 Fe over 20-m of core (1-1820 CE). For the LA-ICP-MS 56 Fe raw data, an average of 570 data points was collected per year for the first 100 years (1720-1820 CE; 3.3 m) and an average of 170 data points per year for the entire 20-m of core (1-1820 CE). The maximum data points collected in 1 year for this record is 1,187 in 1818 CE This research expands on previous studies that validate LA-ICP-MS technique Haines et al., 2016;Mayewski et al., 2013;Sneed et al., 2015;Spaulding et al., 2017). Applying this novel technique to alpine ice cores allowed us to develop the longest continuous summer Saharan dust record and provide detailed interpretations of past atmospheric conditions on subannual to storm-scale event scales.

Study Area
A 71.8-m-deep ice core (N 45.92893, E 7.87627) was recovered in 2013 from the CG glacier located in the Swiss-Italian Alps ( Figure 1) by a joint ice core drilling expedition from the University of Heidelberg, University of Bern, and the Climate Change Institute (University of Maine) discussed further in Bohleber et al. (2018). Details concerning CG glacier flow, englacial temperature, and surface accumulation appear in Alean et al. (1983), Haeberli et al. (1988), Haeberli and Funk (1991), Lüthi and Funk (2000), Hoelzle et al. (2011), and Konrad et al. (2013).

High-Resolution Ice Core Trace Metal and Aerosol Analysis
Approximately 1,600 discrete samples were collected from meltwater-stream of continuous flow analysis at the University of Bern, summarized in Bohleber et al. (2018), then collected clean fractions were also  Schwikowski et al. (1995) and Thevenon et al. (2011) (brown arrow) and marine airmass transport (blue arrow) to the Colle Gnifetti ice core site in the Swiss-Italian Alps (red star). Map created using the Basemap module in Python 3.6 and ESRI_Imagery_World_2D from arcgis. analyzed using the Climate Change Institute's Thermo Electron Element2 ICP-MS instrument to measure concentrations of 26 major and trace elements (Sr, Cd, Cs, Ba, La, Ce, Pr, Pb, Bi, U, As, Li, Al, S, Ca, Ti, V, Cr, Mn, Fe, Co, Na, Mg, Cu, Si, and K) and selected major ions (Cl − , NO 3 − , SO 4 − ) via ion chromatography.

Continuous Ultrahigh-Resolution Laser Ablation Ice Core Analysis
Measurements collected from the LA-ICP-MS system were compiled to produce an ultrahigh-resolution (120 μm per sample) continuously sampled record below the firn ice transition (42 m, 1820 CE). Previous studies have also used LA-ICP-MS systems to observe changes in ice (Della Lunga et al., 2014;Della Lunga et al., 2017). In this study, we present only the first~20 m below the firn ice transition (42 to 61 m, 1-1820 CE). Details regarding method, sample preparation, and calibration protocols appear in Spaulding et al. (2017) and Sneed et al. (2015). The same ice sections were previously used in Bohleber et al. (2018) to establish a chronology for the last~1,000 years of the 2013 CG ice core through annual layer counting of a continuous single-element LA-ICP-MS 44 Ca record. For the LA-ICP-MS analysis reported here, we applied a new multielement method to yield a precise investigation of the relative phasing of elements at submillimeter scales . We applied this new multielement method , to the deepest 30 m of the ice core to create a continuous record for the following elements: 56 Fe, 27 Al, 44 Ca, 32 S, and 28 Si. An example of data collected from multiple 4-cm runs and five concatenated runs in a section appears in Figure 2. An explanation of the postprocessing of the LA-ICP-MS output data appears in Text S1. For each of the five elements, we collected close to 500,000 measurements and in total, over 1,500,000 measurements for the 30 m of core. As discussed in Sneed et al. (2015) and Spaulding et al. (2017) and further investigated within this study, the resulting intensity values from the dust elements ( 56 Fe, 27 Al, and 44 Ca) mirror closely their respective elemental concentrations established by traditional ice core sampling. For this study, we disregard 32 S and 28 Si measurements as they do not capture as definitively as the dust elements their respective Ca (green) in intensity as counts per second (cps) collected simultaneously during a 4-cm run. element concentrations compared to traditional ice core sampling from the ICP-MS system. We address gaps in the data >6 cm at: 60.97-61.04 m (7 cm.), 59.44-59.51 m (7 cm), 58.93-59.01 (8 cm), 52.86-52.95 m (9 m), and 58.6-58.74 m (14 cm).

Time Scale Development
The original CG ice core chronology was established by annual layer counting using a combination of continuous flow analysis Ca 2+ and ultrahigh-resolution LA-ICP-MS single element 44 Ca in deeper sections of the core paired with previously well-known horizons such as SDEs and the 1963 bomb radioactivity layers ( Figure S1; Table S1; Bohleber et al., 2018). Since this work, we found more accurately identified age markers in the original annual layer counting (Table S1, Figure S1), including a decrease in lead emissions to natural levels precisely during the Black Death (~1349-1353 CE;  and a large sulfur/sulfate peak in the 930s, associated with the Icelandic Eldgja eruption (Oppenheimer et al., 2018). The dating uncertainties are adjusted from Bohleber et al. (2018) based on the resolution of ICP-MS measurements at the time marker depth resulting in 5 and 10 years, respectively, for the sudden decrease in lead emissions and the Eldgja sulfur/sulfate peak. In addition, we have refined the original CG time scale with the discovery of a tephra chronological tie point for the 536 CE eruption Loveluck et al., 2018). Rhyolitic tephra was identified in a melted ice sample between the depth of 57.67-57.88 m, corresponding to a large peak of S and SO 4 −2 detected at this depth interval (Loveluck et al., 2018). Based on the geochemical composition of the two tephra particles, Loveluck et al. (2018) concluded that it is similar to the 536 CE volcanic event tephra in the NEEM 2011 S1 ice core (Sigl et al., 2015). Based on the preliminary time scale by Bohleber et al. (2018), this depth interval was approximately dated to 470-500 CE with an uncertainty of 72 years based on the dating uncertainty. By recounting annual layers from the Eldgja time marker at 939/940 CE, the age scale was adjusted by about 50 years at the identified depth interval to 520-550 CE with an established age uncertainty of 35 years, accounting for the resolution of the ICP-MS data and the length of the melted core where the tephra was detected. Using our adjusted age scale, the CG record goes back to at least 1 CE at 61 m (abs) with an age uncertainty of 75 years (based on the latest adjustment) with the possibility to extend back even further Loveluck et al., 2018). Future research will be extending the deeper part of the record to the maximum 14 C date available (4421-3907 cal year BP;~71.73 m) near the bottom of the ice core .

Proxy Saharan Dust Event Record Development
We apply a principal component analysis to the entire ice chemistry (coregistered 26 elements, chloride, nitrate, and a sulfate concentrations) data set (annually-resampled for 1780-2006 CE) to reduce dimensionality and find correlations between major and trace elements measured, allowing the identification of significant patterns or relationships within the chemical time series using the sci-kit (v0.20.2) module in Python3.6. The top three resulting principal components (PCs) and their total percent variance (≥10%) are utilized to distinguish air-mass sources to CG and the European Alps. Anthropogenically emitted pollutants, Pb and Cd, are associated with the third principal component (PC3) capturing 11% of the total variance. PC3 exhibits a notable increase during the 1900s after leaded gasoline was introduced, and a sudden drop after the 1980s attributed to lead abatement as previously noted by Schwikowski et al. (2004) in a Mont Blanc firn core. Detailed analysis of Pb from the 2013 CG ice core is discussed in . Known marine aerosols: Na, Cl, and K are the major constituents to the second principal component (PC2), accounting for 14% of the total variance in the record (Figure 3). This elemental pattern has been recognized previously by a principal component analysis performed on snow chemistry from the French Alps (Maupetit & Delmas, 1994) and at Fiescherhorn glacier in the Swiss Alps, where K + is most highly correlated with the sea-salt elements Na + and Cl − (Schwikowski et al., 1999).
Major crustal elements Fe, Al, Mg, and Ca are associated with PC1 accounting for~50% of the total variance ( Figure 3). In agreement with previous studies documenting the elements Fe, Ca, Al, and Mg as a part of the chemical signature of Saharan dust, we associate the crustal elements in PC1 with a Saharan origin (e.g., Angelis & Gaudichet, 1991;Avila & Penuelas, 1999;Avila et al., 1997;Bohleber et al., 2018;Maupetit & Delmas, 1994;Thevenon et al., 2011;Zhu et al., 1997). The major dust element Fe has the highest percent variance in PC1 with~80% variance decomposition (Table S2) and has a well-studied presence in the chemical composition of Saharan dust (e.g., Zhu et al., 1997;Guieu et al., 2002); therefore, we establish it as the prime indicator for SDEs in our study.
We use the ICP-MS Fe concentration measurements for 1780-2006 CE to calibrate modern CG SDE ( Figure 4a; Table S3). We then correlate the ICP-MS Fe concentration to LA-ICP-MS 56 Fe intensity measurements for the period of 700-1820 at a 20-year resolution (Gaussian filter, 1σ; r = 0.89, p = <0.01; Figure S2). We chose this resolution and time period due to the low resolution of the ICP-MS measurements, below 700 CE, where the ICP-MS Fe data contains less than four measurements per 20-year interval. Using the significant positive correlation to the ICP-MS record and SDEs found in our modern record as support, we created the proxy Saharan Dust Event (pSDE) record using LA-ICP-MS 56 Fe intensity measurements for 1-1820 CE as our primary proxy for Saharan dust transport (Figure 5a). Due to the ultrahigh-resolution record resulting from the LA-ICP-MS, the difference in number of measurements per year ranges is~170 (mean) to 1,200 (max); thus, we have the ability to resample the LA-ICP-MS 56 Fe intensity measurements to 0.02 year or 50 data points per year. Weak pSDEs are defined greater than 0.65 percentile above background dust levels (total mean of pSDE record) and below the 0.9 percentile above background; while strong pSDEs are defined as greater than 0.9 percentile above the background dust levels (Figure 5b). We assess the results of the SDE record in section 3.2 and the pSDE record in section 3.3.  1977, 1943, 1937, 1994, and 1984 while 11 of the 12 lowest annual concentrations (<0.05 percentile, 0.75 μg/L) are between the years 1790-1820, also noted as an extremely cold  Table S3. b. Top 3 SDE years (1984, 1994 post 1979 from compiled ERA-Interim Reanalysis data for MSLP during JJA. c. Comparison of CR20 v2 MSLP (red) smoothed from the Mediterranean region (33-48N;0-18E) to annual smooth ICP-MS Fe (black) for 1870-2007, with a statistically significant r-value of 0.66.

10.1029/2019JD030725
Journal of Geophysical Research: Atmospheres period in Europe (Lamb, 1995). The largest annual peaks (>0.95 percentile, 35 μg/L) interpreted as large SDEs match previously recorded periods of increased dust input to Central Europe, shown in Table S3. Among these years of large SDEs, several coincide with the current Sahara/Sahel drought that started in the 1970s, (Prospero & Lamb, 2003). A dust deposition record from northwest Africa, based on a marinederived grain-size distribution of terrigenous sediments (>10 μm), also shows strong similarities to our record over the past century (Mulitza et al., 2010), apart from the increasing trend since 2000 during where our record is decreasing similar to a wind field controlling most of the North African dust emission variability Northern Africa (Evan et al., 2016).
In order to explore atmospheric circulation systems associated with Saharan dust transported to the European Alps, we describe the relationship of our record to synoptic conditions and factors affecting the North Atlantic/Mediterranean region. Dominant air-mass sources during the summer months to the CG region are Saharan dust-laden winds (Figure 1) via northeastward transport across the Mediterranean (Type 1) and less frequent transport northward along the coastline or over the eastern North Atlantic, carried eastward by the westerlies to CG (Type 2), as discussed in Barkan et al. (2005), and Varga et al. (2013). These two modes of transport are confirmed in back trajectories of CG by Thevenon et al. (2009) andSchwikowski et al. (1995). Based on the synoptics of dust transport in Barkan et al. (2005) and Varga et al. (2013), Type 1 transport is likely to occur when a high-pressure system is located over NW Africa to central Europe, which is indicative of a strong Azores high, a northerly shift of cyclones in the North Atlantic and decreased westerly humidity advection to Morocco (Esper et al., 2007). This pattern is also observed in scaled anomaly composites during Feb-June 500 hPa patterns for the 15 driest years in Morocco from 1659-2001 (Esper et al., 2007). Mean sea level pressure (MSLP) from the 3 largest SDEs in our record after 1979 (1994,2000,1984), show a similar pattern over Sardinia caused by a higher than average pressure system presiding over Tunisia and Northern Algeria and into the Mediterranean (Figure 4b). From 1870-2006, the annually resolved SDE proxy has a statistically significant correlation (r = 0.66, p ≤ 0.05) to the MSLP over the Mediterranean (35-49N; 0-25E) derived from 20CR v2 data set (Figure 4c). The 20CR v2 is a reanalysis dataset from NOAA's Earth System's Research Laboratory and Physical Science Division (ESRL and PSD,  Whitaker and Hamill (2002). The dataset runs from late 19th to the 21st century. The dataset is available online through NOAA's ESRL PSD webpage. This correlation along with MSLP anomalies from climate reanalysis data shows Saharan dust is likely to be transported to the European Alps during a stronger Azores High leading to a higher-pressure system residing over NW Africa and the Mediterranean.

pSDE Variability From 1 to 1820 CE
Using the modern record as an analog for climatological conditions favoring periods of notable SDEs, we extend the CG record back to 1 CE using our pSDE record. Figure 5a demonstrates the 0.02-year (gray) and annual (black) LA-ICP-MS 56 Fe (Gaussian [2σ], normalized), highlighting the four most significant periods (multi-decadal periods with multiple decades greater than 0.9 percentile of strong pSDE). We outline the number of strong pSDEs (above 0.9 percentile above the background dust levels) and weak pSDEs (between 0.65 and 0.9 percentiles above the background) occurring per decade in Figure 5b, while thresholds are shown in Figure 5a. The four periods include (CE): 140-170, 370-450, 870-1000, and 1320-1360, along with the following decades greater than the 0.9 percentile of strong pSDEs: 570s, 60s, 670s, 680s, 710s, 810s, 1180s, 1200s, 1260s, 1460s, 1520s, 1590s, 1770s, and 1780s. The decade with the highest frequency of strong pSDEs occurs in the 430s (193 strong pSDEs), this episode that lasted from 370 to 450 starts with an increase in weak dust events starting in the 400s and strong pSDEs beginning in 410s, then decrease after 430s until the 450s. The longest period of continuous weak and strong pSDEs occurs from 870-1000 CE, consisting of 871 weak pSDEs and 526 strong SDEs. The decades with the most frequent strong pSDEs during the 870-1000 CE period include: 920s (141), 960s (124), and 910s (100). In the 20-year resolution ICP-MS Fe record ( Figure S5), we detect similar above average periods of SDEs (as compared with the LA-ICP-MS 56 Fe record) during 110-170, 360-440, 900-1000 in addition to 1600-1660 (which we do not observe in the pSDE record) and 1880-2000. We additionally observe the increase in pSDEs during the mid-1300s that is not in agreeance with our ICP-MS Fe record.
The variability between the two records might be reflecting methodological differences between the two analyses, therefore the profiles do not mirror each other for every case. The differences include the state of the "ice" when sampling: the ICP-MS measures a melted section of ice, in this case about 1-3-cm resolution while the LA-ICP-MS measures a cleaned surface of solid ice at a much finer resolution. We do not claim they show the same measurements, yet they show similar trends and have a good correlation to each other. This paper is the first time we have used the LA-ICP-MS system to produce a continuous record longer than a meter, and we believe the general comparison to the ICP-MS data shows satisfactory similarities between the two different measurements.
During three of the four periods with the highest number of weak and strong pSDEs (130-170, 370-450, and 870-1000 CE), there is a decrease in annual layer thickness ( Figure S3) that could suggest enhanced snow melt, creating the appearance of abrupt increases and terminations of large pSDE episodes, likely related to accumulation rate or Fe delivery/flux variability. According to Gabbi et al. (2015), high input of Saharan dust during individual years can reduce surface albedo substantially producing a strong influence on snow and ice ablation, while average years of Saharan dust input cause a negligible effect on surface albedo. In addition, enhanced dust concentrations could be a result of reexposed firn layers with high impurity content due to periods of increased melting (Gabbi et al., 2015). The accumulation rate is variable throughout the core but decreases coevally with three of the largest pSDE episodes. Based on our measurements of Saharan dust variability over the past two millennia, we observe variations of Saharan dust at a decadal scale that do not show a consistent pattern, but that do coincide with the onset of the MCA, a known major climate transition in European climate.
attributed to anthropogenic influences such as land use and greenhouse gas emissions (Mann, 2007). In the case of Saharan dust transport to CG, Thevenon et al. (2009) touch on possible forcing mechanisms (solar, volcanic, and anthropogenic) and macroscale ocean-atmospheric teleconnections (NAO, ENSO, and AMO) that have been previously proposed as influences. Here, we attempt to further expand upon the influences of northward SDE transport for multidecadal timescales. To investigate the role of macroscale ocean-atmospheric teleconnections during the modern period, we correlate the CG SDE record (annual ICP-MS Fe record, 1780-2006 CE) to high resolution proxy reconstructions from adjacent regions at 5-and 10-year resolutions (Table S4; Figure S4). We additionally compare our pSDE record to high resolution paleoclimate reconstructions (Table S5; Figure S5), with a focus on the onset of the MCA in section 4.2 ( Figure 6).

Solar and Volcanic Forcings
Statistically significant correlations (Table S4) to the Lean et al. (1995) solar irradiance record for 1780-1995 CE with our SDE record ( Figure S4) demonstrates the strong influence of solar forcing on overall trends. Solar forcing is an important influence on European and North Atlantic climate, notably relating to a stronger Azores High (positive NAO), with a lag of 2-4 years (Gray et al., 2013;Gray et al., 2016;Ma et al., 2018). Solar forcing can also influence NH surface heating (Lean et al., 1995), leading to arid conditions in the Sahara. The relationship between pSDEs and solar activity reconstruction (Steinhilber et al., 2012) from 800-1800 CE ( Figure S5), results in a positive correlation (Table S4) which shows total solar irradiance is above average during periods of strong pSDEs, during 140-170, 800-1000, and mid-1300s.
Although limited research has been conducted connecting SDE and volcanic forcing, we speculate on the possibility of volcanic forcing creating suitable conditions for increased SDEs. Volcanic aerosol loading in the stratosphere is associated with positive NAO patterns, cool sea surface temperature (SST) anomalies, and a southward shift in the ITCZ, which is in turn associated with increased aridity in Northern Africa, cooling over the North Atlantic, and more frequent and/or more intense ENSO events (Birkel et al., 2018;Clement et al., 2000;Haug et al., 2001;Moy et al., 2002;Pausata et al., 2015). NH high latitude eruptions produce changes in atmospheric circulation by creating a blanket of ash particles which block solar radiation thereby causing regional cooling. In the NH summer following an eruption, a weakened African summer monsoon results in decreased precipitation and consequently the flow of the Nile and Niger Rivers (Oman et al., 2006).
While long term impacts of Northern Hemispheric volcanic eruptions on Saharan aridity are under-researched, we observe increases in the average occurrence of pSDEs within the 10 years after the eight largest NH eruptions in the Common Era from Sigl et al. (2015; Figure S6). Specifically, the increase of pSDE events following the Eldgja (939/940; Oppenheimer et al., 2018) and Laki (1783/ 1784; Thordarson & Self, 2003) events (pSDE, Figure 5; SDE, Figure 4), respectively, coincide with reduced Nile River flow due to decreased regional precipitation (Manning et al., 2017;Oman et al., 2006). High peaks in Nile River discharge (stronger monsoon intensity) occur during minima in Saharan dust flux records from Western Africa on a multimillennial timescale over the past 240 ka (Skonieczny et al., 2019). Research by Manning et al. (2017) also reveals a decrease in precipitation across the Sahel resulting from a weakened ITCZ due to 20th century eruptions based on CMIP5 modeling results. Additionally, significant correlations result for comparisons of NH Volcanic Forcing (Sigl et al., 2015) with the SDE record for 1780-2000 ( Table   Figure 6. Onset of Medieval Climate Anomaly Proxy Comparison. Common Era climate forcings/ocean-atmospheric teleconnection proxy comparison with a) NH volcanic forcings based on nssS from the NEEM ice core (Sigl et al., 2015), total solar irradiance record (Steinhilber et al., 2009), reconstructed ENSO variance record (Li et al., 2011), extended reconstructed NAO index (Olsen et al., 2012), reconstructed AMO temperature anomolies (Mann et al., 2009), reconstructed European JJA temperature anomalies (Luterbacher et al., 2016; gaussian smooth applied), North Atlantic August SST (Miettinen et al., 2015), and dust deposition record from northwest Africa, based on a marine-derived grain-size distribution of terrigenous sediments (>10 μm; Mulitza et al., 2010) compared to smoothed LA-ICP-MS 0.5 year resolution 56 Fe data from this study (gray).
The interval 900-1000 CE is highlighted in blue.

10.1029/2019JD030725
Journal of Geophysical Research: Atmospheres S4; Figure S4) and the pSDE record for 800-1800 CE (Table S5) suggesting a possible association between NH volcanic forcing and northward Saharan dust transport on multidecadal scales.

North African Winds
The connection between two Saharan dust transport pathways, eastward across the North Atlantic and northward across the Mediterranean (Moulin et al., 1997), is further supported by positive correlations (Table S4; Figure S4) between our SDE record and a Saharan dust emission proxy record for westward transport over the Atlantic for 1850-2005 CE (Evan et al., 2016). The time series produced by Evan et al. (2016) is derived from NOAA-CIRES 20th Century Reanalysis monthly mean 10-m wind fields accounts for the magnitude and the direction of the wind field controlling most of the North African dust emission variability ( Figure S4). This suggests that atmospheric phenomena such as ITCZ, NAO, and ENSO associated with increased wind across the North Atlantic, may in turn, also cause increased dust transport across the Mediterranean, as previously suggested by Prospero (1996) and observed in Özsoy et al. (2001). In addition, Saharan dust-laden winds transporting northward are associated with warmer temperatures in Europe , further supported by positive correlations between our SDE record and European temperature (Table S4, Table S5; Figure S4; Figure S5) from Luterbacher et al. (2016).

Macroscale Ocean-Atmospheric Teleconnections
Prolonged periods of wet and dry climate in the Sahel and changes in the African summer monsoon on decadal time scales are influenced by global and regional SSTs (Giannini et al., 2003). On a regional scale, warmer (cooler) Mediterranean SSTs increases (decreases) rainfall over Northern Africa as increased moisture is advected southward over the Sahara (Rowell, 2003). On a macroscale, colder SSTs in the North Atlantic are associated with increased dust transport traveling eastward through feedback mechanisms. Wang et al. (2012) state that southward shifts in the ITCZ during periods of cooler North Atlantic SSTs lead to a decrease in rainfall in the Sahel, followed by an increase in dust production which further cools the North Atlantic Ocean during transport, creating a positive feedback loop. The resulting negative correlation to ERA-Interim reanalysis June to August SSTs for the 1979-2006 in the North Atlantic, off the coast of NW Africa ( Figure S7), supports covariability between North Atlantic SSTs and northward transport of Saharan dust into Europe. Contrary to previous work, our pSDE and SDE records do not show significant correlations to long-term changes in the North Atlantic SSTs (Mann et al., 2009); rather, two large pSDE intervals (880-950 and 1330-1360 CE) match with below average North Atlantic SSTs (negative AMO).
Sahelian drought has been connected with increased Saharan dust transport across the Atlantic (Prospero & Nees, 1986;Prospero et al., 1995;Prospero, 1996;Prospero & Lamb, 2003), whereas its impact northward across the Mediterranean is less well understood. To determine the influence along the latitude of western Africa (20W-10N), we performed a moving correlation for precipitation rate (mm/day) using 20CRV2c data segmented into four areas from 10 to 35N ( Figure S8). Our results show that decreased precipitation in both the Sahel and Sahara play a role in Saharan dust transport northward across the Mediterranean. During the 1880s, a significant negative correlation between precipitation rate at 15-30N and CG SDE corresponds with an increase in precipitation during a decrease in SDEs. Note the significant positive correlations to Sahel precipitation from 1912-1924. Lastly, the negative correlation with decreased precipitation rate at 25-35N and increased Saharan dust during 1975-2006, corresponds with the Sahelian drought since the 1970s (Prospero & Lamb, 2003). The running average correlations suggest that over the past 30 years or so, northward Saharan dust transport more closely reflects precipitation in the Sahara compared with the Sahel but this covariability changes over time.
We do not find a statistically significant association between our SDE record and the standard winter NAO reconstructions (Trouet et al., 2009;Table S4; Figure S4) during the modern era , reflecting the strength of the summer signal at CG (Lugauer et al., 1998), and the European Alps are in an area variably influenced by the NAO (Casty et al., 2005). However, the SDE record is significantly correlated to the summer NAO (sNAO) reconstruction ; Table S4, Figure S4). The sNAO is an empirical orthogonal function pattern of observed summertime extratropical North Atlantic MSLP, located farther north than the winter NAO, that exerts a strong influence on European climate and influences summer climate extremes such as flooding, drought, and heat stress . Although the sNAO is well correlated to increased wet weather in the Mediterranean, this pattern has a negative correlation to precipitation in the Sahel, suggesting that a positive sNAO corresponds with Sahelian droughts and a weaker West African Monsoon Linderholm et al., 2009). Summer transport of Saharan dust having a weak connection to the winter NAO index and a stronger correspondence to sNAO supports previous work (Doherty et al., 2008;DeFlorio et al., 2016). The CG pSDE record therefore suggests that the sNAO has a stronger impact on dust transport compared to the NAO.
Although previous studies suggest North African dust concentrations are associated with ENSO (Prospero & Nees, 1986), our SDE record does not yield a statistically significant correlation (Table S4; Figure S4) to the reconstructed ENSO record from Li et al. (2011). ENSO events identified in Prospero and Lamb (2003;1972-73, 1982-83, 1986-87, 1991-92, and 1997, however, occur before major SDEs in our annual post-1780 record, suggesting that although overall trends in ENSO variability disagree, strong winter ENSO events could indirectly affect summer SDE transport through atmospheric teleconnections. The changes associated with strong ENSO circulation patterns can lead to increasing wind speed over tropical Africa and periods of decreased North African/Sahel rainfall (Folland et al., 1986;Janicot et al., 1996). Our data suggests that ENSO variability may have a stronger influence on transport across the North Atlantic compared with the Mediterranean. In relation to ENSO and large-scale atmospheric teleconnections, Thevenon et al. (2009) discovered a potential link between CG SDEs to the South Asian monsoon through a Himalayan ice core (Thompson et al., 2000). Our resulting positive correlations, r-value = 0.59, p-value = 7.9*10 −6 (10-year resolution for 1500-1990, to the Himalayan ice core dust record ( Figure S9; Thompson et al., 2000), supports evidence of a connection between northward Saharan dust transport to large-scale teleconnections affecting the South Asian monsoon including the ITCZ.
Based on our analysis, the CG ice core-derived pSDE record is strongly influenced by solar forcing and the sNAO and is associated with warmer temperatures in Europe. Previous studies and our record provide evidence for a causal climatic relationship between the European Alps, North African hydrology, and North Atlantic SSTs represented by SDEs. The weak correlations resulting from our initial comparisons to other proxies suggest there is more involved in dust deposition than what previous studies have proposed.
Based on comparison to multiple paleoclimate records noted in Figure S5, we further suggest that there is not one specific cause of multi-year frequency for large SDEs, but rather multiple influences that create favorable conditions for increased frequency or magnitude of SDEs. Further research beyond our comparisons is necessary to elucidate the complex influences on Saharan dust variability.

Onset of the Medieval Climate Anomaly
The longest-occurring interval of large SDEs (870-1000 CE; Figures 5 and 6) is coincident with onset of the MCA (900 and 1400 CE; e.g., Lamb, 1965;Bradley et al., 2003;Mann & Jones, 2003;Mann et al., 2009;Berner et al., 2011), a period with widespread anomalous droughts, noted in multiple paleoclimate records in Europe, Africa, Asia, and the mid-latitudes of both Americas (e.g., Bhattacharya et al., 2017;Pederson et al., 2014;Rein et al., 2004). On a regional-scale, paleoclimate reconstructions for Africa yield conditions favoring or indicating increased production and transport of dust events during 800-1050 CE Most notably, the 3,200-year dust deposition record (terrigenous; >10 μm) from the coast of northwest Africa (Mulitza et al., 2010; previously discussed in 3.2) similarly exhibits its largest dust peak (pre-1780) during the mid-900s. Dry conditions between approximately 900-1000 CE are indicated by an oxygen isotope record of Lake Bosumtwi in Ghana (a proxy for Sahel precipitation; Shanahan et al., 2009) and Moroccan lake level fluctuations (Détriché et al., 2009). Based on high-resolution marine records extracted from the Algerian-Balearic basin (Moreno et al., 2012), higher aeolian African input is noted in geochemical proxies during the onset of the MCA. And an extremely low frequency of flooding characterizes the period 930 to 1070 in a Nile flood record, further suggesting dry conditions in northern Africa (Hassan, 2007). The concurrence of intense drought events in multiple paleoclimate records across the North Atlantic and along the midlatitudes indicates that these occurrences are linked through ocean-atmosphere teleconnections.
The centennial-long episode of strong pSDEs during the onset of the MCA suggests there was a major change in macroscale NH atmospheric teleconnections, favoring sustained drought throughout Africa. Aboveaverage temperatures are documented in the European temperature anomaly data (Luterbacher et al., 2016) while the few NAO Index reconstructions that extend back to at least 800 CE suggest a strong Azores High (positive NAO) between 800 and 1200 CE ( Figure S5, Figure 6; e.g., Olsen et al., 2012;Wassenburg et al., 2013;Baker et al., 2015;Faust et al., 2016). The onset of the MCA (870-1000) is indicated as the coldest interval of high-resolution reconstruction documented in diatom-inferred records of August 10.1029/2019JD030725

Journal of Geophysical Research: Atmospheres
SSTs (Miettinen et al., 2015). In addition, the high resolution AMO reconstruction (Mann et al., 2009) shifts from a colder (more negative) to a warmer (more positive) phase occurring at 960 through 1000 (onset of the MCA). North Atlantic SST changes, climatic conditions specified by African hydroclimate records, and a positive phase NAO suggest this is a regionally dry period of the MCA. Based on the time period in which this episode of strong pSDEs occurs, we suggest that more frequent input of large pSDEs starting in the 800s could be part of a feedback loop or trigger that caused the MCA in Europe. The abrupt arrival and decline of the episode of strong and weak pSDEs during the onset of the MCA (also seen in ICP-MS record, Figure S5) could be indicative of a precursor shift that relayed conditions favorable for pSDEs. The appearance of abruptness and prolonged strength of pSDEs could also be influenced by changes in surface albedo due to increased dust deposition, yet another indicator that this period is anomalous (Gabbi et al., 2015). Further research on our pSDE record in concert with high resolution paleoclimate proxies is necessary to fully understand the pSDE relationship to the MCA.
Interestingly, this anomalous period of increased large SDEs also coincides with the period generally identified as the economic takeoff of France, Germany, Britain and neighboring regions (~900 to~1100). Historians and archaeologists have adduced multiple causes and actors (Bartlett, 1993;Loveluck, 2013), but most agree it started with a medieval Agrarian Revolution in the countryside producing more food to feed more people (White, 1962). Historians are now exploring whether improving climate conditions may have contributed to the Agrarian Revolution (Williamson, 2013;Brooke, 2014, pp. 350-351, pp. 358-364, etc.) The chronological convergence of the SDEs anomaly, some of its salient features, and its geographic scope suggest a link with the climate reorganization in the MCA and, through it, with the takeoff of European food production that fueled Europe's demographic and economic expansion. Acknowledging the context of societal changes with exceptional climate-related events is critical to develop more accurate interpretations of increasingly higher resolution paleoclimate data over the Common Era.

Historic Implications of Ultrahigh-Resolution SDEs
Written records that capture evidence of events or interannual changes in climate and pollution are highly valuable when used in concert with paleoclimate data, especially when cutting-edge methods of glaciochemical analysis can reveal environmental signals at a subannual scale, as shown in More et al. (2018). This multidisciplinary approach led to the development of a European lead pollution record for the past two millennia . Using our novel LA-ICP-MS-derived pSDE record, we detect and synthesize environmental signals from the ultrahigh-resolution glacio-chemical time series in concert with the written historical records to reveal new multidisciplinary analysis of paleoenvironmental signals. For example, the years 1325-1365 CE stand out in our record, showing one of the highest concentrations of SDEs ( Figure 5), while historical records show the 1300s were a time of extreme events beginning with the Great Famine (1315-1317) and ending with the Black Death (1346-1353). Further research should seek to answer whether SDEs further exacerbated a decline in human health by decreasing air quality or if they can be used as indicators for extreme meteorological events leading to human suffering. We focus on the period of 1325-1365 CE using our raw 120-μm LA-ICP-MS 56 Fe SDE proxy to further examine the possibility of connecting written historical testimonies to dust events analyzed in our record.
The Great Famine (1315-1325 CE) predated the occurrence of large pSDEs. Historical evidence shows a period of weather deterioration in this decadesuch as torrential rains, floods, mud slides and cold temperatureswhich resulted in the death of at least one tenth of the population in northern Europe (Jordan, 1996;Dawson et al., 2007;Jordan, 2010;Campbell, 2010). Our pSDE record is predominantly flat during this period, suggesting SDEs were scarce (with the exception of a moderate event in 1319-1320 CE). Based on the patterns described previously, this would suggest a period of increased precipitation, which in fact matches the historical evidence.
Following a small break in the CG glacio-chemical record (Figure 7) (Table S6). Blood raina phenomenon whereby red, iron-rich Saharan dust falls with rain, appearing similar to blood to contemporaneous observerswas recorded in 1348-1349 CE in Germany (Table S6).
The shift in occurrence of large SDEs could be explained by a prolonged drought followed by increased winds and atmospheric circulation.
Marking the end of the MCA, several Moroccan climate reconstructions covering the mid-1300s suggest a sudden shift from wetter to drier conditions (Brahim et al., 2017;Esper et al., 2007;Wassenburg et al., 2013). For southwestern Morocco, 1350 CE is the driest peak in the entire speleothem record ranging from 831 ± 27 CE to 1935 ± 25 CE (Brahim et al., 2017). Other indicators of a changing climate during this time interval include below average North Atlantic SSTs (negative AMO) between the interval 1330-1360 CE (Mann et al., 2009), coinciding with the first peak in LIA alpine glacier advances from the Great Aletsch Glacier located in the Swiss Alps in Europe (Holzhauser et al., 2005). Abrupt changes in atmospheric circulation during the period of 1325-1365 CE are reflected in changes in the frequency and magnitude of SDEs in addition to evidence found in historical written records.

Conclusions
This study presents the longest continuous, ultrahigh-resolution ice core record yet produced, utilized as a proxy for Fe-bearing SDEs via southerly air masses to the European Alps derived from North Africa over the past 2000 years. Environmental signals are well preserved in the CG ice core; thus, the combination of multiple ice core sampling techniques creates a reliable record to interpret the long-term behavior of Saharan dust transport in relation to its occurrence, duration, and magnitude. This study substantially extends in temporal resolution and length of record, the knowledge derived from past ice core studies focused on SDEs transported to CG and the European Alps regions and demonstrates the value of ultrahigh-resolution sampling for future ice core records.
Our data suggests the following for SDE variability: 1) interannual to decadal Saharan dust transport to Europe does not vary in a consistent pattern; rather, multiyear episodes of frequent, large dust events occur irregularly; 2) covariation between eastward transported North Atlantic and northward transported Mediterranean SDEs during summer months is evident from correlations with North African wind patterns as originally suggested by Evan et al. (2016); 3) multidecadal Saharan dust transport demonstrates variability similar to key climate transitions in European climate such as the MCA; 4) multidecadal scale variations in Saharan dust transport allow for further analysis of the influence of multiple ocean-atmosphere teleconnections associated with the North Atlantic including the NAO, AMO, ENSO, and the ITCZ; 5) high latitude volcanic activity may result in conditions that are suitable for the amplification of anomalous Saharan dust transport to Europe; and 6) ultrahigh-resolution ice core chemistry measurements can be used in concert with historical written records to improve interpretations of human-climate interactions. These findings offer new insights into annual to decadal scale Saharan dust variations during the modern and pre-industrial periods in the European Alps region.
In the past, major dust episodes have been related to Northern African drought accompanied by increased wind speed caused by changes in the pressure-temperature gradient. Based on our study, we predict that warming will result in less frequent but more intense SDEs in the future. In particular, a weaker Azores High (negative NAO) and increased warming in the North Atlantic, due to increased CO 2 emissions, will (black) 56 Fe intensity (120-μm). Number of historical events that account for drought and heat events in Europe shown in red shown with right axis (Alexandre, 1987;Buisman, 1995;Weikinn, 2002). Orange stars signify historical records of dust events.
lead to fewer SDEs across the Mediterranean and subsequently across the North Atlantic. A warming Arctic and increased loss of sea ice will further amplify this effect by weakening the temperature-pressure gradient between the mid and high latitudes, resulting in an even weaker Azores High (negative NAO state) (Magnusdottir et al., 2004). A warming climate will impact the movement of the ITCZ. On the other hand, intense droughts caused by a reduction in precipitation and increased temperature over northern Africa will lead to increased dust availability and lead potentially to massive periods of dust uplift, causing harm to population centers located in the transport pathways by reducing air quality, as already shown in published case studies of PM 10 , PM 2.5 and other air quality proxies in these areas (De Longueville et al., 2013;Dimitriou & Kassomenos, 2018;Prospero & Mayol-Bracero, 2013).