Volcanic impact on the Atlantic ocean over the last millennium

The oceanic response to volcanic eruptions over the last 1000 years is investigated with a focus on the North Atlantic Ocean, using a fully coupled AOGCM forced by a realistic time series of volcanic eruptions, total solar irradiance (TSI) and atmospheric greenhouse gases concentration. The model simulates little response to TSI variations but a 5 strong and long-lasting thermal and dynamical oceanic adjustment to volcanic forcing, which is shown to be a function of the time period of the volcanic eruptions, probably due to their di ﬀ erent seasonality. The thermal response consists of a fast tropical cooling due to the radiative forcing by the volcanic eruptions, followed by a penetration of this cooling in the subtropical ocean interior one to ﬁve years after the eruption, and 10 propagation of the anomalies toward the high latitudes. The oceanic circulation ﬁrst adjusts rapidly to low latitude anomalous wind stress induced by the strong cooling. The Atlantic Meridional Overturning Circulation (AMOC) shows a signiﬁcant intensiﬁcation 5 to 10 years after the eruptions of the period post-1400 AD, in response to anomalous atmospheric momentum forcing, and a slight weakening in the following decade. In re- 15 sponse to the stronger eruptions occurring between 1100 and 1300, the AMOC shows no intensiﬁcation and a stronger reduction after 10 years. This study thus stresses the diversity of AMOC response to volcanic eruptions in climate models and tentatively points to an important role of the seasonality of the eruptions.

Several studies have also began to point out the specificity of the thirteenth century in terms of intense volcanic activity, and the possibility for a cumulative impact on the ocean (e.g. Zhong et al., 2010). The second half of the thirteen century is indeed the most perturbed half century of the past 1500 years (Jansen et al., 2007). In two out of four simulations, Zhong et al. (2010) found a centennial-scale climate change 15 following the succession of decadally paced eruptions following the 1257-1258 megaeruption. They highlighted a coupled ice-ocean interaction between the subpolar North Atlantic, a reduced extension of the AMOC into the northern North Atlantic, and the Arctic ocean, which maintained significantly expanded sea ice and reduced surface air temperatures for at least 100 years. However, the feedback mechanism depended 20 on other factors since it was only activated in half of the simulations. In this context, it is important to highlight that while both Stenchikov et al. (2009) andOttera et al. (2010) found a significant intensification of the AMOC following volcanic eruptions, the first study was based on sensitivity experiments following single eruptions of different intensities and the other based on composite analysis over the last 600 years of the 25 millennium, thereby excluding the particular succession of events of the thirteen century.
Here, we explore the interannual to decadal oceanic response to volcanic activity in a coupled OAGCM forced by a full set of reconstructed external forcings over the last  Sicre et al. (2011) have shown that the simulated sea surface temperature (SST) in the northern North Atlantic compares well with a recent high resolution SST reconstruction off Iceland. We propose to describe more thoroughly the oceanic response to the major volcanic eruptions of the last millennium and investigate the mechanisms for the oceanic circulation adjustment. The model configuration and the 5 forcings are presented in Sect. 2. The oceanic response to solar and volcanic forcing are compared in Sect. 3 and the temperature response to volcanic eruptions is discussed in Sect. 4. In Sect. 5, we investigate the response of the Atlantic circulation to isolated volcanic eruptions (occurring after year 1400) and in Sect. 6, we highlight the differences with the twelfth and thirteenth century. Conclusions are given in Sect. 7.

The coupled model
We use the IPSLCM4 v2 climate model developed at the Institut Pierre-Simon Laplace (Marti et al., 2010). This model couples the LMDz4 atmosphere GCM (Hourdin et al., 2006) and the ORCHIDEE 1.9.1 module for continental surfaces (Krinner et al., 2005) 15 to the OPA8.2 ocean model (Madec et al., 1998) and the LIM2 sea-ice model (Fichefet and Maqueda, 1997), using the OASIS coupler (Valcke et al., 2000). The resolution in the atmosphere is 3.75 • in longitude, 2.5 • in latitude, and 19 vertical levels. The ocean and sea-ice are implemented on the ORCA2 grid (averaged horizontal resolution 2 × 2 • , refined to 0.5 • around the equator, 31 vertical levels). In all simulations, the 20 vegetation was set to a modern climatology from Myneni et al. (1997). After a 310 year spin up with preindustrial greenhouse gases (GHG) concentrations and tropospheric aerosols, two simulations were run. The first one is a 1000-year control simulation (CTRL) with the same preindustrial conditions as the spin up, also used in Servonnat et al. (2010). The main characteristics of the AMOC in the model and its sensitivity Introduction an excess of freshwater flux over the Labrador Sea was responsible for the lack of deep convection in this region and the relatively weak AMOC (11 Sv) in the model. Deep convection in the northern North Atlantic only takes place in the Nordic Seas and south of Iceland (Marti et al., 2010). The natural variability of the AMOC, its link to deep convection and its impact on the atmosphere have been studied by Msadek 5 and Frankignoul (2009). They showed that the multidecadal fluctuations of the AMOC are mostly driven by the deep convection in the subpolar gyre with a time lag of 6 to 7 years. Convection in the subpolar gyre is itself primarily influenced by anomalous salinity advection caused by the variability of the East Atlantic Pattern (EAP), second dominant mode of atmospheric variability in the North Atlantic region. The lack of 10 Labrador Sea convection in the model probably explains the dominance of the EAP (as opposed to the North Atlantic Oscillation) in forcing multidecadal variations of the AMOC. The second simulation (LM2SV) was forced with a reconstruction of TSI, GHGs concentrations, changes in orbital parameters, and radiative effect of volcanic eruptions over the last millennium, from 850 to 2000 AD. The choice and implementation of the 15 forcings are discussed below. To reduce the influence of the model drift, a quadratic trend was removed from each variable and grid point. As our main focus is on the oceanic response to volcanic eruptions at interannual to decadal timescales, all data are considered in annual mean, or seasonal mean for such variables as sea ice cover and mixed layer depth.

External forcing over the last millennium
A number of different reconstructions for TSI variations have been produced (e.g. Jansen et al., 2007), mostly differing in the estimated reduction of total irradiance during the 17th century Maunder Minimum, which ranges from 0.08 % to 0.65 % (1.1 to 8.9 W m −2 ) of the contemporary value. Ammann et al. (2007) found that a TSI 25 decrease of about 0.25 % during the Maunder Minimum produces a realistic amplitude of the Northern Hemisphere temperature change in climate models. However, recent progress in solar physics (Foukal et al., 2004;Solanki and Krivova, 2006;Gray et al., 2516 Introduction 2010) imply that the TSI variations between the Maunder Minimum and present day value are about 0.1 %. As this scaling is recommended for the third phase of the paleoclimate modelling inter-comparison project (PMIP III, Schmidt et al., 2011), we use the TSI reconstruction by Vieira and Solanki (2009) and Krivova et al. (personal communication, 2009), which follows it. The corresponding variations of the raw shortwave 5 input at the top of the atmosphere is shown in Fig. 1 (top panel). Large volcanic eruptions inject sulfur gases into the stratosphere, which convert to sulfate aerosols with a residence time of about a year. The aerosol cloud has several effects on radiative processes, most notably by backscattering part of the incoming solar radiation, which induces a net cooling at the Earth's surface (e.g. Robock, 2000). Thus, 10 until recent years, most modelling groups (e.g. Jansen et al., 2007) have represented the volcanic forcing by altering the solar constant. Althoughsuch a coarse approach leads to hemispheric averages that compared reasonably well to a "blend" of proxy and/or instrumental reconstructions (e.g. Goosse et al., 2005;Stendel et al., 2006), it does not properly represent regional and seasonal variations. It is indeed known that 15 the climatic impact of volcanic eruptions highly depends on the season and that latitudinal dependence of the cooling in the troposphere (warming in stratosphere) evolves for at least 2 to 3 years after the eruption. Furthermore, the volcanic aerosols serve as surfaces for heterogeneous chemical reactions that destroy stratospheric ozone, which controls solar energy absorption in the stratosphere. Its variations thus alter both the 20 vertical temperature gradient between the troposphere and the stratosphere and the latitudinal temperature gradient in the stratosphere. We implemented in the IPSL model a new radiative module described in Khodri et al. (2011) that mimics the direct radiative effect of sulphate aerosols. The input time series is based on the monthly mean optical thickness latitudinal reconstruction by Ammann et al. (2003) and Gao et al. (2008) 25 from 850 AD to present. The anomalous optical thickness is implemented in the tropical band between 20 • S and 20 • N, and transported poleward within 3 years according to a spreading function as in Gao et al. (2008). Figure 1 ( a change in the global mean optical depth of 0.1 corresponds to a global anomalous radiative forcing of roughly −3 W m −2 . However, as discussed in Khodri et al. (2011) and Timmreck et al. (2009), the volcanic module tends to overestimate the radiative effect of the mega eruptions because of the use, for paleo-eruptions, of aerosol effective radius and optical depth derived from observations over the instrumental period, 5 specifically for the Mount Pinatubo (1991) and El Chichón (1982) volcanic eruptions. The greenhouse gas concentrations are those inferred from ice cores and direct measurements as reported in Servonnat et al. (2010). This simulation does not include the forcing by anthropic aerosols, so that global warming detected over the last decades of the simulation is overestimated (not shown). Hence, this study focuses on 10 the natural external forcings and the period of investigation is limited to years 850 to 1849 AD.

Temperature response to solar and volcanic forcings
The most striking signal in the time evolution of the air temperature at 2m averaged over the Northern Hemisphere ( Fig. 1 third panel) and the SST averaged over the Atlantic 15 ocean ( Fig. 1 fourth panel) are important variations following volcanic eruptions, in particular an abrupt cooling of up to 3 • C in the atmosphere and 1 • C in the ocean. Such signature has also recently been found in temperature reconstructions in the subpolar North Atlantic (Sicre et al., 2011). On the other hand, variations of the solar insolation do not seem to have a strong imprint. The lagged correlation r of the anomalous 20 TSI time series with the averaged surface air temperature in the Northern Hemisphere and with the Atlantic SST have a broad but weak maximum when the TSI leads by 4 years, reaching r = 0.12 and 0.13 respectively (significant at the 5 % level) (Fig. 2, top panel). The corresponding correlation with the volcanic forcing is much larger, peaking when temperature lags by one year, with r = −0.63 and r = −0.52 respectively (Fig. 2 Fig. 2 is due to the use of annual averages, as eruptions might in fact have started during the calendar year preceding the maximum of emission. The stronger influence of volcanic forcing is probably due to our use of a TSI reconstruction with weak variations and to an overestimation of the volcanic radiative effect (Sect. 2.2). The frequency dependence of the solar correlation is illustrated by the cross-5 wavelet coherence spectra in Fig. 3. The wavelet analysis was made with the Morlet wavelet, and the transform performed in Fourier space, using zero padding to reduce wraparound effects (Torrence and Compo, 1998). The parameters were chosen to give a total of 57 periods ranging from 0.5 to 256 years, and the square coherency were calculated using smoothing in the time and space domain (Grinsted et al., 2004), with the 5 % significance level determined from a Monte-Carlo simulation of 1000 sets of surrogate time series. The two temperature time series show episodic coherency with the solar forcing at 11 year period (Fig. 3, top panels), in particular around 1200 and 1600. Meehl et al. (2008Meehl et al. ( , 2009) indeed showed that a peak in the solar activity induces surface cooling in the tropical Pacific. Kuroda et al. (2008) showed that over the historical 15 period, years of anomalously high solar irradiance were associated with a large warming of the lower stratosphere through radiative heating. Such a temperature anomaly in the stratosphere creates anomalous temperature of opposite sign at lower heights. However, these processes require a much higher resolution in the stratosphere to be properly represented. In fact, episodic coherency between SST and TSI variations at 20 11 years timescale is also significant from the control data, suggesting that the signal in Fig. 3 is internal to the data sets and does not indicate physical response of the ocean to the 11-year cycle. The temperature time series also show strong coherency with the TSI variations at multidecadal timescale from 1700, associated to the TSI increase, and at centennial time scales over the whole simulation (Fig. 3, second and 25 third panels).
It is somewhat more difficult to distinguish the response of the Atlantic meridional overturning circulation (AMOC) from its natural variability. As shown in Fig. 1  volcanic activity was intense, peaks around year 1280 and then rapidly decreases, reaching a minimum around year 1320, about 60 years after the major eruption of 1260. There is a hint of a weak response to the eruptive events in the early 1800s. As shown in Fig. 3 (bottom), there is a hint of a weak coherence between the time series of AMOC maximum with the TSI variations at 11-year periods, and a more significant one 5 at about 100-year period, with TSI leading by 15 years. Correlations with the volcanic forcing are barely significant. As will be shown below, this does not imply that there is no AMOC response to natural forcings, in particular volcanic eruptions. Time series of AMOC maximum represents one mode of AMOC variability, namely a basin scale acceleration, as discussed for example in Msadek and Frankignoul (2009). More local 10 AMOC adjustments require more specific analysis. In the following, we concentrate on the response to volcanic forcing, which has a much stronger impact on the atmospheric and the oceanic temperature than the solar forcing in the model.

Anomalous temperature patterns in response to volcanic eruptions
To describe the oceanic response to a volcanic eruption, we construct a composite evo-15 lution based on the oceanic anomalies that follow the major eruptions. Anomalies are computed for each selected eruption as the difference between the time evolution of the field and a reference defined as the average of the field during the two years preceding the eruption. Composites are then defined as the average of these anomalies scaled by the magnitude of each eruption, so that possible non linear effects linked to the 20 eruption magnitude are minimized. Note however that our conclusions are unchanged without this normalization. In order to maximize the signal to noise ratio, we focused on relatively large eruptions, and thus selected eruptions corresponding to an increase of stratospheric aerosol optical depth (AOD) by more than 0.15 (eruptions marked with a star in Fig. 1), which is equivalent to a global radiative forcing of at least −2.8 W m −2 .

25
This corresponds to the 9 strongest eruptions between 850 AD and 1849. As seen in Fig. 1 10 years (1169-1178, 1810-1816). To minimize the interference between successive events, eruptions which precede another one by less than the considered time lag in the composite were omitted. As a consequence, the number of events in the composites may decrease with lag. The composites are displayed for a stratospheric global mean optical depth equal to 0.15. Significativity is tested with a block bootstrap 5 procedure with 500 permutations of the volcanic time series in blocks of 3 years (the maximum residence time of stratospheric aerosols). Figure 4 shows composites of anomalous global surface temperature up to 20 years after a volcanic eruption of AOD of 0.15. The first panel (year 0) shows that the maximum cooling occurs in the tropics and on the lands during the year of the eruption. 10 There is also a meridional dipole in the Southern Atlantic and the Indian oceans, which can be shown to be due to a shift of the westerlies, persisting for a year. An anomalous warming in the polar region over Eurasia is consistent with the observations (e.g. Robock and Mao, 1992) and closely related to tropospheric and stratospheric circulation changes. As the lag increases, the tropical oceanic signal extends in latitude, 15 reflecting the spreading of the atmospheric cooling (e.g. Robock, 2000), while progressively decaying in the tropics. In the subpolar North Atlantic, the cooling peaks at year 3 and decays thereafter. Note the relatively rapid decay of the cooling in the eastern equatorial Pacific at year 1, also present at year 2 (not shown) which could be due to an El Nino-like response. One to two years after the eruption, there is an anoma-20 lous warming in the North Atlantic midlatitudes, the origin of which is discussed below. An anomalous warming near the Drake passage becomes significant at year 3 and reaches its maximum at year 5. Ten years after the eruption, the whole tropical band is still significantly anomalously cold, as well as some land areas such as in Eurasia. In the North Atlantic, the most striking feature is an anomalous warming in the Labrador 25 Sea, which decays thereafter. Figure 5 shows similar composites for the zonally averaged global oceanic temperature response as a function of depth up to 20 years after a volcanic eruption. Consistent with Fig. 4 during the eruption year, together with a warming below 100 m depth in the deep tropics. The latter results from a thickening of the tropical thermocline and a decrease of equatorial ventilation following a weakening of the trade winds, as discussed below. The surface cooling already reaches 60 • N, but its poleward extension is stronger one year after the eruption, consistent with Fig. 4. By year 1, the signal has penetrated 5 in the ocean interior around 30 • N and 30 • S, where oceanic ventilation mostly takes place. In the subtropics, the downwelling is shifted slightly poleward of the climatological ventilation region, indicated by the mean isotherms in Fig 900 m of the cooling is also seen around 60 • N at year 1, reflecting enhanced deep convection. Deep convection also mixes the cooling signal down in the Southern Ocean, reaching its largest depth 2 to 3 years after the eruption (not shown). In the following years, the tropical surface cooling decays, while persisting at depth and deepening further ( Fig. 5, year 5). As the surface cooling reaches greater depths, the subsurface 15 warming deepens and shifts poleward. After 10 years, the cooling signal has reached more than 500 m at 40 • latitude north and south, which is roughly the maximum depth of the subtropical cells, and 700 m in the southern ocean. In the North Atlantic, on the other hand, a warm subsurface anomaly has appeared, reflecting a decrease of deep convection as will be discussed 20 below. At this stage, the response is thus asymmetric in the high latitudes as also found by Stenchikov et al. (2009). Twenty years after an eruption, cooling is still significant in the tropics and at high latitudes, where it reaches 700 to 900 m, while the northern subtropics have warmed, reflecting the dynamical adjustment of the gyres discussed below. 25 In the following, we focus on the response of the Atlantic Ocean, as a case study and in order to investigate the behavior of the AMOC. From Fig. 1, it seems clear that the behavior of the AMOC after the severe and decadally paced eruptions of the twelfth and thirteenth century is peculiar. Figure 6 illustrates the different response of the ocean Introduction to the selected eruptions occurring after 1400, from the ones occurring between 1100 and 1300. In response to volcanic eruptions occurring after 1400 (bottom panels), the initial (in phase) cooling is more clearly limited to the tropics and subtropics, while the mid-and high latitudes are characterized by an anomalous warming, due to anomalous turbulent heat fluxes as discussed below. The anomalous cooling rapidly reaches the 5 higher latitudes (1 yr to 4), except for a small patch of anomalous warming at 45 • N which reflects a northward shift of the North Atlantic Current. After about a decade, the anomalous cooling has disappeared or lost significance in the Atlantic basin, while a strong and persistent warming has appeared in the subpolar gyre, with maximum amplitude south of Greenland, and a coma shape extension in the eastern subtropics 10 with resembles the path of the subtropical gyre. This structure thus is strongly similar to the signature of an AMOC acceleration in the coupled model (e.g. Msadek and Frankignoul, 2009). On the other hand, the anomalous cooling occurring in phase with the intense and decadally paced eruptions between 1100 and 1400 is significant not only in the tropics, 15 but also at subpolar latitudes, in particular in the Irminger Sea and the Nordic Seas, where deep convection in the model takes place. There is also a weak, marginally significant, warming at midlatitude, again probably reflecting a shift in the North Atlantic current, but it is short lived and the entire basin becomes anomalously cold in the years following the eruption. At decadal timescales, the anomalous subpolar warming seen 20 after 1400 can be recognized but it is much weaker and not significant at the 5 % level. The fact that the response response differs as early as in phase with the eruption tends to eliminate the cumulative effect of the decadally-paced eruptions of the twelfth and thirteenth century, as opposed to more isolated eruptions occurring after year 1400. A larger signal to noise ratio in response to stronger eruptions might be an alternative 25 explanation, as discussed e.g. in Shindell et al. (2003) and Schneider et al. (2009). However, the anomalous atmospheric response shown in Fig. 6 (top) is unchanged if the mega eruption of 1258-1259 is omitted for the computation of the composite (not shown). to peak during the cold season while the ones that occurred during the rest of the last millennium mostly peak during the warm season. Investigating the effect of this seasonality requires specific sensitivity experiments and is beyond the point of this study. In the following, we will first investigate the response to eruptions occurring after 1400.
5 5 Interannual to decadal response of the Atlantic ocean to eruptions post 1400 AD In response to the rapid surface cooling, there is a strong anomalous low over the Canadian archipelago and an anomalous high over the northeastern Atlantic. In addition, the sea level pressure (SLP) becomes anomalously high over South America and 10 most of Africa, where the cooling is strongest, and an anomalous low in the western subtropics (Fig. 7). Consequently, the Northern Hemisphere trades and westerlies are reduced during the year of the eruption, and shifted southward. Over the tropical lands, the SLP signal weakens at year 1, but it remains significant for almost 2 decades over the amazonian basin. At mid to high latitudes, the anomalous low quickly disappears 15 but the anomalous anticyclone shifts slightly westward and persists until year 4, resembling a negative phase of the East Atlantic Pattern (EAP). Later, the signal looses significance (not shown), until year 10, where a response resembling a negative phase of the NAO is detected.
At year 0, the wind changes induce a negative wind stress curl anomaly across the 20 basin between 50 and 60 • N and a positive one north and south of it (Fig. 8, left). The depth-integrated oceanic circulation, as described by the barotropic streamfunction, adjusts rapidly to the wind stress curl. At year 0, it is anomalously negative in much of the subtropical Atlantic, reflecting a weakening of the subtropical gyres (Fig. 9, left). A weak positive anomaly is also significant in the subpolar region, where the wind stress  (Fig. 8, right). Two to four years after the eruption, both gyres of the North Atlantic are thus clearly reduced. At longer lags, the response decays in the subtropics while the subpolar gyre stays anomalously weak for more than a decade after the eruption (Fig. 9, middle). Note also the persistent signal in the Labrador Sea where the cyclonic circulation is reinforced.

5
The atmospheric response to the eruption also induces vertical circulation in the ocean, resulting from the anomalous Ekman suction at 30 • N/S and pumping around 50 • N. This appears clearly at year 0 on the meridional streamfunction composite ( Fig. 10, left). The signal is equivalent barotropic, with an upwelling around 30 • N and a downwelling at 50 • N and around the equator. During the following years, the strong 10 negative wind stress curl in the subpolar North Atlantic maintains a positive meridional cell between 20 and 50 • N, which can be viewed as an intensification of the AMOC in the North Atlantic, consistent with previous studies (e.g. Stenchikov et al., 2009;Ottera et al., 2010;Ortega et al., 2011). Note that this positive anomaly is probably also favored by the intensified deep convection that occurs during the year of the eruption 15 (Fig. 11, left), and is associated with strong surface cooling. An intensification of deep convection typically leads by several years an acceleration of the AMOC in the North Atlantic basin (e.g. Mignot and Frankignoul, 2005). However, it is short-lived here, loosing significance by year 1, so that the AMOC intensification does not persist more than a few years (Fig. 10, bottom left). On the other hand, there is a weak reduction of the 20 AMOC north of about 60 • N up to four years after an eruption, which later intensifies and extends to subpolar latitudes as a result of a reduction in deep water formation, as discussed below. Five years after the volcanic eruption, the SLP anomaly decreases (not shown). Nevertheless, as indicated above, a significant SLP anomaly appears again near year [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25] 12, under the form of the dipole in the mid to high latitudes bearing similarity with a negative phase of the NAO. This could reflect the SLP response to the AMOC intensification seen at year 2-4 ( Fig. 7 upper right) to enhanced AMOC in several climate models including the CTRL simulation with IP-SLCM4. In the latter, the SLP response was of similar magnitude and most clearly seen four years after an AMOC intensification. Here, the AMOC intensification indeed remains significant until lag 8 (not shown) Whether there is a link with the weak AMOC intensification seen about 20 years after the eruption cannot be asserted here but could 5 be established in dedicated experiments. As mentioned above, the strong surface cooling rapidly deepens the mixed layer south of Iceland (Fig. 11, left panel) and in the subtropics (not shown). The intensification of deep convection favors the penetration of the cooling signal at depth seen in Fig. 5 at high northern latitudes and a weakly significant retreat of sea ice cover 10 ( Fig. 12, left panel). However, this response looses significance in the following years (Fig. 11, middle panel), and instead, deep convection is reduced both in the Nordic Seas and South of Iceland 4 years after an eruption (Fig. 11, left panel). This persists for about a decade after the eruption. In the Nordic Seas, the reduction of deep convection is due to a persistent sea ice capping of the area during winter, resulting from the 15 strong surface cooling (Fig. 12). This anomaly appears about 1 year after the eruption, peaks after four years and, again, persists for roughly a decade. Such anomalous sea ice extension is consistent with the sea ice reconstruction off Iceland from Masse et al. (2008), showing abrupt events that coincide with the volcanic eruptions after 1300 AD.
South of Iceland, the winter mixed layer shallowing in Fig. 11 (right) is due to a strong 20 negative salinity anomaly (Fig. 13 top). The latter is largely due to anomalous Ekman transport (Fig. 13, bottom), while anomalous atmospheric freshwater fluxes play a lesser role (Fig. 13, middle), consistent with Frankignoul (2003, 2004). Note that the anomalous surface freshwater flux caused by the volcanic eruptions are maximum during the year of the eruptions, and in the tropics (Fig. 13), consistent with 25 Trenberth and Dai (2007). The reduction of the northern trade winds (Fig. 7) is indeed associated to a northward shift in the inter-tropical convergence zone. As a result, precipitation is enhanced near 10 • N and strongly reduced along the equator, including over the continents. In the northern deep tropics, evaporation is reduced, again Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | because of reduced winds and SST. These anomalous surface freshwater fluxes induce the negative salinity anomaly in the northern subtropics in the years following the eruption.
6 Interannual to decadal response of the Atlantic ocean to eruptions between 1100 and 1300 5 Figure 14 shows the response of the AMOC to the volcanic eruptions selected during the period of intense volcanic activity between 1100 and 1300. As during the later period, the in-phase response is essentially characterized by an anomalous downwelling around 30 • N. The associated negative and positive cells south and north of this latitude have nevertheless a much weaker extension in depth (for the tropical one) and 10 in latitude (for the northern one). Indeed, the anomalous sea level pressure induced during the year of an eruption occurring during the earlier period, shown in polar view in Fig. 15, is similar in the tropics and subtropics (not shown) but it has the opposite sign over the canadian archipelago and there is no strong high in the eastern North Atlantic. As a result, the anomalous wind stress curl is much weaker in the subpolar 15 region, inducing a much weaker anomalous Ekman pumping and Ekman transport, and only little salt advection (not shown). The subpolar gyre is thus much less affected during the years following the eruption (not shown). Lacking the large subpolar freshening seen in the later period, the winter mixed layer remains anomalously deep in the subpolar region (Fig. 16, top row middle panel), even further south than deep convec-20 tion locations, due to the surface cooling (Fig. 6), which might might contribute to the broad and shallow AMOC intensification between 20 • S and 45 • N at years 2 to 4. It can be shown that the tropical part of this anomalous cell is associated with an equatorward shift of the subtropical gyre. On the other hand, deep convection is reduced in the Nordic Seas after a few years because of the sea ice capping discussed above. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the stronger cooling (Fig. 6). It is also interesting to note that the weak gyre response also contributes to maintain the oceanic surface cooling in the 1100-1300 A.D. period, since it does not induce the warm anomaly seen in the years following an eruption occurring after 1400 AD (Fig. 6). As a result, the negative AMOC anomaly seen over the ridges two to four years after an eruption is stronger and deeper than for the eruptions 5 occurring after 1400 AD. After a longer delay, when anomalous cooling and convection in the subpolar basin start to vanish, the shallowing of the winter mixed layer in the Nordic Seas seen at years 4-7 in Fig. 16 (top right) persists and is likely responsible for the stronger negative AMOC weakening seen throughout the northern North Atlantic at least two decades after the eruptions (Fig. 14, bottom left). This behavior is consistent 10 with Zhong et al. (2010). That the AMOC weakening extends north of the ridges adds credit to the implied role of Nordic Seas convection and could explain why the AMOC behavior has become nearly opposite to that seen after 1400 AD.

Conclusions and discussion
In this study, we have investigated the oceanic response to volcanic eruptions over the 15 last thousand years, with a focus on the North Atlantic Ocean. We used a fully coupled AOGCM forced by a realistic chronology of volcanic eruptions, variations of the TSI and of the atmospheric greenhouse gases concentrations. The analysis highlighted the multiple timescales of the response, including a fast tropical temperature adjustment to the strong volcanic-induced radiative forcing, dynamical adjustment in response to 20 the associated atmospheric circulation modifications persisting roughly 5 years, and a subsequent adjustment of the AMOC in response to anomalous convection at high latitudes. The analysis also highlighted differences in the response during two distinct periods of the last millennium. The global surface temperature response is maximum one to two years after a vol- well as deep convection at high latitudes. It thus persists globally in the ocean for more than 20 years. During the year of the eruption, anomalous tropical cooling induces an anomalous high over the continents and a reduction of the trades in the Atlantic ocean.
The atmospheric response at mid to high latitudes depends on the eruptions. In this study, in order to investigate apparent discrepancy found in the literature regarding 5 the AMOC response, we only separated eruptions occurring after 1400 from the ones occurring between 1100 and 1300. In the later period, the anomalous atmospheric structure in response to an eruption induces strong wind stress curl anomalies over the North Atlantic ocean, which lead to an important dynamical adjustment in the Atlantic ocean at interannual timescales, namely one to five years after an eruption. This adjustment of the oceanic circulation is equivalent barotropic, with an upwelling around 30 • N and a downwelling at 50 • N and around the equator. The anomalous vertical oceanic circulation can reach the full depth of the ocean. During the following years, the atmospheric structure evolves inducing an anomalous acceleration of the AMOC in the subpolar basin. However, this anomaly does not persist more than a few years, 15 because of reduced deep convection in the high northern latitudes under the effect of anomalous sea ice extension and surface freshening that develop a few years after the eruption. A weak reduction of the AMOC is thus detected a decade after the eruption. In the case of the eruptions occurring between 1100 and 1300, the anomalous SLP structure during the year of the eruption differs over the subpolar region from the one 20 obtained during the later period. It induces much weaker wind stress curl anomalies over the Atlantic basin and thus a much weaker dynamical adjustment of the AMOC in the years following an eruption. On the other hand, the initial reduction of deep water formation is more persistent, as a result of a stronger surface cooling and more persistent sea ice cover anomalies, leading to a stronger negative anomaly of AMOC 25 at high latitudes 2 to 4 years after an eruption, and a stronger reduction of the AMOC in the subpolar North Atlantic 10 to 15 years after the eruption.
As noted in the introduction, the oceanic response to volcanic eruptions is still largely unknown and recent studies based on climate models suggest either an AMOC Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | enhancement or a reduction following volcanic eruptions of the last millennium. The present findings could possibly reconcile these previous studies, suggesting a strong sensitivity of the response to different volcanic eruptions. In particular, the AMOC intensification seen 5 to 10 years after the volcanic eruptions occurring after 1400 AD. bears strong similarity with results of Ottera et al. (2010) using a different coupled cli-5 mate model to investigate this period. On the other hand, the AMOC weakening in the northern North Atlantic and the large sea ice extension following the intense eruptions occurring between 1100 and 1300 can be compared to the response found by Zhong et al. (2010). The present analysis suggests that these different responses involve in fact similar mechanisms, namely an initial dynamical adjustment to anomalous winds 10 and a subsequent thermohaline response to anomalous deep convection. However, the atmospheric response to the volcanic eruptions differ in the two periods and thus plays a large role in modulating the oceanic response. At least three factors could explain why the atmospheric response seems to change in time: the seasonality of the eruption, its intensity, and the cumulative effect in the case of successive eruptions. An 15 analysis of these various effects requires specific experiments that are left for future studies. However, the results presented here suggest that the seasonality is the most plausible explanation. Indeed, the eruptions during 1100-1300 occurred mostly during the cold season while those after 1400 occurred mostly during the warm season. Although cumulative effects may also play a role, they cannot explain why the short term 20 response already differs in the two periods. Non linearities in the response could also be important, even though our results were not changed when the mega eruption of 1258 was omitted from the 100-1300 composite. The oceanic response is also likely to be affected by model biases and experiment design. In particular, the lack of deep convection in the Labrador Sea in the IPSLCM4 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Marti, O., Braconnot, P., Dufresne, J. L., Bellier, J., Benshila, R., Bony, S., Brockmann, P., Cadule, P., Caubel, A., Codron, F., de Noblet, N., Denvil, S., Fairhead, L., Fichefet, T., Foujols, M. A., Friedlingstein, P., Goosse, H., Grandpeix, J. Y., Guilyardi, E., Hourdin, F., Krinner, G., Lévy, C., Madec, G., Mignot, J., Musat, I., Swingedouw, D., and Talandier, C.: Key features of the IPSL ocean atmosphere model and its sensitivity to atmospheric resolution, Clim.  Rev., 28, 1941Rev., 28, -1955Rev., 28, , doi:10.1016Rev., 28, /j.quascirev.2009Rev., 28, .04.008, 2009 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | van der Schrier, G., Weber, S., and Drijfhout, S. S.: Sea level changes in the North Atlantic by solar forcing and internal variability, Clim. Dynam., 19, 435-447, doi:10.1007/s00382-002-0235-y, 2002. 2513 Vieira, L. and Solanki, S.: Evolution of the solar magnetic flux on time scales of years to millennia, arXiv/0911.4396, doi:10.1051/0004-6361/200913276, 2009: Centennial-scale climate change from decadally-paced explosive volcanism: a couled sea ice-ocean mechanism, Clim. Dynam., doi: 10.1007/s00382-010-0967-z, in press, 2010. 2514, 2528, 2530 Fischer-Bruns, I., and Schlese, U.: Climate evolution in the last five centuries simulatis by an ocean-atmosphere model: global temperatures, the North Atlantic Oscillation and the Late Maunder Minimum, Meterol. Z., 13, 271-289, doi:10.1127/0941-2948/0013-0271, 2004 2537 Introduction     Fig. 10