The global mean surface temperature has increased considerably in the last 2 years (NOAA, 2016), which followed on a 15-year period with slow temperature evolution. CMIP5 (Coupled Model Intercomparison Project) models predict stronger-than-observed temperature increases in this period (Fyfe et al., 2013, 2016). Reasons for these differences were sought, and the Interdecadal Pacific Oscillation connected with increased subduction and upwelling (England et al., 2014; Meehl and Teng, 2014), variations in volcanic aerosol (Solomon et al., 2011; Santer et al., 2014) and solar (Myhre et al., 2013) forcings were identified as main causes of the discrepancies. These phenomena are all elements of natural climate variability, thus highlighting the importance of these influences for assessing the human impact on the climate (Ramanathan and Feng, 2008).

Air from the tropical troposphere containing aerosol precursor gases is lifted into the tropical stratosphere in the Brewer–Dobson circulation. Carbonyl sulfide (OCS) is the most abundant sulfur-containing gas in the atmosphere. When lofted into the stratosphere, OCS is converted into sulfur dioxide (SO2) at approximately 25 km altitude aided by UV radiation (Crutzen, 1976) and next into sulfuric acid, giving rise to the background stratospheric aerosol layer. The upward tropical flow is connected with a downward flow in the extratropics, where stratospheric air flows back to the troposphere.

The stratospheric aerosol concentration varies strongly due to volcanic influence, from events like the strong 1991 Mt. Pinatubo eruption, inducing negative radiative forcing in excess of 1 W m-2 (McCormick et al., 1995), to close to background conditions at around the turn of the millennium (Bauman et al., 2003; Martinsson et al., 2005). Overall, anthropogenic influence on stratospheric aerosol is small compared to that from volcanism (Neely III et al., 2013). Recent volcanic eruptions, such as in 2008 and 2009 of the Kasatochi and Sarychev in the extratropics and in 2011 of the Nabro in the tropics, significantly altered stratospheric aerosol (Vernier et al., 2011a; Bourassa et al., 2012). These eruptions also had a profound impact on the Northern Hemisphere lowermost stratosphere (LMS; Martinsson et al., 2009) and the global aerosol optical depth (AOD) of the stratosphere (Andersson et al., 2015). After reaching the LMS, stratospheric aerosol is transported to the upper troposphere (UT). A recent study indicates that volcanic aerosol also has an indirect climate effect by affecting the reflectance of cirrus clouds in the UT (Friberg et al., 2015).

Upper troposphere particulate sulfur, in addition, has tropospheric sources. In the tropics, convection lofts aerosol and aerosol precursor gases from low altitudes into the UT (Hess, 2005). During the monsoon season, primarily the Asian monsoon, appreciable amounts of aerosol and gaseous aerosol precursors are lifted to the tropopause region, extending from the UT to approximately 420 K potential temperature in the stratosphere, forming the Asian tropopause aerosol layer (ATAL; Vernier et al., 2011b). The ATAL is a seasonal and regional phenomenon which, according to atmospheric modeling, is strongly connected with Asian pollution sources (Neely III et al., 2014). In the extratropics, warm conveyor belts (WCBs; Stohl, 2001) and deep convection are the main lofting channels, the latter especially in the summer season. Large amounts of pollutants are transported long-range across the Pacific and Atlantic oceans, where sources are affecting the background concentrations of various species over remote continents. This transport is most efficient above the boundary layer in the pressure interval 700–900 hPa (Luan and Jaeglé, 2013). Warm conveyor belts reaching the UT have their maximum frequency in the winter (Eckhardt et al., 2004) and transport of boundary layer air into the LMS maximizes in the winter and spring (Škerlak et al., 2014), whereas deep convection dominates in the summer (Kiley and Fuelberg, 2006).

Aerosol particles in the UT and LMS contain a large fraction of sulfur compounds, mostly sulfates (Dibb et al., 2000; Martinsson et al., 2001; Xu et a., 2001; Kojima et al., 2004). These particles also contain a considerable fraction of carbonaceous aerosol (Murphy et al., 1998; Nguyen et al., 2008; Martinsson et al., 2009), of which a minor fraction is black carbon (Schwarz et al., 2010; Friberg et al., 2014). In some periods, particularly during spring, crustal particles are also found in this part of the atmosphere (Papaspiropoulos et al., 2002).

In this study the concentration of particulate sulfur in the UT and LMS is investigated based on aerosol samples collected from the IAGOS-CARIBIC platform in the time period 1999–2014. In this period the UT and LMS aerosol was affected by volcanism from several eruptions as demonstrated in our previous studies based on this data set (Martinsson et al., 2009, 2014; Andersson et al., 2013, 2015; Friberg et al., 2014, 2015). The IAGOS-CARIBIC aerosol elemental concentration measurements from the LMS are taken in strong concentration gradients that are affected by mixing tropospheric air into the lowest part of the LMS. Each measurement flight results in a small number of samples, being insufficient to reconstruct the gradient. Therefore, we have frequently relied on concurrent IAGOS-CARIBIC measurements, mostly by relating the particulate sulfur measurements to ozone concentrations to express, for example, volcanic influence on the aerosol concentration (Martinsson et al., 2009). Satellite-based measurements do not usually provide specific chemical information about aerosol particles. This lack of chemical information in, for example, lidar measurements can cause a bias in LMS particle concentrations close to the extratropical tropopause from non-volcanic species such as crustal particles and enhanced signal caused by particle hygroscopic growth. Here we present stand-alone estimates of the stratospheric sulfur aerosol in terms of mass concentration profiles and column concentrations based on a new method, which is then used to study the AOD of the lowest part of Northern Hemisphere LMS and its radiative forcing. This study also comprises a discussion on the relative importance of stratospheric and the tropospheric sources to the UT of particulate sulfur, seasonal dependences and different modes of tropospheric transport involved.

The linear regression methodology described in the previous section was applied to particulate sulfur concentrations (CS) as a function of the distance from the tropopause (Z) for seasons comprising 3 months. Data from different years of one season were grouped according to their concentrations, resulting in four or five groups differing with respect to the CS-Z relationship. The resulting groups of data from each season were modeled by forced linear regression. These model results are used in a second regression step to model seasonal influences from transport, which in turn are used to obtain the response of the LMS and UT particulate sulfur concentrations to changes induced mainly by volcanic eruptions.

This methodology was applied to all seasons, having a duration of 3 months. The final product of the regression methodology, i.e., the second regression step, of all 12 overlapping seasons is shown in Fig. 3. It is clear that slopes and offsets of the season groups obtained in the first step of forced linear regression, differing in particulate sulfur concentration related to distance from the tropopause, can readily be described by linear regressions for all twelve 3-month seasons. The relationship between slopes and offsets has a very strong seasonal dependence. In some seasons small changes in the LMS sulfur concentration slope is connected with a strong change in tropopause sulfur concentration. This is most pronounced for the seasons centered in February, March and April. At the other end we find the seasons centered in September, October and November, where a change in the slope of the first regression step due to varying influence from volcanism has little effect on the tropopause concentration. To further illustrate the connection to the measurements, Fig. 4 shows the four data groups of the seasons most (Fig. 3: FMA) and least (Fig. 3: OND) susceptible to change in the tropopause concentration due to changed stratospheric concentrations. In Fig. 4a the data group of least volcanic influence includes years 1999–2002 and 2014, and the most influenced years are 2009 and 2012 which are the springs after the Kasatochi (7 August 2008) and Nabro (12 June 2011) eruptions (we have no late winter/spring data after the most powerful eruption of the period studied, Sarychev, 12 June 2009, due to maintenance of the measurement aircraft). The season centered in November has the weakest volcanic influence years 1999–2001, 2012 and 2013 (Fig. 4b), whereas the group with strongest slope includes the three strongest eruptions of the period studied (Kasatochi, Sarychev and Nabro), a few months after respective eruption. Despite the very strong volcanic influence of the latter group, the tropopause concentration (Z=0) remains close to that of the closely distributed tropopause concentrations of the other groups of that season.

The sensitivity of the tropopause and UT concentration to changes in LMS concentration slope is denoted k in Eqs. (3) and (4), which thus is obtained for each regression depicted in Fig. 3. These results are collected in Fig. 5a. The salient features of the seasonal dependence can be described by two Gaussian distributions (Fig. 5a). The maximum sensitivity appears in late winter and early spring when the seasonal variation in tropopause altitude has its maximum rate of upward motion (Appenzeller et al., 1996). The maximum in rate of downward motion of the tropopause appears in the fall. This delays transport from the stratosphere to the troposphere, which is reflected by a low-sensitivity k. This sensitivity can, in addition, be affected by the residence time of particulate sulfur in the UT. Interestingly, the downward transport in association with the Brewer–Dobson circulation from deeper, aerosol-rich stratospheric layers through the LMS takes place in the same season as the maximum in k, thus further enhancing the stratospheric influence on the UT by the term a(y,s)×k(s) in Eq. (4).

The offsets of the regression lines shown in Fig. 3 (when a=0) expresses the case when the stratospheric and tropospheric concentrations are equal, implying that this offset expresses the particulate sulfur concentration of tropospheric origin (CS,UTtrop) that is mixed into the LMS; see Fig. 5b. Two Gaussian distributions were used as fits to the seasonal variation of CS,UTtrop. As for k, the seasonal variation of CS,UTtrop is strong. The seasonal variation of CS,UTtrop will be elaborated in the discussion section.

After obtaining CS,UTtrop and k, the data that are needed for conversion of every measurement of the concentration to an estimate of the slope and offset of the LMS concentration are available. Thus, for each individual measurement (i) in the LMS, consisting of the particulate sulfur concentration (CS,i) and the altitude Zi above the tropopause, the slope and offset of Eqs. (3) and (4) are obtained from ai=CS,i-CS,UTtrop(s)Zi+k(s),bi=aiks+CS,UTtrop(s). The estimated slopes and offsets are shown in Fig. 6a and b, where the dots are individual measurements and the histogram monthly averages. Here measurements taken at an altitude of less than 50 m above the tropopause are not shown, because they were judged to have too small a stratospheric character for an estimate of the concentration slope in the LMS. Both the slope and the tropopause concentration are affected by volcanism (Table 1), but the relative response of the slope is much stronger than that of the tropopause concentration; see, for example, the falls of 2008 and 2009 affected by the Kasatochi and Sarychev eruptions.

Observations at various altitudes above the tropopause are difficult to compare, due to the concentration gradient in the LMS. With the estimates of the tropopause concentration and the slope in the particulate sulfur LMS concentration, each measurement becomes an estimate of the total amount of particulate sulfur in the altitude interval investigated through the integration of Eq. (4). However, first the STP concentrations (CS,STP=CS=aZ+b) need to be converted into volume concentrations (CS,V), which are related by CS,V=CS,STPQSTP/QV(Z), where Q is molar volumes. For that purpose, the altitude dependence of the molar volume from the tropopause up to 5 km above the tropopause was extracted from temperatures and pressures obtained from ECMWF for each sample. The molar volume can be expressed as QVZ=QV(0)ewZ, where the tropopause is at Z=0 and w=0.0001535 m-1 obtained as the average of all samples. For a measurement the molar volume is Qm obtained at distance Zm from the tropopause, and the tropopause molar volume is computed by QV0=Qme-wZm. Finally, the column concentration of particulate sulfur for the first 3 km above the dynamical tropopause of 1.5 PVU is obtained by integration of the volume concentration: CS,col=∫03000QSTPQVZCS,STPZdZ=∫03000QSTPQV(0)e-wZaZ+bdZ. The particulate sulfur column is calculated for altitudes above the tropopause (Z) in the range of 0 to 3000 m, the upper limit set where too few measurements (5 %) were taken above that level. After integrating the column, it is also interesting to estimate the total amount of sulfur-connected aerosol. To that end, it was assumed that the aerosol consists of 75 % sulfuric acid and 25 % water, which is a commonly used stratospheric composition (Rosen, 1971; Arnold et al., 1998). This means that the total sulfur column was multiplied by a stoichiometric factor of h=4.084 to obtain the column of the sulfuric acid–water aerosol: CA,col=hCS,col. The measurements were taken in the Northern Hemisphere latitudes higher than 30∘ with the highest latitude of 77∘, with 90 % of the data in the latitude range 32–64∘ N and 70 % of the data between 37 and 57∘ N. The northern midlatitudes sulfur aerosol columns are shown in Fig. 6c as monthly averages with standard errors (the few months with only one measurement available are shown without error bars). The sulfur aerosol column of the LMS shows large variability primarily caused by volcanism. The lowest columns are found in the period 1999–2002, when the volcanic influence on the stratospheric aerosol was small (see Table 1 for relevant volcanic eruptions). The time period mid-2005 to mid-2008 was affected by tropical volcanism (Vernier et al., 2011a), which also caused elevated concentrations in the NH LMS (Friberg et al., 2014). The eruptions of the extratropical volcano Kasatochi in August 2008 placed two volcanic clouds in the stratosphere (Andersson et al., 2015), one in the LMS causing strongly elevated aerosol column of the LMS that ceased by November the same year, and the other above the LMS. The latter cloud was transported downward, causing a rise of the lower LMS aerosol column in December 2008. After some influence from several eruptions of the extratropical volcano Redoubt in the spring of 2009, the eruption of Sarychev strongly affected the Northern Hemisphere stratosphere from June 2009. The eruption of the Icelandic volcano Grimsvötn in May 2011 had a strong and short impact on the northern LMS, which is reflected by a peak in June to July 2011 (Fig. 6c), before the tropical volcano Nabro reached the northern LMS in the early fall of the same year. After that eruption a gradual decrease of the aerosol load can be seen. The concentrations after mid-2013 approach those of the period 1999–2002, which was close to stratospheric background conditions.

The LMS aerosol contains a significant fraction of carbonaceous material (Martinsson et al., 2009), mainly organic in nature (Friberg et al., 2014), which adds to the aerosol columns of Fig. 6c and affect the refractive index of the particles. However, this work is dealing with the sulfurous fraction, the main fraction of the stratospheric aerosol. To put the results presented in Fig. 6c into perspective, the AOD is estimated using a simplified aerosol (which thus likely is an underestimation of the AOD). Thus, the “standard” stratospheric 75 % sulfuric acid, 25 % water composition, 1.669 g cm-3 particle density and 1.44 refractive index will be used. Furthermore, particle size distribution measurements from IAGOS-CARIBIC have been taken since 2010 (Hermann et al., 2016). The changes of the size distribution induced by the moderate 2011 eruptions of Grimsvötn and Nabro were small (Martinsson et al., 2014), and agree well with previous measurements (Andersson et al., 2015) of the stratospheric background particle size distribution by Jäger and Deshler (2002). Thus, for the estimation of the AOD the latter particle size distribution was used for the entire time period studied. For fixed composition and particle size distribution the AOD is obtained as a fixed relationship to aerosol column: AOD=fCA,col, where f contains the relationships between mass and area/extinction, in this case f=3.29×10-6 m2 µg-1. Finally, converting the AOD into radiative forcing (RF) using the global average relation (Hansen et al., 2005; Solomon et al., 2011) of RF=-25×AOD;inWm-2, to obtain an estimate of the climate influence of the sulfate aerosol of the lower LMS. The peak AOD of the Kasatochi and Sarychev eruptions is approximately 0.006, corresponding to -0.15 W m-2 in regional radiative forcing of the lowest 3000 m of the Northern Hemisphere LMS. Although no detailed comparisons will be made here, we find that the AOD and radiative forcing obtained from the particulate sulfur measurements show similar tendencies to satellite-based measurements (Andersson et al., 2015). The findings presented here also corroborate the findings of Andersson et al. (2015) on the importance of the LMS for the total stratospheric AOD and radiative impact.

The results presented here are based on measurements in the extratropical UT and the LMS of the NH, where the latter includes the extratropical transition layer (ExTL). Bi-directional exchange across the tropopause affects strong gradients in the ExTL for species having clearly different stratospheric and tropospheric concentrations (Hoor et al., 2002), such as particulate sulfur (Martinsson et al., 2005). In the previous section, the UT concentration, the gradient in the LMS and column amount of particulate sulfur in the ExTL were investigated, with seasonal dependence (Fig. 5) and influence from volcanism (Fig. 6). In the processing of the data to obtain these results, one feature stands out in particular: the seasonal dependence of the particulate sulfur concentration from tropospheric sources that is mixed into the ExTL (Fig. 5b). It shows a broad maximum from August to December, peaking in September to November and a deep minimum in the late winter and early spring (February and March).

Let us now compare these concentrations in the UT of tropospheric origin (CS,UTtrop) with average concentrations of particulate sulfur in the UT for two cases: “background conditions” dominated by data from mid-1999 to mid-2002 and “moderate influence from tropical volcanism” dominated by data from mid-2005 to mid-2008; see Sect. 2.3 for details. The seasonal dependence of these two categories are shown in Fig. 7a together with CS,UTtrop. It is clear that, in line with the findings of Friberg et al. (2015), the UT particulate sulfur concentration is affected by volcanism. In Fig. 7a we see that the main differences in UT concentrations of the two cases appears from January to July, coinciding with the season of transport down from the Junge layer into to the LMS and the shrinkage of the LMS due to tropopause upward motion (Appenzeller et al., 1996; Gettelman et al., 2011). In the period September to November the influence from volcanism on the UT particulate sulfur concentrations is small (compare the background and moderate volcanism cases; Fig. 7a). When comparing these cases to the concentration of particulate sulfur found to be of tropospheric origin, it is clear that there is strong agreement between all three categories in the fall months, whereas differences are large during the remainder of the year. The UT particulate sulfur concentration of stratospheric origin can be estimated by subtracting CS,UTtrop from the two cases of UT concentrations. The results are shown in Fig. 7b, where peak stratospheric influences of 19 and 36 ng m-3 STP are found in the spring, and minimum contributions of approximately 1 ng m-3 STP are found in the fall for the “background” and “moderate volcanism” cases. Summing the observations up (Fig. 7c), a clear seasonal dependence in the fraction of the UT particulate sulfur concentration originating in the stratosphere was found, from close to 100 % in late winter/spring to approximately 10 % in the fall. On a yearly average the fraction of the UT particulate sulfur that originates in the stratosphere is approximately 50 % during background conditions and 70 % during moderate influence from tropical volcanism.

The tropospheric source of UT particulate sulfur could be transported from the planetary boundary layer either in the form of particulate sulfur or precursor gases, in the latter case primarily sulfur dioxide (SO2). In fall, winter and spring, with maximum in the winter, vertical transport by WCBs from the boundary layer to the UT is strong (Eckhardt et al., 2004), whereas in the summer deep convection is the most important mode (Hess, 2005; Kiley and Fuelberg, 2006) in the extratropics. For sulfate, the most common chemical form of particulate sulfur, the concentration usually shows a rapid decline with altitude in the troposphere (Heald et al., 2011) associated with formation of precipitation. Cloud processing also tends to strongly reduce SO2 concentrations with altitude, where the relative availability of SO2 and hydrogen peroxide (H2O2) is important for the SO2 lifetime in the cloud.

We have found some very distinctive characteristics of the particulate sulfur concentration of tropospheric origin in the UT, with low concentrations in February and March, increasing concentrations during the summer and maximum in the sources of tropospheric origin in the fall. Carbon monoxide (CO) is often used as a tracer of air pollution. Zbinden et al. (2013) found winter/spring maximum in the CO concentration in the UT of the NH in contrast to the tropospheric component of the UT particulate sulfur. This contrast can, at least in part, be explained by the oxidizing capacity in the UT. The summer abundance of the hydroxyl (OH) radical (Bahm and Khalil, 2004) induces a decline in CO (Bergamaschi et al., 2000; Osman et al., 2016). High abundance of this radical, on the other hand, and availability of SO2 can lead to production of particulate sulfur. SO2 measurements by the satellite-based instrument MIPAS have recently become available (Höpfner et al., 2015). These data have large uncertainties (Höpfner et al., 2015), and likely overestimate the SO2 concentration (Rollins et al., 2017). However, here we only qualitatively use the seasonal variation. The MIPAS results indicate a UT seasonal variation in the NH midlatitudes with low concentrations in December to March and the highest concentrations in June to September (Höpfner et al., 2015). Deep convection provides a rapid route upwards in the atmosphere, favoring transport of short-lived species like SO2 (TF-HTAP-2010, 2010; Dickerson et al., 2007). Low abundance of both SO2 and OH in the NH UT thus could explain the weak tropospheric contribution to the UT particulate sulfur during late winter and early spring. The increase in CS,UTtrop during the spring and summer months coincides with increases in convective activity as well as in the concentration of both SO2 and OH, thus offering a plausible explanation. However, the quantitative understanding of the SO2 transport paths and UT seasonality requires further study. Still, we need to consider the late maximum in the fall of the tropospheric source of UT particulate sulfur.

The Asian monsoon is an important feature of global circulation in June to September, which has been found to reach deep into the lower stratosphere. This intrusion of tropospheric air into the tropical transition layer (Sunilkumar et al., 2017) has been manifested by gas-phase components including water and ozone (Gettelman et al., 2004; Randel and Park, 2006) and hydrogen cyanide (Randel et al., 2010). Later an aerosol layer extending from the UT to the potential temperature of 420 K in the lower stratosphere was found at 14 to 18 km altitude (Vernier et al., 2011b; Thomason and Vernier, 2013; Vernier et al., 2015). Figure 8a–e show monthly means of the scattering ratio obtained from the CALIPSO sensor CALIOP for the months June to October 2013, when the volcanic influence was low. Formation of the ATAL can be identified in July (Fig. 8b) with maximum intensity in August (Fig. 8c). Even in September can a weakened ATAL be identified (Fig. 8d), whereas in June and October the scattering in the ATAL region at 14–18 km altitude is very weak (Fig. 8a and e).

The altitude range of the ATAL is above the measurement altitudes of IAGOS-CARIBIC (9–12 km). However, the poleward circulation along isentropes bending downwards, further amplified by an extratropical cross-isentrope, downward component, brings the ATAL down to lower altitudes. The ATAL is too weak to be traced by CALIOP in the downward transport (Fig. 8a–e). We therefore illustrate the subsidence using the eruption of the tropical volcano Nabro in summer 2011 (Table 1), with effluents that occupied approximately the same region as the ATAL (see Fig. 3 in Bourassa et al., 2012). The effluents of this eruption were rapidly transported to the north in the tropical tropopause layer. Figure 8f–j actually includes two volcanic eruptions; besides the tropical volcano Nabro, the Icelandic volcano Grimsvötn injected a volcanic cloud into the tropopause region at midlatitudes which is visible in June and July 2011. The time series of the Nabro volcanic cloud in Fig. 8f–j unambiguously demonstrates the transport down to the IAGOS-CARIBIC flight altitudes in the season of the ATAL. In Fig. 6c we see a clear increase in the particulate sulfur column in September 2011, following the decline of the June–July 2011 short peak from the of Grimsvötn eruption in May 2011, thus demonstrating agreement between CALIOP and IAGOS-CARIBIC measurements of particulate sulfur. The volcanic cloud from Nabro initially resided at somewhat higher altitude than the ATAL, implying that the ATAL can be expected to reach IAGOS-CARIBIC flight altitudes somewhat earlier in the year. The transport in this season contains little particulate sulfur during periods without fresh volcanic aerosol; compare Fig. 7 “tropical volcanism” (the “tropical volcanism” category of Fig. 7 does not include the Nabro eruption – see Sect. 2.3) with “background” and the particulate sulfur of tropospheric origin.

The downward transport allows delayed detection of the ATAL in the fall from altitudes above the measurement altitude range of IAGOS-CARIBIC in the same way as the spring detection of aerosol transported from the Junge layer, with or without volcanic influence, as shown in Fig. 7. A difference between these two seasons is that in the spring the strong source of particulate sulfur is stratospheric, whereas in the fall the source is tropospheric extending from the UT into the stratosphere. We therefore conclude that the UT particulate sulfur concentration of tropospheric origin that starts to increase rapidly in August, peaking in September to November, and is back at low concentration in January (Fig. 5b) is most likely caused by the ATAL formation from the Asian monsoon.

Model studies find similar UT/stratosphere distribution of the ATAL as the experimental studies, that Asian pollution sources strongly contribute to the ATAL and that sulfate is an important component of that aerosol layer (Neely III et al., 2014). Besides this component, primary and secondary organic aerosol constituents are predicted to be important components of the ATAL (Yu et al., 2015). The present study is, as far as we know, the first observation of a chemical component of the ATAL. In agreement with modeling results we find that particulate sulfur is a component of the Asian tropopause aerosol layer. This component affects the tropopause region from July to September at 14 to 18 km altitude (Fig. 8) and the extratropical tropopause region from August to December (Fig. 5b).

Particulate sulfur (usually sulfate) in the upper troposphere (UT) and the lowermost stratosphere (LMS) obtained from the IAGOS-CARIBIC platform was investigated at northern midlatitudes in the time period 1999–2014, which covers several tropical and extratropical volcanic eruptions. The study is based on the use of linear regression models, where individual measurements in the strong gradient of the extratropical transition layer (ExTL) can be converted into an estimate of the column of particulate sulfur in a layer 3000 m above the dynamical tropopause (here defined at 1.5 PVU). The obtained time series in particulate sulfur column concentration shows distinct response to extratropical volcanism and delayed elevation of the column concentration following tropical eruptions. Assuming the stratospheric background particle size distribution and composition (75 % sulfuric acid and 25 % water) the AOD and radiative forcing were estimated; e.g., the peak values following the 2009 Sarychev eruptions were estimated to be 0.006 and -0.15 W m-2, respectively. These estimates refer mainly to the ExTL, i.e., lowest part of the LMS, thus highlighting the importance of the lowest part of the stratosphere for the overall climate impact of volcanism.

As part of this investigation the sources of UT particulate sulfur were explored. A distinct pattern emerges where tropospheric sulfur sources dominate the supply of particulate sulfur to the UT in the fall, whereas stratospheric sources dominate in January to July, the main season of transport from the Junge layer into the LMS. The smallest contributions from the troposphere are found in February and March in conjunction with low sulfur dioxide (SO2) and oxidant concentrations. As the concentrations of these species increase, the UT particulate sulfur concentration of tropospheric origin increases somewhat in April to July. The particulate sulfur concentration shows a threefold increase during the fall, with maximum concentration in September to November. Making use of lidar data from the CALIPSO satellite together with the in situ measurements we find that the Asian tropopause aerosol layer (ATAL) resulting from the Asian monsoon is the cause of the increase. The ATAL is formed at 14–18 km altitude and extends from the UT to approximately 420 K potential temperature in the stratosphere, with main extension in July to September. The ATAL is transported downwards and affects the extratropical tropopause region in August to December. As far as we know, this is the first measurement of a chemical species in particles connected with the ATAL. The stratospheric and tropospheric contributions to the UT particulate sulfur concentrations thus have strong and opposite seasonal dependences. On annual average the stratospheric contribution to the UT particulate sulfur is estimated to be 50 % during stratospheric background conditions. During influence from moderate tropical volcanism the stratospheric fraction rises to 70 %. The particulate sulfur concentration in the UT is thus to a large degree governed by the stratosphere and volcanism.