Model simulations were performed using the global CESM1 using its WACCM module linked to the CARMA module, involving the sulfur cycle with a sectional aerosol scheme . Land, sea ice, and rivers were active modules, whereas oceans were data prescribed. The spatial resolution was a longitude/latitude grid of 144 points by 96, respectively (i.e. approximately 2∘ resolution), and over 88 levels of altitude ranging from the ground to approximately 150 km altitude with approximately 20 levels in the troposphere. Specified dynamics were used, with a nudging towards the Modern-Era Retrospective analysis for Research and Applications (MERRA) meteorological data at every time step (30 min) with a weight factor of 0.1 towards the analysis, for temperature and wind fields. The following surface emissions were prescribed in the model. For SO2, NH3, black carbon, organic carbon, NOx, CH4, and CO emissions, the MACCity data set was used (, ; , ;  ; , ). Anthropogenic CH4 emissions were added from the EDGAR v4.2 database (available at http://edgar.jrc.ec.europa.eu); biogenic CO emissions were added from the Model of Emissions of Gases and Aerosols from Nature – Monitoring Atmospheric Composition and Climate (MEGAN-MACC) database . Carbonyl sulfide (OCS) was prescribed using data from . CH2O was prescribed according to the IPCC RCP8.5 scenario , and for H2 the ECCAD-GFED3 database was used . For CO2, N2O, CCl4, CF2ClBr, CF3Br, CH3Br, CH3CCl3, CH3Cl, CFC11, CFC113, CFC12, and HCFC22 emissions, lower boundary conditions were prescribed following CCMI/RCP8.5 data.

Simulations were started on 1 January 2009, using the CESM1(WACCM) initial atmosphere state file at that date. This enabled a 6-month model spin-up period before the eruption injection on 15 June 2009, after which the simulations were continued for 1 year, ending on 31 May 2010. The Sarychev Peak eruption was simulated by injecting volcanic SO2 (and HCl) gases into model grid boxes corresponding to the location of the volcano (48∘ N, 153∘ E), over the duration of 15 June 2009, spread evenly between 11 and 15 km altitude a.s.l.. The model's 2.5∘ longitude ×1.875∘ latitude grid resolution means that the volcanic plume is initially too dilute in the model compared to reality. This is nevertheless a common methodology; see, e.g. . The vertical distribution of our SO2 injection follows previous model studies. It is a somewhat coarse approximation given that report lidar observations of fine-scale aerosol layers shortly after the eruption. Nevertheless, these were subsequently observed to collapse into a single layer in the lower stratosphere. For the magnitude of the SO2 injection, we use a revised estimate in contrast to previous studies, as discussed below.

A detailed chronology of the Sarychev Peak 2009 eruption can be found in , which identified three explosive periods: on 12–13 June, repeated explosions occurred, reaching heights ranging from 5 to 10 km; an isolated, high-altitude explosion occurred on 14 June, reaching 21 km altitude; finally, on 15 June, a series of consecutive explosions reached altitudes ranging between 10 and 15 km (all times are in UTC). The first eruptive clouds on the 11–14 June period were mainly ash . We neglected the minor, low-altitude (inferior to 5 km) explosions reported on 11 and 16 June, and injected SO2 continuously for a 24 h period on 15 June spread evenly between 11 and 15 km altitude a.s.l.. The timing of the SO2 emissions is based on SO2 satellite retrievals from IASI , MODIS (MODerate-resolution Imaging Spectroradiometer; ), and OMI (Ozone Monitoring Instrument; ), which all show that the majority of high altitude SO2 was released on 15 June (and possibly in the early morning of 16 June). used a total injection mass of 1.2 TgSO2, which was the SO2 total mass value retrieved on 16 June with IASI. An update of the SO2 algorithm found a maximum SO2 mass value of around 0.9 Tg, a value which was confirmed with subsequent updates of that algorithm . It is also consistent with retrievals from OMI and MODIS . In contrast, the IASI retrievals reported in found that the transient SO2 burden reached only up to 0.6 TgSO2. We consider though that 0.9 Tg of SO2 is the best estimate for the mass of SO2 injected by the Sarychev eruption into the UTLS. We did not consider any ash emissions.

In a second simulation, 27 GgHCl was co-injected alongside the 0.9 Tg of SO2. This initialization follows the recent identification of a localized stratospheric HCl enhancement following the Sarychev eruption , based on analysis of Microwave Limb Sounder (MLS) satellite observations, reporting a HCl/SO2 mass ratio of around 3 %. Since the low vertical resolution of MLS in the lower stratosphere makes it difficult to infer the precise injection altitude of HCl, we assumed an HCl injection altitude identical to that of SO2. A control run without the volcanic gas injection was also performed, enabling anomalies to be calculated. In the present paper, we will refer to control runs as “volcano-off” simulations and to runs including the eruption as “volcano-on” simulations.

The CESM1(WACCM) atmospheric chemistry scheme includes a detailed sulfur cycle and key stratospheric nitrogen (NOy), and halogenated (i.e. chlorine and bromine) and hydrogenated (in particular HOx radicals) compounds. The formation and microphysics of sulfuric acid aerosol particles simulated by the CARMA module are described in detail in .

The CARMA module in sectional configuration yields particle concentration across 30 size bins ranging from approximately 0.68 nm to 3.25 µm in dry diameter. Effective radius is also provided as a direct model output. Post-processing of the model output was used to determine wet particle size distributions, extinctions, and optical depth. In each model grid cell, the wet diameter of each size bin was calculated using a (hygroscopic growth) parameterization of H2SO4(aq) particle volume as a function of acid weight percentage (wt%H2SO4), ambient humidity, and temperature following . Extinctions at 750 and 550 nm were calculated by combining the particle concentrations across the sectional size bins with the corresponding wet radii and particle refractive indices following , using a Mie scattering code at the desired wavelength . The aerosol extinctions were integrated with altitude over the stratosphere to yield SAOD.

The model SO2 output (from simulations with and without HCl co-injection) is compared to vertical columns of SO2 and total (northern hemispheric) SO2 burden derived from IASI. The Infrared Atmospheric Sounding Interferometer is an instrument that has been present aboard the MetOp-A satellite since the end of 2006. It is a spectrometer measuring infrared light spectra at nadir. Its primary goal is to assess temperature and water vapour content of the atmosphere, but it can also be used to retrieve the atmospheric concentrations of various gases, including SO2 . IASI provides global coverage twice a day, and its footprint ranges from circular (12 km diameter at nadir) to elliptical (up to 20 km by 39 km at the end of the swath). For this comparison, we use the IASI retrieval of SO2 by . The IASI data set and retrieval algorithm used for this precise eruption can be considered as showing a lower threshold of around 0.3 DU, and SO2 loads can be expected to have a 10–20 % uncertainty. IASI altitude retrievals have a typical sensitivity of 1–2 km. We also compare our results to HadGEM2 model simulations of SO2 and earlier IASI SO2 retrievals reported by .

Comparisons of the modelled aerosols with in situ measurements are two-fold. First, we compare the model's output with size-resolved aerosol measurements carried out with the balloon-borne Stratospheric and Tropospheric Aerosol Counter (STAC) optical particle counter (OPC) instrument over Kiruna, Sweden, on 2, 7, and 18 August 2009, and 18 May 2010. STAC can be borne under stratospheric balloon gondolas and can measure low concentrations in aerosols down to approximately 10-4 cm-3µm-1 . Particles are classified by their diameters into tuneable size bins ranging from a few tenths of micrometre to a few micrometres. The counts in each size bin are normalized by the bin width to yield a size distribution. The uncertainty, defined as the relative standard deviation (SD), is 60 % for aerosol concentrations of 10-3 cm-3, 20 % for 10-2 cm-3, and 6 % for concentrations higher than 10-1 cm-3. STAC was operated successfully on eight different balloon flights throughout the August–September 2009 period over Kiruna, Sweden (68∘ N, 20∘ E), as part of the StraPolÉté campaign (French acronym for Stratosphère Polaire en Été), and also in May 2010, as part of the AEROWAVE project (acronym for AEROsols, WAter Vapour and Electricity). Measurements of the STAC instruments are available online at https://cds-espri.ipsl.upmc.fr. During these flights, it was demonstrated that STAC passed through the Sarychev plume , as explored further in the present paper. Our comparison focuses on the submicron range between ≈0.3 and 1 µm diameters. We have performed an interpolation of the counts from the model's size bins to the STAC size bins (and from the model pressure levels to the observed pressure of the balloon payload) in order to enable a direct comparison.

Second, we also compare the model's aerosol output with in situ measurements carried out by the OPC of the University of Wyoming , flown on stratospheric balloons launched from Laramie, USA (41∘ N, 105∘ W), on 22 June 2009 and 7 November 2009 . For comparison to the model, total particle number above two diameter threshold sizes is considered here: d> 20 nm (condensation nuclei, CN) and d>0.5 µm (nuclei, N). Uncertainties are 85, 25, and 8 % for concentrations of 10-3, 10-2, and 10-1 cm-3, respectively . These data are available from ftp://cat.uwyo.edu/pub/permanent/balloon/Aerosol_InSitu_Meas/US_Laramie_41N_105W/. They have been derived by the University of Wyoming as follows: the measurement of N is calculated directly from the OPC instrument. The CN is derived from a condensation nuclei counter co-deployed on the balloon payload. Note that both STAC and University of Wyoming OPCs have been compared by .

Model SAOD was compared to that derived from extinction measurements by the OSIRIS aerosol instrument (aboard the Odin satellite). OSIRIS is a limb sounder that is able to provide information on the vertical distribution of atmospheric aerosols from the upper troposphere up to the lower mesosphere through the analysis of scattered sunlight. This Canadian instrument has been active since November 2001 aboard the Swedish satellite Odin . Its global coverage reaches up to 82∘ in latitude. Odin evolves on a Sun-synchronous orbit, and therefore the availability of OSIRIS's measurements is latitude and time dependent. Our analysis focuses on extinction measurements from OSIRIS version 5.07, available from http://odin-osiris.usask.ca/. Importantly, a novel aspect of our study is that our analysis specifically accounts for instrument errors or limitations as reported by . Model output data have been degraded accordingly. First, the modelled extinctions have been made to saturate at an upper threshold of 2.5×10-3 km-1; then, extinctions have been only integrated above a certain altitude, dependent on the latitude: a linear variation of this lower limit was assumed from 0.5 km above the tropopause at the Equator up to 5.5 km above the tropopause at the poles. Further details are given in Sect. .

Figure  shows vertical column densities of SO2 from the CESM1(WACCM) simulation in which both volcanic SO2 and HCl were injected (Fig. 1b) and a comparison with IASI retrievals (Fig. 1a). Both sets of maps are shown with the same lower threshold in terms of Dobson units, corresponding to an estimated lower threshold of 0.3 DU in IASI's retrievals for this precise eruption and for the IASI retrieval algorithm used . The spatial and temporal evolution of the Sarychev SO2 plume is reasonably well simulated by the CESM1(WACCM) runs throughout the first fortnight following the eruption. There are some notable discrepancies, for instance, on 16 June 2009 south-west of Alaska: this is likely due to our simulation not accounting for the small amount of SO2 that was emitted before the main eruption on 15 June 2009. Also, Asian pollution (close to 0.3 DU) is evident in the simulations shown in Fig.  but not observed by IASI, likely due to the reduced sensitivity of the IASI retrievals to SO2 below 5 km altitude. To quantify the spatial-amplitude match between the two sets of data, we chose a colocation index calculated as ρ=E[(P1-μ1)(P2-μ2)]σ1σ2, where P1 and P2 are the two-dimensional matrices representing the spatial SO2 loads (for model and satellite retrievals), sampled over the same spatial grid, and stacked into one-dimensional vectors; μ{1;2} and σ{1;2} are their respective means and SDs. It is expected that the index drops quite quickly after the eruption due to greater dispersion in the model on the 2×2 degree grids than in the finer-scale (tens of kilometres) IASI observations. Colocation indices were calculated over the first fortnight following the eruption (Table ) for the simulation with SO2 injection only and with HCl co-injection alongside SO2. As can be noted from Table , colocation indices show comparable values for both model runs. This indicates broadly similar SO2 dispersion in the model runs.

The spatial and temporal evolution of the plume in our study is consistent with the results of , where AIRS data are presented along with results of simulations by a particle dispersion model.

Figure  shows the modelled northern hemispheric SO2 burden in teragrams, calculated by integrating the model anomalies from CESM1(WACCM) simulations with SO2 injection only and with SO2 and HCl co-injection (anomaly denotes a volcano-on simulation from which the volcano-off control run has been subtracted). Two adjusted CESM1(WACCM) model results are also presented that only include data over columns with >0.3 DUSO2 to enable a better comparison to the IASI observations. Alongside is shown the observed evolution in northern hemispheric SO2 burden derived from the IASI retrieval by (which has a lower threshold of around 0.3 DU; see Sect. ). Finally, we also show the northern hemispheric SO2 burden as simulated using the HadGEM2 model and the IASI retrieval reported in that same study, both of which estimated 1.2 Tg SO2 injection in contrast to the revised IASI analysis that yielded 0.9 TgSO2 used in our study.

A notable result is the slower decline in SO2 burden for the model run with volcanic SO2 and HCl co-injection than volcanic SO2 (only). There is also a corresponding slower increase in the sulfate aerosol burden (Fig. ).

The presence of HCl slows down the oxidation of SO2 to sulfuric acid aerosol particles and hence lengthens the e-folding time of SO2 in the stratosphere by about 2 days (see calculations below). This occurs as a result of the competition between their two main oxidation reactions involving OH. These are HCl+OH→Cl+H2O and the trimolecular reaction (where M is a third body, e.g. N2 or O2) SO2+OH+M→HSO3+M, where HSO3 subsequently leads to the formation of H2SO4 through the reaction sequence described by . This conversion of SO2 to H2SO4 is limited by the rate of Reaction () below 40 km in altitude. Competition between Reactions () and () results in a slower rate of oxidation of volcanic SO2 in the presence of co-injected HCl.

A second notable result is that all the unadjusted model outputs overestimate the SO2 burden following the eruption compared to IASI measurements. The HadGEM2 model SO2 exceeds the IASI observations for the whole period. The unadjusted CESM1(WACCM) outputs also exceed the IASI observations after a few days. This behaviour contrasts with the two adjusted CESM1(WACCM) model outputs that correct for the 0.3 DUSO2 lower value of the particular IASI retrievals used. The adjusted CESM1(WACCM) model outputs remain in close agreement with the observed post-eruption SO2 burden for the first 1–2 weeks, after which the model-simulated SO2 burdens decline more rapidly than the IASI 2012 observations. This evolution can be expected: a greater dispersion in the 2∘×2∘ model grid cells than in reality (and than observed by the IASI footprint of tens of kilometres) would cause an underestimation of the model SO2 burden compared to IASI. This effect will become more pertinent with dilution over time as the SO2 column approaches the 0.3 DU limit.

In summary, we find that the CESM1(WACCM) model run (adjusted output) with SO2 and HCl co-injection gives the best agreement with the IASI SO2 observations. The simulation with SO2 with HCl injection therefore forms the basis for further analysis in Sect. .

Our model–observation comparison of SO2 burden trends can also be quantified in terms of the e-folding time. The definition of the e-folding time τ is the following: let M(t) be the concentration of a species through time; if we assume it follows an exponential decay over a certain period of time t>t0, then τ is such as ∀t>t0,M(t+τ)=M(t)/e; i.e. τ corresponds to the time by which the concentration falls to 1/e of its initial value. For these calculations, we choose the SO2 burden maximum as the initial value (0.9 Tg in our study). The e-folding time constant for SO2 is approximately 17.0 days for the simulation including HCl, about 2 days longer than the approximately 15.0 days for the simulation that was run without HCl. When these CESM1(WACCM) model outputs are adjusted to correct for the 0.3 DUSO2 lower value of the particular IASI retrievals used, they yield e-folding time constants of 11.5 and 10.0 days, respectively. For the IASI SO2 retrieval of , we calculate 12.0 days, i.e. very similar to the adjusted model simulation with SO2 and HCl co-injection (11.5 days). For comparison, report that the HadGEM2 model yields a 13- to 14-day SO2 e-folding time (assuming a higher SO2 injection of 1.2 Tg and no HCl co-injection). Regarding IASI observations, report an IASI SO2 e-folding time of 10–11 days, whilst using our method we calculate 9.0 days for the IASI retrieval of 2010. This is summarized in Table .

Here, we compare size-resolved aerosol concentrations from our simulations with in situ measurements from balloon-borne OPCs over Laramie, USA (June and November 2009), and Kiruna, Sweden (August and September 2009). It should be emphasized that the instruments are likely to detect a wider range of particles and particle compositions present in the stratosphere, i.e. internally and externally mixed particles with some organic and meteoric components , whereas our model simulations provide pure sulfuric acid aerosol particles only. Nevertheless, these are expected to be the dominant source of aerosol in the lower stratosphere in the months following the Sarychev Peak 2009 eruption.

First, we compare the model to measurements carried out by the University of Wyoming OPC during balloon-borne flights over Laramie, Wyoming (USA, 41∘ N, 105∘ W), on 22 June and 7 November 2009. These observations were made 1 week and nearly 5 months after the Sarychev eruption, respectively. previously suggested that a significant volcanic influence can be seen in the data from 7 November but not on 22 June, based on comparison with balloon flights from other years. Here, we compare the data directly to aerosol simulated by our model runs.

Figure  shows both the model and measured aerosol particle number concentrations over Laramie for two particle size ranges: d>20 nm (noted CN, for condensation nuclei) and d>0.5 µm (noted N). Overall, there is good general agreement between simulated and measured values in terms of number concentrations and in the general trend with respect to altitude and size range separation. Note that model–measurement differences are greater in the troposphere since only sulfuric acid particles are simulated.

The upper panel of Fig.  (22 June 2009) shows that the volcano-off simulation reproduces the in situ observations of particle number with a very good agreement, supporting the hypothesis of that there was no significant volcanic influence on this day. However, the volcano-on simulation in fact simulates the presence of a volcanic plume, as can be seen by enhancements in CN and N between 13 and 15 km altitude. We note that the precise geographical location of plume structures is difficult to simulate using low-resolution simulations just 1 week after the eruption. Remote sensing observations suggest the initial presence of multiple aerosol layers in the stratosphere that subsequently collapsed into a single layer , whereas our CESM1(WACCM) model study assumes injection over 11–15 km. We also suggest that model horizontal resolution effects are a further possible source of error in the volcano-on simulation that might have led to anomalous sulfate plume structure over the measurement location. A geographic 2-D map of the vicinity of Laramie that shows model-simulated sulfuric acid aerosol particles at 13 km altitude (Fig. ) show that the location of the measurements lies on the edge of an aerosol plume structure simulated by the model. Diffusion on the model grids (2∘×2∘ resolution) or uncertainties in the initialization altitude could therefore lead to modelled plume structure over Laramie that is not evident in the observations.

Conversely, on 7 November, the volcanic plume is simulated to be much more homogeneous (and dilute), covering a larger area that encompasses Laramie. The lower panel of Fig.  shows modelled and observed aerosol particle number concentrations for 7 November 2009. For the d>0.5 µm size range (N), the agreement between the volcano-on simulation and the in situ measurements below 17 km indicates that volcanic aerosol particles were still present and detectable over Laramie nearly 5 months after the eruption, and their presence can be quantitatively reproduced by the CESM1(WACCM) model. The profiles from both volcano-on and volcano-off simulations are very close in the d>20 nm size range (CN) indicating the progressive return of the simulated concentrations to background conditions for this size range.

Next, we compare the CESM1(WACCM) simulations to in situ aerosol measurements made by the STAC instrument on a balloon gondola in northern Sweden. Figure  compares the particle counts observed by STAC and the sulfate particle concentrations simulated by the WACCM model for the same location (Kiruna, Sweden, 67∘ N, 20∘ E), and times: 2, 7, and 18 August 2009. A comparison is also shown for 18 May 2010 when the stratosphere can be considered to be close to background conditions. The model outputs have been interpolated to the pressure observed by the balloon payload and to the specific size bins of the STAC instrument covering 0.325 to 0.885 µm mean diameter.

In Fig. , a volcanic sulfate aerosol plume can clearly be identified between 11 and 19 km altitude for all flights in August 2009. This is demonstrated in the third column of the figure by an important difference in modelled particle number over the size bins of the STAC for the volcano-on and volcano-off simulations: total particle number on the STAC diameter range is enhanced by the volcanic eruption by between 1 and 2 orders of magnitude depending on the altitude. Total number simulated in the volcano-on simulation is in good general agreement with the STAC observations.

There are some discrepancies between model and observations at higher and lower altitudes: at lower altitudes, the model yields lower counts than the instrument's counts: this is likely due to the presence in the troposphere of non-sulfate aerosols unaccounted for by the model. For the discrepancies above the plume's altitude, the radiometer MicroRADIBAL (French acronym for Micro RADIomètre BALlon) , flown alongside STAC, identified the presence of some light-absorbing particles around 20 km altitude : these might have affected the STAC measurements (STAC is designed for sulfate particle detection) and were also not included in the model. Their origin is still to be determined. Nevertheless, the good agreement in total number between model and observations in the lower stratosphere (corresponding to the main influence of the volcanic plume) confirms the strong impact of the Sarychev eruption on aerosol number.

Comparing these aerosol observations above Kiruna in August to those above Laramie in November on an order of magnitude basis, the Laramie measurements have ≈1 cm-3 particles of diameter greater than 0.5 µm at 14 km altitude in November, whereas measurements over Sweden in August of the same year show approximately 10 to 100 times more particles of size greater than 0.4 µm in diameter. This indicates the result of coagulation, condensation, sedimentation, and transport and dilution processes: 2 months after the eruption there is a strong volcanic impact, but few submicrometer-size volcanic particles are left in the stratosphere 5 months after the eruption. The volcanic aerosol evolution is discussed further in Sect. .

Figure  shows the particle size distributions measured by the STAC, separated in 1 km layers of altitude, for the same four flights as Fig. , and compares these to size distributions simulated by the CESM1(WACCM) model. The size distributions are displayed in terms of number, surface, and volume, and should be read by pairs, comparing the STAC observations to the control run (volcano-off) on the one hand, and the simulations including the volcano eruption (volcano-on) on the other hand.

The figure highlights that the control run underestimates the particle number (area or volume) size distribution curves by orders of magnitude compared to the STAC observations. A much better agreement is found when the volcanic emission is included in the model simulations, showing a good ability of the model to reproduce volcanic aerosol plumes in terms of aerosol size distribution. For 18 May 2010, nearly 1 year after the eruption, the difference between the volcano-on simulation and control run is much less noticeable. This comparison reflects the ability of the model to simulate stratospheric aerosol size distributions in background (or near-background) conditions.

Extinction data from OSIRIS have been used for model and observational assessment of stratospheric aerosol impacts following the Sarychev 2009 eruption . However, as mentioned in the introduction, biases in the OSIRIS measurement following volcanic eruptions can affect the reported model–observation comparisons.

In , a detailed analysis of OSIRIS's limitations was carried out. These authors have shown that two main factors affect the derivation of SAOD by OSIRIS: (i) an upper detection limit on the value of extinctions, above which the measured values saturate, and (ii) a latitude dependence in the minimal altitude above which extinctions are integrated to yield the SAOD.

As pointed out by , it is impossible to reverse this process of data degradation; the best we can achieve to perform consistent model-to-observation comparisons is to degrade the extinctions derived from the model in order to derive SAOD “as OSIRIS would detect it”. It must nonetheless be emphasized that such a comparison is not a complete evaluation of the model performance: any agreement found cannot fully validate aspects of the model output that are removed in the degradation process. Nevertheless, such a comparison of the degraded model to (biased) satellite observations is highly valuable: it enables an assessment of model performance on a global scale, which cannot be achieved using local-scale in situ observations.

We use the following method: first, we allow the extinctions calculated in the model to saturate, with an upper threshold of 2.5×10-3 km-1 corresponding to the detection limit described in . Second, extinctions are integrated over truncated vertical columns of the atmosphere, introducing a lower altitude limit dependent on the latitude and the local tropopause height. Following , we define the minimal altitudes Zmin above which the extinctions are integrated as Zmin(λ,ϕ,t)=Ztrop(λ,ϕ,t)+Δ(ϕ), where Ztrop is the local tropopause height, λ is longitude, ϕ is latitude, t is time, and Δ is a positive offset function, which was taken in our case as linearly varying with latitude from 0.5 km at the Equator to 5.5 km at the poles. These were chosen as a trade-off between the histogram of values in and actual minimum altitudes reached by OSIRIS over the 2009–2010 period (see Appendix Fig. ). For this series of calculations, thermal tropopause heights were diagnosed in the model. We verify the broad consistency of these altitude limits for OSIRIS data during the 2009–2010 Sarychev post-eruption period in Fig. . Integrating the model-simulated saturated extinctions at 750 nm over the truncated altitude columns as defined above gives SAOD values that can be considered reasonably consistent with the measurements performed by OSIRIS.

Figure a shows the zonally averaged stratospheric sulfate AOD, through time, over the Northern Hemisphere, as computed by CESM1(WACCM) in the volcano-on simulation (with co-injection of HCl). A degradation of the model data was then performed following the method described above. The resulting estimation of the sulfate SAOD “as would be detected by OSIRIS” is shown in Fig. b. Figure c shows the observed SAOD measured by OSIRIS. Over the winter months, there is a lack of observational data from mid-October 2009 until the beginning of 2010, particularly at high latitudes, that coincides with the polar night. A precise comparison for these months is therefore not possible. CESM1(WACCM) suggests that the sulfate SAOD remains at a fairly constant level over the Northern Hemisphere over the October–December 2009 period, then decreases quite quickly from February to April 2010.

The degraded model SAOD shows reasonable agreement with the SAOD observed by OSIRIS, whilst the non-degraded model simulates much higher SAOD. This demonstrates that OSIRIS's limitations are crucial to the interpretation of its data. In Fig. , the observed (OSIRIS) SAOD shows, however, a slightly stronger maximal magnitude than the degraded SAOD from the model. A possible explanation may be that CESM1(WACCM) yields extinctions for sulfuric acid particulates only, whereas OSIRIS's observations account for a more comprehensive SAOD that can include non-sulfate compounds in the lower stratosphere.

To place a greater emphasis on sulfuric acid particulates due to the volcanic eruption, we convert all three data sets to anomalies. These anomalies were calculated by subtracting background conditions from the SAODs, for which averages calculated in the first week of June 2009 were used as an approximate reference. Figure  presents the same layout as Fig.  but now displays the SAOD anomalies over the same period (1 June 2009 until 31 May 2010). Again, a good accordance is found between the degraded model compared to OSIRIS (with the non-degraded model showing higher SAODs). The agreement in SAOD anomalies in Fig.  is better than for the absolute SAODs in Fig. . This indicates that differences in the background aerosol content prior to the eruption may explain some of the model–measurement discrepancy in terms of SAOD maximum amplitude as highlighted in Fig. .

Integrating anomaly data presented in Fig.  yields the Northern Hemisphere SAOD anomaly calculated at 750 nm over the year following the eruption, shown in Fig. . The dashed red line is SAOD simulated by the model. The plain red line indicates the same data after the OSIRIS bias degradation, and the blue line indicates SAOD from OSIRIS observations. Note that missing data in OSIRIS's measurements during winter were taken into account in the integration of the degraded model data, as shown by the shaded area in Fig. . Figure  along with Fig.  point out very clearly that taking into account OSIRIS's limitations gives a very good match between simulated and measured AOD values. Figure b shows the modelled temporal evolution in 550 nm extinction coefficients, zonally averaged for the Northern Hemisphere, again highlighting maximum aerosol content around mid-July 2009.

Comparing the direct output of the model to OSIRIS in both Figs.  and  highlights a much stronger and faster formation of sulfuric acid aerosols in the model than can be detected by the OSIRIS instrument, which experiences strongest measurement biases shortly after the eruption due to the saturation effect. Further quantification is given below, including e-folding times.

Analysing the temporal evolution of the (non-degraded) model SAOD identifies a peak in the Northern Hemisphere 750 nm SAOD of ≈0.018 on 12 July 2009 (Fig. ), followed by a long decay with an e-folding time of ≈169 days. Conversely, OSIRIS shows a much fainter and later peak (≈0.004 on 1 September 2009), with a quicker decay (e-folding time of 52 days). The degraded model SAOD yields an e-folding decay time (51 days) that is very comparable to that from OSIRIS; the peak value is also similar in amplitude to OSIRIS's and is reached on 17 October. This is slightly later than in the OSIRIS data, although the plateau in SAOD during that period can account for this delay.

Table  summarizes the SAOD e-folding times calculated in this study, along with values from previous studies , and calculated using OSIRIS data. Different bands of latitude are explored. One can note that the e-folding times vary quite significantly between authors, including those computed from OSIRIS's data (it is likely that different versions of the OSIRIS data – v.5.05 up to v.5.07 for the present study – were used). The main point to be highlighted here is the fair consistency obtained between e-folding times computed on the CESM1(WACCM)'s degraded data and on OSIRIS retrievals in our study, as evident from the last two columns of Table . We find that both the saturation limit and the fact that extinction profiles may terminate well above the tropopause are significant sources of measurement bias that need to be taken into account in comparison of OSIRIS data to model studies.

Discrepancies in the magnitude and e-folding times between model and OSIRIS SAODs have been previously mentioned (e.g. ) and are summarized here in Table . This led to a consequent questioning of the models' reliability to simulate sulfuric acid particle formation accurately in terms of timing, thought to be caused by the absence of nucleation of new particles in the model . Conversely, our study using the CESM1(WACCM) model, whose aerosol microphysics includes nucleation, finds very good agreement with OSIRIS retrievals of SAOD in terms of magnitude and temporally when the model SAOD is degraded to account for both saturation and minimum altitude limitations on the SAOD derived from OSIRIS measurements. The maximum in our (non-degraded) model SAOD is significantly higher (by a factor of ≈4.5) than estimated by both OSIRIS and earlier modelling studies of the 2009 Sarychev Peak eruption . A key unconstrained parameter in these earlier studies was the stratospheric particle size distribution that exerts a strong influence on SAOD. It was set to yield an effective radius of around reff=0.13–0.15 µm in , with the model results from also adjusted to represent this size. Previous studies suggested higher reff for large-magnitude eruptions that injected SO2 higher into the stratosphere (yielding longer-lived sulfate clouds): derived reff of 0.22±0.06 µm around 1 month after the Mt. Pinatubo 1991 eruption, whilst suggest post-eruption reff grew from around 0.2–0.3 to 0.4–0.5 µm over the timescale of 1 year. Conversely, a lower reff was thought to be reasonable for the moderate-magnitude 2009 Sarychev Peak eruption that injected into the lower stratosphere (yielding relatively fresh and shorter lived sulfate cloud), and appeared consistent with ground-based remote sensing at Mauna Loa (Hawaii, USA) .

Here, the sectional aerosol representation with full aerosol microphysics in CESM1(WACCM) enables to freely simulate the post-eruption evolution in particle size, without any a priori assumptions. Sulfuric acid is first produced by the oxidation of volcanic SO2, which leads to formation of new sulfuric acid particles by nucleation. Processes such as particle coagulation and condensation of sulfuric acid onto the existing particles causes particle growth. Particles are removed from the stratosphere by sedimentation and tropopause folding . The balance between these processes determines the overall size distribution and its effective radius. Figure  shows the zonally averaged effective radius simulated by the model for three latitude bands (20 to 40, 40 to 60, and 60 to 80∘ N). Particle growth occurs in regions with elevated sulfate concentrations following the volcanic eruption (Fig. , Appendix). Particle size grows to reach a maximum in zonal mean reff of up to 0.2 µm in the lower stratosphere. The greatest enhancement in reff occurs at high latitudes as expected given the poleward atmospheric transport in the stratosphere. At midlatitudes, a temporary decrease in reff can also be seen immediately following the eruption: this is due to new particle formation (nucleation) of particles of a few nanometres' size. The latitudinal trend in reff simulated by our model is broadly consistent with the trend reported from ground-based remote sensing at Eureka (Nunavut, Canada) that found reff=0.29 µm , and with ACE measurements, which report reff=0.1–0.3 µm . Modelled absolute values of reff are also globally consistent with balloon-borne observations in August 2009 . Aerosol size or reff exerts a strong influence on SAOD e.g.. A priori assumptions in stratospheric particle size are thus a major source of uncertainty in model studies that do not freely simulate the aerosol size evolution, and that will tend to cause an underestimation of SAOD in cases where the assumed reff is lower than reality.

Most studies investigating the impacts of modern-day eruptions on stratospheric chemistry have focused on the role of sulfuric acid particles in reducing NOx levels and activating pre-existing chlorine and bromine (ClOx, BrOx) in the stratosphere . One must note that halogens from the 1991 Mt. Pinatubo eruption were efficiently washed out and therefore did not reach the stratosphere , though the washout for the Sarychev case was not necessarily as efficient . Observational evidence of stratospheric NO2 depletion following moderate-magnitude volcanic eruptions is provided by based on satellite remote sensing, and by balloon-borne observations following the Sarychev Peak eruption. Our study builds on these recent works in two aspects. First, we use the CESM1(WACCM) model with sectional aerosol representation. We freely simulate the aerosol surface area (SAD) (a function of particle number and size) that is a key control on stratospheric chemistry impacts. Second, we investigate the stratospheric chemistry influence of volcanic HCl that observations show was co-injected alongside SO2 . The simulated anomalies in ozone and NO2 for the latitudinal bands 40 to 60 and 60 to 80∘ N are shown for simulations with SO2 injection only, and for HCl co-injection with SO2 in Fig. . In summer, greater depletions of up to -60 % for NO2 are found at higher latitudes. This is primarily due to the higher aerosol loadings in these regions, and to favourable solar illumination conditions for which the catalytic ozone loss cycles (through OH radical production) are enhanced . A more detailed description of the involved chemical processes is provided in .

Our results are broadly consistent with the observations of , who reported that stratospheric NO2 abundances were reduced by up to ≈45–55 % over 40–80∘ N as consequence of the Sarychev eruption. showed maximum NO2 depletion of ≈50 % in the summertime lower stratosphere above Kiruna (Sweden) following the Sarychev eruption, based on REPROBUS model simulations with SAD prescribed from observations and without HCl injection. They predicted that ozone depletion reached up to 4 % in the lowermost stratosphere in summertime/early fall. Our CESM1(WACCM) simulation with a sectional aerosol scheme and injection of volcanic SO2 (only) finds similar high-latitude maximum NO2 reduction (≈50 %) and maximum ozone depletion (5 %). Interestingly, our simulations suggest somewhat greater maximum depletions in the simulation with co-injected volcanic HCl (7 and 60 %, for ozone and NO2, respectively) compared to the simulation with SO2 injection only (5 and 50 %, for ozone and NO2, respectively). As a result of the enhanced stratospheric HCl budget throughout the season, more ozone-depleting chlorine radicals are expected to be formed due to Reaction () even at midlatitude conditions, though the impact on ozone appears limited. The impact on summer and fall NO2 is negligible. In the polar winter, cold temperatures lead to further chlorine activation through heterogeneous processes enhancing some NOx and ozone reduction. This study highlights the potential for volcanic HCl to complement and enhance SO2–sulfate impacts on stratospheric chemistry, for eruptions where there is a significant HCl injection to high altitudes. The influence of Sarychev Peak eruption on stratospheric chemistry is nevertheless relatively modest, due to the moderate eruption size, in terms of both the SO2 and HCl injected amounts.