The solar (and lunar) occultation instrument SAGE III/M3M was launched in December 2001 and was operating and providing measurement data between February 2002 and November 2005 . It is the successor of the SAM II, SAGE I, and SAGE II satellite instruments and was later followed by the SAGE III/ISS instrument. SAGE III/M3M gathered information on essentially the same range of different atmospheric constituents as SAGE-II (see ): water vapor, O₃, NO₂, and stratospheric aerosol.

Mounted on the Meteor-3M weather satellite, SAGE III/M3M was in a Sun-synchronous orbit with an inclination of 99.5° . Due to the platform's orbit around 15 sunrise and 15 sunset events were observed by the instrument in 24 h, and the sunrise and sunset measurements ended up happening in different hemispheres, as depicted in Fig. . Sunrise and sunset here refer to orbital sunrise and sunset, i.e., as seen from the instrument, rather than local sunrise and sunset. Sunrise and sunset here refer to the Sun rising or setting behind the Earth from the point of view of the satellite during its orbit. The sunrise measurements oscillate between roughly 35 and 60° S, while the sunset measurements cover the northern latitudes between approximately 40 and 80° N. The data quality of sunrise and sunset measurements is likely very similar.

For a solar occultation measurement a scanning mirror was used to follow the Sun's perceived ascent or descent from or towards the horizon as well as to scan across the Sun's surface and produce multiple samples at each tangent altitude. After passing the neutral density filter and the science aperture, which defines the instantaneous field of view (IFOV) of the instrument, the solar radiation reaches the grating spectrometer and the detectors. Those consist of a CCD array covering the wavelengths from 280 to 1040 nm with a spectral resolution of 1 to 2 nm and an indium gallium arsenide (InGaAs) infrared photodiode around 1550 nm with a 30 nm bandwidth . Aerosol extinction is retrieved from the calculated line-of-sight transmission profiles by removing the effects of Rayleigh scattering due to air molecules, ozone absorption, and NO₂ absorption . In the SAGE III/M3M dataset, the aerosol extinction coefficients are then provided between 0 and 40 km altitude in 0.5 km steps for nine different wavelength channels between 384 and 1545 nm. This is the main dataset used for the stratospheric aerosol size derivation of this work. Version 4 of the SAGE III/M3M level 2 solar occultation data is used .

evaluated the aerosol extinction coefficient channels of SAGE III/M3M based on comparisons to other satellite instruments as well as internal consistency. The channels at 449, 520, 755, 869, and 1021 nm are assessed to be reliable and have accuracies and precisions of around 10 % between 15 and 25 km. The same applies to the 1545 nm channel, although a bias below 15 km cannot be excluded. The 385 nm channel is only recommended for use above 16 km altitude. The 601 nm channel has a much lower precision (∼ 20 %) due to the influence of ozone absorption, and it is generally recommended not to use the 676 nm extinction coefficients. In this work, the channels 449, 755, and 1545 nm are used for the stratospheric aerosol size parameter derivation, as will be explained in more detail in Sect. .

The SAGE III/M3M multiple-wavelength aerosol extinction dataset is used in this work to derive parameters describing the size distribution of stratospheric aerosol particles. This is done by comparing the aerosol extinction data provided in the measurement dataset with a lookup table that was calculated using Mie theory . This aerosol size derivation method is similar to the original methods for deriving size parameters from SAGE-II (see ), adapted for the larger range of wavelengths covered by the SAGE III aerosol extinction channels. It was used also in and is described in detail in .

There are a number of assumptions within the PSD parameters derived this way from SAGE III/M3M, which must be considered when interpreting the results. Firstly, it is assumed that the PSD of stratospheric aerosol can be described well by a monomodal lognormal size distribution, which is defined as follows: 1n(r)=dN(r)dr=N02π⋅r⋅ln⁡σg⋅exp⁡-ln⁡2r/rmed2ln⁡2σg.

In this form, there are three parameters that together describe the aerosol particle size distribution: the median radius rmed, the geometric standard deviation σg, and the total number density N0. At least after the strong Pinatubo eruption from 1991 a bimodal lognormal function yielded a better representation of the observed aerosol size distribution than a monomodal lognormal function in balloon-borne in situ measurements . Despite this, monomodality is a common assumption when deriving stratospheric aerosol size distributions from satellite measurements because it can often be a good model of the true conditions. Additionally, showed that deriving bimodal lognormal PSD parameters from SAGE data is infeasible, stating that “[...] there is not enough information within the SAGE extinction spectra to sufficiently constrain the solution space to accurately infer bimodal PSD parameters”. Still, a monomodal lognormal distribution can be inadequate in some cases. The effect this would have on the size derivation of this work is described in Sect. .

Furthermore, it is assumed that the individual droplets of stratospheric aerosol are spherical, which is a prerequisite for Mie theory to be applicable. Since stratospheric aerosol particles are mostly liquid droplets of sulfuric acid and water with a very small size, usually well below 2 µm , the assumption is realistic.

The basis of the method used to derive aerosol size in this work is the lookup table that is calculated using a Mie code . It consists of aerosol extinction ratios at the wavelengths 449 / 755 nm and 1545 / 755 nm, both calculated for many combinations of different theoretically plausible values of the median radius and the geometric standard deviation σg. The value range covered this way by the lookup table is 1 to 1000 nm for the median radius and 1.05 to 2.0 for σg. The real and imaginary refractive indices necessary for the Mie calculations are taken from and adjusted for temperature using Lorentz–Lorenz corrections as described by . The three wavelengths that are used by the two extinction ratios are chosen to match three of the spectral channels of the SAGE III/M3M instrument, at which aerosol extinction coefficients are provided. As already stated in the previous section, these SAGE III/M3M spectral channels (449, 755, 1545 nm) in turn were chosen because they make use of the wide spectral range of the instrument while also only including reliable aerosol extinction channels . Now that the lookup table is prepared, the median radius and geometric standard deviation value can be derived from a given set of extinction ratios from the measurements via interpolation of the lookup table values. More detailed information on the size derivation approach is provided in .

The median radius and geometric standard deviation can be derived independently of the total number density N0 because the latter cancels out when forming the extinction ratios. The number density can be calculated after obtaining rmed and σg through the following relation: 2N0=kext(λ)σext(λ). Here, kext(λ) is the measured extinction coefficient and σext(λ) is the extinction cross section at wavelength λ simulated from the derived median radius and geometric standard deviation values. The number density values shown in this work were calculated at λ = 755 nm; differences between this and results calculated at 449 or 1545 nm are negligible, however.

Another useful quantity that can be calculated from the derived PSD parameters is the aerosol particle effective radius (reff), which is the area-weighted mean radius . Since larger aerosol particles scatter light more strongly in the relevant size range it is more indicative of the part of the PSD that is responsible for most of the aerosol radiative effect. It can be calculated from rmed and σg or from the aerosol volume density V and surface area density A in the following way: 3reff=rmed⋅exp52⋅ln⁡2(σg)=3VA.

Lastly, the absolute mode width ω, as introduced by , is a very useful parameter to understand how wide the PSD is in linear radius space. It provides the variation in nanometers, with variations then easier to interpret than the geometric standard deviation σg, which only provides information about the width of the size distribution in logarithmic radius space and can therefore easily be misinterpreted. This is illustrated and explained in more detail in . The absolute mode width ω is calculated as follows: 4ω=rmed2⋅expln⁡2(σg)⋅exp⁡(ln⁡2(σg))-1.

Additionally, an accuracy parameter for the derived PSD parameters has been defined to be able to exclude less reliable data from the derived parameters. It relates the lookup table values to the propagated uncertainties of the extinction ratios and is described by . For all SAGE III/M3M aerosol size data shown here, data points with respective accuracy parameter values below 16 are excluded, as was done in . Over the whole dataset this filters out 35.9 % of the derived values overall, although it filters much less in the maximum of the Junge layer, e.g., only 11 % of data at the 18 km altitude level.

SAGE II (Stratospheric Aerosol and Gas experiment II) was the predecessor to SAGE III/M3M. It was mounted on the Earth Radiation Budget Satellite (ERBS). Its mission spanned October 1984 to August 2005, i.e., having considerable overlap with the SAGE III/M3M mission. The platform was in an orbit at 610 km altitude with an inclination of 57° .

Like SAGE III/M3M, SAGE II also measured stratospheric aerosol, ozone, NO₂, and water vapor through the use of the solar occultation geometry. Aerosol extinction coefficients are provided for the spectral channels at 386, 452, 525, and 1020 nm in 0.5 km steps between the Earth's surface and 40 km altitude. The aerosol extinction profiles were found to be reliable and robust during the time period covered by SAGE III/M3M with a bias lower than 20 % at 525 nm below 25 km altitude . Due to the orbit parameters and the solar occultation geometry, roughly 30 profiles were measured per day. The instrument covered all latitudes between 80° N and 80° S, with a gradual daily shift in latitude of the sunrise and sunset measurements, respectively. In this work, version 7.0 of the SAGE II data in binary format was used.

An optical particle counter (OPC) measures the number of aerosol particles with optical scattering signals above pre-set size thresholds. Such in situ observations have provided the ground truth for stratospheric aerosol satellite measurements (e.g., ) since the advent of the SAM-II and SAGE instruments (e.g., ). Although PSD parameters can be derived from satellites, comparisons to in situ OPC measurements are required to validate and refine aspects of the techniques (e.g., ). The University of Wyoming carried out balloon-borne in situ measurements using an OPC between 1990 and 2004 from Kiruna (68° N, 21° E), Sweden. Both the time frame and the location of the measurements open up the possibility of comparisons with the northern high-latitude sunset measurements of SAGE III/M3M.

The in situ measurements in Kiruna were carried out with two instruments: a size spectrometer and a total aerosol counter. The size spectrometer used was the Wyoming white-light optical particle counter (WOPC) . Due to Rayleigh scattering and phototube background, only particles with radii larger than 150 nm can be measured. For the measurement, individual pulses of forward-scattered white light from single aerosol particles are measured at an angle of 40° relative to the incident light of the incandescent lamp. The scattered light is focused onto a pair of photomultiplier tubes (PMTs), which convert the photons to a voltage. Each voltage signal is compared to 12 voltage discriminators, each discriminator summing all particles producing voltages above that discriminator level. The discriminator levels are set to represent specific sizes based on the instrument counter response function and calibration using laboratory aerosol . Thus each size bin measures all particles above a certain size threshold. To reduce noise, only coincident PMT pulses from two independent but symmetrical photon paths were counted towards the integral particle concentrations. Air was pulled through the inlet with a flow rate of 10 L min⁻¹. All Kiruna WOPC soundings analyzed here also included a separate measurement of the total aerosol number concentration of particles larger than 10 nm, often referred to as condensation nuclei (CN). This is achieved through the use of ethylene glycol vapor and a growth chamber, forcing aerosol particles larger than 10 nm to grow to optically detectable sizes via condensation of ethylene glycol. This effectively gives a measure of the total aerosol number density . By coupling the total aerosol number to the WOPC measurements, the measurements can be fitted with an assumed PSD to provide a means of comparison with satellite measurements that derive a smaller number of parameters. For this the WOPC data are fitted with both unimodal and bimodal lognormal PSDs . For this work the unimodal or monomodal data of UWv2.0 are used .

In Figs.  and  the temporal evolution of some of the derived stratospheric aerosol particle size distribution parameters is illustrated. Figure  depicts the data from the Northern Hemisphere, i.e., the sunset measurements, while Fig.  corresponds to the Southern Hemisphere and the sunrise measurements. The data shown are daily zonal mean profiles combined over the whole SAGE III/M3M mission time span from 27 February 2002 to 12 November 2005. The topmost plot shows the mean latitude of each day to which the profiles of the color plots correspond. The derived parameters shown in the color plots are, from top to bottom, the median radius, the absolute mode width ω, the effective radius, and the number density.

The oscillation of the vertical extent of the aerosol layer over time in the color plots is in large part explained by the latitudinal oscillation of the SAGE III/M3M measurements (see Fig.  or topmost plots of Figs.  and ). The red line in each color plot shows the tropopause height, which is provided in the SAGE III/M3M data. Some outliers have been excluded from the different parameters in the color plots due to their infrequency and the consequent widening of the displayed range of values, which would obscure the important signatures and structures within the plot. Values directly above the tropopause are often missing due to the filtering of data using the accuracy parameter described in Sect. .

Vertical dashed lines mark the times of volcanic eruptions that were strong enough to potentially have had an effect on the stratospheric aerosol layer. The first two were small (in terms of emitted SO₂) but significant eruptions by two tropical volcanoes in late 2002: namely, Ruang (2.3° N) on 25 September 2002 and Reventador (0.08° S) on 3 November 2002 . According to the Darwin Volcanic Ash Advisory Center, the plume of Ruang rose up to 17 km . The Reventador plume reached 17 km as well and emitted 0.06–0.08 Tg SO₂ . The third dashed vertical line marks the peak of the eruptive phase of the Manam volcano (4.08° S) that lasted from October 2004 until the end of January 2005. The peak occurred on 27–28 January 2005, with the volcanic plume reaching into the stratosphere, up to 21–24 km altitude and emitting approximately 0.14 Tg of SO₂ . It should be mentioned that prior to this, between 24 October and 27 November 2004, Manam also emitted around 0.15 Tg of SO₂ at altitudes between 6 and 18 km, making it unclear how much of it reached and impacted the stratosphere before the peak in January 2005 .

In the Northern Hemisphere (Fig. ), the strongest signatures in the SAGE III/M3M time series for each parameter are found in the months after the three tropical volcanic eruptions. Without necessarily claiming causality, the median radius, absolute mode width ω, and effective radius show decreased values starting 2 to 3 months after the volcanic eruptions in the lower Junge layer and lasting for over half a year, as indicated by the darker colors. In contrast, the number density, which is shown with a logarithmic color scale, shows a very strong increase in the same regions. In the context of the assumed monomodal lognormal shape of the PSD this means that its peak is shifted towards smaller radii and its width is reduced; i.e., smaller particles than before dominate the optical signal measured by the instrument. The effect of the monomodal assumption is discussed further in Sect. . The observed overall pattern in the derived parameters is strikingly similar to the results shown in , where eruptions of the Ambae, Ulawun, and La Soufrière volcanoes between 2018 and 2021 were shown to also have led to a decrease in median radius, absolute mode width, and effective radius and to a strong increase in number density. This was backed up by both data derived from satellite measurements and model simulations, which generally showed good agreement. Similar to Ruang, Reventador, and Manam, these volcanic eruptions all emitted less than 1 Tg of SO₂ into the stratosphere. Also, all of these eruptions were tropical. Both the emission strength and the latitudinal location of the volcano could be important causal factors for the observed reduction in average aerosol size . The low temperatures in the tropical upper troposphere/lower stratosphere (UTLS) region may be especially important due to their strong positive effect on nucleation of new particles from gaseous H₂SO₄ formed from the emitted SO₂. The results of model simulations indicate that lower temperatures result in a shift in the aerosol size distribution towards lower radii .

Another notable signature found after the volcanic eruptions, particularly in the Northern Hemisphere, is an increase in average aerosol size in a roughly 5 km thick layer above the discussed layer of reduced aerosol size. This increase is seen in the median radius and effective radius, but not necessarily in the absolute mode width ω; i.e., the derived monomodal lognormal PSD remains at a similar width but its peak shifts towards larger radii. This size increase is much more pronounced in 2005 after the Manam eruption.

In the derived parameters of the Southern Hemisphere, which are shown in Fig. , a size reduction after the Ruang and Reventador eruptions can be seen as well. Both the increase in number density and the reduction in median radius and effective radius are much weaker than in the Northern Hemisphere. However, the signals are visible earlier here, which makes sense, since the measurements in the Southern Hemisphere were taken at lower latitudes, i.e., closer to the tropical volcanoes, than in the Northern Hemisphere. Increased aerosol size, best visible in the median radius and effective radius, and decreased number densities are found throughout the year 2004. The strongest signals in the Southern Hemisphere are the increased number density and the reduced median radius in the lowermost 4 to 5 km of the Junge layer throughout the year 2005 and starting around the eruption of Manam. It therefore appears that more of Manam's emissions were transported to the Southern Hemisphere than Ruang's and Reventador's emissions before. Factors that may have contributed to this are the larger amount of SO₂ injected into the stratosphere by Manam, its location further south, and the timing of its main eruption at the start of the year rather than towards its end.

The SAGE III/M3M measurements did not cover the tropics and lower latitudes, as shown in Fig. . Therefore we have to look at the question of whether the observed aerosol size reduction signals can plausibly be traced back to the tropical volcanic eruptions of Ruang, Reventador, and Manam or if other causes have to be considered.

In terms of the chronological sequence it seems plausible that these signatures can indeed be of volcanic origin, since the strongest size reduction signals are found a few months after the eruption dates; i.e., there was enough time for the plumes to be transported from their possible tropical origin to the observed higher latitudes. However, there may also have been non-volcanic factors influencing the evolution of aerosol size and leading to an average size reduction in 2003 and 2005. An important factor may be the seasonally occurring new particle formation events in the polar winter stratosphere , also known as the springtime condensation nuclei (CN) layer that is linked to the onset of photolysis in polar spring .

Additional phenomena possibly also linked to the polar springtime CN layer are meteoric smoke particles (MSPs) that seasonally accumulate in the polar winter stratosphere . They originate from meteoroids continuously entering Earth's atmosphere and being vaporized in the upper mesosphere and lower thermosphere in a process called meteoric ablation. Through this, layers of neutral metal atoms and ions are formed. The vaporized material recondenses to form the meteoric smoke particles , which by coagulation likely grow up to sizes of 40 nm . As part of the general atmospheric circulation the MSPs in the mesosphere are transported to the winter hemisphere and then downwards in the polar vortex . Through the lowering of the H₂SO₄ saturation vapor pressure due to the combination with meteoric material, sulfuric acid is suggested to be able to condense above the usual confines of the Junge layer . This way, meteoric sulfate aerosols may be formed that coagulate and grow into measurable sizes during their continued descent in the stratosphere . The signal of low values in, e.g., the median radius and effective radius that seemingly descends from the upper end of the Junge layer in January 2003 (see Fig. ) could theoretically be related to this.

These seasonal nucleation–condensation events that include the MSPs accumulating in the polar winter stratosphere could be an important factor in the observed size evolution and reduction of average aerosol size in northern hemispheric winter. If that were the case, there would be an unknown error in the derived aerosol size due to the refractive indices used in the Mie calculations of the lookup table that are based on the assumption of pure sulfate aerosol. Unfortunately, little is known about the actual properties of meteoric smoke particles, mainly because of the difficulties of performing in situ measurements at such high altitudes .

However, the size reduction signal is visible well before spring, especially in 2003. Also, in particular after the Manam eruption, a size reduction signal is found in the Southern Hemisphere as well, although it is weaker, i.e., in the summer hemisphere where no seasonal nucleation event would be expected. Adding to this point, the Southern Hemisphere measurements were taken at relatively low latitudes. Additionally, no comparable size reduction is found in 2004, when notable volcanic eruptions strong enough to influence the stratosphere were absent, except for a very weak possible signal in the Southern Hemisphere around January and February 2004. These points argue for the signals being at least in part of volcanic origin.

As an additional piece of information, the measurements of the SAGE II instrument, the predecessor of SAGE III/M3M, can be used to shed some light on the temporal and spatial dispersion of the volcanic plumes. Its mission started in 1984 and ended in August 2005, therefore temporally covering most of the SAGE III/M3M mission. The reason for the additional information gained on the volcanic plume dispersion by using the SAGE II instrument's data is its different orbit. Because of it a wider latitude range is covered, between up to 80° N and 80° S, depending on season. The tropics are included as well, facilitating the tracking of the volcanic plumes, albeit with a relatively poor temporal resolution due to the sampling of solar occultation measurements.

Monthly zonal means of the aerosol extinction coefficient at 525 nm, as provided in the SAGE II v7.0 dataset, are shown in Fig.  and the top row of Fig. . Extinction coefficients with associated relative uncertainties of 30 % or higher were excluded. The extinction data were averaged onto a latitude grid with 5° bin size and plotted logarithmically. Even though monthly averages were calculated there are still data gaps, which is mostly due to the sparse sampling of satellite solar occultation measurements. Furthermore, the SAGE II sampling causes different latitude bins to represent different days within the considered month. Both figures show characteristic months around the volcanic eruptions in question, i.e., Ruang and Reventador for Fig.  and Manam for Fig. . In both cases, the leftmost plot shows the atmospheric conditions before the respective volcanic eruptions, while the other two show the emergence and dispersion of the aerosol extinction enhancement due to the volcanic plume.

Before the Ruang and Reventador eruptions, which happened on 25 September and 3 November, respectively, there are data gaps preventing a good look at the pre-volcanic conditions. The best overview is possible in September 2002, shown in the first panel of Fig. , where at least some northern high-latitude measurements are included. From this to the other panels for November 2002 and January 2003 a clear increase in the extinction coefficient can be seen in the Junge layer between the tropics, where the volcanoes are located, and the northern midlatitudes, where the SAGE III/M3M measurements happened. It can be concluded that the size reduction signal found in early 2003 in the Northern Hemisphere (see Fig. ) is very likely connected to the volcanic eruptions of Ruang and Reventador in late 2002.

For Manam the picture is not as clear. In the top row of plots in Fig. , a strong and clear increase in aerosol extinction coefficient is visible in the tropics after the Manam eruption in March 2005. However, whether the volcanic plume reached higher latitudes or not is not as easily discernible. For this reason an additional plot was added to Fig. . It shows the change in extinction coefficient at 525 nm between December 2004 and March 2005, i.e., the difference between the first and third plot of the top row, in percent. To distinguish it from the extinction coefficient data another color scheme is used. The data are presented logarithmically. The plot suggests at least some transport of the Manam volcanic plume from the tropics towards midlatitudes. Transport to the Southern Hemisphere shows up as a distinct enhancement of the typical Junge layer, with increases between roughly 10 % and 30 %. In the Northern Hemisphere, however, an increase of a similar magnitude can be seen not in a layered shape but throughout the lower stratosphere. Especially since this time frame covers mostly the northern hemispheric winter, a strong seasonal component may be at play here, with increasing aerosol extinction coefficients due to a lowering of the temperature. There is also a cluster of high aerosol extinction coefficient values (first row of plots) in the lowermost stratosphere between roughly 40 and 60 ° N that exists even in December 2004, before the main volcanic eruption on 27–28 January 2005. This cluster of values is, however, associated with high relative uncertainties and may therefore simply be a retrieval artifact rather than a real feature. It is therefore challenging to determine from the SAGE II data to what degree the aerosol size reduction in the Northern Hemisphere after the Manam eruption (see Fig. ) is attributable to the eruption, seasonal variation, or a combination of both.

Taking everything discussed above into account it is likely that the average aerosol size reduction signals that were observed after the Ruang and Reventador eruptions in both hemispheres and after the Manam eruption in at least the Southern Hemisphere are indeed volcanic events; they may be comparable to the ones discussed in . In the case that the signals are indeed of volcanic origin, the fact that the size reduction signals are stronger in the Northern Hemisphere (Fig. ) than in the Southern Hemisphere (Fig. ) can likely be explained by the tropical location of the volcanoes, the timing of the eruptions in northern winter, and the tendency of the Brewer–Dobson circulation to enhance transport from the tropics towards the winter hemisphere .

Polar stratospheric clouds (PSCs) can form in the lower stratosphere and play an important role in both the activation of halogen species, resulting in ozone destruction and the ozone hole, and the denitrification and dehydration of certain regions of the polar stratosphere . Since the stratosphere is typically very dry, the very low temperatures in the stratospheric polar winter are a precondition for them to form . There are multiple known types of PSC, whose particles have different compositions and layers and which can be liquid or frozen, based on the conditions in which they form .

Solid PSC particles, such as nitric acid trihydrate (NAT) and ice particles, as well as liquid supercooled ternary solution (STS) droplets, exist , the latter of which are the most commonly occurring type of particle in a PSC. Also, liquid and solid particle types can often coexist . The formation of these different kinds of PSC particles depends foremost on temperature, with NAT particles having the highest equilibrium temperature. At temperatures roughly 3 K above the ice frost point (and ∼ 3 K below the NAT frost point), STS particles form. Ice particles then nucleate 3 to 4 K below the frost point . Other factors, like the preexistence of other particles, may also play a role. For example, ice particles can provide surfaces for the heterogeneous nucleation of NAT particles , and meteoric particles may play a role in the heterogeneous nucleation of both NAT and ice particles .

Since the SAGE III/M3M measurements in part covered high latitudes, mostly in the Northern Hemisphere, there was a potential to encounter polar stratospheric clouds (PSCs) in the dataset. In Fig. , the top plot again shows the latitudinal coverage of the measurements in the Northern Hemisphere. Below this, time series of the aerosol extinction coefficient at 449 nm and the temperature are shown, both parameters being provided in the SAGE III/M3M dataset. For visibility of the relevant signals, the extinction coefficient is plotted with a reverse color scale; i.e., darker colors in this case mean higher values. To facilitate referencing of the same areas in the different plots, extinction coefficient values are only shown where aerosol size parameters could be derived and temperature values are only plotted where extinction coefficients are shown as well.

The two main areas of interest are immediately obvious when looking at the extinction coefficient in Fig. : in the winters of 2002/2003 and 2004/2005 there are multiple clusters of only a few days of strongly elevated aerosol extinction at the uppermost altitudes of the Junge layer (and in part above it) including some of the highest values of the whole time series. The short duration of the signals and the position at the uppermost end of the aerosol layer basically rule out a volcanic origin. Looking at the lowermost plot, both winters included the lowest temperatures over the longest times in the SAGE III/M3M time series. More specifically, the clusters of high extinction coefficient in question also exhibit very low temperatures, well below 200 K, with the lowest temperature in these daily average profiles being 186 K. It therefore is very likely that the observed sporadic enhancements in the extinction coefficients come from PSCs that had been forming due to the cold winters in 2002/2003 and 2004/2005.

In the cases in which the aerosol size derivation method described in this work does obtain results for the measurements containing a PSC signal, median radii of up to roughly 350 nm are found. However, these results are in no way reliable. Firstly, the actual refractive indices of the PSC particles may differ strongly from the values assumed for pure sulfate aerosol. Secondly, the particles may not even be liquid and therefore not spherical, in which case Mie theory would be inaccurate. For probably the same reasons there are many measurements where the size derivation method does not work and produces no results at all, since the extinction ratios formed from the measurements fall outside of the lookup table value range.