The measurements took place at the Chacaltaya (CHC) Global Atmosphere Watch (GAW) station which is located east of the Altiplano close to the main ridge of the Bolivian Andes (16∘21′ S, 68∘07′ W, 5240 m a.s.l.; see Fig.  in Appendix  for the elevation data) and 330 km away from the Pacific coast. The area is semi-arid and partly covered by snow, especially after precipitation events. The vegetation on the Altiplano is dominated by tufts of hard grass on which local farmers let their llamas graze. The metropolis La Paz and El Alto, located 17 km south of Chacaltaya GAW station with an elevation ranging from 3300–4100 m a.s.l., is an important source of anthropogenic emissions impacting CHC.

The southern part of Lake Titicaca (Lago Menor) lies about 75 km away to the west and northwest of the station, as can be seen in Fig. . Lake Titicaca, which is the largest freshwater lake in South America with a surface area of 8372 km2, experienced its first algal bloom in 2015. The algal bloom occurred due to anthropogenically increased nutrient levels after climatologically extreme rain events . These rain events and subsequent algal blooms will likely become more frequent in the future and impact especially the shallow and nitrogen-rich Lago Menor at the south end of Lake Titicaca. The algal bloom in the year 2015 involved green algae of the Carteria species , known to produce DMSP and DMS . Other lagoons on the Altiplano that might host algae and thus emit DMS lie a few hundred kilometers to the south. To evaluate the possible impact of these potential local sources on signals of marine trace gases, we used backward Lagrangian dispersion calculations . Figure  shows the horizontal radial extent (1600 km) of the model domain; a detailed description follows in Sect. . A description of the continuous long-term measurements from the GAW station that we used in this study can be found in Sect. .

Advanced field measurements with an extensive suite of novel gas and particle composition instruments at CHC were conducted between December 2017 and the beginning of June 2018 (SALTENA, ). The aim of the campaign was to determine the chemical species of anthropogenic and biogenic origin impacting the station and contributing to new particle formation.

Atmospheric ions were detected by the Atmospheric Pressure interface-Time of Flight mass spectrometer (Positive-APi-TOF, Sect. ) and the gas-phase composition by a nitrate chemical ionization atmospheric-pressure-interface time-of-flight mass spectrometer (nitrate-CIMS, Sect. ), a filter inlet for gases and aerosols (FIGAERO) coupled to a high-resolution time-of-flight chemical ionization mass spectrometer using iodide chemical ionization (I--FIGAERO-CIMS, Sect. ), as well as a proton transfer reaction time-of-flight mass spectrometer (PTR3, Sect. ). The Quadrupole Aerosol Chemical Speciation Monitor (Q-ACSM, Sect. ) and the I--FIGAERO-CIMS further analyzed the aerosol particle composition. Table  summarizes which of the analyzed species are detected by which instrument. An overview of the operation times of the instruments and the herein presented periods is given in Fig. .

The data from the SALTENA campaign were analyzed in the context of the data from the continuous measurements taking place at the GAW station, which is equipped with standard instruments to measure trace gases like CO and CO2, for example. An automatic weather station (AWS) deployed at the station records air temperature, relative humidity, radiation, wind direction, and wind speed at a 1 min resolution. Aerosol physical, chemical, and optical properties have been measured nearly continuously in the past 10 years. The aerosol particle size distribution is monitored with a mobility particle sizer spectrometer (MPSS), assembled with the TSI Inc. 3772 Condensation Particle Counter and a custom-made differential mobility analyzer. The MPSS measures the particle size distribution in the range of 10 to 650 nm with a 5 min time resolution. A multi angle absorption photometer (MAAP) monitors the equivalent black carbon (eBC) mass with a time resolution of 1 min . The data are reported at standard temperature and pressure conditions, following the GAW network recommendations , and the data are corrected considering the effective wavelength of the instrument in accordance with .

Two interchangeable digital impactors are used to collect the aerosol particulate matter with aerodynamic diameters lower than 10 and 2.5 µm (i.e., PM10 and PM2.5, respectively). The collectors are equipped with pre-baked quartz fiber filters (Ø 150 mm) collecting samples at a flow rate of 30 m3 h-1. The filters are exchanged once a week and are analyzed by a Dionex Ion Chromatograph, Sunset Laboratory total organic carbon analyzer (applying the EUSAAR_2 protocol; ) and by high-performance liquid chromatography coupled to a pulse amperometric detector to determine the ions, organic and elemental carbon, and anhydrous monosaccharides. We use data of methanesulfonate from December 2011 to August 2017. A more detailed description of the site and further long-term measurements can be found in previous studies .

used a high-resolution meteorological model coupled with Lagrangian dispersion calculations to identify where air masses sampled at CHC originate from and to quantify the relative influence of the surface and the free troposphere. Here we utilize, and further analyze, the output from the same simulations .

performed a 6-month-long simulation with the Weather Research and Forecast (WRF) model. The simulation had four nested domains. The outermost and largest domain had a grid spacing of 38 km, whereas the innermost domain, covering the area closest to CHC, had a grid spacing of 1 km. The meteorological output of the WRF simulation was saved every 15 min and used as input to the Lagrangian dispersion model FLEXible PARTicle dispersion model (FLEXPART). The high temporal and spatial resolution of the driving meteorological data is a key advantage over dispersion or trajectory simulations driven by reanalysis data – typically only available every 1 to 6 h. Backward dispersion simulations were performed with FLEXPART. For every hour, 20 000 particles were randomly distributed and then released from a 10 m deep (a.g.l) layer covering a 2×2 km square centered on CHC. The trajectories of all particles were tracked back in time for 4 d. It should be noted that the particles were treated as passive tracers that do not undergo any chemical or microphysical processing.

The output of FLEXPART is the source–receptor relationship (SRR), which is related to the particles' residence time in the 3-dimensional grid cells of the FLEXPART output grid. transformed the SRR output from the Cartesian grid to a log-polar grid so that the size of the grid cells gradually increases with the distance from CHC (the receptor); additional details of this transformation, and its motivation, are provided in . The domain of the log-polar grid is a cylinder covering an area with a radius of 1600 km centered on Chacaltaya (Fig.  shows its horizontal extent) and reaching from the surface to 15 km a.g.l.

As the FLEXPART domain is limited in size, it is possible that during the 4 d of backwards travel particles can leave the domain. Therefore, unlike in , we also estimated the amount of air masses arriving at CHC originating from outside the FLEXPART domain (herein referred to as “out of domain”). Similar to , we identified air masses arriving from different pseudo layers within the FLEXPART domain (referred to as “within domain”): surface layer (air masses from the lowest FLEXPART output layer 0–500 m a.g.l.), boundary layer (air masses from below 1500 m a.g.l.), and free troposphere (air masses from above 1500 m a.g.l.).

The out-of-domain air mass contribution is computed as follows: theoretically, if all particles remained in the domain, the domain-integrated SRR would be 4 d. However, in the case when particles leave the domain, the domain-integrated SRR is less than 4 d. The out-of-domain SRR is therefore computed as the theoretical maximum SRR minus the actual SRR. No information is available about the trajectories of the particles once they have left the domain.

To identify periods dominated by free tropospheric (FT) air masses versus periods dominated by boundary layer (BL) emissions, we present a compilation of important measurements (all reported for standard conditions) in Fig. . The time series of gas-phase MSA is shown in Fig. a), as well as the sum of C8H10Ox compounds, which are likely oxidation products of aromatic xylenes or ethylbenzene, and we can thus use them as indicators for anthropogenic emissions (e.g., ). It is noticeable that gas-phase MSA is high whenever C8H10Ox is low, as shown more clearly in Fig.  in Appendix .

Figure b shows the amount of air coming from the free troposphere (>1500 m a.g.l.) and influenced by surface emissions (0–500 m a.g.l.) according to the results of the FLEXPART Lagrangian dispersion analysis. Furthermore, we show the fractions of free tropospheric air from out of the FLEXPART domain and within the domain (defined in Sect. ).

We also present the water vapor mixing ratio (WVMR) and equivalent black carbon mass (eBC) (Fig. c), shortwave radiation (SWR) and temperature (Fig. d), the Q-ACSM total particle mass and its composition (Fig. e), followed by the particle number size distribution (Fig. f).

Figure b reveals that free tropospheric air masses reach the CHC GAW station all the time during the period shown. However, neglecting the varying boundary layer height in these traces (see , Sect. 4 for a justification of the differentiation using constant pseudo layer heights instead of a time-resolved boundary layer height) leads to a tendency to underestimate the surface influence during daytime and the free tropospheric air during nighttime (when the inversion is below 1500 m a.g.l.). Especially during the daytime, free tropospheric air mixes with local air masses from the surface layer (spikes in the surface layer influence in Fig. b). This leads to enhanced concentrations of the anthropogenically emitted C8 compounds (Fig. a), as well as a larger nitrate fraction in the particle mass (Fig. e). We observe particle formation events nearly every day as previously reported by , typically followed by particle growth that regularly extends into the night and sometimes without interruption (e.g., 15–16, 16–17 May 2018) until the next particle formation event the next day (Fig. f). However, during certain periods (marked by blue shadings in Fig. ), at late night or early morning, when the temperature reaches its daily minimum, eBC (Fig. c) and the particle mass determined by the Q-ACSM (Fig. e) show a sudden drop, coinciding with overall lower particle numbers measured by the MPSS (Fig. f; see also Fig.  in Appendix  for the number size distribution) and particularly low water vapor mixing ratios (Fig. c). These are strong indications that the station was above the shallow boundary layer of the Altiplano and that free tropospheric air masses had subsided from higher altitudes.

The FLEXPART dispersion simulation can resolve this, as shown in Fig.  for 11 May; as an example: during the whole day, the main wind direction was west-southwest (Fig. a and b for the 08:00 case). Figure c and d clearly shows that the air reaching CHC descended from higher altitudes and had not been in contact with the surface before detection (<5 % of the air had been in contact with the surface according to FLEXPART). In contrast, during the afternoon of the same day, shown in Fig. e and f, a still small but larger fraction of the air traveled uphill close to the surface (ca. 10 % of the air had been in contact with the surface according to FLEXPART).

To summarize, a shallow and strong inversion close to the ground likely did not allow the mixing of the different air masses during the marked times. This gives us the unique opportunity to characterize the free tropospheric air masses reaching the station from the westerly sector (as typical for the Altiplano dry season) with our array of state-of-the-art mass spectrometers. To avoid uncertainties caused by the FLEXPART time resolution and the potential daytime-dependent over- and underprediction of the boundary layer influence, we also determine when the station is in the free troposphere above the shallow boundary layer by applying the conditions from Table .

These FT identifiers are chosen after careful data analysis to be very strict so that we certainly exclude all periods affected by the boundary layer. We assume that all periods during which the conditions are not fulfilled are likely influenced by the local boundary layer to at least some extent, and we refer to them as BLday between 07:00 and 19:00 local time and BLnight otherwise. The FLEXPART analysis provides us with additional information on the horizontal and height-resolved transport of the air masses. Furthermore, the FLEXPART analysis can be considered a priori to data analysis, while the FT identifiers are set a posteriori and might vary with season (such as the water vapor mixing ratio).

Due to the limits of the FLEXPART domain, we cannot know exactly where the free tropospheric air masses had the last contact with the boundary layer, but chances are high that the last surface interaction occurred in a marine, remote setting over the south to the southwestern Pacific Ocean or in the intertropical convergence zone (ITCZ), as convection typically occurs in these areas, while it is typically unlikely in the southeastern Pacific due to the cold Humboldt upwelling. The enhanced gas-phase MSA concentrations during the marked undisturbed free tropospheric events, detected by the nitrate-CIMS and shown in Fig. a), are the first indication of this proposition. We also observed higher-than-median concentrations of MSA outside the marked periods, indicating that free tropospheric air masses mix in regularly. With our FT identification rule, we made sure to select events that are least impacted by surface emissions. That does not exclude the influence of FT air masses during other periods.

The Q-ACSM identifies the aerosols in the free troposphere as composed of sulfate and organics with only little ammonium and no nitrates (see also Fig. f in Appendix ). The relative absence of nitrate observed in the particles by the Q-ACSM is reasonable, as it likely links to ammonium nitrate that forms from traffic emissions in the La Paz and El Alto metropolitan area and is thus solely emitted into the local boundary layer. The large sulfate fraction we observed probably has a regional origin and is not solely attributed to marine DMS oxidation, because CHC is influenced by volcanic emissions in the area , especially under westerly wind direction.

The Positive-APi-TOF can give us information on the positive ion composition of the free tropospheric air masses. Using our previously gained understanding of the fingerprint of the FT events, we found an FT period under the westerly wind influence in February, when the instrument was operating.

In Fig.  we show a free tropospheric time period (19 February, 03:00–10:00) highlighted in green. A detailed air mass origin description for the FT event on 19 February is shown in Fig.  in Appendix , which is close to the conditions of the event on 11 May.

As described above, the low water vapor mixing ratio (dropping from previously 4 to 0.8 g kg-1), low equivalent black carbon concentration (below 0.1 µg m-3), and low total number concentration of particles in the diameter range 10–650 nm (Fig. c) are indications of such FT events. The wind direction measured at the peak of the Chacaltaya mountain is shown as barbs in Fig. a, together with the time series of different DMSO clusters. DMSO has a proton affinity of 884.4 kJ mol-1. The signals of the typically dominating amines (proton affinities above 900 kJ mol-1) in the Positive-APi-TOF are significantly smaller during FT than during BL times (see Fig. b), while other compounds (unidentified, likely organics) take most of the charge (see Fig.  in Appendix ).

The signals of DMSOH+ and (H2O)DMSOH+ show an increase during the free tropospheric air event. This does not necessarily imply a higher DMSO concentration in the free troposphere; after all, (NH3)DMSOH+ was regularly present throughout the whole period (17–21 February) with maxima during nighttime and minima during daytime. This does indicate, however, the general presence of DMSO and additionally the lower abundance of species with higher proton affinities in FT air, which do typically affect the ionization likelihood of species with lower proton affinities . Therefore, the strong signals of DMSOH+ and (H2O)DMSOH+ and the decreased signal of (NH3)DMSOH+ show that ammonia, amines, and pyridine are significantly less available during FT times. Even though the water vapor mixing ratio decreased during the FT event, the water concentration in the atmosphere always remains high in contrast to other trace gases, which explains the slight increase of water clusters.

Comparing Fig. a and b shows that the signals of DMSO clusters are small compared to the total ions. The total ion counts remain high during the FT event, because the total ion concentrations in the atmosphere are determined by solar radiation, altitude, and other parameters. In our case, the ions switch from the species with the highest proton affinity to ones with lower proton affinity during air mass transition, which can explain the increase of the signal of unidentified organics, as shown in Fig. , and proves a significantly lower availability of high proton affinity compounds such as amines.

The molecular formulae of the organosulfur compounds detected in the gas phase by the PTR3 and nitrate-CIMS indicate that they are, in fact, oxidation products of DMS. In particular, we detected all presently known DMS oxidation products, except the recently observed HPMTF , which might be below the detection limit.

We show their time series in Fig.  in Appendix  and present boxplots of the detected concentrations in the different air masses in Fig. . DMS (Fig. a) and gas-phase MSA (Fig. g) show significantly enhanced concentrations in the free troposphere. MSA is in fact one of the few condensible vapors which are higher in FT air than during BLday and dominates the nitrate-CIMS' FT – BLday spectrum shown in Fig.  (Appendix ). When comparing with Fig. , the sporadic high concentrations of DMS occur during the first analyzed FT air period on 11 May. One rare subtropical storm, which was observed from 4 to 9 May only a few hundred kilometers off the Chilean coast , might have impacted our data and could explain the surprisingly high DMS concentration in such a large distance from the typically convective regions over the southern Pacific substantially further west.

Gas-phase H2SO4 (Fig. h) is strongly enhanced in the FT compared to the BLnight but is maximal in BLday. Since H2SO4 formation is driven by photochemistry, a diurnal maximum in the H2SO4 concentration can be explained, if local production, e.g., from SO2 oxidation occurs, although enhanced losses onto particles during BLday (Fig. d in Appendix ) have a decreasing effect on the H2SO4 gas-phase concentration. Due to the influence of the metropolitan area of La Paz and El Alto and volcanic emissions, we expect H2SO4 to have a large OH-initiated production rate by oxidation of anthropogenic and volcanic SO2 during daytime. The high concentration of H2SO4 in the FT is partly explained by an increase in sulfuric acid directly after sunrise as some of the FT periods extended into the morning hours. Additionally, the lower condensation sink during the FT periods enhances the sulfuric acid lifetime during these events. This is especially evident in the night of 20 May: although in the middle of the night, where we expect the production term to be small, a substantial amount of sulfuric acid is observed that behaves anticorrelated to the low condensation sink during the event, as shown in Fig.  in Appendix .

The other oxidation products show a significantly weaker trend – but none shows features of a typical boundary layer product (compare to Fig. ). However, as seen in Fig. , there is always an influence of free tropospheric air. A compound with little to no trend between those conditions could thus still have its origin in the free troposphere, but in such cases, additional local sources cannot be excluded and should be checked. We do this for MSIA (Fig. d) in Sect. . Afterwards, we relate their detected concentrations in Sect. .

The gas-phase MSA concentration in the FT is an order of magnitude higher than in the BL air (Fig. g), but, because of the low vapor pressure of MSA, we expect it to reach CHC mainly in the particle phase. We show the signal of particulate methanesulfonate (MSA), measured by the I--FIGAERO-CIMS in Fig. a above the time series of the influence of the different air masses according to FLEXPART in Fig. b.

First of all, the FLEXPART analysis shows that the station is under the influence of free tropospheric air throughout the described measurement period. The total (mainly organic) signal of the particle phase detected by the I--FIGAERO-CIMS (gray in Fig. ) shows strong spikes corresponding to the daily influence of the boundary layer at the station. In contrast to the total signal, the methanesulfonate (light blue) stays relatively constant throughout the week, and its fraction in the particles (dark blue, circles) is anticorrelated to the boundary layer influence (25–28 May). From 29 May onwards, we observed a higher fraction of the methanesulfonate signal, likely due to the increased fraction of air masses from outside the domain. Wind velocities are significantly higher in higher altitudes in the free troposphere, so air masses get transported over large distances. On the night of 30 May, the fraction of air parcels from the free troposphere reaches nearly 90 %. This includes the fraction of air parcels from outside the domain that reached up to even 40 % during this period. At this time, the methanesulfonate fraction in the particles also reached its maximum. This strongly suggests that the methanesulfonate reaches CHC via transport in the free troposphere and that there is no significant local source.

As the absolute methanesulfonate signal appears independent of the surface influence, we can also use sampling filter measurements with a very coarse time resolution. While the filter measurements do not allow a differentiation between the free troposphere and boundary layer air masses, they allow for investigating the methanesulfonate more quantitatively.

The long-term particle composition measurements using sampling filters cover the years 2012–2017 (Fig. a) and show that approximately 4±1 ng m-3 methanesulfonate is found in the particles from April to November, which corresponds to the time when westerly wind conditions occur regularly (Fig. b). The particulate mass concentration of methanesulfonate corresponds to an average number of roughly 2.5×107 molecules cm-3, which is 1–2 orders of magnitude higher than the detected gas-phase methanesulfonic acid concentrations, in line with its low vapor pressure.

The strongly FT-dependent gas-phase MSA signal observed from the nitrate-CIMS (compared to the nearly constant methanesulfonate signal from the I--FIGAERO-CIMS) could be explained in two ways: firstly, MSA is known to participate in new particle formation and growth to larger aerosol particles , and the condensation sink in surface-affected air masses is much higher compared to within free tropospheric air masses. In FT air, the mild condensation sink leads to a longer lifetime for gas-phase MSA, which can explain the higher concentration of gas-phase MSA in the periods of undisturbed free tropospheric air reaching CHC. Upon mixing of the free tropospheric air masses with the boundary layer, the stronger condensational loss of MSA to the aerosols fully scavenges the gas-phase MSA. Secondly, it has been suggested in earlier publications that MSA might degas from methanesulfonate aerosol in air masses with low relative humidity. The herein analyzed FT events always involved very low relative humidities, so this effect might occur as well. stated that parallel measurements of both MSA and DMSO are necessary to prove or disprove their MSA evaporation hypothesis. We relate MSA to the other oxidation products in Sect.  after considering the impact of local sources.

Local sources certainly impact H2SO4, as already discussed in Sect. . This is confirmed by comparing methanesulfonate and sulfate concentrations: sulfate fractions of 50 % were detected by the Q-ACSM (Fig.  in Appendix ) even during the marked free troposphere periods, which means that approximately 100 ng m-3 of non-refractory PM1 would be made up of sulfate (see Fig. e and f).

In comparison, methanesulfonate from the filter measurements is only about 4 ng m-3 in May in PM2.5 and PM10 (Fig. a). The ratio between methanesulfonate and sulfate is thus below 0.04, which is significantly lower than in remote marine areas , laboratory experiments , or at similarly cold temperatures in free tropospheric measurements of marine air . This suggests that the remote marine particle composition is altered due to the volcanic SO2 emissions in higher altitudes prior to reaching CHC as mentioned in . While we might be able to limit the impact of the anthropogenic emissions on our dataset by using the FT indicators discussed in Sect. , we are hesitant to exclude the volcanic impact on sulfuric acid and sulfate data even in the undisturbed FT periods (defined by the specifications in Table ) due to the higher emission altitude of volcanoes. Therefore, our dataset does not allow any conclusions for the production of H2SO4 by DMS oxidation.

To understand which sources impact the DMS oxidation products' time series measured at CHC, we show the source regions (“footprints”) of MSIA on a map together with the height-resolved domain lateral boundaries outside the white ring (see Sect.  for an explanation), determined from FLEXPART Lagrangian dispersion calculations, as an example.

It is important to note that this analysis is based on the full MSIA time series – not distinguishing between the free troposphere and the boundary layer. It shows source regions on the surface (inside the white ring) from which air masses likely traveled close to the surface and height-resolved source regions at the lateral boundaries (outside the white ring). We chose to show the footprint analysis of MSIA for multiple reasons: first, we wanted to show a compound with a short atmospheric lifetime on the order of a few hours, which gives more weight to local sources compared to distant ones, thus giving an upper limit for the influence of local sources. Second, the time series had to be above the limit of detection for most of the time to have as many valid data points as possible, which unfortunately excluded DMS. Finally, the PTR3's sensitivity to MSIA is at the kinetic limit, thus its time series is independent of humidity changes.

One of the resolved local source regions affecting the MSIA time series is the southern part of Lake Titicaca, which also shows enhanced chlorophyll concentrations in May 2018, as shown in Fig. . More details on Lake Titicaca are given in Sect. . The DMS flux from Lake Titicaca has not been determined so far, but DMS fluxes from other lakes are roughly 10 times lower than the average marine flux of DMS .

Compared to Lake Titicaca, the coast of Peru is a more important MSIA source according to the FLEXPART analysis – also showing enhanced chlorophyll concentrations in satellite images (see Fig. ). Convection is very unlikely to occur at this cold and typically overcast coast of Peru in May, but the air masses might still move upslope thermally driven, thereby staying close to the mountain slope surface.

However, the FLEXPART backward dispersion analysis locates the major fraction of the MSIA source regions in the domain at the lateral boundaries in heights between 6000–8000 m a.s.l., suggesting that the air masses that carry DMS and DMS oxidation products are advected into the domain from the west (southwest to the northwest) and thus originate from the (remote) Pacific. For the other DMS oxidation products, the signal fraction explained by long-range advection is larger than for MSIA, as summarized in Table .

The differences in the allocation of source regions by FLEXPART for the various oxidation products likely originate from the different chemical transformation processes, which the Lagrangian dispersion model neglects. Despite the associated uncertainty, however, one result of the FLEXPART analysis is clear: local sources contribute around 20 % to the full time series of DMS oxidation products, while most of the signal is attributed to long-range transport with the largest fraction from outside of the domain. We can imply for our case – having even only a small impact of Lake Titicaca on the total time series – that the concentrations of DMS oxidation products measured in the FT air masses are a result of marine DMS emissions, long-range transport, and oxidation along the trajectory.

However, as described by , the algal blooms in Lake Titicaca are likely to become more frequent in the future. Increasing DMS fluxes from Lake Titicaca would probably enhance especially the concentrations of DMS and its intermediates at CHC when the station is within the boundary layer.

The observed DMS concentrations in comparison with the concentrations of its reaction products relate to a long reaction time during the transport from the source regions to Chacaltaya: the DMS signal is very variable, ranging from below detection limit (3×105 cm-3) to maximally 1×108 cm-3. We observed its highest concentration during the FT event on the morning of 11 May 2018. During this time, the air mass was coming from the west with high wind speeds. In fact, the air masses only require about 30 h to travel from the 1600 km distant lateral boundary of the simulated domain to CHC. This period is similar to the DMS lifetime of ca. 1.5 d in the atmosphere with respect to its main oxidant OH under halogen- and NOx-free conditions . Typical DMS concentrations in 6000–8000 m a.s.l. over the ocean are on the order of 5×107-1×108 cm-3 but can reach up to 4×108 cm-3 close to convectional updrafts . Consequently, the maximum value measured here is plausible if transport from a convective cell to CHC was on the order of the DMS lifetime. Interestingly, an extremely rare subtropical cyclone has been observed in the southeastern Pacific Ocean, only hundreds of kilometers away from the Chilean coast shortly before (9 May) , which most likely impacted the DMS concentrations at CHC. During all other FT events, the DMS concentration was ≤2×107 cm-3 so that the major part of the DMS was already oxidized due to long transport times compared to the DMS lifetime.

All observed oxidation products except DMSO2 show gas-phase concentrations typically between 105-5×106 molecules cm-3. DMSO2 is the only compound with gas-phase concentrations regularly in the 107-108 molecules cm-3 range. At air temperatures of approximately -10 ∘C at 6000 m a.s.l. (assuming a dry adiabatic temperature decrease with altitude above the Altiplano), DMSO2 is volatile to semivolatile (see Appendix ). In the free tropospheric air masses, DMSO2 will be far more abundant in the gas phase than in the particle phase. The measured mean gas-phase DMSO2 concentration of approximately 2×107 cm-3 in the free troposphere equals a mass concentration of 3.1 ng m-3. This is comparable to the total MSA mass concentration when taking the particle-phase data from the filter measurements (see Fig. a) into account (Sect. ). Due to the close molar masses of DMSO2 and MSA, this implies similar amounts of sulfur in these two chemical states.

Such high DMSO2 yields are in contrast to chemical models, where DMSO2 is simulated to be only a minor oxidation product of the DMS addition pathway , due to previously reported low concentrations of DMSO2 in the marine boundary layer . However, the colder temperatures and the long reaction time with negligible losses likely favor its accumulation in the free troposphere. High gas-phase DMSO2 concentrations were observed in lab experiments and were attributed to the reaction of DMSO with O3 at interfaces and in the aqueous phase . The persistently high DMSO2 concentrations compared to other DMS oxidation products observed at Chacaltaya might therefore also indicate oxidation at air–liquid interfaces of aerosols or cloud droplets.

All other oxidation products were significantly less abundant, which likely has to do with their shorter lifetimes: the DMSO-to-DMSO2 ratio is relatively low, approximately 0.1. This is because the DMSO lifetime is short, typically hours, although DMSO is the precursor of DMSO2 under low-NOx conditions. The ratio between MSIA and DMSO2 is 0.03. Both compounds are products of DMSO oxidation, but MSIA has a similarly short lifetime like DMSO, while DMSO2 is photochemically inert with a lifetime of weeks and therefore builds up with reaction time along the trajectory. MTF (might be underestimated to some extent due to fragmentation), formed via H abstraction and reaction with HO2 and OH, thus requires two-step oxidation and is an intermediate product. It has an at least 5 times higher concentration than MSIA. This is reasonable because MTF also has a longer lifetime than MSIA: its lifetime is on the order of 6–12 h due to its lower photolysis and reaction rates (by OH radicals), while it is around 3 h for MSIA. MSA is detected at the kinetic limit in the nitrate-CIMS. DMSO, DMSO2, and MSIA are detected at the kinetic limit in the PTR3 (see the collision-induced dissociation ramps of DMSO, DMSO2, and MSIA in the H3O+-CIMS in ) so that this discussion is not impacted by uncertain calibration factors.

Furthermore, we measured signals close to the detection limits of our instruments at the exact masses of CH3S(O)2OOH (in nitrate-CIMS) and CH3SOH (in the PTR3) that both have unknown reaction rate constants and unfortunately also uncertain sensitivities. According to , the latter reacts quickly with ozone.