Evaluation of the Sources, Precursors, and Processing of Aerosols at a High-Altitude Tropical Site

This work presents the results from a set of aerosol- and gas-phase measurements collected during the BIO-MAÏDO field campaign in Réunion between March 8 and April 5, 2019. Several offline and online sampling devices were installed at the Maïdo Observatory (MO), a remote high-altitude site in the Southern Hemisphere, allowing the physical and chemical characterization of atmospheric aerosols and gases. The evaluation of short-lived gas-phase measurements allows us to conclude that air masses sampled during this period contained little or no anthropogenic influence. The dominance of sulfate and organic species in the submicron fraction of the aerosol is similar to that measured at other coastal sites. Carboxylic acids on PM10 showed a significant contribution of oxalic acid, a typical tracer of aqueous processed air masses, increasing at the end of the campaign. This result agrees with the positive matrix factorization analysis of the submicron organic aerosol, where more oxidized organic aerosols (MOOAs) dominated the organic aerosol with an increasing contribution toward the end of the campaign. Using a combination of air mass trajectories (model predictions), it was possible to assess the impact of aqueous phase processing on the formation of secondary organic aerosols (SOAs). Our results show how specific chemical signatures and physical properties of air masses, possibly affected by cloud processing, can be identified at Réunion. These changes in properties are represented by a shift in aerosol size distribution to large diameters and an increased contribution of secondary sulfate and organic aerosols after cloud processing.

Figure S2 shows that most inorganic species were neutralized as ammonium sulfate , ammonium ( 4 ) 2  4 nitrate Or ammonium chloride .However, the 0.57 slope shows that at higher ( 4 ) 3  ( 4 ) concentrations, there is a deviation away from the slope of 1, which means that there is more SO 4 or NO 3 than needed to neutralize NH 4 , from which we can suppose that SO 4 must be present in other forms than ammonium sulfate, such as H 2 SO 4 .

Section S2. Comparison between DMPS and ToF-ACSM
DMPS aerosol size distribution measurements are converted to mass concentrations and compared with ToF-ACSM mass concentrations as follows; . Where i is a given measurement, D p its diameter, N i is the number of the particles (cm -3 ), and ρ the particle's density (1.5 g.cm -3 ). A)

Aerosols' size distribution
Daily particle number concentration varied from 2000 cm -3 to 6000 cm -3 , whereas nighttime concentrations were always less than 400 cm -3 .Concentrations increased during morning early hours (R = -0.45,p = <0.001),which illustrates the influence of atmospheric dynamics and the rising of the boundary layer, bringing other atmospheric layers.Aitken mode aerosols (25 nm to 90 nm) contributed on average 57 14 % to the total ± aerosol size distribution, accumulation mode aerosols (90 -600 nm) contributed on average up to 21 ± 9%, while the nucleation mode (13.7-25 nm) contributed to 21 16% (Figure S4).Maximum nucleation ± contributions were mostly prevalent during the early morning hours (Figure S4b), resulting from the daily new particle formation events triggered by convective uplifting of boundary layer air masses to the MO site 38,82.
From 28th March onwards, we observe an increase (up to 25% ± 9.7%) in the contribution of accumulation mode aerosols.This corresponded to periods when air masses traveled from lower altitudes, passing over the forested areas and being more influenced by cloud events than during the first part of the field campaign.More discussion on the aerosol size distributions can be found in section 3.5.

Section S4. Meso-NH model and backward trajectories
Meso-NH simulates small scale (Large-Eddy Simulations (LES) type, horizontal resolution from a few meters) up to synoptic-scale atmospheric parameters (horizontal resolution of several tens of kilometers) and can be run in a two-way nested mode involving several nesting stages.Surface fluxes and cloud microphysics are computed using the surface model SURFEX 13 combined with the ICE-3 module 14 .
The Meso-NH model has already been used to characterize air masses above Reunion Island at 500 m resolution during the FARCE campaign in 2015 and in the present work, the Meso-NH configuration used in the present work includes three nested domains at the horizontal resolutions 100, 500, and 2000 m.
All three domains include 72 vertical points, with a vertical stretched resolution decreasing with height from a few meters (near the ground level) up to 1 km (from 8 km above sea level to the top of the domain).The model results show an excellent agreement between the cloud formation on the slope between the Indian Ocean and the Maïdo observatory reproduced by Meso-Nh with the coincident observation by tethered balloon.
A high-resolution back trajectory study of this campaign focused on Volatile Organic Compounds observations has already been published in JGR-Atmosphere 15 .

Section S5. Aerosol acidity
Acidity is estimated using the ion balance equations, which describe protons loading [H + ] (in µeq/m 3 ).The major difference between aerosol pH and the proton loading method is that pH is the H + concentration per liquid water volume.In contrast, aerosol proton loading is the H + concentration per unit volume of air.Aerosol pH is the parameter of interest for atmospheric phenomena, but the proton loading is often used instead as a simpler non-quantitative surrogate for pH 8 . - Relevant Pearson's correlation values were achieved in this study between [H + ] and malonic acid (R=0.4,p-value=0.02)and succinic acid (R=0.52,p-value=0.002).In contrast, oxalic acid did not show a significant correlation (R=0.28,p-value=0.12),suggesting preferential oxalic acid production in the aqueous phase.
Equivalent NH 4 + / (NO 3 -+ SO 4 2-) ratios were also calculated (figure 5).A ratio of 1 would indicate that HNO 3 and H 2 SO 4 were completely neutralized by NH 4 + 9 .At our site, the average ratio is 0.24, indicating strong SO 4 2- concentrations prevalence over other species (From 2nd to 4th April in particular).These results suggest that acidity may enhance the uptake of dicarboxylic acids, gaseous precursors, onto aerosols, as also reported by Feng et al., 10 .

Figure S1 .
Figure S1.Average diurnal profiles (LT) of meteorological parameters (temperature, relative humidity (RH), solar flux, and wind speed).The bold lines represent the average hourly values and the surrounding shaded areas the confident intervals at 95% during the whole campaign.

Figure S2 .
Figure S2.Scatterplot of measured and predicted ammonium (NH 4 + , yellow dots) and average diurnal profile of NH 4 + measured/predicted ratio.

Figure
Figure S4.a) DMPS time series observed during the field campaign at MO, b) inset showing a zoom of the m diurnal variability from the 25th and 27th of March.

Figure S9 .
Figure S9.Average diurnal profiles (LT) of meteorological parameters (temperature, relative humidity (RH), solar flux, and wind speed), observed during the two periods of the campaign: period 1 from 8th March to 21st March (left) and period 2 from 22nd March to 4th April (right).The bold lines represent the average hourly values and the surrounding shaded areas the confident intervals at 95%.

Figure S10 .Figure S11 .
Figure S10.Diurnal backward trajectories obtained from MO during different period of the field campaign, A) Period1 (15 th to 22 nd Mach), B) Period2a (23 rd to 28 th March) and C) Period 2b (29 th March to 3 rd April).

Figure S12 .
Figure S12.Relative diurnal contribution of NR-PM1 aerosols during the whole field campaign at MO.

Figure S16 .
Figure S16.PMF factors diurnal variability (LT), MOOA (green), IEPOXOA (dark green), and POA (grey), separated into two sampling periods, on the Period 1 (left panels): from 13 th March to 21 st March (average of 9 days) and on the Period 2 (right panels): from 22 nd March to 4 th April (average of 14 days).Shaded areas represent hours with solar radiation.

Table S1 . Instrument and measurement details performed at MO during the BIO-MAIDO campaign
Below is an example of the CDCE used for this data set; during high ammonium sulfate mass fractions (ASMF) periods, a CE of > 0.45 was applied to the data.When ASMF (and Ammonium nitrate MF (ANMF) contributed <0.5 to the total PM1, a CE of 0.45 was applied to all species.