The composition of cluster ions was measured by an atmospheric-pressure-interface time-of-flight mass spectrometer (APi-TOF, Aerodyne Research Inc. & Tofwerk AG). The APi-TOF mass spectrometer consists of an atmospheric-pressure-interface (APi) module and a time-of-flight (TOF) mass spectrometer. The APi module allows the instrument to sample ions from ambient air directly. The positive or negative ions within the sampled airflow are focused and guided by two quadrupoles and an ion lens, with a gradually decreasing pressure (from atmospheric pressure to ∼ 10-4 mbar), before entering the TOF mass spectrometer (∼ 10-6 mbar). A more detailed description of this instrument is given in Junninen et al. (2010). In this study, ambient air was sampled through a ∼ 1.5 m stainless-steel tube with a total sample flow of 14 standard liters per minute (SLPM) to ensure laminar flow during sampling, and 0.8 SLPM of the total flow entered APi-TOF.

During the wet season, APi-TOF was first operated in negative mode to measure negative cluster ions (January) and then switched to positive mode to measure positive cluster ions (February to March). During the wet-to-dry transition period (April) and the dry season (May), the instrument was changed back to negative mode to investigate the potential seasonality of negative ion composition. It is important to note that, similarly to Frege et al. (2017), for better characterization of the connection between cluster ions and NPF, we only included the ion data observed under the cloud-free condition in this study (to avoid influence from, e.g., lightning activity). CHC was considered to be affected by clouds when relative humidity (RH) exceeded 95 %, as suggested by a previous study at the same location (Rose et al., 2015a). From January to May 2018, the proportions of cloud-free hours in the total measurement time were 72 %, 78 %, 79 %, 98 %, and 98 %, respectively (Fig. S1 in the Supplement).

Concentrations of H2SO4 and oxidized organic molecules (OOMs) were measured using a nitrate ion (NO3-)-based chemical-ionization atmospheric-pressure-interface time-of-flight mass spectrometer (CI-APi-TOF, Aerodyne Research Inc. & Tofwerk AG; Jokinen et al., 2012). The instrument is a combination of APi-TOF and a CI unit, which has been widely used to measure H2SO4 in the atmosphere (Jokinen et al., 2012; Bianchi et al., 2016; Zha et al., 2018). In this study, a soft X-ray source (L9490, Hamamatsu) was used to charge nitric acid (HNO3) in a sheath flow of 20 SLPM to produce the reagent ion, NO3-. H2SO4 and OOMs in the sample flow (10 SLPM) were then charged by either proton transfer or the formation of an adduct with the reagent ion during the ∼ 200 ms residence time in the CI unit. A calibration factor of 1.5 × 1010 cm-3 for H2SO4 was determined (with sampling loss corrected) according to the approach by Kürten et al. (2012). The same calibration coefficient was adopted for determining OOM concentrations in this study, which could result in an underestimation of their concentrations due to a lower charging efficiency of OOMs than H2SO4 by NO3- (Hyttinen et al., 2015).

The number concentration and size distribution of atmospheric ions and neutral particles were measured with a neutral cluster and air ion spectrometer (NAIS, Airel Ltd.; Mirme and Mirme, 2013). The instrument can detect naturally charged air ions and total particles with mobility diameters from 1.4–50 and 3–50 nm, respectively. It is important to note that the size range of the NAIS was corrected based on a side-by-side comparison with an updated version of NAIS (designed for measurements under low-pressure environments; Mirme et al., 2010) at CHC. Thus, the size range of detection was slightly different from the traditionally reported ranges (i.e., 0.8–42 and 2–42 nm, respectively; Manninen et al., 2010). The details of the instrument used can be found in Rose et al. (2017).

Particle number size distributions between 10 and 500 nm were measured by a mobility particle size spectrometer (MPSS; Wiedensohler et al., 2012), and the data were used for calculating the CS, which represents the loss rate of condensing vapors and cluster ions on preexisting particles (Kulmala et al., 2001).

Meteorological parameters, such as temperature, RH, and global radiation, were also measured simultaneously at CHC. Detailed descriptions can be found in Bianchi et al. (2022).

Six air mass pathways (PWs) representing air masses arriving at CHC were determined from the clustering analysis. They are named based on their clock positions from CHC (e.g., 03_PW indicates the pathway with its centroid located in the three-o'clock direction (east, 90∘) of CHC, Fig. 2a). The characteristics of these air mass pathways, such as source region, transport distance, and transport time, were distinct from each other (Table 1). A detailed description of the air mass pathways and their characteristics can be found in Aliaga et al. (2021).

The influence of each air mass pathway on CHC varied with time and was estimated by its SRR percentage (SRR[%]pathway) as in Eq. (1): 1SRR[%]pathway=SRRpathwaySRRtotal×100, where SRRpathway and SRRtotal are the residence times of a specific air mass pathway and the total (96 h = 345 600 s) in the simulation, respectively.

Air mass history analysis shows that the air sampled at CHC was typically a mixture of multiple pathways. Thus, the cluster ion composition observed during the study period was often influenced by multiple source regions concurrently. To characterize the influence of every single pathway on cluster ion composition, periods when an air mass pathway exerted its largest impact on CHC (the highest 10 % of its SRR[%]pathway values; Fig. 2b) during the whole study period are identified as the representative periods of the specific pathway. For instance, the representative periods of 03_PW (covering, e.g., the Amazon basin) are more frequently seen during the wet season (highest in January), whereas 08_PW (covering, e.g., the Altiplano region) has most of its representative periods during the dry season (highest in May). Note that the SRR[%]pathway of any individual pathway rarely reached 40 % due to the relatively short study period (see Fig. S2), and thus the representative periods cannot be directly identified via SRR[%]pathway values. In contrast, such representative periods were determined by using a certain threshold (e.g., > 70 %) in a previous study at CHC, which was based on a more than 6-year dataset (Koenig et al., 2021).

A number of negative ions were consistently observed at CHC throughout the study period (i.e., in the wet, wet-to-dry transition, and dry seasons). Based on the chemical composition of the observed negative ions, we classified them into eight groups as follows: sulfuric acid ((H2SO4)0–3⚫HSO4-), nitric acid ((HNO3)0–2⚫NO3-), SO5-, sulfuric acid–ammonia ((NH3)1–6⚫(H2SO4)3–7⚫HSO4-), malonic-acid-derived (MA-derived, including C3H3O4-, C3H4O4⚫NO3-, and C3H4O4⚫HSO4-), oxidized organic molecules (CHO / CHON⚫(HSO4- / NO3-)), others (other identified negative ions, such as IO3-), and unidentified ions. The campaign-average diurnal variations of these eight negative ion groups are shown in Fig. 4. Different diurnal patterns of each negative ion group were observed, mainly due to changes in concentrations of the parent neutral species and their unique physicochemical properties (e.g., the EA and PA of molecules; Ferguson and Arnold, 1981; Bianchi et al., 2017; Hirsikko et al., 2011).

(HNO3)0–1⚫NO3- and (H2SO4)0–3⚫HSO4- were among the highest in the signal of all negative ion groups in all seasons, making up 37 % (whole day) and 20 % (daytime: 07:00–19:00 and hereafter) of negative ions at CHC during the study period, respectively. The EA of their neutral molecules (HSO4 and NO3) is higher than that of most of the neutral species in the atmosphere and thus hinders the direct electron transfer from HSO4- and NO3- to other molecules through ion-molecule reactions (Ferguson and Arnold, 1981). As a result, these ion groups were found to dominate negative cluster ions at CHC and other locations, such as a number of remote sites in the United States (Eisele, 1986), a boreal forest site in Finland (Ehn et al., 2010; Bianchi et al., 2017), and the JFJ in Switzerland (Frege et al., 2017).

Distinct diurnal patterns were observed for the (HNO3)0–1⚫NO3- and (H2SO4)0–3⚫HSO4- ion groups. (HNO3)0–1⚫NO3- exhibited a relatively flat diurnal pattern (see Fig. 4), with similar fractions at daytime (35 %) and nighttime (19:00–07:00; 39 %). Such a diurnal pattern could result from the high EA of the NO3 molecule (4.01 eV and an additional ∼ 1 eV per HNO3; Ferguson and Arnold, 1981) and its relatively abundant (usually several parts per billion by volume) parent neutral species (e.g., HNO3 and N2O5) with multiple sources in the atmosphere (e.g., anthropogenic emission and lightning; Martin et al., 2007). In contrast, (H2SO4)0–3⚫HSO4- exhibited a strong diurnal variation. While the fraction of (H2SO4)0–3⚫HSO4- remained low (2 %) during nighttime, it started to increase after sunrise (shortly after 07:00) and reached a maximum (30 %) at around 10:00. Despite an EA comparable to that of the NO3 molecule (4.75 eV for HSO4; Wang et al., 2000), the strong diurnal variation of (H2SO4)0–3⚫HSO4- is a result of the photochemical production of neutral H2SO4. The influence of neutral H2SO4 on (H2SO4)0–3⚫HSO4- is indicated by their similar diurnal patterns (R: 0.52; see Fig. S4a). Similarly, a higher level of (NH3)1–6⚫(H2SO4)3–7⚫HSO4- was only observed with the presence of abundant (H2SO4)0–3⚫HSO4- during daytime. It is also important to note that the decreases in (H2SO4)0–3⚫HSO4- and (NH3)1–6⚫(H2SO4)3–7⚫HSO4- at around noontime (12:00; see Fig. 4) coincided with an enhanced CS, indicating the influence of a higher ion sink in addition to the decrease in neutral H2SO4 concentrations (Boulon et al., 2010; Frege et al., 2017).

The MA-derived ion group is distinguished from the CHO / CHON⚫(HSO4- / NO3-) ion group by its abundance in total organic ions. This group of ions, mainly composed of C3H3O4-, is formed from the deprotonation of malonic acid, with a higher EA (∼ 4.60 eV) than that of the NO3 molecule (4.01 eV; Ravi Kumar et al., 2005). Malonic acid, similarly to HNO3, has multiple origins in the atmosphere, such as primary and secondary anthropogenic sources, biogenic sources, and the degradation of larger organic compounds (Braban et al., 2003). It is also one of the main dicarboxylic acids which make up a substantial fraction of total carbon in aerosol particles (Kawamura and Bikkina, 2016). However, the concentration of gas-phase malonic acid is less well documented and is estimated to be in the range of 107 to 109 cm-3, up to 3 orders of magnitude higher than the typically reported ambient H2SO4 concentration (Fang et al., 2020). Thus, the fraction of the MA-derived ion group was high during nighttime (12 %; see Fig. 4). In contrast, its fraction decreased significantly (to 5 %) during daytime due to the increase in (H2SO4)0–3⚫HSO4-, which has an even higher EA (4.75 eV for HSO4; Wang et al., 2000).

The CHO / CHON⚫(HSO4- / NO3-) ion group, with an overall molecular formula of C2–15H2–26O2–13N0–2⚫NO3- / HSO4-, constituted a significant fraction of negative ions (31 %) at CHC. These organic ions are formed through the adduction between primary charge carriers, such as HSO4- and NO3-, and neutral OOMs. These OOMs are likely the oxidation products of the volatile organic compounds (VOCs) from the Amazon, the Altiplano, and the adjacent La Paz–El Alto metropolitan area. Their chemical composition is potentially affected by the changing air pathways covering different VOC source regions (Aliaga et al., 2021) and by the different conditions during daytime/nighttime due to the evolution of different atmospheric layers (Beck et al., 2022). While the diurnal variation was relatively small (34 % for the nighttime and 27 % for the daytime), the ion composition of CHO / CHON⚫(HSO4- / NO3-) could also be significantly different between daytime and nighttime due to the availability of the charging ions (see more discussions in Sect. 3.2.2). A previous study from a boreal forest shows that organic ions are mainly composed of CHO / CHON⚫NO3- during nighttime and that the fraction of CHO / CHON⚫HSO4- increases with the HSO4- signal during daytime (Bianchi et al., 2017). This is also shown by the slightly positive correlation between the CHO / CHON⚫HSO4- signal fraction and the total neutral OOM concentration during daytime (R: 0.25; see Fig. S4b), whereas no clear dependence was found between CHO / CHON⚫(HSO4- / NO3-) and the total neutral OOM concentration.

The SO5- ion group, consisting of SO5- ions and/or O2⚫SO3- cluster ions (Bork et al., 2013; Frege et al., 2017), exhibited a lower fraction (< 5 %) than the aforementioned ion groups during the study period. Similar to that of the (HNO3)0–1⚫NO3- ion group, no diurnal pattern was evident for the SO5- ion group. This may be the result of its different major formation pathways during daytime and nighttime (Bork et al., 2013; Frege et al., 2017). Daytime production of SO5- ions is likely associated with photooxidation of SO2 (similar to the formation pathway of H2SO4; Ehn et al., 2010; Schobesberger et al., 2015). This is shown in the positive correlation (R is 0.46 for daytime data in Fig. S4c) between the neutral H2SO4 concentration and the signal fraction of the SO5- ion group. During nighttime, however, the SO5- ion group is mainly composed of O2⚫SO3- cluster ions, which are possibly formed via the oxidation of SO2 with O3- (producing SO3-) and subsequent addition of O2 (Bork et al., 2013).

For better seasonality comparison at the high-altitude CHC, we calculated the average mass spectra of the negative ion groups for each season (Fig. 5 for daytime and Fig. S5 for nighttime). Distinct seasonalities (wet season, wet-to-dry transition period, and dry season) were found for the majority of the negative ion groups at CHC, including (H2SO4)0–3⚫HSO4-, (NH3)1–6⚫(H2SO4)3–7⚫HSO4-, SO5-, and organic cluster ions, as shown in the averaged daytime mass spectra (Fig. 5; a more detailed reason will be discussed below). However, the signals of some other negative ion groups, e.g., MA-derived ions and (HNO3)0–2⚫NO3-, were generally stable (with differences ≤ 20 %) across the seasons. Such unclear seasonalities can be attributed to the high EA (Ferguson and Arnold, 1981; Ravi Kumar et al., 2005) and/or the stability of the parent neutral species (Martin et al., 2007; Kerminen et al., 2000; Bikkina et al., 2021). Similar patterns can also be found in the average nighttime mass spectra among the seasons (Fig. S5).

The (H2SO4)0–3⚫HSO4- group exhibited a much higher contribution in the dry season (May) than in the wet season (January) and the wet-to-dry transition period (April). The daytime fraction of (H2SO4)0–3⚫HSO4- increased continuously from 16 % in the wet season to 20 % in the wet-to-dry transition period and to 30 % in the dry season. The maximum number of H2SO4 molecules increased concurrently from two to four in the cluster ions (i.e., from (H2SO4)2⚫HSO4- to (H2SO4)4⚫HSO4-). Similar trends were also found for other H2SO4-related ions, such as (NH3)1–6⚫(H2SO4)3–7⚫HSO4- and SO5- during daytime (Fig. 5).

The seasonal variations of the aforementioned H2SO4-related ion groups are likely due to the changes in neutral H2SO4 (Fig. S6) linked to the changing synoptic-scale wind patterns carrying different air masses with varying SO2 (Bianchi et al., 2022). The air mass pathways 07_PW and 08_PW, covering the western and northern Altiplano (see Table 1 and Fig. 2a), where active volcanic degassing of SO2 has been reported (Moussallam et al., 2017; Carn et al., 2017), had their largest influence on CHC in the dry season (i.e., May; see Fig. 2). The corresponding daytime fractions of (H2SO4)0–3⚫HSO4- from these two pathways (Fig. 6a) were also the highest (27 % and 32 %, respectively). In contrast, air mass pathways 03_PW and 12_PW, originating in the Amazon basin and the eastern/southeastern lowlands, exerted their most significant impact on CHC in the wet season (i.e., January), with lower daytime fractions of (H2SO4)0–3⚫HSO4- (13 % and 14 %, respectively). The low fractions of H2SO4-related cluster ions in the wet season are also consistent with the lower SO2 level in the Amazon basin compared to the Altiplano (Andreae et al., 1990). As for the wet-to-dry transition period (i.e., April), 05_PW covering both the southeastern lowlands and the southern Altiplano (where volcanic degassing is also significant; Carn et al., 2017) had an evident influence on CHC, resulting in a substantial level of H2SO4-related cluster ions (21 % for daytime). It is also noted that, because of the much lower nocturnal neutral H2SO4 concentrations, the nighttime fractions of H2SO4-related cluster ions in all air mass pathways (Fig. 6b) were generally low (< 3 %), and no clear seasonality was found.

The organic cluster ion group exhibited a distinct seasonal variation from (H2SO4)0–3⚫HSO4-. The signal fraction of organic cluster ions was higher in the wet season (31 % for daytime and 32 % for nighttime) than in the dry season (23 % for daytime and 27 % for nighttime; Figs. 5 and S5), but it was highest for the wet-to-dry transition period (46 % for daytime and 52 % for nighttime; see Fig. 6).

The seasonal changes in organic cluster ions could be due to the combined effect of different meteorological conditions and VOCs from different air mass origins (see Fig. 2). The air masses that originated from the Amazon basin and the lowlands (03_PW and 12_PW) showed their largest impact on CHC in the wet season (i.e., January). They contained higher fractions of organic cluster ions, which were 35 % and 34 % for daytime and 50 % and 45 % for nighttime, respectively (Fig. 6). In the dry season (i.e., May), however, the changes in air mass origin towards the Altiplano and the Pacific Ocean led to a lower content of organic cluster ions. The organic cluster ion fractions for 07_PW and 08_PW (largest influence on CHC in the dry season) in May were 23 % and 19 % for daytime and 29 % and 27 % for nighttime, respectively. As for the wet-to-dry transition period (i.e., April), due to the combined influences of biogenic and anthropogenic VOC sources from 05_PW (evident impact on CHC in April), covering the southeastern lowlands and the southern Altiplano, the corresponding organic cluster ion fractions from this air mass pathway were also the highest (41 % for daytime and 53 % for nighttime; see Fig. 6).

A further investigation of the organic ion group shows that the seasonal trends of the individual organic ions also varied (Fig. 7 for daytime and Fig. S7 for nighttime). Although the majority of the organic cluster ions at CHC were more abundant during the wet season (Figs. 7a and S7a), fractions of CHO / CHON⚫HSO4- increased during the dry season. The observed increases in CHO / CHON⚫HSO4- cluster ions could be associated with the increased HSO4- / NO3- ratios in the dry season (Figs. 5 and S5). Similar increases in CHO / CHON⚫HSO4- cluster ions were also found to relate to the ratio of HSO4- / NO3- in a boreal forest environment (Bianchi et al., 2017). In addition, changes in OOM composition between the wet and dry seasons may also play a role (Figs. 7b and S7b), as NO3- tends to cluster with OOMs containing hydroxyl and hydroperoxyl functional groups (Hyttinen et al., 2015), while some other observed OOMs may be more efficiently charged by HSO4-.

The seasonal variations of the individual organic cluster ions are likely caused by different air masses (Figs. 7b and S7b). The air masses influenced by tropical rainforest vegetation from the Amazon basin are dominated by isoprene (C5H8) emissions and isoprene oxidation products (Bianchi et al., 2022). This region corresponds to 03_PW and 12_PW (largest impact on CHC in the wet season in January) consisting of relatively higher fractions of organic cluster ions with OOMs containing four to five carbon atoms (50 % and 46 % for nighttime and 29 % and 32 % for daytime, respectively). In contrast, when the air masses were more influenced by the Altiplano (i.e., 05_PW, 07_PW, and 08_PW, with more anthropogenic emissions and less vegetation) in the wet-to-dry transition period and the dry season, organic cluster ions with six to eight carbon atoms, potentially originating from anthropogenic sources (e.g., toluene (C7H8); Huang et al., 2019; Cai et al., 2022), had higher contributions. The signal fractions of these organic ions were thus the highest in these air mass pathways, accounting for 36 %–39 % for nighttime and 37 %–39 % for daytime. For all the air mass pathways, fractions of organic ions with more than nine carbon atoms were relatively low (< 10 %). This might be due to their lower volatilities compared to OOMs with smaller carbon numbers (Donahue et al., 2012), resulting in a larger probability of them being removed during their transport to CHC (e.g., condensing on preexisting particles).