Water-soluble dicarboxylic acids, oxoacids and α-dicarbonyls in the tropical aerosols in coastal megacity Mumbai: molecular characteristics and formation processes

Daytime and nighttime PM10 samples were collected during summer (June) and winter (February) at a representative urban site in Mumbai, located on the western coast of Indian subcontinent. Samples were studied for molecular distribution of water-soluble dicarboxylic acids, oxoacids and dicarbonyls as well as total carbon (TC), water-soluble organic carbon (WSOC), inorganic ions along with specific markers (levoglucosan, K+) to better understand sources and formation processes of organic aerosols in Mumbai. The distribution of water-soluble organics was characterised by high abundance of oxalic acid (C2), followed by phthalic (Ph), terephthalic (tPh), azelaic (C9), malonic (C3), and succinic acids (C4). Positive correlations between C2, sulfate and glyoxal (ωC2) suggest secondary production of C2 predominantly via aqueous phase chemistry. C2 also showed positive correlation with K+ and levoglucosan indicating that biomass/biofuel burning is the potential source of diacids in the Mumbai aerosols. In addition, higher average contributions of total diacids to WSOC and OC in winter than in summer suggest that aerosols were aged i.e., photochemically well processed in winter in Mumbai. On the other hand, diurnal change in their ratios is observed with higher ratio in daytime samples than that of previous and succeeding nighttime samples, suggesting diacids are also influenced from local sources in both the seasons. This study demonstrates that biomass burning as well as biogenic sources are important sources influencing the distributions of aerosols in Mumbai.


Introduction
Diacids and related organic compounds comprise a substantial fraction of water-soluble organic carbon (WSOC) (Saxena et al. 1995), which can act as cloud condensation nuclei and significantly alter hygroscopic properties of aerosols (Novakov and Penner 1993;Carrico et al. 2005;Kaku et al. 2006;Lee et al. 2006;Aggarwal et al. 2007). Therefore, the knowledge of primary source emissions and secondary production of diacids as well as other organic acids in the atmosphere is important for evaluating atmospheric processes and regional air quality. South and Southeast Asia are potentially important regions for studying the source regions of atmospheric aerosols globally because of extensive economic development and growing anthropogenic pollutants (Lelieveld et al. 2001). The Indian Ocean Experiment (INDOEX) in 1999 was conducted in order to study the influence of long-range transport of continental aerosols from South and Southeast Asia over Indian Ocean (Lelieveld et al. 2001). The experiments indicated significantly high pollution load over the northern Indian Ocean influenced by extensive fossil fuel combustion and biomass burning in India and Southeast Asia, and subsequent influence on air quality deterioration at local, regional and global scales. Also, meteorological parameters such as high ambient temperature, intense solar radiation and high levels of oxidants in the atmosphere (O 3 , NO X etc.) (Pulikesi et al. 2006) favour photo-induced formation of aerosols, further influencing the physical and chemical properties of tropical Indian aerosols.
A few studies on diacids and related organic compounds have been conducted in the South and Southeast Asian atmosphere (Lelieveld et al. 2001;Carrico et al. 2003;Rengarajan et al. 2007;Miyazaki et al. 2009;Pavuluri et al. 2010Pavuluri et al. , 2015Fu et al. 2012Fu et al. , 2016Deshmukh et al. 2016) however the importance of primary source emissions and secondary production of diacids is limited in tropical Indian aerosols. Recently, Aggarwal et al. (2013) investigated the bulk carbonaceous and inorganic components along with specific markers in PM 10 samples to better understand the photochemical processing of aerosols in Mumbai. Here, we report molecular distributions of water-soluble diacids, oxoacids and α-dicarbonyls in PM 10 samples collected in two different seasons; June 2006 (summer) and February 2007 (winter) on daynight basis from coastal megacity Mumbai of the Indian subcontinent. Seasonal changes in concentrations of diacids and related compounds, their relative abundance, potential sources and photochemical formation were explored to fully understand atmospheric processing and sources of aerosols.

Aerosol sampling and chemical analysis
PM 10 samples were collected in summer (8-14 June 2006;n = 14) and winter (13-18 February 2007; n = 10) during daytime (06:00-18:00 h, local time) and nighttime (18:00-06:00 h) at the rooftop (12 m above the ground) of the building of the Centre for Environmental Science and Engineering at the Indian Institute of Technology Bombay (IITB) campus in Mumbai, which is a representative urban site about 10 km interior from the coast. A highvolume air sampler (∽1.2 m 3 min −1 ) was used with pre-combusted quartz filters (8″ × 10″) to collect PM 10 particles at the sampling site. The filter samples were stored in clean precombusted glass jars with Teflon-lined screw cap, and stored at -20 °C until analysis. During the sampling period, the average ambient temperature and humidity in summer were recorded as 31 ± 2 °C and 77 ± 12%, while 25 ± 3 °C and 37 ± 5% in winter.
Samples were analysed for water-soluble dicarboxylic acids, oxocarboxylic acids, and α-dicarbonyls in the filter samples by the methods described elsewhere (Kawamura and Ikushima 1993). Briefly, filter aliquots (∽6 cm 2 ) were extracted with 12-15 ml Milli-Q water using an ultrasonic bath for about 3-10 min. The extracts were reacted with 14% BF 3 in n-butanol to derive carboxyl and aldehyde groups to esters and acetals respectively. The derivatives were then extracted with n-hexane and determined using a capillary gas chromatograph (GC) (HP 6890 GC) equipped with a flame ionization detector (FID). The analytical errors (repeatability) in diacids determination were within 6% for major diacids based on duplicate analysis. The concentrations reported here have been corrected for the field blanks. In this paper, full datasets of dicarboxylic acids and related compounds are discussed.
Inorganic ions were analyzed in the filter samples using a Metrohm-761 ion chromatograph (IC) coupled with an autosampler. The detailed methodology is discussed elsewhere (Agarwal et al. 2010). By duplicate analysis of filters, analytical error was estimated to be 4%. In this paper, we use concentrations of SO 4 2− and K + to understand sources of diacids. For measurement of water-soluble organic carbon (WSOC), an aliquot of the filter sample (2 cm disc) was extracted with Milii-Q water (7 ml) followed by ultrasonication for 15 min. Using a disc filter, the particles in the extracts were removed. In 5 ml water extracts, 0.1 ml of 2 M HCl solution was added. After purging 10 min with nitrogen gas (80 ml min −1 ), 100 µl of solution was were injected into TOC analyzer (Shimadzu TOC-5000A) (Agarwal et al. 2010). Analytical error was within 6% by the duplicate analyses.
Levoglucosan was determined by another aliquot of filter with a mixture of dichloromethane and methanol (2:1) derivatized with BSTFA (N,O-bis-trimethylsilyl trifluoroacetamide) and pyridine as catalyst followed by determination with a GC/mass spectrometry (GC/MS) (HP 6890 GC, 5973 MSD) (Agarwal et al. 2010). Recoveries determined during the analytical procedure were > 90%, and the analytical errors (repeatability) were within 10% based on duplicate analyses.
For analysis of organic carbon (OC), a filter punch was placed in a quartz tube inside the chamber of the semi-continuous carbon analyzer (Sunset laboratory Inc., Model 4L). The Interagency Monitoring of Protected Visual Environments (IMPROVE) thermal protocol was used assuming that the carbonate carbon in the sample is negligible (Agarwal et al. 2010). The analytical errors were within 9% by duplicate analyses. For the determination of total carbon (TC) content in the aerosol samples, elemental analyser (EA) (Carlo Erba, NA 1500) has been used. A 1.4 cm diameter filter cut was packed in a tin cup and for analysis loaded on an autosampler. The analytical errors for TC in the samples were determined to be within 4% by duplicate analyses.

Dicarboxylic acids
The mass concentration range of total diacids was 194-1101 ng m −3 (average 433 ng m −3 ). Among dicarboxylic acid compounds, oxalic acid (C 2 ) was found to be dominant (average, 185 ng m −3 ) followed by phthalic (Ph) (55 ng m −3 ), terephthalic (tPh) (53 ng m −3 ), azelaic (C 9 ) (41 ng m −3 ), malonic (C 3 ) (23 ng m −3 ), and succinic acids (C 4 ) (22 ng m −3 ). Further we observed that the abundance of individual homologues of diacids (C 2 -C 12 ) generally decreased with an increase in carbon chain length, although C 7 and C 9 are the exceptions that are more abundant than C 6 and C 8 , respectively. The relative abundance of C 2 (the most abundant diacid) in total diacids, fluctuated from 20 to 61% (average, 39%). The mass concentration of C 2 detected in this study (average, 185 ng m −3 ) is relatively similar than that observed at other urban sites in Asia, such as Beijing (220 ng m −3 ) (Huang et al. 2005) and Tokyo (270 ng m −3 ) (Kawamura and Ikushima 1993). Although, higher C 2 concentrations have been reported at urban sites in India such as in Delhi (average, 1430 ng m −3 ) (Miyazaki et al. 2009) and Chennai (440 ng m −3 ) (Pavuluri et al. 2015). Phthalic (aromatic) acid was the second most abundant diacid, accounting for 5-21% (average, 11%) of total diacids. This feature is less common in the available studies, although a previous study has reported high abundance of phthalic acid accounting for 2-12% (7%) of total diacids in aerosol samples collected at urban site in Los Angeles (Kawamura and Kaplan 1987). It is suggested that phthalic acid which is a tracer for anthropogenic emissions, is formed in the atmosphere as a result of photochemical oxidation of polynucleated aromatic hydrocarbons (Kawamura and Ikushima 1993). It is also suggested as one of the tracers for open waste burning in the Delhi aerosols influenced with maximised plastic waste burning (Kumar et al. 2015).

α-Dicarbonyls
They were also detected in aerosol samples at much lower concentrations (average, 7 ng m −3 ) than diacids (433 ng m −3 ) and oxoacids (23 ng m −3 ). This is possibly because dicarbonyls are mainly present in the gas phase as their vapour pressures are much higher than those of diacids . They are produced from photochemical oxidation products of benzene and toluene as indicated in laboratory experiments (Bandow et al. 1985) and also from the atmospheric oxidation of isoprene Bikkina et al. 2014).

Seasonal changes in the concentrations of diacids, oxoacids, and α-dicarbonyls
Because photochemical formation of diacids is expected to be influenced by meteorological parameters like ambient temperature and solar radiation, studying seasonal variation of diacids is important to better understand their sources and distribution in the atmosphere. Figure 1 presents the variations in the total concentrations of diacids, oxoacids and α-dicarbonyls in the Mumbai aerosol samples in both summer (n = 14) and winter (n = 10) seasons. All the three components showed maximum average concentrations in the winter season, with significant day and nighttime variations predominantly in winter.

Diacids
In both the seasons oxalic acid was the dominant species with an average concentration of 85 ng m −3 in summer and 325 ng m −3 in winter ( Table 2). Concentration of C 2 was less abundant in summer while maximised in winter with the highest concentration in the February 17 daytime sample (Fig. 2a) Similar trend was observed for C 3 and C 4 (Fig. 2b, c) except that their peak magnitudes were smaller (10 times) than that of C 2 . The abundance of shorter chain diacids in winter as compared to summer is possibly due to (i) enhanced continental/regional anthropogenic emissions of aromatic hydrocarbons in 1 3 winter followed by their subsequent photo-oxidation (ii) pronounced scavenging and wet removal of particles in summer due to high temperature and humidity in the tropics. The meteorological conditions at the sampling site in Mumbai have been mentioned elsewhere . Briefly, in summer higher temperature and humidity (average, 31 ± 2 °C and 77 ± 12%, respectively) were recorded than in winter (average, 25 ± 3 °C and 37 ± 5%, respectively), suggesting higher wet deposition rates in summer. On the other hand, longer chain diacids (C 6 -C 12 ) showed significant peaks in summer as well as in winter i.e., in 8 th June and 14 th February and/or 17 th February samples. Also, the day to night variations in concentration levels of long-chain diacids appeared to be similar in both the seasons, unlike small chain diacids (C 2 , C 3 , C 4 ). Interestingly, the average mass concentration of C 7 (shown in Table 2 and Fig. 2d) was higher in summer (1.2 times) than winter. The abundance of C 7 and other longer-chain diacids during summer is possibly due to the influence of long-range air mass transport in Mumbai from the continent across the Arabian sea . These air masses are proposed to have abundant biogenic unsaturated fatty acids, which are oxidised at a double bond in the atmosphere thereby resulting in the formation of longer-chain diacids (Kawamura and Gagosian 1987) (further explained in Section 3.3). Interestingly, it is worthy to note that, among straight-chain diacids (C 2 -C 12 ), C 9 (23-68 ng m −3 ) was the second most abundant diacid after C 2 , with maximum concentration in the daytime 8 th June sample (68 ng m −3 ) and 14 th February sample (72 ng m −3 ) (Fig. 2e). Such high abundances of C 9 have been reported in previous studies (Kawamura and Ikushima 1993;Pavuluri et al. 2015). It is proposed that C 9 is produced as a result of atmospheric oxidation of biogenic unsaturated fatty acids, which contains a double bond predominantly at C-9 position (Kawamura and Gagosian 1987), suggesting that biogenic organic species might have an influence on aerosol composition in the Mumbai air. The concentrations of branched chain diacids, iC 4 and iC 5 varied differently from rest of the diacids, such that they showed a different daytime-maxima in summer i.e., on 10 th June (1.34 ng m −3 and 3.98 ng m −3 respectively) and 17 th February (1.45 ng m −3 and 8.17 ng m −3 respectively) (Fig. 2g, h). These results suggest that individual diacids have different sources in Mumbai.
Aromatic diacids, e.g., Ph and tPh are abundantly present in PM 10 in Mumbai. The seasonal average concentration of Ph (Fig. 2i) was similar in both summer and winter i.e., about 55 ng m −3 , while that of tPh (Fig. 2j) was higher in winter (92 ng m −3 ) than summer (24 ng m −3 ). Ph is the second most abundant diacid accounting for about 15% in total diacids, which is predominantly emitted from automobiles (Kawamura and Kaplan 1987) and/or formed by photo-oxidation of aromatic hydrocarbons emitted from incomplete combustion from vehicles (Kawamura and Kaplan 1987;Kawamura and Ikushima 1993) and from open waste burning emissions (Kumar et al. 2015;Kalogridis et al. 2018), suggesting that anthropogenic source strength is a significant factor influencing Mumbai aerosols. Fairly high concentration of Ph has been reported in other urban sites like Chennai (40-87 ng m −3 ) (Pavuluri et al. 2015) and Hong Kong (84 ng m −3 ) (Ho et al. 2006).
Oxoacids Both glyoxylic (ωC 2 ) as well as pyruvic acid (Pyr) showed maximum average concentrations in winter i.e., 31 ng m −3 and 12 ng m −3 , respectively, but were less abundant in summer (4.5 ng m −3 and 3 ng m −3 respectively). Also, the seasonal variations of ωC 2 (Fig. 2k) were similar to that of Gly (Fig. 2l). The predominance of ωC 2 in winter has been reported in previous study by Kawamura (1993) in the urban atmosphere, further suggesting that it is produced extensively in winter by photo-oxidation of Gly and then serve as precursor for oxalic acid. α-Dicarbonyls Glyoxal (Gly) and methylglyoxal (MGly) showed similar peak concentrations in winter (average = 6.5 ng m −3 and 5.8 ng m −3 respectively), which is 3 to 4 times higher than in summer. Previous studies (Kawamura 1993;Volkamer et al. 2001) proposed that gas-phase oxidation of biogenic and anthropogenic precursors (VOCs) produced Gly and MGly in the atmosphere, which subsequently could act as potential precursors of C 2 acid via heterogeneous reactions (Bikkina et al. 2014;Kawamura and Bikkina 2016). Interestingly, we found a good positive correlation (r 2 = 0.82) between Gly and levoglucosan (an important tracer of biomass burning) (Simoneit 1999) during the sampling period, which was found relatively higher in winter (r 2 = 0.87) than summer (0.67). Similarly, MGly and levoglucosan were found to be more positively correlated in winter (0.85) than summer (0.69). This indicates that biomass burning activities which are more common in winter, are potential sources of α-dicarbonyls in Mumbai.

Mass fractions of WSOC, OC, TC and Diacid-C
Contributions of diacids to water-soluble organic carbon (WSOC), organic carbon (OC), total carbon (TC) were evaluated to better understand the atmosphere processing of water-soluble organic aerosols in Mumbai. The average concentrations of WSOC, OC and TC in the samples were 3.4 µg m −3 , 9 µg m −3 and 13 µg m −3 respectively. WSOC fractions comprise 21-49% (average, 37%) in OC and 17-41% (27%) in TC. These values are similar to those (44% of OC and 39% of TC) reported in Sapporo (Aggarwal et al. 2007), where photochemical processing of diacids is prominent during long-range transport. These comparisons suggest aerosols in Mumbai are photochemically processed. Interestingly, similar WSOC/OC ratios were observed in daytime and nighttime samples in summer (36% and 38%, respectively) and winter (39% and 38%, respectively) (Fig. 3), indicating that aerosols in Mumbai are not significantly influenced by local emissions, a point to be discussed in detail in the next section. High mass fraction ratios of diacid-C in TC and WSOC indicate that diacids and related compounds are significantly produced by photochemical processing in the atmosphere (Kawamura and Yasui 2005). In this study, the mass fraction of total diacid-C to TC ranged from 0.8% to 3% with an average of about 2%. These values are twice higher than those reported in Tokyo (range, 0.8% to 1.8%, average, 0.95%) (Kawamura and Yasui 2005) but similar to those reported in Sapporo (range = 0.74% to 3.6%, average = 1.8%) (Aggarwal and Kawamura 2008), where diacids were reported to be largely produced by photochemical processing during long-range transport of aerosols. Further, the mass fraction of total diacid-C to WSOC varied from 4.8% to 14.2% with an average value of 7%, which is also higher than that in Tokyo (average = 3%, Kawamura and Yasui 2005) and in Sapporo (average = 4.8%, Aggarwal and Kawamura 2008). Thus, the higher contribution of total diacids-C Fig. 3 Temporal variations in mass fraction (%) of WSOC to a OC, and b TC in summer (8-14 June, 2006) and winter (13-18 February, 2007) to TC and WSOC obtained in this study indicate the high abundances of secondarily produced diacids in tropical aerosols collected in Mumbai, i.e., coastal city in the Indian subcontinent, further suggesting that tropical aerosols are photochemically well aged.
Further, in order to better understand seasonal trend of secondary production of diacids and related compounds, contributions of total diacid-C to WSOC and OC in summer and winter are shown in Fig. 4. We observed higher values in winter than summer, indicating that aerosols are more photochemically processed or aged in winter. Also, diurnal change in their ratios is observed with higher ratio in daytime samples than nighttime samples. This suggests that diacids are also possibly influenced from local sources in both the seasons. The results are consistent with previous study in Mumbai which demonstrate that Mumbai aerosols are photochemically processed on the basis of stable C-isotopic ratios of TC ).

Emission source contributions-local versus continental
To better understand the specific sources of individual diacids in the atmosphere, their relative abundances in total straight-chain diacids in different seasons are evaluated. Figure 5 shows the relative abundances of individual diacids in total straight chain diacids (C 2 -C 12 ) for summer and winter samples. The overall contribution of smaller-chain diacids (C 2 -C 4 ) was predominantly higher in the total straight-chain diacids than longer-chain diacids (C 5 -C 12 ) in winter than in summer. For example, C 2 , C 3 and C 4 collectively contributed to about 80% of total straight chain diacids in winter, which lowered to 62% in summer. In contrast, the relative abundance of longer chain diacids (C 5 -C 12 ) was higher in summer (38%) than in winter (20%). Particularly, C 9 showed two times higher abundance in summer (20%) than winter (10%). The predominance of longer-chain diacids in summers can be explained by an air mass 5-day back trajectory performed in a previous study conducted in Mumbai . It showed that the summer air masses originating from southwest subcontinental sites travelled longer distances over the Arabian Sea and Indian Ocean before arriving at Mumbai. Accordingly, the air masses may contain abundant biogenic unsaturated fatty acids, which are subsequently oxidised in the atmosphere predominantly at C-9 position to form C 9 acid (Kawamura and Gagosian 1987), and other longer chain diacids (C 5 -C 12 ) in summer, indicating a prominent continental influence.
Higher abundances of shorter-chain diacids in winter may be interpreted by enhanced anthropogenic emissions (which may have local or continental sources, to be discussed further), for example, aromatic hydrocarbons emitted by biomass burning/ fossil fuel combustion, and their subsequent atmospheric oxidation (Pavuluri et al. 2010). To study the effect of biomass/biofuel burning on the sources of organic compounds at the sampling site, potassium (K + ) is used as tracer (Andreae 1983). In winter samples, C 2 correlated well with K + (r 2 = 0.55) (average: 312 ng m −3 , range: 22 ng m −3 to 1250 ng m −3 ). Levoglucosan (average: 233 ng m −3 , range: 29 ng m −3 to 715 ng m −3 ), which is an important tracer of biomass/biofuel burning, showed a strong correlation with K + (r 2 = 0.82) and C 2 (r 2 = 0.74). In addition, strong correlations were observed among C 2 , C 3 and C 4 diacids (r 2 = 0.87 to 0.96), indicating that these were likely to have similar sources. Many previous studies have reported such good correlation (Huang et al. 2005;Kawamura and Bikkina 2016;Chen et al. 2021). These results suggest that biomass/biofuel burning is likely to be an important source of smaller diacids in winter.
Further, to better understand whether aerosols in Mumbai are influenced from local sources or continental air mass, we plotted water soluble potassium (K + ), levoglucosan and oxalic acid (C 2 ) concentrations for day and nighttime samples in Fig. 6. We found that their concentrations do not show a diurnal pattern i.e., their concentrations do not always show higher values during daytime, suggesting that the aerosols are not likely to be freshly emitted from local sources but continental influence may be more significant. This inference is further supported by the similar WSOC/OC ratios in daytime and nighttime samples in summer (36% and 38% respectively) and winter (39% and 38% respectively), indicating that aerosols are not likely fresh, being emitted from local sources. Interestingly, excellent correlation was observed between C 2 and K + during daytime (r 2 = 0.93) and nighttime (r 2 = 0.68). These results demonstrate important contributions of biomass/biofuel burning to small chain diacids (C 2 , C 3 , C 4 ) in Mumbai, which are largely involved with continental or regional sources.

Preferential production and relative abundance of diacids
The C 3 /C 4 ratio can be used as an important indicator of enhanced photochemical formation of diacids in the atmosphere, because the C 3 acid is largely produced by the photoinduced oxidation C 4 acid in the atmosphere (Kawamura and Ikushima 1993). In our study, C 3 /C 4 ratio varied from 0.4 to 1.7 with an average of 1.1, showing a clear seasonal  (2007) trend with maximum value in summer (average, 1.2). This value is much higher than that reported in motor exhaust in Los Angeles (average, 0.35, Kawamura and Kaplan 1987), but comparable with those reported in Chennai, India (1.4, Pavuluri et al. 2010), China (1.1, Wu et al. 2015), Hong Kong (1.3, Ho et al. 2006 and Chichijima island (1.2, Mochida et al. 2003) in the western North Pacific, where aerosols were reported to have significant photochemical modifications. The high C 3 /C 4 ratio in the present study thus indicates enhanced photochemical production of diacids, suggesting that aerosols are significantly aged in Mumbai. Also, diurnal change in C 3 /C 4 ratio is observed with higher average ratio in daytime than nighttime samples, suggesting that diacids are also possibly produced by local photochemical reactions in both the seasons. It is of interest to note that C 3 /C 4 ratio values are substantially similar to those in ambient urban aerosols which are photochemically aged (0.14-1.0) (Kawamura and Bikkina 2016).
Among aliphatic unsaturated diacids, the relative abundance of maleic acid (M, cisconfiguration) and fumaric acid (F, trans-configuration) is important to better understand secondary formation of diacids and related compounds in the atmosphere. It is proposed that the photooxidation of organic precursors like benzene and toluene results in the formation of maleic acid (cis configuration), which may further be photochemically isomerised into fumaric acid (trans configuration) (Kawamura and Ikushima 1993). Thus, M/F ratio is controlled by two factors: (1) formation of maleic acid, and (2) subsequent cis-to-trans isomerization to form fumaric acid. In our study, M/F ratios varied from 0.86 to 7 with an average value of 2.3. Also, the averaged M/F ratio is twice higher in summer (average, 3.01) than in winter (1.44) samples. This suggests that in summer, there is an enhanced significant production of maleic acid, which is later subjected to the cis-to-trans isomerization.
The C 6 /C 9 ratio can be used as an important tracer for evaluating the contribution of anthropogenic and biogenic sources to the aerosol diacids (Hatakeyama et al. 1987;Kawamura and Ikushima 1993) because C 6 is produced by oxidation of anthropogenic cyclohexene whereas C 9 is a tracer for biogenic sources, i.e., oxidation of unsaturated fatty acids. The average C 6 /C 9 ratio in Mumbai (0.17) is relatively similar to that in Raipur (0.10-0.26, Deshmukh et al. 2016) and Chennai (0.34, Pavuluri et al. 2010) while substantially lower than in Sapporo (0.78, Aggarwal and Kawamura 2008), Tokyo (0.69, Kawamura and Ikushima 1993) and Hong Kong (0.9, Ho et al. 2006), where contribution 1 3 of anthropogenic sources is more significant. Therefore, lower ratio in Mumbai indicates more influence of biogenic unsaturated fatty acids possibly derived from biomass burning, which are further oxidised in the atmosphere to produce C 9 diacid.
We studied the correlation of oxalic acid and sulfate to better understand the formation processes of oxalic acid because reaction mechanism of both species is mainly via aqueous phase reactions (Yao et al. 2004;Huang et al. 2005;Liu et al. 2012). The C 2 and sulfate concentrations (average: 5 µg m −3 , range: 2 µg m −3 to 16 µg m −3 ) showed a strong positive correlation (r 2 = 0.90) during the sampling period, indicating significant secondary production of C 2 in Mumbai aerosols via aqueous phase reactions. Also, a more positive correlation between C 2 and sulfate was observed during winter (r 2 = 0.76) than summer (r 2 = 0.60), suggesting preferential secondary C 2 production during winter. Further, previous studies (Lim et al. 2005;Legrand et al. 2007;Ervens et al. 2011) have proposed that α-dicarbonyls like glyoxal (Gly) and methylglyoxal (mGly) are important intermediates in glyoxylic acid (ωC 2 ) formation which further produces oxalic acid via aqueous phase reactions. We found a strong positive correlation (r 2 = 0.88) between glyoxal and oxalic acid, as well as between methylglyoxal and oxalic acid concentrations (r 2 = 0.84) in Mumbai (Fig. 7). These results possibly suggest that Fig. 7 Correlation of a glyoxal, and b methylglyoxal with oxalic acid in Mumbai α-dicarbonyls undergo aqueous phase oxidation and result in formation of glyoxylic and oxalic acid in the Mumbai aerosols.
Further, in order to better understand the above idea, we studied the concentrations of possible product/precursor ratios. As shown in Fig. 8, oxalic acid/glyoxal ratios showed weak diurnal trends in summer with usually higher ratios in daytime than nighttime, suggesting aqueous phase production of oxalic acid by glyoxal. Similar diurnal trends were observed for ωC 2 /Gly ratios indicating oxidation of glyoxal to from glyoxylic acid. We found relatively higher product/precursor ratios and more prominent diurnal trends in winter, again indicating preferential secondary formation of C 2 in winter. Therefore, these results together with meteorological conditions such as strong solar radiation and high relative humidity in tropical regions, support aqueous phase oxidation formation of C 2 .
It is of interest to note that biomass burning activities emit large amounts of oxalic acid and α-dicarbonyls in the atmosphere (Wittrock et al. 2006;Fu et al. 2008;Kawamura et al. 2013). In fact, glyoxal and methylglyoxal showed a good positive correlation with levoglucosan as discussed in Section 3.2. Further, Gly and MGly concentrations were higher in winter when biomass burning is extensive, than in summer. This further suggests that biomass burning is an important source of diacids and related compounds at sampling site in Mumbai.

Conclusions and summary
Tropical aerosol (PM 10 ) samples collected at an urban site in Mumbai in summer (June 2006) and winter (February, 2007) were studied for molecular characteristics of watersoluble dicarboxylic acids and related compounds. Irrespective of the season, oxalic acid was detected as the most abundant diacid (49% to 66% of total diacids). Total mass concentrations of diacids (194-1101 ng m −3 ), oxoacids (5-87 ng m −3 ) and α-dicarbonyls (3-22 ng m −3 ) were higher in winter than summer, except for C 7 , which showed opposite trend. C 9 that is a tracer for biogenic sources showed high concentrations in both seasons, suggesting that biogenic organic species seriously influence the aerosol composition in the Mumbai atmosphere. Moreover, the concentration of Ph (tracer for anthropogenic emissions, especially from open waste burning activities) was the second most abundant organic species in both seasons, following C 2 . Such high abundance of Ph suggests that anthropogenic source strength is a significant factor influencing Mumbai aerosols, which is further supported by high C 6 /C 9 ratio.
Good correlations among C 2 (and low molecular weight diacids), K + and levoglucosan supported that biofuel/biomass burning is a potential source of aerosols in Mumbai; however, their concentrations do not show a diurnal pattern. Therefore, the aerosols were not likely to be freshly emitted from local sources. This inference is further supported by high mass fraction ratios of total diacid-C to TC and WSOC, as well as similar WSOC/OC ratios in daytime and nighttime samples. Also, higher mass fraction of total diacid-C relative to WSOC and OC in winter as compared to summer indicates aerosols are more photochemically processed or aged in winter. Further, positive correlation of C 2 with sulfate and ωC 2 especially in winter suggest significant secondary production of C 2 via aqueous phase reactions. Secondary production of diacids was further supported by high C 3 /C 4 ratios obtained in our study.