Molecular composition of dicarboxylic acids, ketocarboxylic acids, alpha-dicarbonyls and fatty acids in atmospheric aerosols from Tanzania, East Africa during wet and dry seasons

Atmospheric aerosol samples of PM 2.5 and PM10 were collected during the wet and dry seasons in 2011 from a rural site in Tanzania and analysed for water-soluble dicarboxylic acids, ketocarboxylic acids, α-dicarbonyls, and fatty acids using a gas chromatography/flame ionization detector (GC/FID) and GC/mass spectrometry. Here we report the molecular composition and sources of diacids and related compounds for wet and dry seasons. Oxalic acid (C2) was found as the most abundant diacid species followed by succinic and/or malonic acids whereas glyoxylic acid and glyoxal were the dominant ketoacid and αdicarbonyl, respectively in both seasons in PM 2.5 and PM10. Mean concentration of C 2 in PM2.5 (121± 47 ng m−3) was lower in wet season than dry season (258 ± 69 ng m−3). Similarly, PM10 samples showed lower concentration of C2 (169± 42 ng m−3) in wet season than dry season (292± 165 ng m−3). Relative abundances of C 2 in total diacids were 65 % and 67 % in PM 2.5 and 65 % and 64 % in PM10 in the wet and dry seasons, respectively. Total concentrations of diacids (289–362 ng m −3), ketoacids (37.8– 53.7 ng m−3), andα-dicarbonyls (5.7–7.8 ng m −3) in Tanzania are higher than those reported at a rural background site in Nylsvley (South Africa) but comparable or lower than those reported from sites in Asia and Europe. Diacids and ketoacids were found to be present mainly in PM 2.5 in both seasons (total α-dicarbonyls in the dry season), suggesting a production of organic acids from pyrogenic sources and photochemical oxidations. Averaged contributions of total diacids to aerosol total carbon were 1.4 % in PM 2.5 and 2.1 % in PM10 during wet season and 3.3 % in PM 2.5 and 3.9 % in PM10 during dry season whereas those to water-soluble organic carbon were 2.2 % and 4.7 % in PM 2.5 during wet season and 3.1 % and 5.8 % in PM 10 during dry season. The higher ratios in dry season suggest an enhanced photochemical oxidation of organic precursors probably via heterogeneous reactions on aerosols under strong solar radiation. Strong positive correlations were found among diacids and related compounds as well as good relations to source tracers in both seasons, suggesting a mixed source from natural biogenic emissions, biomass burning, biofuel combustion, and photochemical production.


Introduction
Low molecular weight dicarboxylic acids, ketoacids and α-dicarbonyls are important components that contribute to water-soluble organic carbon (WSOC) in aerosol particles (Simoneit et al., 2004;H. Wang et al., 2006;Kawamura et al., 2010).Dicarboxylic acids and related compounds have been reported to influence on human health (Highwood and Kinnersley, 2006), hygroscopic property of aerosols (McFiggans et al., 2005), and contribute to the cloud condensation nuclei (CCN) activity (Gierlus et al., 2012).Changes in chemical composition and CCN activity can alter the optical properties of aerosol particles (Reid et al., 1998) and affects cloud microphysical properties and hence precipitation patterns and cloud albedo (Reid et al., 1998;Ramanathan et al., 2001).
Dicarboxylic acids and related compounds are derived from primary sources and secondarily produced from different precursor species via photochemical reactions.Biomass burning (Gao et al., 2003;Falkovich et al., 2005;Kundu et al., 2010a) and fossil fuel combustion (Kawamura and Kaplan, 1987;Ho et al., 2006) are major primary sources.Photochemical-oxidation of biogenic unsaturated fatty acids (Kawamura and Gagosian, 1987), volatile organic compounds (VOCs) from biogenic and anthropogenic emissions, and aromatic hydrocarbons and cyclic olefins (Kawamura et al., 1996a) are important secondary sources.Atmospheric loadings of organic aerosols are significantly influenced by primary emissions from biomass burning and secondary productions (Claeys et al., 2004;G. Wang et al., 2006;Carlton et al., 2006).Recently, emission of isoprene and other biogenic VOCs followed by the subsequent oxidation in the atmosphere has been proposed as an important source of oxalic acid (Myriokefalitakis et al., 2011).Isoprene is the most important biogenic VOC emitted from higher plants (Guenther et al., 2006).
In Tanzania, biomass/wood fuels accounts for major source of energy providing up to 90 % of the total national energy consumption with 8 % from petroleum products and 1.2 % from electricity (URT, 2003).Burning of biofuels such as wood, charcoal, and agricultural waste is the main energy source in this country.Charcoal burning for cooking is very common not only in rural areas but also in cities.On the other hand, dumping of domestic and municipal solid waste into open landfills and the subsequent uncontrolled open burning are common in the country.Therefore, photochemical reactions together with anthropogenic biofuel combustion produce various organic species including dicarboxylic acids and related compounds to form atmospheric particles, leading to the deterioration of the air quality in Tanzania.

Site description
Aerosol sampling was carried out at a typical rural site in Morogoro (06 • 47 40.8 S, 37 • 37 44.5 E, altitude 504 m a.s.l), located about 200 km west of the Indian Ocean, and the city of Dar es Salaam, a business capital in Tanzania.The site lies at the base of Uluguru Mountains, which rise to 2648 m a.s.l. and are characterised as residential area with small-scale agricultural fields (cereal crops), and cattle grazing fields.Like in most other developing countries where poverty is concentrated in rural areas, Tanzania with a population of 42.7 million people (July 2011, estimates) has 25 % of its population living below basic needs poverty line (2008 estimate).Their main sources of fuel for domestic cooking and heating are wood and charcoal.

Aerosol sampling
Aerosol samples of PM 2.5 and PM 10 were collected using low volume samplers (Gent type, flow rate 17.0 L min −1 ) in parallel (Maenhaut et al., 1994) in the 2011 wet and dry seasons.The samplers were placed at a fenced meteorological observatory located at Solomon Mahlangu campus of Sokoine University of Agriculture.Aerosol collection was performed approximately at 2.7 m above the ground level using quartz fibre filters (Pallflex 2500QAT-UP, 47 mm), which were pre-baked at 450 • C for 4 h in a furnace to eliminate adsorbed organics before use.A total of 21 sets of actual samples were collected using each sampler on approximately 24 h basis (exchange of filters was done at 07:30 a.m.).Two field blank sets were collected without pumping air.Before and after sampling the filters were placed in a pre-heated glass vial with a Teflon-lined screw cap and kept frozen at −20 • C during storage.The samples were transported to the atmospheric chemistry laboratory at the Institute of Low Temperature Sciences (ILTS), Hokkaido University (Japan), where the samples were stored at −20 • C prior to analysis.All the analytical procedures were strictly quality-controlled to avoid any possible contamination of the samples.

Chemical analysis
Filter samples were analyzed for water-soluble dicarboxylic acids, ketocarboxylic acids, α-dicarbonyls and fatty acids using the method reported by Kawamura and Ikushima (1993) and Ho et al. (2010).Briefly, a 1.54 cm 2 punch of each quartz fibre filter was extracted three times with 10 mL ultra pure organic-free water (resistivity of > 18.2 M cm) under ultrasonication for 10 min.To remove insoluble particles and filter debris, the extracts were passed through a glass column (Pasteur pipette) packed with quartz wool.The pH of the extracts were adjusted to 8.5-9.0 with 0.1 M KOH (potassium hydroxide) solution, concentrated almost to dryness using a rotary evaporator under vacuum and then derivatized to dibutyl ester (for carboxyl group) and dibutoxy acetals (for keto group) with 14 % boron trifluoride (BF 3 )/n-butanol at 100 • C for 1 h.The derived esters and acetals were dissolved in n-hexane, washed with pure water three times and the extracts were again concentrated using rotary evaporator under vacuum.After nitrogen blow down to near dryness, n-hexane (100 µL) was added and the derivatives were analyzed using a capillary gas chromatograph (GC; HP 6890) installed with a flame ionization detector (FID).Peaks were identified by comparing GC retention time with authentic standards and confirmed by mass spectral examination using a gas chromatography/mass spectrometer (GC/MS).
Recovery experiments were performed by spiking authentic standards to a pre-combusted quartz fibre filter.The recoveries were 81 to 88 % for oxalic acid and more than 92 % for malonic, succinic, glutaric, and adipic acids.Following the same analytical procedure in our laboratory, recoveries for glyoxylic acid, pyruvic acid and methylglyoxal were reported to be 88 %, 72 % and 47 %, respectively (Kawamura and Yasui, 2005).Reproducibility of filter sample was within 4 % for major species.GC chromatograms of field blanks showed small peaks for oxalic, malonic, phthalic, and glyoxylic acids, however, they were less than 1 % (oxalic acid), 3 % (malonic acid), 8 % (phthalic acid) and 5 % (glyoxylic acid) of real samples.All the reported concentrations of diacids and related compounds are corrected for the field blanks.Details for measurement of gravimetric aerosol mass and analyses of total carbon (TC), organic carbon (OC), elemental carbon (EC), water-soluble organic carbon (WSOC), levoglucosan and water-soluble inorganic ions are described elsewhere (Mkoma et al., 2012).

Meteorology and air mass trajectories
Morogoro, where the sampling site locates, experiences a humid tropical savanna climate (TMA, 2011).The ambient temperature during the campaigns varied from 22.9 to 29.1 • C (avg: 26.0 ± 0.8 • C) in the wet season and 21.1-28.2• C (avg.: 24.6 ± 0.7 • C) in the dry season.The site is sensitive to frequent phenomenon of temperature inversion events due to its proximity to the foothill of the Uluguru Mountain ranges (Mkoma et al., 2009).The daily average relative humidity ranged from 65 to 96 % (avg.: 81 ± 0.7 %) in the morning hours and from 41 to 60 % (avg.: 50 ± 0.6 %) in the afternoon.In the wet season campaign (30 May to 13 June), there were only 4 days without rain.In contrast, there were only 2 rainy days in the dry season (28 July to 8 August).However, the entire sampling period was rather dry and few aerosol samples were met with a very weak rain.The prevailing winds during both campaigns were the southeasterly (SE) monsoons with daily average wind speed of 4.1 m s −1 and 12 m s −1 in the wet and dry seasons, respectively.
To find out the possible source regions of air masses at Morogoro during the campaigns, we computed 5-day backward air mass trajectories at an altitude of 500 m for every 24 h using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model from NOAA/ARL (Draxler and Rolph, 2012).The trajectory sectors showed similar transport pathways and source regions in both wet and dry seasons (Fig. 1).The air masses that arrived in Morogoro were mainly from the Indian Ocean over the continents in Madagascar, Mozambique or Tanzania.

Molecular composition and seasonal variations of diacids and related compounds
We determined diacids, ketoacids, α-dicarbonyls, and fatty acids during the wet and dry seasons in PM 2.5 and PM 10 at a rural site in Tanzania (Table 1).The detected compounds were saturated straight chain diacids (C 2 -C 9 ), unsaturated aliphatic (maleic, fumaric, and methylmaleic), brached chain (methylmalonic, iC 4 methylsuccinic, iC 5 ) and aromatic (phthalic, isophthalic, and terephthalic acids) diacids, ketocarboxylic acids (ωC 2 -ωC 9 and pyruvic acid), α-dicarbonyls (glyoxal and methylglyoxal), and fatty acids (C 14 , C 16 , C 18 , C 18:1 , C 20 , C 22 and C 24 ).Longer-chain diacids (C 10 -C 12 ) were below the detection limit (ca.0.005 ng m −3 ) in all the samples.Size distributions and average molecular compositions of diacids, ketoacids, α-dicarbonyls and fatty acids in wet and dry seasons are shown in Figs. 2 and 3. Total concentrations of diacids ranged from 97-547 ng m −3 in PM 2.5 and 165-959 ng m −3 in PM 10 .Oxalic acid (C 2 ) was the most abundant diacid followed by succinic (C 4 ) and/or malonic (C 3 ) acids in two seasons and sizes.Mean concentrations of C 2 in PM 2.5 were 121 ± 47 ng m −3 during the wet season and 258 ± 69 ng m −3 during the dry season whereas those in PM 10 were 169 ± 42 ng m −3 and 292 ± 165 ng m −3 during the wet and dry seasons (Table 1).The relative abundances of individual diacids to total saturated straight-chain (C 2 -C 9 ) diacids in PM 2.5 and PM 10 during the wet and dry seasons are shown in Fig. 4 as pie diagrams.The relative abundances of C 2 in total diacids were 65 % and 67 % in PM 2.5 and 65 % and 64 % in PM 10 in the wet and dry seasons, respectively (Table 2 and Fig. 4).Other studies suggested that higher relative abundances of oxalic acid are caused by enhanced biomass burning activities (Narukawa et al., 1999), photochemical production (Ervens et al., 2004) and ageing of organic acids (Kawamura and Sakaguchi, 1999;Aggarwal and Kawamura, 2008).Predominance of oxalic acid in aerosol samples from background sites in Africa (Limbeck and Puxbaum, 1999;Limbeck et al., 2001) and continental sites affected by biomass burning (Gao et al., 2003;Kundu et al., 2010a) have been reported.
Concentrations of longer-chain diacids (C 6 -C 9 ) varied in different seasons and size fractions with suberic acid (C 8 ) being the most abundant in the wet season in PM 2.5 , adipic acid (C 6 ) and azelaic acid (C 9 ) in the dry season in PM 2.5 and C 9 in both seasons in PM 10 (Fig. 4).Previous studies reported an enhanced production of C 6 from the oxidation of anthropogenic cyclohexene (Kawamura and Ikushima, 1993) and of C 9 from photochemical oxidation of biogenic unsaturated fatty acids (Kawamura and Gagosian, 1987).Phthalic acid (Ph) and terephthalic acid (tPh) had high abundances in PM 2.5 during the wet season and in PM 10 during the dry season.The observed high abundances of these aro-matic diacids suggest anthropogenic effect from combustion sources (Kawamura and Kaplan, 1987) and/or atmospheric photochemical degradation of polycyclic aromatic hydrocarbons (Kawamura and Ikushima, 1993).In Tanzania, dumping of municipal solid waste (large amounts of plastics) into open landfills is very common and 60 % of daily domestic solid waste are disposed and subjected to open burning (Kassim, 2006).Plastic burning under open-fire conditions and local anthropogenic emissions in both seasons should be responsible for these aromatic acids (Yassaa et al., 2001;Simoneit et al., 2005;Kawamura and Pavuluri, 2010).
Mean concentrations of total α-dicarbonyls in PM 2.5 were higher in wet season (6.4 ± 3.0 ng m −3 ) than dry season (4.4 ± 1.7 ng m −3 ) whereas those in PM 10 were equivalent (7.3 ± 1.9 ng m −3 in wet season and 8.0 ± 3.5 ng m −3 in dry seasons) (Table 1).In both seasons and size fractions, glyoxal (Gly) was found to be more abundant than methylglyoxal (Table 2).Gly may be produced by the photooxidation of anthropogenic hydrocarbons (Ho et al., 2006) and p-xylene (Volkamer et al., 2001).In contrast to diacids and ketoacids, lower concentrations of α-dicarbonyls in PM 2.5 were obtained during the dry season, which may be caused by photochemical degradation under stronger solar radiation in dry season.The aldehyde group in dicarbonyls is highly susceptible for nucleophilic addition of oxygen forming carboxylic acid (Ervens et al., 2004;Fick et al., 2004).Thus, α-dicarbonyls may be in part converted to organic acids such as glyoxylic acid and oxalic acid in aerosol phase and the carbonyl concentrations may be decreased in dry season.
Homologous series of straight chain fatty acids (C 14:0 -C 24:0 ) and unsaturated fatty acid (C 18:1 ) were detected (Table 1).Mean concentrations of total fatty acids in PM 2.5 were 26 ± 19 ng m −3 and 32 ± 15 ng m −3 during the wet and dry seasons, respectively, whereas those in PM 10 were 102 ± 44 ng m −3 and 117 ± 72 ng m −3 .Average molecular compositions of fatty acids and their size distributions in the  (Lechevalier, 1977;Simoneit, 1988).Palmitic (C 16:0 ) and stearic (C 18:0 ) acids showed higher concentrations in wet season in both sizes probably due to enhanced emissions from vegetation sources.In contrast, longer-chain fatty acids, behenic (C 22 ) and lignoceric (C 24 ), which are specific to terrestrial higher plants (Kawamura et al., 2003), were not detected in PM 2.5 , but were abundantly detected in PM 10 .They showed higher concentrations during the dry season, suggesting a long-range atmospheric transport of lipid compounds.Higher concentration of azelaic acid (C 9 ) and higher ratios of C 9 to oleic acid (C 18:1 ) were found in the dry season for both sizes.Mean C 9 /C 18:1 ratios in PM 2.5 were higher in dry season (4.5) than wet season (0.63).Similar trend was observed in PM 10 : 8.6 (dry season) and 1.5 (wet season).Photochemical oxidation of C 18:1 to C 9 via ozone and/or OH radicals (directly or via its intermediate precursor ωC 9 ) is likely in the atmosphere during the long-range transport (Kawamura and Gagosian, 1987;Stephanou and Stratigakis, 1993).The highest C 9 /C 18:1 ratio in the dry season for PM 2.5 and PM 10 may be due to an enhanced photochemical degradation of C 18:1 .On other hand, mean C 18:1 /C 18:0 ratios in PM 2.5 were 1.2 and 1.9 in the wet and dry seasons, respectively whereas those in PM 10 were 1.1 and 1.6, respectively.The lower mean C 18:1 /C 18:0 ratios in the wet season compared to dry season in both sizes may suggest that photo-chemical degradation of C 18:1 may be accelerated in aqueous aerosols under wet conditions.more production and/or accumulation of the diacids in dry season.The C 2 /total diacids ratios in PM 2.5 were 0.65 ± 0.06 (range: 0.49-0.72) in wet season and 0.67 ± 0.04 (range: 0.59-0.72) in dry season whereas those in PM 10 were 0.65 ± 0.05 (range: 0.59-0.76) in wet season and 0.64 ± 0.04 (range: 0.54-0.69) in dry season.

PM 2.5 to PM 10 ratios
Mean PM 2.5 /PM 10 ratios for total diacids, ketoacids, αdicarbonyls and fatty acids in aerosol samples are shown in Fig. 7 for the wet and dry seasons.The ratios were calculated on the basis of the data for PM 2.5 and PM 10 samples taken in parallel and averaged overall samples from the campaign.The results indicate that total diacids, ketoacids and α-dicarbonyls were mostly present in PM 2.5 in both seasons (except for α-dicarbonyls in the dry season).This suggests a larger contribution of pyrogenically and photochemically produced organic acids and related compounds, which are most likely enriched in PM 2.5 particles.Strong correlation was found between Ca 2+ (crustal tracer) with the PM 10 mass (r 2 = 0.81) and with α-dicarbonyls (r 2 = 0.33) in the dry season (in contrast r 2 = 0.17 and r 2 = 0.08 in the wet season, respectively).Mean PM 2.5 to PM 10 ratios for total diacids and related compound are larger than 60 % and larger ratios above 80 % were obtained for total diacids and total ketoacids in the dry season and total α-dicarbonyls in the wet season (Fig. 7).Other studies have reported associ-  ation of diacids and related compounds with fine particles (Narukawa et al., 2003;Kawamura et al., 2007;Wang et al., 2012).Interestingly α-dicarbonyls showed a significantly high PM 2.5 /PM 10 ratio (ca.85 %) in wet season.Because glyoxal and methylglyoxal that are mostly present as gas in the atmosphere can form hydrated forms in the presence of moisture, it is reasonable that α-dicarbonyls are detected as PM 2.5 particles during wet season.
In contrast, fatty acids were found mostly in PM 10 in both seasons (Fig. 7).Because fatty acids with carbon number > 14, in particular C 16 , are known to be mostly in particulate phase (Cheng et al., 2004), they are likely associated with PM 10 .This result suggests that fatty acids at our site are from mixed sources (Alfarra et al., 2004), which are derived from primary sources such as viable microbiota in the ambient particles, marine biological activity and/or terrestrial higher plants.Fatty acids in ambient aerosols from rural, urban and oceanic regions have been reported to show the predominance of C 16 acid (Simoneit et al., 1988;Limbeck and Puxbaum, 1999;Cheng et al., 2004).

Seasonal contributions of diacids, ketoacids, and α-dicarbonyls to the PM mass, TC, and WSOC
Temporal variations in the contributions of total diacids to the aerosol mass, total carbon (TC), and water-soluble organic carbon (WSOC) in PM 2.5 and PM 10 during the wet and dry seasons are given in Fig. 8a-c.We generally found that the mean ratios are higher in the dry season than in the wet season for both size fractions.The mean contributions of total diacids to PM 2.5 mass were 0.65 % (range: 0.38-0.89%) in the wet season and 1.04 % (range: 0.57-1.39%) in the dry season whereas those to PM 10 mass were 0.97 % (range: 0.50-1.52%) in the wet season and 1.20 % (range: 0.59-2.47%) in the dry season.Total diacid-C/TC ratios ranged from 0.73 to 5.0 % (mean: 2.4 ± 0.70 %) in PM 2.5 and 1.3 to 3.2 % (mean: 3.0 ± 1.4 %) in PM 10 .The averaged ratios of 1.4 % in PM 2.5 and 2.1 % in PM 10 in the wet season are twice lower than 3.3 % in PM 2.5 and 3.9 % in PM 10 in the dry season (Fig. 8b).The higher ratios (e.g., diacid-C/TC) in PM 10 than in PM 2.5 may suggest that diacids are more enriched in coarse fraction due to the heterogeneous production on the coarse aerosol particles.The mean contributions of total diacids to TC at our site (2.4 % in PM 2.5 and 3.0 % in PM 10 ) are much higher than those reported in Sapporo (1.8 %) (Aggarwal and Kawamura, 2008) and in Tokyo (0.95 %) (Kawamura and Ikushima, 1993), in Chennai, India (1.6 %) (Pavuluri et al., 2010) and in Mongolia (0.60 %) (Jung et al., 2010).These comparisons may suggest that photochemical production of diacids is more significant in tropical Tanzania than in mid latitudes    via gas-to-particle conversion of precursor organics as well as heterogeneous reactions on aerosols under the conditions of stronger solar and higher humidity.However, the ratios at our site are lower than that (8.8 %) reported in remote marine aerosols including the tropical Pacific (Kawamura and Sakaguchi, 1999), where photochemical processes are more enhanced during long-range atmospheric transport.Contributions of total diacids to WSOC in PM 2.5 during the wet and dry seasons were 2.2 % (range: 1.1-3.0%) and 4.7 % (range: 2.1-6.9 %), respectively, whereas those in PM 10 were 3.1 % (range: 2.0-5.8%) during the wet season and 5.8 % (range: 2.7-14.3%) during the dry season.Higher ratios in the dry season further support enhanced pho-  tochemical oxidations of organic precursors and production of water-soluble diacids in the dry season.In fact the contributions of total diacids to WSOC in the dry season were 2 times higher than those for wet season in both sizes; 6.5 % and 14.4 % in PM 2.5 and 9.3 % and 17.4 % in PM 10 for wet and dry seasons, respectively.These results again demonstrate that photochemical production of diacids is more significant in dry season than wet season.
On the other hand, mean contributions of total ketoacids to TC in the wet and dry season were 0.29 % and 0.65 % in PM 2.5 and 0.54 % and 0.81 % in PM 10 , respectively.Their contributions to WSOC were 0.43 % and 0.95 % in PM 2.5 and 0.86 % and 1.15 % in PM 10 , respectively.These results are consistent with the trend obtained for diacids as described above.Higher contributions of total ketoacids to TC and WSOC in the dry season suggest an enhanced photochemical production of organic acids and photochemical aging of organic aerosols in rural Tanzania under higher solar radiation condition.In contrast, total α-dicarbonyls show similar contributions to TC (0.06 % and 0.05 % in PM 2.5 and 0.08 % and 0.09 % in PM 10 ) and to WSOC (0.09 % and 0.07 % in PM 2.5 and 0.11 % and 0.13 % in PM 10 ) during both seasons.

Comparison of molecular composition of diacids and related compounds with other studies
Table 3 compares mean concentrations of diacids, ketoacids, and α-dicarbonyls in Tanzania with those reported from other sites in Africa, Asia and Europe.Concentrations of total diacids in Morogoro (289-362 ng m −3 ) are lower than those reported from Sapporo (390 ng m −3 ), Tokyo (446 ng m −3 ), Chennai (588 ng m −3 ), Hong Kong (671 ng m −3 ) and Jeju Island (648 ng m −3 ) but comparable to that (359 ng m −3 ) from Nainital and higher than those from Nylsvley (158.1 ng m −3 ) and Salzburg (61.7 ng m −3 ).Predominance of oxalic acid (C 2 ) found at our site is consistent with other studies (Table 3) except for Salzburg where malonic acid (C 3 ) was reported as the most abundant diacid.Oxalic acid is the end product of the photooxidation of aromatic hydrocarbons, isoprene, ethylene, and acetylene (Kawamura et al., 1996a;Lim et al., 2005) and may be emitted from biomass burning (Legrand and de Angelis, 1996;Kundu et al., 2010b).Concentration of C 2 in Morogoro (PM 10 ) is several times higher than that of Nylsvley in South Africa, but comparable to those from Tokyo and Nainital (Table 3).Concentrations of C 3 and C 4 are comparable to those from Nainital, but lower than Chennai, Hong Kong and Jeju Is.Concentrations of C 6 and C 9 are comparable to those from megacities such as Tokyo, Chennai and Hong Kong, but lower than those from Sapporo and Jeju Island.Adipic acid (C 6 ) and azelaic acid (C 9 ) are tracers for anthropogenic and biogenic emissions, respectively (Kawamura and Ikushima, 1993).On the other hand, predominance of fumaric acid over maleic acid at our site, in contrast to literature values in Table 3, suggests that the Morogoro aerosols contain aged oxidation products of aromatic hydrocarbons emitted from regional pollution sources.Concentrations of phthalic (Ph) acid (mean: 13 ng m −3 ), a tracer for vehicle emissions (Kawamura and Ikushima, 1993) are about 2 folds lower than that from Chennai (mean: 21 ng m −3 ) and 7 folds lower than that from Hong Kong (mean: 84 ng m −3 ).We found that concentration of tPh acid in PM 2.5 is similar to that reported from Sapporo (mean: 2.6 ng m −3 , Aggarwal and Kawamura, 2008) but is lower than those from Chennai (mean: 52 ng m −3 , Pavuruli et al., 2010) and Nainital (mean: 4.3 ng m −3 , Hegde and Kawamura, 2012).tPh could be produced from open burning of solid waste (plastic) (Simoneit et al., 2005;Kawamura and Pavuluri, 2010), which occurs commonly in Tanzania.Concentrations of total ketoacids in Morogoro (mean: 38-54 ng m −3 ) with a predominance of ωC 2 are comparable to those from other sites whereas concentrations of total α-dicarbonyls (mean: 5.7-7.8ng m −3 ) are lower than those reported in the literature (Table 3).α-Dicarbonyls may be more oxidized to result in oxalic acid (end product) in the rural site of Tanzania.The C 3 /C 4 ratio has been used to understand the photochemical processes and atmospheric production of diacids in the atmosphere (Kawamura and Ikushima, 1993;Aggarwal and Kawamura, 2008;Kundu et al., 2010b).At Morogoro the average C 3 /C 4 ratios were 0.72 and 0.81 in PM 2.5 and 1.3 and 1.0 in PM 10 during the wet and dry seasons, respectively.The higher C 3 /C 4 ratio in PM 10 than in PM 2.5 during wet season (see Table 1) may suggest that atmospheric oxidation of C 4 to C 3 is more significant under wetter conditions where heterogeneous processing may occur in aqueous aerosol phase of PM 10 .Although the differences in the ratios are not significant, wetter conditions under strong solar radiation may enhance the photochemical aging of organic aerosols in the tropical atmosphere.Based on the filed experiments, aqueous aerosol phase production of oxalic acid was reported in urban and suburban sites (Miyazaki et al., 2009;He and Kawamura, 2010).However, when compared to literature values, our C 3 /C 4 ratios are slightly lower than those (1.5) reported in Tokyo (Kawamura and Ikushima, 1993) and 1.4 in Chennai (Pavuluri et al., 2010) but comparable to those (0.84) in Nainital, India (Hegde and Kawamura, 2012) and those (1.3) in Jeju Island in the East China Sea (Kawamura et al., 2004).

Source identification of diacids, ketoacids and α-dicarbonyls
Correlation matrix for diacids and related compounds in PM 2.5 and PM 10 during the 2011 wet and dry seasons are given in Tables 4 and 5 Levoglucosan (1,6-Anhydro-β-D-glucopyranose), watersoluble K + and organic carbon (OC) are used as source tracers for biomass burning and elemental carbon (EC) for biofuel combustion (Cachier et al., 1991(Cachier et al., , 1995;;Andreae and Merlet, 2001;Puxbaum et al., 2007;Zhang et al., 2008).Table 6 shows correlation coefficients for selected diacids and related compounds against those source tracers in PM 2.5 and PM 10 during the wet and dry seasons.We found strong correlations between the source tracers and diacids (C 2 −C 4 ), iC 5 and ωC 2 (in both season), iC 4 , F, Pyr and MeGly (in wet season) and Ph (in dry season) in both PM 2.5 and PM 10 .The source tracers also strongly correlated with Gly in PM 2.5 and ωC 3 in PM 10 during both seasons, F in PM 2.5 during the wet season, M, F, mM, Gly and MeGly in PM 10 during the dry season.These correlations suggest that the diacids, ketoacids and α-dicarbonyls are partly produced from biomass and biofuel burning in both seasons.Previous studies have reported a good correlation between biomass/biofuel burning tracers (K + and EC) and diacids and related compounds (Graham et al., 2002;Kundu et al., 2010a).
Mean ratios of C 2 to C 4 and C 5 at our site can be compared with the ratios in biomass burning aerosols.The C 2 /C 4 ratios in PM 2.5 were 7.0-8.3whereas C 2 /C 5 were 8.8-56.3during both seasons.Those values from this study are much higher than those (C 2 /C 4 : 1.6 and C 2 /C 5 : 2.5) reported in aerosols associated with savannah fires in Southern Africa (Gao et al., 2003).These comparisons suggest a secondary formation of C 2 from C 4 and C 5 diacids.Concentration ratios of C 2 and C 4 to non-sea-salt potassium (nss-K + ) and levoglucosan (LG) were generally higher (C 2 /K + ; 0.15-0.31,C 2 /LG; 0.81−1.02,C 4 /K + ; 0.02-0.04,and C 4 /LG; 0.10-0.14)than those (C 2 /K + ; 0.05, C 2 /LG; 0.05, C 4 /K + ; 0.03, and C 4 /LG; 0.03) reported for the smoke aerosols collected in southern Africa savannah fires (Gao et al., 2003).These comparisons suggest that, in addition to the emission of C 2 and C 4 diacids from biomass burning (Gao et al., 2003;Sillanpää et al., 2005), photochemical production of these diacids is significantly important in the Tanzanian aerosols.
We assessed the relations between total diacids with Na + , EC, nss-K + and LG in PM 2.5 and PM 10 during the wet and dry seasons.Although the air masses often originated from the Indian Ocean during the campaigns, total diacids were poorly correlated with Na + (tracer for sea-salt) in both seasons (Fig. 9a, b), suggesting that contributions of diacids from sea-salt aerosols are insignificant.Biomass burning has been reported to be an important source of diacids and related compounds (Kundu et al., 2010a).Strong correlations were found between total diacids with EC (Fig. 9c, d) and nss-K + and LG especially in the dry season (Fig. 10a-d).These relations suggest that biofuel combustion and biomass burning Fig. 9  significantly contribute to the water-soluble organic species in the aerosols from Tanzania.On the other hand, nss-K + showed no correlation with total diacids in PM 2.5 during the wet season (Fig. 10a), although a weak correlation was found for PM 10 (Fig. 10b).However, levoglucosan showed a positive relation with total diacids in both wet and dry season (Fig. 10c, d).The above apparent discrepancy may suggest that nss-K + can only be produced by flaming process during dry season but not by smoldering process during wet season whereas levoglucosan can be produced by both modes of the burning processes to result in aerosol particles.

Summary and conclusions
We determined low molecular weight diacids, ketoacids and α-dicarbonyls, and fatty acids in atmospheric aerosol samples collected from a rural site in Tanzania during the wet and dry seasons.The results on both PM 2.5 and PM 10 showed that oxalic acid (C 2 ) was the dominant diacid species whereas glyoxylic acid (ωC 2 ) and glyoxal (Gly) were the most abundant ketoacid and α-dicarbonyl, respectively.Fatty acids with even carbon number were detected with a peak at myristic acid (C 14:0 ) in PM 2.5 and palmitic acid (C 16:0 ) in PM 10 in both seasons, indicating significant influences from biological sources.The mass concentrations of diacids, ketoacids and fatty acids were higher during the dry season than the wet season.They are mainly present in fine particles (PM 2.5 ), suggesting a larger contribution of pyrogenically produced organic acids and enhanced photochemical processing in the dry season due to strong solar radiation.Relatively high ratios of diacid-C/TC and diacid-C/WSOC further indicate a strong influence of photochemical oxidation of organic precursors in the atmosphere.Nss-K + (biomass-burning tracer) showed different regression patterns with total diacids in PM 2.5 during the wet and dry seasons, but levoglucosan (another biomass-burning tracer) showed a positive correlation with total diacids in both seasons.The difference between the two tracers may be caused by the different behaviour during biomass burning processes; K + may be produced only by flaming process whereas both flaming and smoldering processes produce levoglucosan.Strong correlations between organic components and source tracers in PM 2.5 and PM 10 during both wet and dry seasons suggest common sources (e.g., biomass and biofuel burning) and/or similar formation pathways for water-soluble dicarboxylic acids in the rural site of Tanzania, East Africa.

FigureDiacidsFig. 4 .
Figure 5a-e show temporal variations of total aerosol mass, diacids, ketoacids, α-dicarbonyls and fatty acids in PM 2.5 and PM 10 during the wet and dry seasons.Temporal variations of selected diacids (C 2 -C 9 ) are given in Fig. 6.Concentrations of C 2 , C 3 and C 4 diacids showed similar temporal variations with higher concentrations during dry season compared to wet season in both sizes (Fig. 6a-c), suggesting

Fig. 9 .
Fig. 9. Correlation between total diacids with Na + and EC in PM 2.5 (a, c) and PM 10 (b, d) at Morogoro during the wet and dry seasons.

Table 1 .
Average concentrations and concentration ranges (ng m −3 ) of diacids, ketoacids, α-dicarbonyls and fatty acids in PM 2.5 and PM 10 during 2011 wet and dry seasons in Morogoro.
Table 1 for abbreviations.

Table 2 .
Relative abundances (%) of individual compound in total diacids, ketoacids, and α-dicarbonyls in PM 2.5 and PM 10 during wet and dry seasons in Morogoro.

Table 3 .
Comparison between average concentrations of diacids, ketoacids, and α-dicarbonyls in atmospheric aerosols from Morogoro, Tanzania and different sites around the world.

Table 4 .
Correlation coefficients (r 2 ) for selected diacids and related compounds in PM 2.5 during wet season (upper diagonal triangle) and dry season (lower diagonal triangle) at Morogoro.Positive correlation coefficients ≥ 0.55 are indicated in bold.

Table 5 .
Correlation coefficients (r 2 ) for selected diacids and related compounds in PM 10 during wet season (upper diagonal triangle) and dry season (lower diagonal triangle) at Morogoro.Positive correlation coefficients ≥ 0.55 are indicated in bold.
, respectively.Many combinations of different compounds exhibit strong correlations in both seasons and sizes, suggesting common sources and/or similar formation mechanisms.In PM 2.5 , C 2 showed a strong positive correlation with its precursors C 3 , C 4 , iC 5 and ωC 2 in both seasons, with iC 4 , Pyr, and MeGly in the wet season, and with Ph, ωC 3 , and Gly in the dry season.Glyoxylic acid (ωC 2 ) correlated with Pyr in the wet season and with Gly in the dry season.These correlations suggest a possible production of C 2 from aromatic hydrocarbons because Pyr and Gly may be oxidation products of aromatics.In PM 10 , C 2 correlated with its precursor compounds C 3 , C 4 , iC 5 , ωC 2 and ωC 3 in both seasons, with ωC 4 and MeGly in wet season, and with Pyr and Gly in dry season.ωC 2 correlated with C 2 , C 3 , C 4 and iC 5 in both seasons, with Pyr in the wet season, and with Gly and MeGly in the dry season.These correlations further suggest that C 2 may be formed through chain reactions of other diacids and related compounds.