Understanding aerosol composition in an inter-Andean valley

14 Agro-industrial areas are frequently affected by various sources of atmospheric pollutants that negatively impact public health 15 and ecosystems. However, air quality in these areas is infrequently monitored because of their lower population density 16 compared to large cities, especially in developing countries. The Cauca River Valley (CRV) is an agro-industrial region in 17 Southwest Colombia, where a large fraction of the area is devoted to sugarcane and derivatives production. CRV is also 18 affected by road traffic and industrial emissions. This study aims to elucidate the chemical composition of particulate matter 19 fine mode (PM2.5) and to identify the main pollutant sources before source attribution. For this, a sampling campaign was 20 carried out at a representative site of the CRV region, where daily-averaged mass concentrations of PM2.5 and the 21 concentrations of water-soluble ions, trace metals, organic and elemental carbon, and various fractions of organic compounds 22 (carbohydrates, n-alkanes, and polycyclic aromatic hydrocarbons – PAHs) were measured. Mean PM2.5 was 14.38 ± 4.35 g 23 m, and the most abundant constituent was organic material (52.99% ± 17.79%), followed by ammonium sulfate (16.12% ± 24 3.98%), and elemental carbon (6.95% ± 2.52%), which indicates secondary aerosol formation and incomplete combustion. 25 Levoglucosan was present in all samples with a mean concentration of (113.8 ± 147.2 ng m) revealing biomass burning as a 26 persistent source. The diagnostic ratios applied to organic compounds revealed the influence of petrogenic and pyrogenic 27 sources. Principal component analysis identified the influence of traffic-generated road dust, secondary aerosol formation, 28 gasoline and diesel combustion vehicle exhaust, vegetative detritus, and resuspended agriculture soil. However, no single 29 component was dominant nor explained the CRV PM2.5 chemical species variance. Many components had equally important 30 roles instead. Likewise, sugarcane pre-harvest burning, a frequent activity in CRV, was not identified as an independent 31 https://doi.org/10.5194/acp-2021-601 Preprint. Discussion started: 3 August 2021 c © Author(s) 2021. CC BY 4.0 License.

Buenaventura on the Pacific Ocean is one of the busiest ports in Colombia, thus significant diesel combustion emissions occur 143 along the Buenaventura highway. 144

2.2.
Sampling protocols 145 The sampling campaign was conducted between 25 th July and 19 th September 2018. PM2.5 aerosol particles (aerodynamic 146 diameter <2.5 m) were simultaneously collected on Teflon and quartz fiber filters for 23 h (from 12:00 local time -LTto 147 the following day at 11:00 LT), using 2 in-tandem low-volume samplers (ChemComb speciation samplers, R&P). Each 148 sampler used an independent pump set at a flowrate of 14 L min -1 . Quartz filters were pre-baked at 600 °C for 8 h before 149 sampling to eliminate contaminant trace hydrocarbons. In total, 45 samples were collected. Prior to and after exposure, the 150 filters were conditioned at constant humidity (36±1.5% relative humidity) and temperature (24 ± 1.2 ºC) for 24 h before 151 weighing them on a microbalance (Sartorius, Mettler Toledo) with 199.99 g capacity and 10 g resolution. Particulate matter 152 loaded filters were stored at -20°C until analysis. Mass concentrations were determined from the Teflon filters by differential 153 weighting. It is worth mentioning that 1888 sugarcane pre-harvest burning events took place during the sampling period. The 154 vast majority of these events were intentional, controlled, size-limited (~6 ha median area), and short (~25 minutes median 155 duration) (Fig S1). 156

2.4.
Mass closure and diagnostic ratios 191 PM2.5 main components were estimated from the concentrations of EC, OC, water-soluble ions (NO3 -, SO4 2-, NH4 + and Na + ) 192 and tracer metal concentrations (Ca, Ti, Fe, Ni, Cu, Zn, As, Se, Sb, Ba and Pb). The main components considered were organic 193 material (OM), elemental carbon (EC), ammonium sulfate ((NH4)2SO4), ammonium nitrate (NH4NO3), crustal material (dust), 194 other trace elements oxides (TEOs), particle-bounded water (PBW), and sea salt (SS), reckoned as sodium chloride. PM2.5 195 closure is described by Eq 1 (Dabek-Zlotorzynska et al., 2011). Except for EC, these components were not directly determined 196 by chemical analysis but calculated from measured species. For these, we used the Interagency Monitoring of Protected Visual 197 Environment (IMPROVE) equations (Chow et al., 2015). See Table 1. Also, this reconstruction was instrumental towards the 198 identification of the main fine airborne particle sources. 199 The aerosol particle bounded water content was estimated from measured ionic composition, relative humidity, and 200 temperature following the aerosol inorganic model (AIM) described by (Clegg et al., 1998) (Chow et al., 2015), which depend on the OM 225 oxidation level and the secondary organic aerosol formation and aging during transport. Turpin and Lim, (2001a) 226 recommended a ratio of 1.6 and 2.1 for urban and non-urban areas, respectively. However, biomass burning aerosols can have 227 an even higher f values (2.2-2.6), due to the presence of organic components with higher molecular weight, e.g., levoglucosan. 228 We believe that traffic is the dominant OCprim source at out site, therefore used an f1 = 1.6 to estimate OMpri. 229

230
We used a factor of 2.2 to estimate OMsec from OCsec fraction. This factor was chosen based on i) recommended ratios of 231 2.1±0.2 for aged or non-urban aerosols and ii) the molecular weight to carbon weight ratio for levoglucosan of 2.2. 232 Levoglucosan is taken as component of reference due to its abundance in samples collected where the biomass burning happens 233 often and as shown in section 3.6, levoglucosan was a tracer present in whole samples collected in this study (Schauer, 1998). 234 235 Concentration ratios among distinct species were used to chemically characterize and infer the main sources of fine particle 236 matter at Palmira. PM2.5 acidity was assessed through cation and anion charge balances and then by comparison of cation 237 equivalent (CE) and anion equivalent (AE) concentrations (Eq (4) and Eq (5)). Parent PAH ratios are widely used to identify     There is very little information in the literature on the composition of several of the aerosol emission sources deemed important 259 in CRV. This is particularly true for sugarcane pre-harvest burning and sugarcane bagasse combustion. Because of this, instead 260 of directly jumping into a source attribution effort, using receptor modeling methods, we deemed it more important at this stage of our research to apply multivariate statistical techniques to unravel correlations among the various aerosol components, 262 and to potentially identify various aerosol sources. For this, we applied principal component analysis (PCA). We consider this 263 useful in our case, even if PCA is nowadays considered an outdated technique for source attribution in regions with reasonably 264 characterized sources (Hopke, 2016). The species Br -, C19H40, COR, and manosan were excluded from these analyses because 265 more than 80% of their concentrations were below the detection limit (BDL). Data were organized into a matrix of 45 PM2.5 266 samples (rows) times 73 chemical species (columns). BDL "missing" values were replaced by corresponding species detection 267 limit. To reduce skewness and order of magnitude effects, the concentration dataset was log10-transformed, mean-centered, 268 and scaled to unit variance. Principal components were derived from the correlation matrix. We applied varimax rotation PCA 269 as rotated components have easier-to-interpret loadings. Principal components (PC) were selected to explain at least 60% of 270 the total variance. Calculations were made with the Psych (2.0.9) R package. Niña conditions. This synoptic feature is one the most important determinants of atmospheric circulation in Colombia, with 280 prevailing east-to-west winds in the lower troposphere along with upper troposphere return winds (Mesa S. and Rojo H., 2020). 281 The Andean Cordilleras are nevertheless effective barriers to the Walker circulation near the CRV surface (Lopez and Howell, 282 1967). The elevated humidity in the Pacific Ocean watershed and the closeness of the two Andes branches drive a zonal 283 regional circulation pattern, consisting in west-to-east anabatic winds over the Pacific slope of the Western Cordillera during 284 daytime followed by rapid katabatic winds late afternoon (Lopez and Howell, 1967). These winds rapidly ventilate CRV during 285 the late afternoonearly evening period on an almost regular basis. CRV is wide (~22 km) and long (~248 km) enough to 286 develop a valley-mountain wind circulation pattern during daytime. Winds are very mild during this time period and expectedly 287 highly dispersive, i.e. with high turbulence intensities (Ortiz et al., 2019). The arrival of the katabatic "tide" at late afternoon 288 wipes the valley-mountain wind pattern out. 289 One year prior to the sampling period, we monitored the local meteorology, first at 14.5 m, a few meters over the mean canopy to ~8 m/s at the sampling site elevation, peaking at ~17:00 LT. Wind speeds were a factor ~2-3 slower at ground level. The 292 wind runs at the sampling height were typically over ~200 km per day ( Fig S3) indicating that the samples had quite a large spatial coverage of CRV, much larger than it would have been at ground level. This also implies that the samples were 294 frequently and significantly influenced by emissions coming from Yumbo's industrial hub (northwest of Palmira), and also by 295 Palmira and Yumbo urban and highway emissions, along with pre-harvest sugarcane burning and sugarcane mill emissions. 296 The wind rose (Fig 2a) suggests that the influence of urban emissions from Cali, CRV's largest city by far, was minor. Other 297 meteorological variables are reported in the Supplementary Material (SM) (Fig S2). Temperature (24.2ºC on average) and 298 relative humidity (71.6%) were very likely controlled by solar radiation (350 W m -2 on average). The pressure daily profile 299 (~763 hPa on average) clearly showed the influence of the katabatic tide, with a ~3 hPa drop during its arrival at late afternoon. 300 Overall, we believe our measurements at the Palmira site are quite representative of the regional air quality. 0.62 g m -3 , and 0.51 ± 0.30 g m -3 , respectively (12.7 ± 2.8%, 3.7 ± 1.1% and 2.6 ± 1.3% of mass concentration, respectively). 327 Mean concentrations of other water-soluble ions, such as Na + , Ca + , and C2O4 2-, were around 0.1 g m -3 , while those of K + , 328 PO4 3-, CH3O3S -, Mg 2+ , and Clranged within 10-80 ng m -3 (Table 3). 329

330
The predominant elements were Ca (0.42 ±0.33 g m -3 ), K (0.13 ± 0.08 g m -3 ), and Fe (88 ± 65 ng m -3 ), followed by Zn (34 331 ± 33 ng m -3 ), Pb (18 ± 19 ng m -3 ), Sn (52 ± 37 ng m -3 ), Ti (5 ± 4 ng m -3 ), Ba (9 ± 13 ng m -3 ), Sr (2 ± 5 ng m -3 ). Mn, Ni, Cr, and 332 Se concentrations were below 2 ± 1 ng m -3 . Tracer metals such as Ti, Cr, Mn, K, Ca, Fe, Ni, Cu, Zn Sr, Pb and Se were found 333 in all PM2.5 samples, while V was not found in any sample. Other tracer metals such as As and Sb were detected only at a 334 reduced number of samples with concentrations below 20 ng m -3 . Table 3   The mass closure (Figure 3) shows the crucial contribution of organic material (52.99% ± 17.79%) and the secondary inorganic 354 fraction, represented by ammoniated sulphate (16.12 ± 3.98%) and ammonium nitrate (3.19 ±1.71%). EC constituted 6.95 ± 355 2.52% of PM2.5. The mineral fraction corresponded to dust (8.67 ± 5.71%) and TEO (0.82 ± 0.44%). The sea salt was 0.80 ± 356 1.28 % and PBW 5.20 ± 1.20%. A mass closure of 93.40 ± 33.38% was achieved. Although the PM2. The mineral fraction, quantified as the sum of the oxides present in the crustal material (dust) and other trace element oxides 371 (TEO) contributed 9.1 ± 5.5% and 0.9 ± 0.4%, respectively. Despite the non-quantification of highly abundant mineral dust 372 elements such as Si, the concentrations of Ca, Ti, and Fe indicated the impact of soil resuspension on the PM2.5 mass 373 concentration. 374 375 Particle-bound water (PBW) depends on the concentration of hygroscopic compounds embodied in the particulate matter and 376 relative humidity of the weighing room where the PM2.5 mass collected on the filters was determined. In this study, it was 377 assumed that (i) NH4 + , SO4 2and NO3were the main compounds responsible for the absorbed water and (ii) thermodynamic 378 equilibrium is dominated by these ions that allow calculating the H + molar fraction as a difference of (SO4 2-+ NO3 -) and NH4 + 379 required to establish the charge neutrality. Polar organic compounds and other water-soluble ions were not considered in the 380 present study. The PBW content was estimated using the mean measured concentrations of NH4 + , SO4 2and NO3in the AIM 381 Model, where a multiplier factor was found equivalent to 0.32 as a proportion between the concentrations of summatory of 382 theses ions and the water fraction contained in the PM2.5. As a result, the PBW was 5.3% of PM2.5 mass concentration. 383

Ions 387
Anion-and cation-equivalent (AE and CE, respectively) charges were compared to estimate the acidity of PM2.5 (Figure 4). 388 AE and CE displayed a tight Spearman linear correlation (r 2 =0.99). The AE to CE ratio of 1.2 ± 0.1 suggests that cations were 389 generally well balanced by anions and that PM2.5 was nearly neutral. Just a few samples displayed AE/CE ratios significantly 390 higher than 1, i.e. slightly acidic, which might be attributed to the sulfate dianion (SO4 2-) abundance. The ratio between the 391 two main water-soluble ions, ammonium cation (NH4 + ) and  including Palmira. They found significant enrichment of Fe and K metals at locations exposed to PHB. It must be bear in mind 425 that PM10 samples included coarse mode aerosols, of which dust might have been a significant fraction. Also at Palmira is much closer to the chemical profile ratio of leaves than that of bagasse. Moreover, ambient air samples in 459 Araraquara and Piracicaba showed levoglucosan/mannosan ratios of 9 ± 5 and ~33, respectively. For comparison, the 460 levoglucosan/mannosan ratio in particulate matter from rice straw and other crops burning were ~26.6 and~23.8, respectively 461 (Engling et al., 2009). This indicates that the levoglucosan/manossan ratio is sensitive to the type of biomass burned but also to burning conditions. The large levoglucosan/mannosan ratio variability in our study suggest that Palmira was impacted by 463 sugarcane pre-harvest burning most of the time but also by bagasse combustion in sugar mills to a lesser extent. Levoglucosan 464 and mannosan emissions factors from bagasse combustion have not been reported so far. We hypothesize that, even if these 465 were very small, levoglucosan and mannosan combustion emissions might not be negligeable as CRV sugarcane biomass 466 yields are very high and most of the harvested sugarcane bagasse is combusted for electric power and steam production. 467  Figure 5a shows the PAH concentration variability during the sampling campaign (mean and standard deviation 473 are available on Table S2). The most abundant PAH were FLE (44.2%±11.9% total concentration share), ANT (9,10) 474 (10.0%±4.5%), BbF (7.4%±2.3%), BghiP (6.7%±2.4%), IcdP (6.4%±1.9%), CPY (6.0%±2.3%), FLO (9H) (5.4%±3.1%), 475 BeP(4.6%±1.3%), and BaP(4.4%±1.6%), which accounted for 95.1% of the total PAH concentration (Figure 5b). Three-ring 476 PAHs were the most abundant (59.04% of total PAH). Put together, five-and six-ring PAHs accounted for an additional 477 38.44%. The less abundant PAH group was the four-ring (2.52%). A previous study in CRV, carried out by Romero et al. 478 (2013), but on PM10 samples, showed higher FLT, PYR and PHE concentrations in areas highly exposed to sugarcane pre-479 harvest burning compared to other locations. In contrast, PM2.5 FLE concentrations in this research were significantly higher 480 than those in PM10 by Romero et al. (2013), while PYR and PHE levels were similar . 481

482
The carcinogenic species BaP, BbF, BkF, BaA, BghiP, FLE, CPY and BeP were identified in all the PM2.5 samples. BaP is a 483 reference for PAH carcinogenicity (WHO, 2013a) that is used as PAH exposure metrics, known as the BenzoaPyrene-484 equivalent carcinogenic potency (BaPE). We calculated BaPE using the toxic equivalent factors (TEF) proposed by Nisbet 485 and LaGoy (1992) and (Malcolm and Dobson, 1994). PM10-bound BaP-TEQ and BaP-MEQ levels for areas not directly exposed to sugarcane burning were 0.16 ng m -3 and =0.21 497 ng m -3 , respectively. Toxicity and mutagenicity due to PM10-bound PAHs were a factor 4 higher at areas directly exposed to 498 sugarcane burning. It is reasonable to assume that PAHs are largely bound to fine aerosol (<2.5 µm), thus that our 499 measurements are comparable to (Romero et al., 2013). If so, our site at Palmira would be at an intermediate exposure 500 condition, higher than areas not directly exposed to sugarcane burning but lower than exposed zones. ~36% came from non-traffic sources, like wood, grass, or coal (n = 15). 517 518 Also, the structure and size of PAHs are indicative of their sources. PAHs with low molecular weight (LMW) (two or three 519 aromatic rings) has been reported as tracers of wood, grass and fuel oil combustion, while the PAHs of medium molecular 520 weight (MMW) (four rings) and high molecular height (HMW) (five and six rings) are associated with coal combustion and 521 vehicular emissions. The ratio between LMW ratio to the sum of MMW and HMW, LMW/(MMW+HMW), is used for source 522 identification. Ratios lower than one are indicative oil products combustion, while ratios larger than one are associated to coal 523 and biomass combustion (Tobiszewski and Namieśnik, 2012). The ratio at Palmira, LMW/(MMW+HMW) = 1.43 ± 1.00, was 524 rather variable but suggests that a large fraction of PAHs in CRV (82.2% of samples) were generated by biomass burning or 525 combustion, as coal combustion is quite limited nowadays. Just one in five samples (17.8%) have PAHs attributable to oil 526 products combustion. A total of 16 alkanes ranging from C20 up to C34 were analyzed in this study and used to identify the presence of fossil fuel 548 combustion and plant fragments in the PM2.5 samples. The abundance of total n-alkanes during the whole sampling period was 549 in the range of 13.0 to 88.45 ng m -3 with an average concentration of 40.36 ng m -3 ± 18.82 ng m -3 . In general, the high molecular 550 weight n-alkanes such as C29 -C31 were the most abundant. These are characteristic of vegetative detritus corresponding to 551 plant fragments in airborne particle matter (Lin et al., 2010). The most abundant n-alkanes were C29, C30 and C31 (Fig 6.).
Likewise, the carbon number maximum concentration (Cmax) was C29 in 43% of samples and C31 in 28% of them. This result 553 is consistent with the chemical profile of sugarcane burning reported by (Oros et al., 2006) with Cmax of C31. 554

555
The carbon preference index (CPI) and wax n-alkanes percentage (WNA%) are parameters used to elucidate the origin of the 556 n-alkanes and infer whether emissions come from biogenic or anthropogenic sources. The CPI represents the ratio between 557 odd and even carbon number n-alkanes. The equation used to calculate CPI in the present study is shown in Table 2, following 558 the procedure reported by (Marzi et al., 1993). Values of CPI ≤ 1 (or close to 1) indicate that n-alkanes are emitted from 559 anthropogenic sources, while values higher than 1 indicate the influence of vegetative detritus in the PM2.5 samples (Mancilla 560 et al., 2016). In this study, mean CPI was always greater than 1, with an average value of 1.22 ± 0.18 (min:1.02max:1.8) 561 that is between the CPI for fossil fuel emissions of ~1.0 (Caumo et al., 2020) and sugarcane burning of 2.1 (Oros et al., 2006), 562 revealing the influence of several sources over the PM2.5 in the CRV. 563 564 Likewise, WNA% represents the preference of odd n-alkanes in the sample. The odd n-alkanes, especially of higher molecular 565 weight, are representative of plant wax related emissions. The waxes are present on the surface of plants, especially on the 566 leaves, and they become airborne by a direct or indirect mechanism like wind action or biomass burning (Kang et al., 2018;567 Simoneit, 2002). In this research, the samples analyzed showed a preference for odd carbon on C27, C29, C31 and C33, which 568 have higher concentrations than the next higher and lower even carbon number homologs, proving the biogenic contribution 569 over the PM2.5 in the CRV. The WNA% was calculated using the equation shown in Table 2 described by Yadav et al. (2013). 570 A larger WNA% represents the contribution from emissions of plant waxes or biomass burning. Otherwise, a smaller value 571 represents that n-alkanes from petrogenic sources, known as petrogenic n-alkanes (PNA)%. The mean WNA% calculated for 572 the PM2.5 samples collected from the CRV was 12.65% ± 5.21% (min: 4.71%max: 29.92%) and can be defined as petrogenic 573 inputs (PNA%) that were 87.35% during the sampling period. The correlation between CPI and WNA was moderate (r 2 =0.53) 574 supporting a consistent meaning between these two parameters, and they are useful for assessing the plant wax contribution 575 on PM2.5. We applied a PCA to the chemical composition data to assess the latent factors controlling the PM2.5 concentrations in the 595 CRV. This statistical tool was used to find the chemical species that describe each component and qualitatively associate these 596 to potential sources of fine aerosol particles. In order to extract the number of components in a PCA many procedures exist, 597 while one of the most common ones is the scree plot of successive eigenvalues for several components from which it is possible 598 to identify the point where the proportion of the variance explained by each subsequent component drops off abruptly. Fig S7  599 shows the inflection point in component number four, explaining 45% of the chemical composition data variance. The addition 600 of two following components allows describing 61% of the variance. Therefore, this study was conducted taking into account 601 six components. Table 4 shows the loading for each chemical component assessment for the six components, where the 602 loadings higher than 0.6 were considered in the discussion interpreted as a source that contributed to the formation of PM2.5.   and moderate for others such as C2O4 2-, Na + , K + . Particularly, the ions Cl -, NO3 + , K + increase during biomass burning (Ryu et 652 al., 2004). In addition, PC6 strongly explained the variance of calcium as water soluble ions and a fraction of the trace metal, 653 therefore the erosion of soil could be considered as an activity that explained PC6. After preharvest biomass burning and fires 654 as a tool to prepare the land for the next crops the soil erosion can increase because of the reduction of vegetation. Therefore, 655 compounds associated with soil erosion and derivates of biomass burning can simultaneously affect the soil erosion and the 656 chemical composition of PM2.5. 657

658
The PCA results showed there was no dominant component that explained the variance of chemical species contained in PM2.5 659 in CRV. Instead, many components have roles equally important that are associated with road dust derived from traffic, the 660 formation of secondary aerosol particles and biomass combustion, petroleum combustion associated with vehicular exhaust, 661 and the presence of vegetative detritus and agriculture soil resuspended by wind erosion. Sugarcane burning was not identified 662 as an individual component that can be explained because the open sugarcane burns happened continuously during the 663 sampling, so they became a background source for this study that very likely was included in the secondary formation as 664 another background source. However, the carbohydrates contained in PM2.5 was linked to the characteristic species of 665 secondary aerosol formation and vegetative detritus. Therefore the secondary pollutants could also originate from the burning 666 of sugarcane in the CRV, similar to the results reported by (Vasconcellos et al., 2007)  components of fine aerosol particles and to qualitatively identify aerosol sources using ratios and principal component analysis 673 (PCA). The main PM2.5 components were organic material (52.99%here), followed by ammonium sulfate (16.12%here) and 674 elemental carbon (6.95%here). The contribution of secondary organic material and inorganic salts was found to be significant 675 and likely related to biomass burning and agricultural practices and estimated secondary aerosol formation was estimation of. 676 EC and PAHs concentrations confirm the presence of incomplete combustion process in CRV. Diagnostic ratios applied to 677 organic compounds indicate that PM2.5 was emitted locally and had contributions of pyrogenic and petrogenic sources. In 678 addition, levoglucosan and mannosan levels showed that biomass burning was ubiquitous during the sampling period. 679 Fluoranthene (FLE) was the most abundant PAH, confrieming the strong influence of sugarcane burning. Five-and six-ring 680 PAH associated with vehicular emissions were also abundant in PM2.5. Our measurements point to sugarcane pre-harvest 681 burning as the main source of PAHs in CRV. The comparison of PM2.5 concentrations and mutagenic potentials suggest that 682 year-long sugarcane pre-harvest burning in CRV, which is also conducted on less than half of the harvested area (34% in 2018) 683 and over limited plots sizes (~6 ha median), leads to lower atmospheric pollutant burdens and mutagenic potentials compared 684 to those at locations where the harvesting period is shorter (zafra) thus with higher burning rates. 685 686 Several sources were identified through PCA, including road dust, secondary aerosol particles, and biomass combustion, 687 vehicle exhaust, vegetative detritus and resuspended agricultural soil likely induced by pre-harvest burning. Not one of these 688 sources was dominant nor explained the chemical species variance of measured PM2.5. Sugarcane burning was not identified 689 as an independent source, but it was found related to the secondary aerosol formation component on PCA. This link between 690 sugarcane burning emissions and secondary aerosol formation requires further investigation. We found that the effects of 691 agriculture on CRV's air quality, particularly of sugarcane preharvest burning are non-trivial. Besides primary particles, this 692 activity generates SOA precursors, induces soil resuspension and is closely tied to diesel emissions during harvesting. 693