Spatial and seasonal variations of isoprene secondary organic aerosol in China: Significant impact of biomass burning during winter

Isoprene is a substantial contributor to global secondary organic aerosol (SOA). The formation of isoprene SOA (SOAI) is highly influenced by anthropogenic emissions. Currently, there is rare information regarding SOAI in polluted regions. In this study, one-year concurrent observation of SOAI tracers was undertaken at 12 sites across China for the first time. The tracers formed from the HO2-channel exhibited higher concentrations at rural sites, while the tracer formed from the NO/NO2-channel showed higher levels at urban sites. 3-Methyltetrahydrofuran-3,4-diols exhibited linear correlations with their ring-opening products, C5-alkenetriols. And the slopes were steeper in the southern China than the northern China, indicating stronger ring-opening reactions there. The correlation analysis of SOAI tracers with the factor determining biogenic emission and the tracer of biomass burning (levoglucosan) implied that the high level of SOAI during summer was controlled by biogenic emission, while the unexpected increase of SOAI during winter was largely due to the elevated biomass burning emission. The estimated secondary organic carbon from isoprene (SOCI) exhibited the highest levels in Southwest China. The significant correlations of SOCI between paired sites implied the regional impact of SOAI in China. Our findings implicate that isoprene origins and SOAI formation are distinctive in polluted regions.

Large-scale and long-term survey of SOA I tracers can provide a full picture of SOA I fate on a continental or global scale. SOA I tracers over global oceans have been measured by round-the-world cruises [13][14][15] . In the continents, Lewandowski et al. 16 analyzed the characteristics of SOA tracers at 15 field sites across the United States during the spring-summer period. Our previous study made a snapshot of SOA tracers at 14 sites in China during the summer of 2012 17 . These existing large-scale observations provided unique information on SOA formation mechanisms and spatiotemporal distribution in the ambient atmosphere. However, all these studies were undertaken within one season or several months. As a major biogenic VOC, isoprene emission is driven by temperature and light 18 , and exhibits a typical seasonal trend, highest in summer and lowest in winter 3 . NOx emission is mainly driven by anthropogenic activities, and presented higher levels in winter and summer over China 19 . As mentioned above, the HO 2 -channel and the NO/NO 2 -channel reactions coexist in the atmosphere. The variations of the two pathways in SOA I production depend on the relative abundances of isoprene and NOx emissions that both vary from place to place and season to season. Long-term concurrent observation over a national or continental scale is still rare and urgently needed.
China is the largest developing country and has undergone very rapid economic growth during the past decades. At present, particulate pollution is a serious environmental problem in China. The concentrations of fine particles (PM 2.5 ) exceed the national air quality standards in most cities 20 . And haze events occur nationwide 21 . OA is a major component of PM 2.5 and an important light extinction substance in China 22 . During the extremely severe haze pollution in China, SOA contributed up to 35% of PM 2.5 and 71% of OA 23 . Thus, SOA plays an important role in particulate pollution in China. Previous modeling studies predicted that SOA was mainly from biogenic precursors in China [24][25][26] . Our ground-based observation in the 2012 summer illuminated that isoprene was the major precursor of SOA in China 17 . In this study, the observation at 12 sites across 6 regions of China extended to one year (Fig. 1). We focused on SOA I tracers formed from the HO 2 -channel and the NO/ NO 2 -channel, and characterized their spatial and seasonal trends. Moreover, the spatial homogeneity of the estimated secondary organic carbon from isoprene (SOC I ) was checked between the paired sites within each region to understand the regional impact of SOA I .

Results and Discussion
Spatial distribution of SOA I tracers. The sum of the SOA I tracers ranged from 6.67 to 122 ng m −3 among the 12 sites (Supplementary Table S1). The highest concentration was observed at the rural BN site in Southwest China, and the lowest level occurred at the desert SPT site in Northwest China (Fig. 2). Southwest China (BN and KM) exhibited the highest concentrations among the 6 regions, followed by East China (WX, HF, and QYZ), Northeast China (HL and TYU), North China (BJ and TY), South China (SY) and Northwest China (DH and SPT). MTLs were the major compounds with the annual average of 37.2 ± 28.4 ng m −3 , followed by     3-MeTHF-3,4-diols are produced by the acid-catalyzed intermolecular rearrangement of IEPOX on acidic particles 9 . The annual average ranged from 0.08 to 0.70 ng m −3 with the highest level at the rural QYZ site in East China (Fig. 3b). Our measurements in the summer (0.61 ± 0.46 ng m −3 ) were one order of magnitude lower than those during summer in the southeastern United States (6.0 − 19.0 ng m −3 ) 27,28 . As a modeling study predicted, IEPOX concentrations in China were much lower than those in the southeastern United States (approximately 0.1 vs. 0.5 ppb) in summer 6 . Compared with the southeastern United States, China had lower isoprene emission 3 but higher NO x levels ( Supplementary Fig. S1) during summer. Since IEPOX yield drops rapidly with increasing NO, high NOx levels could significantly suppress IEPOX formation 6 . Thus, relatively low isoprene emission and the suppression of IEPOX formation due to high anthropogenic emissions resulted in low 3-MeTHF-3,4-diols levels in China.
The two 3-MeTHF-3,4-diol isomers exhibited significant correlations with each other at all sites (p < 0.01, Supplementary Fig. S2), indicating that the relative formation rate of these two isomers was constant at each site. It is worth noting that the slopes of cis-vs. trans-3-MeTHF-3,4-diols exhibited a spatial difference. At the most sites, the slopes were close to 0.5. However, the slopes exceeded 1.0 in Northwest China (2.80 at DH and 1.25 at SPT). The GC-MS chromatogram also demonstrated the apparent change from trans-majority at SY to cismajority at DH ( Supplementary Fig. S3). As reported by Lin et al. 9 29 pointed out that β -IEPOX accounted for ~70% of total IEPOX (β -IEPOX + δ -IEPOX) in the isoprene photo-oxidation. Thus, the dominance of trans-3-MeTHF-3,4-diol over cis-3-MeTHF-3,4-diol is expected in the ambient air, such as at the most sites in this study, and in Yorkville, southeastern United States 9 . The majority of cis-3-MeTHF-3,4-diol observed at DH and SPT suggested that the δ -IEPOX chemistry might play more important role in the isoprene photo-oxidation in Northwest China than other regions of China.
There was no significant correlation between the slope and the annual mean humidity (p = 0.121). Higher humidity could produce more aerosol water and favor the partitioning of gas-phase species to aerosol water, as driven by Henry's Law 30,31 . On the other hand, high aerosol water could dilute the concentrations of hydrogen ion, and reduce aerosol acidity, which would not favor the acid-catalyzed reactions in the particle phase 32 . Thus, we did not see a significant correlation between the slope and humidity in this study (Fig. 4b).
The slopes were steeper in the southern China (Southwest China, East China, and South China) than the northern China (Northeast China, North China and Northwest China, Fig. 4c). The broad areas in the south of Yangtze River were the acid-rain zone of China, as a result of substantial sulfur dioxide emission from large coal consumption 33 . The acidic particles over the southern China would favor the acid-catalyzed ring-opening reactions. Moreover, the high temperature in the southern China could also favor the acid-catalyzed reactions in the particle phase and enhance the conversion efficiency of C 5 -alkenetriols from 3-MeTHF-3,4-diols. Thus, it is expected that the acid-catalyzed ring-opening reactions were stronger in the southern China.
Increasing NOx level changes the isoprene oxidation from the HO 2 -channel to the NO/NO 2 -channel 4 . Figure 5a presents the spatial distribution of MGA, the tracer formed from the NO/NO 2 -channel 7,11 . Unlike the tracers formed from the HO 2 -channel, the highest level of MGA occurred at the urban BJ site in North China (annual average 5.99 ng m −3 ). And the urban sites had higher MGA concentrations than the rural sites. Since NOx emission in China was mostly from anthropogenic sources (power plants, industry and transportation) 34 , high anthropogenic emissions in urban areas would favor SOA I production from the NO/NO 2 -channel. The ratio of MGA to MTLs (MGA/MTLs) can reflect the influence of NOx on SOA I formation 13 . Previous chamber studies witnessed the elevated MGA/MTLs ratio with the increasing NOx/isoprene ratio 12,[35][36][37][38] . The BJ site showed the highest ratio of MGA/MTLs (annual average 1.05) among the 12 sites (Fig. 5b). Previous studies combining model simulations and satellite observations demonstrated that North China was the hot spot of NOx emission in China 19 with an annual emission trend of 4.76 ± 1.61% yr −139 . The highest MGA level and the highest MGA/ MTLs ratio at the BJ site implied that the NO/NO 2 -channel had more influence on SOA I formation in North China compared with the other regions. Figure 2 presents the seasonal trend of SOA I tracers. The highest concentrations existed during summer at the most sites. The elevated levels were surprisingly observed from fall to winter, especially at the DH site.

Seasonal variation of SOA tracers.
Isoprene emission rate (E I ) depends on light and temperature 18 : where EF I is the basal emission rate at 30 °C leaf temperature and 1000 μ mol m −2 s −1 photosynthetically active radiation (PAR). C L and C T are the factors representing the influences of light and temperature, respectively. C L and C T can be simply estimated as: where α = 0.0027, C L1 = 1.066, R = 8.314 J K −1 mol −1 , C T1 = 95000 J mol −1 , C T2 = 230000 J mol −1 , T s = 303 K, T m = 314 K, L is PAR (μ mol m −2 s −1 ) and T is leaf temperature (K) 18 . In this study, daily mean temperature and maximum solar radiation during each sampling episode were downloaded from China Meteorological Data Sharing Service System (http://cdc.nmic.cn/home.do). PAR can be estimated by solar radiation multiplying photon flux efficacy of 1.86 μ mol J −1 40 . With daily mean temperature and maximum PAR, C T and C L were roughly  Fig. S5 h-l), implying that there should be other factors determining seasonal trend of SOA I in these regions. Figure 6 presents monthly variation of MGA/MTLs ratios in China. The ratios raised from fall to spring, and dropped in summer. As mentioned above, MGA/MTLs can reflect the influence of anthropogenic emissions (e.g. NOx) on SOA I formation. The increase in MGA/MTLs in Fig. 6 implied the enhancement of anthropogenic influence on SOA I formation during fall to spring. The biomass burning tracer, levoglucosan 41 also presented higher levels from fall to spring, and dropped in summer (Fig. 6). This indicated that there was significant enhancement of biomass burning emission during fall to spring in China. Previous emission inventory studies recorded large amounts of isoprene and NOx emissions from all biomass burning types 42,43 . Thus, biomass burning might have significant influence on seasonal trend of SOA I during fall to spring.  Considering the two facts: (1) elevated levels of SOA I tracers from the fall to the spring (Fig. 2), and (2) high values of MGA/MTLs and levoglucosan during the fall to the spring (Fig. 6), we split the whole year samples into the warm period (May to September, 2013) and the cold period (October 2012 to April 2013). And the correlations of SOA I tracers with the factor determining biogenic emission (C L × C T ) and the tracer of biomass burning emission (levoglucosan) were analyzed in these two periods.
As Table 1 showed, SOA I tracers exhibited positive correlations with C L × C T but not with levoglucosan during the warm period in all regions except Southwest China. This indicated that biogenic emission was the main factor determining monthly variation of SOA I during the warm period in China. During the cold period, the drop of temperature from the summer maximum to the winter minimum would lead to the biogenic emission of isoprene decreasing by 66%-99% (C T drop in Table 1). And SOA I tracers exhibited poor correlations with C L × C T during the cold period (p > 0.05) at all sites except QYZ. Thus, the unexpected increase of SOA I tracers in winter could not be explained by the biogenic emission of isoprene. Differently, the enhanced biomass burning emission in the cold period would release large amounts of isoprene. And SOA I tracers exhibited positive correlations with levoglucosan during the cold period (Table 1). This suggested that biomass burning emission had significant influence on monthly variation of SOA I during the cold period in China. Southwest China was the exception that SOA I tracers showed no dependence on biogenic emission or biomass burning emission (p > 0.05, Table 1). Biogenic emission is expected to be high all the year in Southwest China, due to high annual mean temperature and small temperature variation (e.g. 23.0 ± 2.7 °C during the whole year and 25.4 ± 0.9 °C during the warm period at BN). The spike of biomass burning emission would produce additional SOA I , which might compromise any significant changes in SOA I levels with respect to biogenic emission in Southwest China. SOC estimation and correlation between sites. These SOA I tracers were further applied to estimate SOC I over China using the SOA-tracer method developed by Kleindienst et al. 44 . The researchers performed chamber experiments to obtain the mass fraction of the tracers in SOC ( f SOC ) for individual precursor: is the total concentration of the tracers for a certain precursor.
[SOC] is the mass concentration of SOC. With these f SOC values and the measured SOA tracers in the ambient air, SOC from different precursors can be estimated in the atmosphere, with the assumption that the f SOC values in the chamber are the same as those in the ambient air 44 .
For isoprene, the f SOC was reported as 0.155 ± 0.039 μ g μ gC −1 using the total concentration of MGA and MTLs 44 . The same f SOC and tracers were applied for SOC I estimation in our previous studies in the Pearl River Delta, South China 45 , Tibetan Plateau 46 , and 14 sites across China 17 . The uncertainty in the SOA-tracer method is induced from the analysis of organic tracers and the determination of the conversion factors. The major limitation of SOA tracers is the lack of authentic standards for quantification. Instead, surrogate standards are applied to quantify SOA tracers, which induces uncertainties in the results of tracer concentrations. The errors in the tracer analyses due to surrogate quantification were within 40% for SOA I tracers 46 . The uncertainty of f SOC was reported to be 25% for isoprene 44 . Considering these factors, the uncertainty of estimated SOC I was calculated as 47% through error propagation. Figure 7a shows spatial distribution of SOC I over China. The annual average of SOC I ranged from 0.03 to 0.63 μ gC m −3 with the highest concentration at BN and the lowest level at SPT. Southwest China exhibited the highest concentrations among the 6 regions. Since there were at least two sites in each region except South China, the correlations of SOC I between the paired sites in the five regions were examined to investigate whether SOC I was homogenously distributed on a regional scale. Figure 7b shows the correlation coefficients (r) of SOC I and the

Region Site
Warm period (May-September) Cold period (October-April) distance between the paired sites in the five regions based on the one-year data. In North China and East China, SOC I presented strong correlations between the paired sites (r = 0.8, p < 0.01). The correlations within Southwest China and Northeast China were not strong but still significant (r = 0.4, p < 0.05). Since the distance between the two sites within these regions was about 400 km, the significant correlations indicated that the regional impact of SOA I could be within a scale of ~400 km. When the distance between the sites increased to ~1000 km in Northwest China, the poor correlation implied the lack of large spatial homogeneity for SOA I . The SOA I formation depends on isoprene emission, oxidant concentrations, gas-and particle-phase reactions, partitioning, and meteorology. The above factors within a certain scale, such as ~400 km would tend to be similar. However, it might not be true for a large scale, e.g., the DH and SPT pair.

Methods
Field sampling. Sampling was simultaneously undertaken at 12 sites in 6 regions of China, including five urban sites, three sub-urban sites, and four rural sites (Supplementary Table S2). Size-segregated particle samples were collected using Anderson 9-stage cascade impactors equipped with pre-baked quartz fiber filter (Whatman, 450 °C for 8 h) at an airflow rate of 28.3 L min −1 . The airflow was calibrated before and after sampling using an airflow meter to keep the sampler working at the specified flow rate. One set of 9 size-fractionated filters were collected for 48 hours every two weeks at each site. In this study, 294 sets of field samples were collected from October, 2012 to September 2013. Additionally, one set of field blank were collected at each site as the same way as ambient samples for 5 minutes when the sampler was turned off.