Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan
State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institutes of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China
Corresponding author: X. L. Pan, Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, 3173-25 Showa-machi, Kanazawa-ku, Yokohama 236-0001, Japan. (firstname.lastname@example.org)
Corresponding author: X. L. Pan, Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, 3173-25 Showa-machi, Kanazawa-ku, Yokohama 236-0001, Japan. (email@example.com)
 Intensive open crop residue burning (OCRB) has a great impact on regional air quality and climate. A field observation campaign in a rural area of the Yangtze River Delta Region (YRDR) was performed during the harvest season, and Elemental carbon (ECa), organic carbon (OC), black carbon (BCe), carbon monoxide (CO), carbon dioxide (CO2) and PM2.5mass were concurrently measured. During the observation period, urban pollution and OCRB-impact episodes were classified. The emission ratio of ECa mass (defined as the ΔECa/ΔCO ratio) from OCRB was estimated to be 18.2 ± 4.6 ng/m3/ppbv, much higher than that (3.0 ± 0.3 ng/m3/ppbv) of urban pollution from the YRDR. A significant amount of OC was emitted from OCRB with ΔOC/ΔCO ratio of 101.3 ± 41.6 ng/m3/ppbv. The value found in the present study was near the upper limit of OC emission ratios in the literature, implying great impacts from combustion conditions, types of biomass burned and subsequent evolution. Regarding urban pollution episodes, the ΔOC/ΔCO ratio was found to be 23.7 ± 2.4 ng/m3/ppbv, and secondary organics accounted for the major fraction of OC mass. Combustions phases of OCRB were classified according to a modified combustion efficiency (MCE, defined as ΔCO2/(ΔCO + ΔCO2)). Our results support the view that ECa tend to be produced in flaming combustions (MCE > 0.95) than in smoldering combustions (MCE < 0.95), whereas OC is emitted preferentially from smoldering combustions. Based on our observed carbonaceous aerosol correlations, we estimate that the ECa and OC emissions from OCRB in East Asia might be underestimated by at least 50%.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Open crop residue burning (OCRB), a widespread practice for disposing of crop residues in agricultural regions, has caused great concern recently for its significant emissions of black carbon (BC) and organic carbon (OC). These emissions can result in severe degradation of regional air quality [Crounse et al., 2009] and increasing respiratory disease morbidity [Arbex et al., 2007]. Tropospheric ozone in the western Pacific [Kondo et al., 2004] and inter-annual variability of greenhouse gases in the Northern Hemisphere[Simmonds et al., 2005] have been found to be related to emissions from biomass burning (BB). In China, agricultural residues were normally used as biomass fuels in rural regions, and over 30% was directly burned in the field during harvest seasons [Yamaji et al., 2010]. This practice has impacts on atmospheric composition, ozone production and regional air quality [Kanaya et al., 2008; Li et al., 2010; Yamaji et al., 2010; Yuan et al., 2010; Feng et al., 2012].
 Accurate estimation of spatiotemporally resolved carbonaceous aerosol emissions is very difficult because the emission factor normally vary significantly under different combustion conditions, and determination of the amounts and types of biomass burned from bottom-up statistical methods is still questionable. For instance, field burning of crop residue estimated from a top-down remote sensing approach was less than 1% of that estimated from ground survey data [Yan et al., 2006], and the range of uncertainty for BC estimations in China is a factor of two [Zhang et al., 2009]. The authors of these studies emphasized the importance of narrowing the uncertainties in the emission factor. From the perspective of observations, correlation of carbonaceous aerosols (e.g., ΔBC/ΔCO, BC emission ratio with respect to CO) provides better constraints on the uncertainties of the emissions, and the definition has been explicated in literature [Andreae and Merlet, 2001]. Higher ΔBC/ΔCO ratios have been reported for BB plumes than for urban plumes. For example, observations using a single-particle soot photometer (SP2) indicated that the ΔBC/ΔCO ratio was 13.8 ng/m3/ppbv in open biomass burning plumes during the TexAQS campaign, and it was reported to be 25 ng/m3/ppbv in India [Badarinath et al., 2007]. Recent observations at a background site in East China showed ΔBC/ΔCO ratios of 10–12 ng/m3/ppbv when the site experienced OCRB plumes [Pan et al., 2011]. Kondo et al.  reported that the ΔBC/ΔCO ratios were even as low as 8.5 ± 5.4 ng/m3/ppbv in Siberia and Kazakhstan and 2.3 ± 2.2 ng/m3/ppbv in North America. In a recent review of the physical properties of BB particles [Reid et al., 2005], the variations were attributed mostly to differences in biofuel type, combustion phase, diffusion conditions and dry deposition processes during transport. Substantial amounts of OC are also produced from OCRB owing to condensation and coagulation of vaporized organic matter, especially in the smoldering combustion phase, when temperature is relatively low. Previous studies have indicated that the ΔOC/ΔCO ratio varies significantly owing to differences in emission scenarios [Maria et al., 2003] and evolution processes, and the ratio can decrease dramatically as a result of cloud scavenging and dry deposition of OC during transport [Pan et al., 2012]. Thus it is of great importance to obtain in situ measurements in the vicinity of OCRB sources.
 In the YRDR, carbonaceous aerosol correlations with emissions from urban sites have been investigated in several studies [Li et al., 2007; Zhang et al., 2008; Li et al., 2009]; however, quantitative investigations near OCRB sources have been limited. In response to this demand, an intensive field campaign in Rudong town, Jiangsu province, was performed during OCRB season in May and June 2010. The major crops at Rudong are wheat, and some commercial rape plants are also cultivated for edible oil production. During Rudong field campaign, we found that residues of wheat straw were the dominant biomass burned. Some branches and leaves of rape plants were also open burned in the farmland at the same time; however the impact seemed to be very limited. This study reported the emission characteristics of carbonaceous aerosols (ΔECa/ΔCO, ΔOC/ΔCO, ΔOC/ΔECa, ΔPM2.5/ΔCO etc.) from OCRB and the impact of the combustion phase. Uncertainties in the ECa and OC emissions inventory from OCRB in East Asia are also discussed in this paper.
2.1. Site and Instrumentation
 The observation site is located in the Dong'An economic development zone (Rudong town, 32.25°N, 121.37°E, Jiangsu province of the People's Republic of China), about 5 km west of the shore of the Yellow Sea and about 50 km east of Rudong town. The site is surrounded by acres of farmland with few inhabitants, and local sources of aerosols are quite limited. The Yangtze River Delta Region is about 180 km to the south of the site. Figure 1 shows the geographic location of the observation site. Concurrent measurements of ambient mass concentrations of BCe, ECa, OC, and PM2.5 and mixing ratios of CO and CO2 were performed from June 1 to June 24, 2010. Descriptions of the instruments are summarized in Table 1. During the campaign, all the instruments were placed in two containers with sampling inlet height of ∼3 m above ground. Ambient equivalent BC mass concentration (BCe, equivalent to the amount of pure black carbon that absorbs the same amount of light) was measured using a multiple-angle absorption photometer (MAAP) at 1 min time intervals after the air had passed through a 1.5 m long, 1/2 inch (inner diameter) wide conductive tube. The loss of BC particles with typical diameter of ∼200 nm on the inner walls of the tube is negligible [Kondo et al., 2009].
Table 1. Description of Instruments During Campaigna
Units: ppmv, parts per million by volume; ppbv, parts per billion by volume.
Light scattering photometry and beta radiation attenuation
<0.5 μg m−3 (1 h)
 The mass-absorption cross section (MAC), an essential factor for BC measurements by filter-based optical techniques, is normally location-specific. The low MAC value (6.6 m2 g−1) at a wavelength of 637 nm which is suggested by manufacture was found to result in ascertain overestimation, and many recent studies suggested to increase MAC value on the basis of laboratory experiments [Bond and Bergstrom, 2006] and field measurements [Kanaya et al., 2012]. In the present study, we adopted a MAE value of 8.6 m2 g−1 to convert the absorption coefficient to mass concentrations according to the field measurement at YRDR [Xu et al., 2002]. Overall uncertainty for BCe measurements was estimated to be less than 15% [Kanaya et al., 2008].
 Mass concentrations of “apparent” EC (ECa, not strictly “elemental” carbon in a chemical sense but rather an operationally definition) and OC in PM2.5were determined by a Sunset semi-continuous ECOC analyzer with a PM2.5 cyclone (2.5 μm diameter cutoff, URG-2000-30EH, URG Inc.). The flow rate of sampling was 8.5L/min, and the detection limit of ECa was estimated to be less than 0.4 μg/m3. Before June 16 17:30 LST (Local Standard Time, UTC + 8 h), 12 samples each day (starting at 00:00 LST with a time interval of 2 h and a sampling period of 90 min) were collected and analyzed. After that date and time, 16 samples each day with a time interval of 90 min were analyzed each day in view of the high ambient aerosol loadings from the great impact of OCRB. The IMPROVE-like temperature protocol (oven temperature from 350°C [holding 90 s] and 550°C [holding 120 s] in a pure He environment and then 550°C [holding 75 s], 700°C [holding 75 s], 850°C [holding 90 s] and oven off [30 s] in an O2/He environment) [Kanaya et al., 2008] and a transmittance correction method were adopted for the whole observation period. It was worth note that the presence of potassium and sodium in aerosols may catalyze the oxidation of BC particles at relatively low temperature [Martins et al., 1998; Sciare et al., 2003] and affects the ECa determination. Mineral oxides might supply oxygen to neighboring carbon particles at high temperature, causing ECa to be falsely classified into OC [Chow et al., 2001]. the overall uncertainty of the ECa measurements was estimated to be ∼31% [Kanaya et al., 2008].
 Ambient CO mixing ratio was measured at a time resolution of 1 min using a gas filter non-dispersive infrared carbon monoxide gas analyzer (Thermo 48C, Precision: 10.0 ppbv; Zero Noise: 5.0 ppbv RMS in 120-s averaging time) with a Teflon particulate filter. Sample air was drawn from the ambient atmosphere by passing it through a 2 m, 1/4 inch (inner diameter) Teflon pipe. The zero point (baseline of the instrumental signal) was checked each hour using purified air produced by passing ambient air through a heated Pt catalyst (Thermo Electron Co., Model 96). Span calibrations were performed in the ambient environment on June 2, June 9 and June 24, 2010, by injecting standard span gas [2.03 ppmv], produced by Nissan-Tanaka Corp., Japan. Good consistencies were found for the differences between the span and zero points. The uncertainty of the instrument was estimated to be less than 5%. The ambient CO2mixing ratio was measured with a Cavity Ring-Down Spectroscopy analyzer (Picarro, Model G1301, Precision: less than 150 ppbv with 5 s average). Mass concentrations of particles in PM2.5 (aerodynamic diameter less than 2.5 μm) were measured using a Synchronized Hybrid Ambient Real-time Particulate Monitor (SHARP, Model 5030, Thermo scientific Inc., hourly precision: ±2μg/m3 (<80 μg/m3) and ±5 μg/m3 (>80 μg/m3)) with a 2.5 μm cyclone (URG-2000-30EH, URG Inc.). The ambient air was pumped into the instrument with a flow rate of 16.67 L/min and passed through a 1.5 m, 1/2 inch wide conductive tube, and the loss of particles was negligible. The overall uncertainty was estimated to be ±5%. Besides, a commercial PTR-MS instrument (Ionicon Analytik GmbH, Innsbruck, Austria) was used during the campaign. The detailed information about principle, configuration and calibration were in another paper (S. Saito et al., manuscript in preparation, 2012). Aerosol particles of PM2.5were collected with a high-volume sampler (SIBATA, flow rate of 500L/min) with Quartz filters from May 15 to June 23. The filters were changed twice a day (at 0900 LST in the morning and 1830 LST in the afternoon respectively) in consideration of shifting of land-sea breeze, and the sampling time periods were averagely 8.0 and 14.0 h in the daytime and nighttime, respectively. All Quartz filters were heated for 3 h at 900°C temperature to vaporize the organics matters before sampling. The samples were kept at site in the refrigerator at temperature below zero. Water-soluble inorganic species (SO42−, NO3−, NH4+, K+, Mg2+, Ca2+, NO2−, Br−, F−, Cl−, PO43−) were measured at laboratory with ion chromatography and trace metals (Al, Ca, Cu, Fe, Mg, Mn, Ni, Pb, Sr, Zn, Ti, and V) were analyzed with ICP-AES.
2.2. Meteorological Conditions and Footprint
 Meteorological parameters (ambient temperature, relative humidity, air pressure and wind direction, shown in Figure 2) were measured with Visala macro weather station (Model WXT520) at site during the observation period. Surface wind at the site was responsive to meso-scale synoptic conditions in East Asia. From June 1 to June 8, easterly wind was prevailing owing to the lasting influence of the high-pressure system over the Korean peninsula region. Afterwards, the site was subjected to torrential rains (gross rainfall of 20 mm) from June 9 10:00 LST to June 10 04:00 LST due to strong air convection at the front of the easterly moving low pressure system in East China. After the transit, northerly winds prevailed at the observation site from June 11 to June 14, and then southerly winds prevailed from June 15 to June 17 due to the impact of a high pressure system on the western Pacific Ocean. From June 19 to June 24, two obvious pollution episodes were observed, and a southerly wind prevailed, indicating an impact of transport from the YRDR. Two hours of rain occurred on June 19, 1700 LST to 1800 LST. There was no clear decrease of ambient carbonaceous aerosol loadings at that time, and the impact on field measurements seemed to be slight.
 The footprint region (altitude below 100 m height in the planetary boundary layer (PBL)) of sampling at the site was determined by five days' backward simulations of air mass movement on the basis of the Flexpart-WRF model. The Flexpart model is a Lagrangian particle dispersion model that has been widely applied in calculating long-range and meso-scale dispersion of air pollutants. A detailed description and validations of this model can be found in the literature [Stohl et al., 1998]. In the present study, the Flexpart model (Version 6.2) was driven by the high-resolution Weather, Research and Forecasting Model (WRF, version 3.2) with spatial resolution of a 75 km×75 km grid in Lambert conformal projection and time resolution of 1 h. Initial meteorological field data were NCEP FNL analysis data (GRIB2 format, spatial resolution of 1.0 by 1.0 degree and vertical resolution of 26 mandatory levels from 1000 mb to 10 mb). In simulations, 30,000 tracer particles were released from the site at a height of 10 m each hour and the model was run backward. The residence time (in seconds) of tracer particles over the footprint region was computed, and the longer the tracer stayed in that cell, the greater the impact from surface emission sources. The results show that there were significant impacts from the YRDR from June 9 to June 11, from June 14 to June 16 and from June 20 to the end of the campaign because of the long residence time of the air mass. On June 12 and June 13, the air masses came mostly from inner China regions. From June 17 to June 19 the footprints were predominantly in the marine regions owing to the prevailing easterly wind.
3. Temporal Variations of Carbonaceous Aerosols
Figure 2 shows the temporal variations of carbonaceous aerosols (ECa, OC and BCe), CO and CO2. All aerosol species have consistent temporal variations, and four significant pollution episodes were clearly captured by the field measurements, from June 10 10:00 LST to June 12 00:00 LST, June 14 19:00 LST to June 17 12:00 LST, June 18 18:00 LST to June 19 20:00 LST and June 21 19:00 LST to June 24 06:00 LST. The local environmental protection bureau reported that the air quality on these days at Rudong was inferior to National Secondary Standard III, with air pollution indexes (API, conversion of mass concentration of respirable suspended particulate to a scale of 0–500 on daily basis) of 148, 147, 118 and 117, respectively. The intermittent decreases of aerosol concentrations at the site were mostly related to cessation of OCRB or alterations of easterly winds, which brought relatively clean air masses from marine regions. The severest pollution episode occurred from June 14 18:00 LST to June 15 09:00 LST, with mean values (±1 standard deviation) of 213.9 ± 134.8 μg/m3, 8.1 ± 3.6 μg/m3, 5.7 ± 1.8 μg/m3 and 42.9 ± 37.7 μg/m3 for PM2.5, BCe, ECa and OC, respectively. Dramatic increases of CO and CO2mixing ratios (with means of 969.4 ± 412.1 ppbv and 434.1 ± 14.2 ppmv) were also observed. The build-up of pollution was mostly attributable to low wind speeds (mean value 2 ± 0.5 m/s) and subjection to active OCRB plumes. The WRF-FLEXPART model demonstrates that the anthropogenic emissions from the YRDR also had a great impact. The diurnal variations of ECa and OC mass concentrations (Figure 3) show bimodal distributions with two peaks, one at 18:00 LST–22:00 LST and the other at 05:00 LST–09:00 LST, with maximum values of 3.4 μg/m3 (ECa) and 23.6 μg/m3 (OC), respectively at 21:00 LST. In general, mass concentrations of ECa and OC in the daytime (from 09:00 LST to 18:00 LST) were about 30∼50% lower than they were at night, and the substantial increase of ambient carbonaceous particulates at night was mostly related to depression of the PBL and OCRB activities. The timing of OCRB was suitable for studying burning emissions without photochemical losses. The diurnal variation of ambient CO mixing ratio was consistent with that of carbonaceous particulates (Figure 3a). The maximum occurred at 21:00 LST, with value of 470 ppbv, and the large standard deviation (451 ppbv) implies that there was a great impact from discrete local combustion sources. A slight increase of ambient CO mixing ratio in the morning (from 06:00 LST to 12:00 LST) was also observed, with a mean of 414 ± 316 ppbv. With regard to CO2, a smooth unimodal distribution was found, with a maximum value of 413 ± 23 ppmv at 05:00 LST, and the value gradually decreased to 401 ± 12 ppmv with intensifying solar radiation and the effect of plant photosynthesis (Figure 3b).
4.1. Definition of Carbonaceous Correlations
 The correlation of a carbonaceous aerosol is expressed as the ratio of the excess concentration of a trace species [x] to the excess concentration of a simultaneously measured reference gas. CO, a byproduct of incomplete combustion, was normally chosen as the reference gas for its long lifespan. Excess concentration is defined as the difference between the concentration measured in situ and the background concentration. The formula is as follow:
The ratio can also be obtained from the slope of the best linear fit to an [x]-CO scatterplot. The ratios ΔECa/ΔCO, ΔBCe/ΔCO and ΔOC/ΔCO are in units of ng/m3/ppbv and the ΔPM2.5/ΔCO ratio is in μg/m3/ppbv.
4.2. Identification of Urban Pollution-Dominant and OCRB-Impact Episodes
 It is a common understanding that approximately 90% of combusted carbonaceous matter is directly released in the form of CO or CO2 [Andreae and Merlet, 2001]. Therefore, OCRB episodes could be roughly determined on the basis of sudden, dramatic increases of CO concentrations (e.g., ΔCO > 400 ppbv/h) when wind direction did not apparently change. As seen in Figure 2, the influence of OCRB was clearly observed from June 13 to June 17, and sharp increases of CO mixing ratio occurred on June 13 05:00 LST (Max. 659.1 ppbv), June 14 21:00 LST (1992.9 ppbv), June 15 20:00 LST (1536.2 ppbv), June 16 07:00 LST (1096.0 ppbv) and 20:00 LST (776.3 ppbv). In the present study, OCRB-impact episodes were objectively identified according to Positive Matrix Factorization (PMF) analysis, which included 75 filter-based samples collected during the field campaign and containing 22 species, including water-soluble ions and trace metals. Detailed description will be published elsewhere (Y. Kanaya et al., manuscript in preparation, 2012). According to PMF analysis, contributions of OCRB to ambient ECa mass loading ranged from 0 to 75% from June 12 to the end of the campaign (Figure 4). The contribution (mean: 33.4%) at night (from 18:00 LST to 08:00 LST) was much larger than that (3.7%) in the day (from 09:00 LST to 15:00 LST) because OCRB activities were extensive. The episodes during which OCRB contributed to over 10% of the ambient ECa masses were confidently defined as due to OCRB impact. The nighttime samples of June 12, June 14 and June 16 (shown in Figure 4) were identified as due to OCRB-dominant episodes because the OCRB source accounted for 69%, 68% and 75% of ambient ECa mass, respectively. Footprint analysis indicated that the urban sources from the YRDR also had great impact on 15 June (Figure 5). Ambient mass concentrations of non-sea-salt K+ion (nss-K+ = [K+] − 0.0355*[Na+], a marker of biomass burning) also sharply increased according to filter-based analysis, with a maximum value of 5.3μg/m3 on June 15. Nighttime events on June 16 and June 18–June 22 were classified as mixing pollution episodes in view of comparable contributions of OCRB (48%) and urban sources (52%) to ambient ECamass. The contribution of urban sources in the daytime of June 10–June 24 was significant (∼95%) according to PMF analysis, and we classified this period as urban-pollution-dominant.
4.3. Emission Ratio of Carbonaceous Aerosol From OCRB
4.3.1. ECa Mass
 As seen in Figure 6, ΔECa/ΔCO and ΔOC/ΔCO ratios clearly increased when observations at the site were influenced by OCRB plumes at night; the ΔECa/ΔCO and ΔOC/ΔCO ratios had means of 7.1 ± 0.4 and 41.7 ± 3.2 ng/m3/ppbv, respectively. The ΔCO/ΔCO2 ratios during urban pollution dominant and OCRB-impacted episodes were found to be 17.0 ± 1.0 ppbv/ppmv and 11.0 ± 0.7 ppbv/ppmv. Note that these values were just overall characteristics of ambient carbonaceous aerosols when urban pollution was mixed with OCRB emissions (contribution of OCRB to ambient ECa mass ranging from 13–75%, estimated from PMF analysis). As seen in Figure 7, positive correlations were found for ΔECa/ΔCO and ΔOC/ΔCO ratios with increasing OCRB contributions. Based on the best linear regression fit, carbonaceous aerosol correlations with a sole OCRB-source (contribution of OCRB to ECa = 100%) episode could be statistically determined. We found that ΔECa/ΔCO ratio had means of 18.2 ± 4.6 ng/m3/ppbv. Table 2 lists the ΔBC/ΔCO ratios of biomass burning plumes measured in different studies, Airborne measurement of BC with a Single Particle Soot Photometer (SP2, in size range of 90–900 nm, assuming a soot particle density of 2 g/cm3) during the TexAQS 2006 campaign [Schwarz et al., 2008] indicated that the OCRB plume had a ΔECa/ΔCO ratio of 13.8 ng/m3/ppbv (converted from a CO/CO2 ratio of 102 ± 25 ppbv/ppmv and a BC/CO2ratio of 1770 ± 400 ng-BC/kg-air/ppmv).Kondo et al.  reported a ΔECa/ΔCO ratio of 8.5 ± 5.4 ng/m3/ppbv for wildfire plumes from North Asia and a low ΔECa/ΔCO ratio of 3.4 ± 1.4 ng/m3/ppbv for open biomass-burning plumes in North American regions.Andreae and Merlet  reported a ΔECa/ΔCO ratio (converted from units of g(EC)/g(CO), assuming a molar volume of 22.4L at STP for CO) of 9.4 ng/m3/ppbv for agricultural residue burnings. Our estimation was generally higher than their results. Diversity of biomass types being burned was mainly responsible for the difference. Field measurements conducted at Mt. Huang in East China showed ΔECa/ΔCO ratios of 10.3 ± 0.5 ∼ 11.6 ± 0.6 ng/m3/ppbv for OCRB plumes [Pan et al., 2011], and the authors suggested that coating/mixing effects of water-soluble compounds on soot particles might have boosted dry deposition, leading to lower ΔECa/ΔCO ratios when the air masses had been long transported.
Table 2. Comparison of ΔBC/ΔCO Ratios Measured in Different Studies for Biomass Burning Plumes
 In general, mass concentrations of OC were highly correlated with CO mixing ratios, with r> 0.8 for all OCRB-impacted episodes, although OC comprised a variety of anthropogenic sources and biogenic origins. In the present study, the ΔOC/ΔCO ratio of a sole OCRB source was estimated to be 101.3 ± 41.6 ng/m3/ppbv (Figure 7b). This result is over twice the value (44.8 ng/m3/ppbv) reported by Andreae and Merlet  for burning of agricultural residues and near the upper limit of the ΔOC/ΔCO ratio (range: 70.5 ∼ 105.8 ng/m3/ppbv, derived from OA (organic aerosol)/ΔCO ratio of 148 ng/m3/ppbv assuming OA/OC = 1.4∼2.1) reported for the plumes of pine-savanna fires [Yokelson et al., 2007]. DeCarlo et al. suggested that evolution of SOC will add mass to primary OC by a factor of 32%∼100%. During the Rudong campaign, OCRB normally occurred at night (18:00 LST to 09:00 LST next day), and the high ΔOC/ΔCO ratio probably resulted from condensation of semi-volatized organics at night, though atmospheric photochemical activity was very weak. Significant variation of the ΔOC/ΔCO ratio was attributed mostly to variations in fire intensity and in combustion phase and its evolution during OCRB. The ΔOC/ΔECa ratio for a sole OCRB source was reported to be 10.1 ± 7.4 (Figure 7c). Field measurement of OCRB plumes in Central East China indicated that the ΔOC/ΔECa ratio in PM1 was 7.7 [Pan et al., 2012], comparable to the values found in the present study. ΔOC/ΔECa ratios were found to range from 1.7–6.6 for biomass burning in northern Europe [Iinuma et al., 2007; Saarikoski et al., 2008], 6.25 in southern Africa [Kirchstetter et al., 2003] and 5.3 ± 1.6 for indoor biomass burning in India [Rehman et al., 2011]. Our relatively high estimate is reasonable because crop residues burned in the field were normally well bundled up, tending to produce more organic matter owing to an oxygen-limited interior and a lower flame temperature. Large variations of the ΔOC/ΔECa ratio were also attributed to variations of combustion evolution. Impacts of combustion phase on carbonaceous aerosol correlations will be discussed in section 4.7.
4.4. Comparison With Urban-Dominant Pollution
 The ΔECa/ΔCO, ΔOC/ΔCO and ΔCO/ΔCO2ratios during urban-pollution-dominant periods are shown inFigures 6b, 6d, and 6f. Statistical information is summarized in Table 3. As shown, the ΔECa/ΔCO ratio was 3.0 ± 0.3 ng/m3/ppbv when pollution was mostly from the YRDR. Our result is slightly lower than the observation (4.9 ng/m3/ppbv) at an industrial region in North America [Schwarz et al., 2008]. Source apportionment of the emission inventory [Zhang et al., 2009] indicated that about 51.5% of the CO could be attributed to the transport sector of Shanghai mega-cities, implying a great impact of on-road vehicles on the ΔECa/ΔCO ratio of urban plumes. Studies have shown that the ΔECa/ΔCO ratio normally varies with residential activity patterns at urban sites. It tends to reach its maximum during rush hours and subsequently decreases till the afternoon according to studies at urban sites in Tokyo [Kondo et al., 2006], Beijing [Han et al., 2009] and Guangzhou [Verma et al., 2010]. Recently published results [Pan et al., 2011] from Mt. Huang showed that the ΔECa/ΔCO ratio was 8.8 ± 0.9 ng/m3/ppbv in the urban plume from eastern China (Jiangsu and Zhejiang provinces and Shanghai city), and the relatively high values implied a possible impact from domestic and industrial contributions. The correlation between OC and CO in urban-dominant pollution was quite significant, withr = 0.81 and the ΔOC/ΔCO ratio was found to be 23.7 ± 2.4 ng/m3/ppbv. A study in the Nagoya urban area found that the ΔOC/ΔCO ratio was 14.7 ng/m3/ppbv [Kadowaki, 1990]; however, aircraft measurements during the ACE-Asia campaign [Maria et al., 2003] showed that fresh urban plumes from the YRDR have a high ΔOC/ΔCO ratio of 84.5 ng/m3/ppbv (converted from a ΔCO/ΔOC ratio of 14.8 in units of μg m−3/μg m−3assuming a molar volume of 22.4 L at STP for CO), implying a great impact from the residential sector. The large discrepancy could be attributed to the respective contributions from different sources, evolution of OC, cloud scavenging and precipitation. Nevertheless, alteration of the energy-consumption structure of the YRDR in the past decade since ACE-Asia in 2001 could be responsible for a large bias. In the present study, the transport time of air masses sampled at the observation site was estimated to be 13 ± 9 h (range: 4–46 h) according to FLEXPART simulation with release of tracer particles each hour from Shanghai mega-city (latitude 30.5°N–31.5°N; longitude 120.5°E–122.0°E, at heights between 5 and 10 m). Condensation and evolution of secondary organic carbon (SOC) as well as transport loss from dry deposition processes seem to have a great impact on our observed ΔOC/ΔCO ratio.
Table 3. Summaries of Detailed Information for Three Different Subgroups
where prim × [ECa] accounts for the fraction of POC from combustion-related sources, and parameter stands for the aggregate of non-combustion sources and observation artifacts. The prim ratio was statistically determined according to linear fits to data for which the primary emission of carbonaceous aerosol was dominant [Turpin and Huntzicker, 1995]. We used the high benzene(C8+C9)/benzene ratio determined by PTR-MS as the indicator of predominance of freshly produced aerosol because C8- and C9-benzenes reacted more quickly with OH radicals than benzene [Calvert et al., 2002]. We just identified several short episodes (13:00 LST and 16:00 LST on June 20; 13:00 LST, 14:00 LST and 16:00 LST on June 21; 16:00 LST on June 22) when the benzene (C8+C9)/benzene ratio (mean: 8.4) was ±2σ (2.6) larger than the overall mean value of 2.8 during urban pollution episodes. In Figure 8, linear fitting demonstrates that prim was 1.5 ± 0.8, comparable to the measurements performed around Taiwan (1.5–2.0) [Chou et al., 2010], the urban area of Tokyo (mean: 1.33, ranging 1.24–1.48) [Miyazaki et al., 2006] and the Pearl River Delta Region (1.3–1.4) [Cao et al., 2003]. Mean POC and SOC mass concentrations were estimated to be 3.6 ± 1.5 μg/m3 and 6.2 ± 7.1 μg/m3during urban pollution episodes, and the fraction of SOC was about twice that of POC, indicating a significant contribution from photo-oxidation processes. It is worth noting that the ΔOC/ΔECa ratio normally varied in obvious diurnal patterns. In the present study, altering of primfrom 0.5 to 2.0, which resulted in 5∼30% variations of SOC, would not change our conclusions. Besides, SOC generally was well correlated with water-soluble organics, which would make the SOC-enriched particles more easily incorporated into hygroscopic growth and quickly removed from the atmosphere, leading to a certain underestimation.
4.5. Fractions of ECa and OC in PM2.5
 Ambient PM2.5 mass is not always necessarily related to primary emissions because of the presence of a large fraction of secondary aerosols; however a very good consistency between PM2.5 mass with CO mixing ratio was observed during the campaign, and ΔPM2.5/ΔCO ratios were 0.19 μg/m3/ppbv (r = 0.97) for urban pollution episodes and 0.30 μg/m3/ppbv (r= 0.89) for OCRB-dominant episodes.Yokelson et al.  reported that the initial ΔPM2.5/ΔCO ratio was ∼0.06 μg/m3/ppbv (converted from 0.071 g/g) for biomass burning during the MILAGRO campaign, and formation of SOC and water-soluble inorganic aerosols (WSIA) could increase the ΔPM2.5/ΔCO ratio by 2.6 times for biomass-burning plumes within several hours [Yokelson et al., 2009]. In the present study, a high fraction of WSIA present in PM2.5, with means of 45% for OCRB-dominant episodes, accounted for the major differences.
 Variation of total carbon matter (TCM = OC*1.6 + ECa) mass correlated well with mass concentrations of PM2.5 during the observation period (Figure 9a), and TC accounted for an average of 30% of PM2.5 mass. The fraction of ECa in PM2.5mass was stable, with a mean value of 3.4%, and it did not show clear increases, even during the OCRB-dominant episodes (Figure 9b). Nevertheless, the fraction of organic matter (OM = OC*1.6) in PM2.5 increased significantly when the site was impacted by OCRB, and the OM/PM2.5ratio during OCRB-dominant episodes had a mean of 0.41 with a maximum value of 0.97, comparable to the results (OC/PM2.5 ratio: 0.38) of laboratory experiments with wheat-straw burning [Li et al., 2007] and 0.42 of field measurements [Li et al., 2009]. In the present study, refractory OC (OC lost on heating to >550°C based on the Thermal-optical-transmittance approach) was abundant in all our samples, with a mean fraction of over 50% in OC mass. The maximum value (65%) occurred during OCRB periods. At Rudong, the fraction of volatile OC (OC lost during heating to 450°C) in OC mass decreased slightly with the occurrence of OCRB, especially when OCRB was dominant. During urban-pollution-dominant periods, mass concentrations of OM and ECa accounted for 10% and 5% of PM2.5 mass, respectively.
4.6. ΔBCe/ΔECa Ratio
 A tight correlation between BCe mass and CO mixing ratio was found during all the observation periods, and ΔBCe/ΔCO ratios were 8.3 ± 0.3 ng/m3/ppbv and 5.9 ± 0.3 ng/m3/ppbv, respectively, for urban-pollution-dominant and OCRB-impacted episodes (Figure 10). As suggested by previous studies [Reisinger et al., 2008], MAAP tends to overestimate the mass concentration of BC because of the lens effects of organic matter and the light-absorbing humic-like substances deposited on the filter matrix. Inter-comparison of BC measurements in central eastern China indicated that BCemass measured by MAAP was systematically 45–54% higher than ECa mass determined by the thermal-optical-transmittance method [Kanaya et al., 2008]. In the present study, we found that ΔBCe/ΔECa ratios in PM2.5mode were 0.99 ± 0.1 and 1.3 ± 0.2 for OCRB-impacted and urban-pollution-dominant periods. The larger ΔBCe/ΔECa ratio for urban pollution episode was mainly due to the overestimation by optical method which thick coatings of inorganic matters which could increase the lensing effects of aerosols sampled on the filter matrix. Furthermore, after rechecking the filter measurements of trace metals, we found that the mass concentration of Ca and Al during urban pollution dominant period were 0.38 μg/m3 and 0.28 μg/m3, about 2–4 times higher than those (Ca: 0.08 μg/m3; Al: 0.14 μg/m3) in OCRB-dominant period. It implies a possible artifact in the ECa analysis in that mineral oxides might supply oxygen to neighboring carbon particles at high temperature, causing underestimate of ECafor urban pollutions. The mass concentration of nss-K+ does not seem to explain the differences of ΔBCe/ΔECa ratio for different pollution episodes since its mean mass loadings on the filters were similar.
4.7. Impact of Combustion Phase
 The modified combustion efficiency (MCE, defined as ΔCO2/(ΔCO + ΔCO2), no units) was calculated to investigate the impact of the combustion phase on carbonaceous aerosol correlations. Generally, combustion with MCE values over 0.95 is regarded as flame-dominant, while MCE less than 0.95 indicate smoldering combustion in which significant CO is produced from incomplete oxidation [Reid et al., 2005]. On the basis of MCE values, the OCRB episode from June 12 18:00 LST–June 13 08:00 LST was mostly dominated by flaming (MCE: 0.99), while the OCRB episode from June 17 04:00 LST–June 17 09:00 LST was dominated by smoldering (MCE: 0.93). Table 4 lists the ΔCO/ΔCO2, ΔECa/ΔCO, ΔOC/ΔCO and ΔOC/ΔECa ratios during different combustion phases. As seen, the ΔCO/ΔCO2 ratios were 3.8 ± 1.2 and 69.9 ± 9.8 ppbv/ppmv for the flaming and smoldering phases, respectively. The ΔECa/ΔCO ratio of smoldering combustion had a mean of 11.8 ± 2.3 ng/m3/ppbv, about 40% lower than that (17.4 ± 5.2 ng/m3/ppbv) of flaming combustion. The ΔOC/ΔCO ratio of smoldering combustion was found to be 141.4 ± 40.7 ng/m3/ppbv, about 40% higher than that (101.3 ± 10 ng/m3/ppbv) of flaming combustion. These results are reasonable because condensation processes were overwhelming in the presence of un-combusted condensates at the lower temperature, leading to a majority of particles being in the accumulation mode. As expected, the ΔOC/ΔECa ratio (12.3 ± 2.0) for smoldering combustion was determined to be about twice that (5.9 ± 5.3) of flaming combustion, supporting the argument that significant formation of particulate OC takes place during smoldering combustion because of the condensation of volatilized organics on any available particles or surfaces [Reid et al., 2005].
Table 4. Comparison of Carbonaceous Correlations During Flaming and Smoldering Dominant Episodes
 To illustrate the evolution of OCRB on correlations of carbonaceous aerosols, an OCRB-dominant case (from 12 June 18:00 LST to 13 June 09:00 LST) was investigated.Figure 11 shows the temporal variations of MCE, CO, CO2, ECa, OC mass concentration and ΔOC/ΔECa ratio. To calculate MCE, ΔCO2 was derived from the ambient CO2 mixing ratio by subtracting its averaged diurnal variation (illustrated in Figure 3) plus 3 ppmv. We deemed that background CO (217.9 ppbv, the minimum value during this episode) remained constant since the observation site was dominant by weak northerly winds (less than1.0 m/s) during this episode. Interferences from different sources due to shift of wind direction were negligible. The temporal variation of MCE clearly reflected the evolution of the OCRB episode in the surrounding area. As shown in Figure 11, the MCE was lower, with an hourly mean of 0.94, before 12 June 22:00 LST, and then it had a sharp increase to 0.99 (on 12 June 23:00 LST), indicating an influence of flame-dominant OCRB in the vicinity of the site. As expected, the ambient CO2 mixing ratio also had an apparent increase from 398.5 ppmv (on June 12 21:00 LST), reached its maximum of 464.6 ppmv (on June 13 05:00 LST). The temporal variation of CO mixing ratio showed a quite different pattern: a moderate increase before June 13 03:00 LST and a sharp increase on June 13 04:00 LST to its maximum value of 659.0 ppbv on June 13 05:00 LST. After June 13 03:00 LST, the MCE gradually decreased with increase of atmospheric CO loadings. The gradual decrease of MCE value indicated that combustion phase of OCRB probably have started to change; however flaming combustion was still dominant since the ambient CO2 mixing ratio continued to increase till 05:00 LST on June 13. After 06:00 LST on June 13, both CO and CO2 mixing ratios decreased quickly with increase of northerly wind, and the mixing ratios of CO (277 ppbv) and CO2(401 ppmv) at 09:00 LST on June 13 seemed back to the values before OCRB episode. We deemed that the sudden decrease of MCE below 0.95 at 08:00 LST on June 13 was due to dilution of non-OCRB plumes from north.
 The variation of ambient mass concentration of ECa (maximum of 2.4 μg/m3 on June 13 05:00 LST) was mostly consistent with that of CO2, highlighting its preferential production from flaming combustion. It is notable that enhancement of ambient CO2 mixing ratio at the beginning of OCRB did not lead to a clear increase of atmospheric OC loading; instead, ambient OC dramatically increased at the end of the OCRB episode (40.2 μg/m3 on June 13 06:00 LST). The good consistency of variation patterns for CO and OC mass concentrations implies that OC and CO were likely emitted simultaneously from the same combustion processes.
 The relationships of the ΔECa/ΔCO2 and ΔPM2.5/ΔCO2 ratios with the MCE are illustrated in Figure 12. As shown, the state of combustion had a great impact on the correlation of carbonaceous aerosols. Both the ΔECa/ΔCO2 ratio and the ΔPM2.5/ΔCO2 ratio showed clear linear decreasing trends (r = −0.97, r = −0.62) with increase of the MCE, suggesting that carbonaceous matter of biomass tended to be emitted as its final oxidation product (CO2) at high temperature flaming conditions. Linear regression fitting indicated that the slope of the ΔECa/ΔCO2 ratio with respect to the MCE was −3.5 μg/m3/ppmv, lower than the observation (−2.2 μg/m3/ppmv) of biomass burning during the ARCTAS campaign [Kondo et al., 2011]. Correlation of the ΔPM2.5/ΔCO2 ratio with the MCE was −68.8 μg/m3/ppmv.
4.8. Application to OCRB Emissions
 Emissions of carbonaceous aerosol from crop residue burning at the provincial level or grid resolution in China have been reported in literature. For instance, the Regional Emission Inventory in Asia (REAS) suggested that CO, ECa and OC emissions from field burning of crop residue in 2000 were 12.6 Tg/yr, 95 Gg/yr and 455 Gg/yr, respectively [Yamaji et al., 2010].
 Recently lower emissions of CO (4 Tg/yr), ECa (30 Gg/yr) and OC (100 Gg/yr) masses in 2006 for OCRB was estimated based on MODIS Thermal Anomalies/Fire products [Huang et al., 2012]. By dividing the emission amount of EC by that of CO, it is obvious that the emission ratio for BC was the same, with a value of 7.5 g/Kg for their studies. If the emission ratio obtained from our field campaign were applied, our estimation of EC emission from OCRB could be ∼50% higher. Without doubt, the different types of crops (rice, wheat, corn cotton, rapeseed etc.) in China and case-to-case combustion efficiency must also introduce large uncertainties in EC estimations. In the present study, the ΔOC/ΔCO ratio (101.3 ± 41.6 ng/m3/ppbv) was near the upper limit of previous archived results since formation of SOC might be included in the total OC mass. We preferred to use the lower estimation of the ΔOC/ΔCO ratio (74.6 g/Kg, converted from 59.7ng/m3/ppbv assuming the gas volume is 22.4L at standard conditions). In this scenario, the ΔOC/ΔCO ratio (25 ∼ 35 g/Kg) from the bottom-up method seemed to underestimate OC emission by 50 ∼ 70% for OCRB.
 In this study, we performed a comprehensive field campaign at a rural area of the YRDR during harvest season to investigate the impact of OCRB on correlations of carbonaceous aerosols. Mass concentrations of elemental carbon (ECa), organic carbon (OC), black carbon (BCe) and PM2.5were measured with a Sunset semi-continuous ECOC analyzer by an IMPROVE-like temperature-control protocol with MAAP and SHARP instruments, respectively The mixing ratios of carbon monoxide (CO) and carbon dioxide (CO2) were concurrently measured with an infrared absorption CO analyzer and by wavelength-scanned cavity ring down spectroscopy. During the campaign, the observation site was frequently subject to intensive OCRB occurring in the nighttime, and ΔECa/ΔCO and ΔOC/ΔCO ratios were found to be 7.1 ± 0.4 ng/m3/ppbv and 41.7 ± 3.2 ng/m3/ppbv. In the daytime, urban pollution was predominant, and ΔECa/ΔCO and ΔOC/ΔCO ratios were relatively low, with means of 3.0 ± 0.3 ng/m3/ppbv and 23.7 ± 2.4 ng/m3/ppbv, respectively. By combination with PMF analysis, emission ratios of ECa and OC masses for pure OCRB were estimated to be 18.2 ± 4.6 ng/m3/ppbv and 101.3 ± 41.6 ng/m3/ppbv. Our results in the present study were mostly near the upper limit of emission ratios measured in previous studies. The contribution of primary organic carbon (POC) and secondary organic carbon (SOC) to OC mass concentration for urban pollution were estimated according to a semi-empirical method. The mean SOC mass concentration was twice that of the POC mass concentrations; the means were 6.2 ± 7.1μg/m3 and 3.6 ± 1.5 μg/m3, respectively, reflecting the great importance of photochemical formation of organic matter. Evolution of OCRB had great impact on the correlation of carbonaceous aerosols. This study demonstrated that flaming combustion (MCE > 0.95) of biomass tended to produce EC and CO2 but that smoldering combustion (MCE < 0.95) tended to produce OC matter and CO. The ΔOC/ΔECa ratios (12.3 ± 2.0) for smoldering combustion were about twice as high as those (5.9 ± 5.3) of flaming combustion. Our estimated emission ratios of ECa and OC from OCRB were about twice as large as those adopted in the emission inventory, implying that their emissions from OCRB might be underestimated by >50% because of low emission factors.
 The authors would like to thank all of the officers in the Dong'an science and technology zone and the local environmental protection bureau for their support during the field campaign. We gratefully acknowledge the anonymous reviewers for their helpful comments and suggestions. This work was supported by the Global Environment Research Fund (S-7, C-081, B-051) from the Ministry of the Environment, Japan.