Observations of SO 2 and NO 2 by mobile DOAS in the Guangzhou Eastern Area during the Asian Games 2010

F. C. Wu, P. H. Xie, A. Li, K. L. Chan, A. Hartl, Y. Wang, F. Q. Si, Y. Zeng, M. Qin, J. Xu, J. G. Liu, W. Q. Liu, and M. Wenig Key Laboratory of Environmental Optical and Technology, Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei, China School of Energy and Environment, City University of Hong Kong, Hong Kong, China Meteorological Institute, Ludwig-Maximilians-Universität Munich, Munich, Germany


Introduction
NO 2 is an important trace gas in the atmosphere because it readily undergoes photochemical reactions with other atmospheric constituents.In the atmosphere, NO reacts rapidly with ozone to NO 2 , which during the day photolyzes back to NO.The tropospheric production of ozone is thus greatly influenced by NO x (sum of NO and NO 2 ).NO 2 also has an important influence on human health.For example, longterm exposure to NO 2 increases the symptoms of bronchitis in asthmatic children (WHO, 2006).The major sources of anthropogenic emissions of NO 2 are combustion processes such as heating, power generation, and engines in vehicles and ships (Finlayson-Pitts et al., 1999).SO 2 is a colorless gas that adversely affects the respiratory system, especially the lungs.The main anthropogenic source of SO 2 is the burning of sulfur-containing fossil fuels for domestic heating and power generation.SO 2 and NO 2 tend to form sulfuric and nitric acids, respectively, which in the form of acid rain, are one of the causes of deforestation (WHO, 2006).
Population growth, industrial development, and heavy traffic lead to higher energy consumption and, therefore, an increase in the emission of pollutants such as SO 2 , NO 2 , and volatile organic compounds (VOCs) into the atmosphere, if no measures are taken to counteract this development.In recent years, China has experienced a significant increase in atmospheric pollutant concentrations because of rapid industrial development, which has an important impact on ecosystems and human health.The Pearl River delta (PRD) in the Published by Copernicus Publications on behalf of the European Geosciences Union.
south of China is one of the three major economic areas (the other two are the Yangtze River delta and the Beijing-Tianjin-Hebei economic region).It includes many populated and strongly industrialized cities such as Guangzhou and Shenzhen, and has experienced an extremely fast economic development (Zhang et al., 2008a, b;Wang et al., 2008).PRD has a total land area of 42 794 km 2 and a population of over 38 million (Cao et al., 2004).As a result, emissions of SO 2 , NO 2 , and other pollutants have also largely increased in this area (Zhang et al., 2007).
The 16th Asian Games were held in the city of Guangzhou from November 2010 to December 2010.The pollutant sources were identified in order to alleviate air pollution for this occasion.In addition, strategies including emission control for factories, vehicle limitation, and so forth were employed by the Guangzhou government to reduce the air pollution problem during the Asian Games.The Guangzhou eastern area (GEA) (Fig. 1) was considered the most seriously affected region of the city because of the many pollutant sources present, such as the Guangdong Yuehua Power Plant (the largest power plant in Guangzhou according to the Guangzhou Environmental Center), the Guangzhou Hengyun thermal power company, and the Guangzhou Zhujiang steel company, where air pollutants such as SO 2 , NO 2 , VOCs, and fine particulates are emitted.Regional studies investigated that area sources have a very strong influence on air quality through the regional transport of air pollutants, possibly causing severe pollution events to the area and its neighbors (Melamed et al., 2009;Takashima et al., 2011).The temporal and spatial scale of transport can range from a few days to several weeks and from a few kilometers to hundreds of kilometers.Therefore, understanding the spatial and temporal distribution as well as the emission sources of air pollutants in GEA was important for environmental management during the Guangzhou Asian Games.Air pollutants were routinely monitored by the local environmental protection agency using a network of ground-level monitors.However, data from this network were insufficient for spatial distribution and transportation processes as well as emission sourcing.
Previously, regional studies in the PRD combined aircraft measurements and modeling studies to examine spatial and vertical distributions of O 3 , SO 2 , NO x , PM 10 , and PM 2.5 (Wang et al., 2008); or used a bottom-up approach to estimate emissions of NO x , SO 2 , CO, VOCs, and fine particulates (Zheng et al., 2009;He et al., 2011).These studies focused on the larger area of the PRD but did not consider smaller-scale distributions of air pollutants or area sources like the GEA.In this study, passive differential optical absorption spectroscopy (DOAS) on a mobile platform (referred to as mobile DOAS in the following) was used to detect spatial and temporal distributions and emissions of SO 2 and NO 2 related to the GEA.This technique was first used to measure volcanic emissions (Edner et al., 1994;Galle et al., 2003) and subsequently applied to determine the emis-sion of point sources (e.g., power plants, oil refineries, etc.) and area sources (e.g., cities and industrial areas).Johansson et al. (2008Johansson et al. ( , 2009) ) and Rivera et al. (2009) examined the outflow of SO 2 , NO 2 , and HCHO in Mexico, Beijing and the Tula industrial area in the city of Mexico.NO x emissions in Mannheim and Ludwigshafen, Germany, using mobile MAX-DOAS (multi-axis differential optical absorption spectroscopy) were investigated by Ibrahim et al. (2010).The same method was used by Shaiganfar et al. (2011) to quantify emissions in Delhi.In China, several measurements based on mobile DOAS were also carried out (Li et al., 2005(Li et al., , 2007a;;Wu et al., 2011).However, in previous field measurements that used a zenith viewing DOAS instrument on a mobile platform, emissions were estimated by first taking a "cleanair" spectrum at some point along the measurement path and using it as a reference spectrum to obtain differential vertical columns relative to this "clean-air" measurement (Li et al., 2005(Li et al., , 2007a;;Wu et al., 2011).For the field campaign reported in the present study, concurrent measurements of a MAX-DOAS instrument at a fixed location were used to obtain estimates of absolute vertical columns for NO 2 and SO 2 , making the analysis of data from the mobile measurement relatively simple (see detailed description in Sect.2.4.1).
In this paper, we present mobile DOAS measurements carried out from 9 November to 26 December (each day from 10:00 to 13:00 local time) during the Asian Games 2010 in Guangzhou, South China.We derived vertical columns of SO 2 and NO 2 along a closed path around the GEA in combination with data from stationary MAX-DOAS to estimate emissions from this area.We study the variation of these pollutants for different wind fields and emission periods during the games, thereby identifying individual sources.Furthermore, we compare NO 2 vertical columns from our mobile measurements to those from Ozone Monitoring Instrument (OMI).Finally, the influence of GEA emissions on the venue of the games at Guangzhou University Town is explored.
The paper is organized as follows: in Sect.2, the measurements in GEA, our measurement instrument, and principle are introduced.In Sect.3, our results and discussions, including the distributions of SO 2 and NO 2 around GEA, the comparison with OMI NO 2 , and emissions of SO 2 and NO x from GEA with their influence on the downwind region are all presented.In Sect.4, the results of our study are summarized.

Description of the measurements in GEA
GEA is located southeast of Guangzhou City and west of Dongguan City (Fig. 1).The measurement route shown in Fig. 1    approximately 2.5 h to complete.Our measurements took place for seven weeks, from 9 November to 26 December 2010, starting every odd day at 10 a.m. and encircled the GEA once during each day of measurement.The Guangzhou Municipal Government attempted to control traffic-induced air pollution for the Asian and Asian Para Games by limiting the number of vehicles in the city area during these events.Vehicles with odd number plates were only allowed to drive on odd days from 1 November to 29 November and from 5 to 21 December 2010, whereas vehicles with even number plates were allowed only on even days.Our van has an odd number plate; therefore, we had a total of 25 days of measurements.During the entire measurement period, the temperature varied from 22 to 28 • C, and the wind blew predom-inantly from the north/northeast (see Sect. 3.1.1for further details).

Description of the instrument
Our mobile DOAS instrument records sunlight scattered into the telescope pointing to the zenith.The mobile DOAS system was developed at the Anhui Institute of Optics and Fine Mechanics (AIOFM).The components of our AIOFM mobile DOAS instrument are the telescope, a UV/VIS detector spectrometer unit, a computer, and a global positioning system (GPS).The lens telescope with 80 mm diameter and 170 mm focal length is used to collect scattered sunlight.Light collected by the telescope is transmitted to the spectrometer (Ocean Optics HR2000) through a 3 m length optical fiber with diameter of 400 µm.The field of view of the telescope is about 0.1 • .The spectrometer has a spectral resolution of about 0.6 nm (FWHM) and a spectral range of 290 nm to 420 nm.It is placed in a small refrigerator to stabilize the temperature at +25 ± 0.5 • C. The GPS tracks the coordinates of the measurement route and provides the car speed and acquisition time for each spectrum.A miniature weather station is used to record meteorological data such as wind speed, wind direction, temperature, humidity, and pressure along the route at street level.The telescope, GPS and weather station are mounted on top of the measurement van.
The entire system is automatically controlled by our AIOFM mobile DOAS software and is equipped with a DC (direct current) +12 V battery.The +12 V direct current is converted to alternating current (AC) by a power converter to power the entire system (Li et al., 2005).The detection limits of SO 2 and NO 2 column densities for the instrument are about ∼ 3-5 × 10 15 molecules cm −2 (the "molecules cm −2 " is abbreviated to "molec cm −2 " and the "molecules cm −3 " is abbreviated to "molec cm −3 " in the following).
The long-path DOAS (LP-DOAS) instrument used to explore the influence of GEA on the downwind region (see Sect. 3.4) uses a xenon lamp as light source and a UV/VIS spectrometer (QE65000, Ocean Optics).The detailed setup of the long-path DOAS instrument has been provided by Qin et al. (2006).The instrument is mounted on the third floor (about 15 m above ground) of one of the office buildings at Guangzhou University Town (Fig. 1), southeast of Guangzhou City, about 20 km away from GEA.The retro reflector is installed on the roof of a teaching building, southeast of the office building.The total optical path length of LP-DOAS is about 788 m pointing to the southeast (Fig. 1c).The wavelength coverage of this instrument, ranging from 260 nm to 369 nm, allows monitoring of SO 2 , HCHO, NO 2 , and O 3 among others.During this campaign, the instrument is mainly used to monitor SO 2 , NO 2 , and O 3 .
In this study, the MAX-DOAS instrument is used to obtain the vertical columns for NO 2 and SO 2 .This system is also from the AIOFM, with parts such as a stepper motor used to rotate a telescope, a miniature spectrometer (Ocean Optics HR2000) with a spectral resolution of 0.6 nm.Scattered sunlight is collected and focused by the telescope and is led into the spectrometer unit through 3 m optical fiber.The spectrometer unit is placed in a small refrigerator to stabilize the temperature at +25 ± 0.5 • C. The MAX-DOAS telescope is equipped with a stepper motor, therefore allowing for pointing at different elevation viewing angles.During our measurements, scattered sunlight was collected sequentially at 5, 10, 20, 30, and 90 degree elevation viewing angles.The azimuth viewing direction of the MAX-DOAS is north (Fig. 1b).The detailed description of our MAX-DOAS system is given in Li et al. (2007b).Furthermore, lidar observations (Fig. 1) are carried out by AIOFM.For the vertical columns calculations based on the MAX-DOAS, the height of the boundary layer and aerosol extinction coefficients are deduced from the lidar (Sect.2.4.1).Light source of the lidar is a pulsed Nd:YAG laser with a repetition rate of 20 Hz and 6 ns of pulse duration.It is operating at wavelengths of 532 and 355 nm.A Cassegrainian telescope is designed to receive backscattered light.The detailed descriptions of the lidar system can be found in Chen et al. (2011).

Description of OMI
The Ozone Monitoring Instrument (Levelt et al., 2006) onboard the NASA Earth Observing System (EOS)-Aura satellite was launched on 15 June 2004.It is capable of monitoring global atmospheric NO 2 via observation of backscattered sunlight in the wavelength range of 270 nm to 500 nm.The crossing time for OMI is 13:45 (local time) on the ascending node.Compared with NO 2 satellite observations from GOME, GOME-2, and SCHIAMACHY, OMI can provide a data set with higher spatial (up to 13 km × 24 km) and temporal resolution (daily global coverage).The OMI retrieval of NO 2 vertical columns is based on the DOAS method and consists of three steps: determination of slant column densities, conversion to vertical column densities (VCDs) using the so-called air mass factor (AMF), and estimation of the stratospheric contribution.The detailed description of this retrieval process can be found in Bucsela et al. (2006).
Ground-based measurements cover a limited spatial area, where vertical column densities may not completely represent the whole spatial extent of the satellite ground pixel (Louisa et al., 2008).Airborne measurements (Dix et al., 2009;Heue et al., 2011;Sluis et al., 2010) are costly and usually cannot be carried out on a regular basis.Mobile DOAS observations, on the one hand, can provide more data points within a given satellite ground pixel and be used to detect variability and gradients (Wagner et al., 2010).Furthermore, they can easily be conducted regularly.In the current study, the OMI tropospheric NO 2 data product of NASA is used (Bucsela et al., 2006).To achieve a better comparison between OMI NO 2 and mobile DOAS data, the OMI tropospheric NO 2 VCDs are gridded onto a 0.1 • × 0.1 • grid (Wenig et al., 2008;Chan et al., 2012).

Retrieval of vertical columns for tropospheric trace gases
The DOAS technique has been employed in numerous applications that use artificial light sources or sunlight with instruments mounted on various fixed or mobile platforms (see Platt and Stutz, 2008, for a comprehensive overview).The DOAS evaluation procedure is described in this section in relation to our mobile observation of sunlight intensities in a zenith viewing direction, and the retrieval of tropospheric SO 2 and NO 2 from these intensity spectra.Details of the DOAS analysis are presented in Platt and Stutz (2008).The spectral evaluation applied to each recorded spectrum while the van moves along the measurement route starts with dark current and offset corrections, followed by the division with a Frauenhofer reference spectrum.Subsequently, a high-pass filter is applied to the logarithm of this ratio to separate the broad and narrow band spectral structures.Differential slant column densities (DSCD, the difference between the slant column density of a measurement spectra and the reference; slant column density, SCD, defined as the concentration integrated along the light path), are then obtained by fitting narrow band spectral absorption cross sections to the processed measurement spectra.Figure 2 illustrates such a DOAS fit for NO 2 and SO 2 using the WinDOAS software package (Van Roozendael et al., 2001) for mobile DOAS measurement.

Atmos
During our retrieval process, a spectrum is first selected arbitrarily on the upwind path as the reference spectrum to obtain the concentration distribution trends along the route.The minimum concentration of SO 2 and NO 2 along each driving route is then chosen as the Frauenhofer spectra to re-retrieve the measurement spectra.The spectrum with the lowest measurement value for each day is then chosen as the Frauenhofer reference to minimize the effect of potential instrumental changes on different days.
The wavelength range of 310 nm to 324 nm with three strong absorption peaks is selected for the SO 2 fit.Absorption cross sections of SO 2 , NO 2 , HCHO and O 3 at 293 K are taken from Bogumil et al. (2003).Apart from the Frauenhofer reference spectrum, a Ring spectrum is also included in the fit as described in Hönninger et al. (2004).All absorption cross sections are convoluted with the instrument's slit function to adapt to the instrument resolution during the fit process.The emission peak from mercury lamp at 334 nm is recorded to yield instrument's slit function.The synthetic Ring spectrum is generated from the measured Frauenhofer reference spectrum using the DOASIS software (Kraus, 2006).For the analysis of NO 2 , the spectral window of 345 nm to 365 nm is selected, and the cross sections of O 4 at 298 K (Greenblatt et al., 1990) and HONO at 298K (Stutz et al, 2000) are also included, except for NO 2 , HCHO, O 3 at 293 K (Bogumil et al., 2003) and Ring spectrum.The wavelength calibration is performed using a highly resolved solar spectrum (Kurucz et al., 1984) convoluted by the instrument's slit function.An example for such a spectral fitting is shown in Fig. 2, where the fit residual of NO 2 (Fig. 2a) and SO 2 (Fig. 2b) are 6.54 × 10 −4 and 1.37 × 10 −3 , respectively, due to the unknown structures and system noise.The fit uncertainties of retrieved values from these two spectra for NO 2 and SO 2 are about 1.80 and 2.30 %, respectively.For all measured spectra the fit uncertainties are less than 15 % for NO 2 and 20 % for SO 2 .
The procedure described so far yields differential slant column densities relative to the reference spectrum.The assumption by Johansson et al. (2008) and Rivera et al. (2009) that around noon the slant column densities for the zenith viewing direction approximate those of the vertical columns is now adopted.The tropospheric vertical columns obtained from the MAX-DOAS observation at the time the mobile DOAS passes its location (Fig. 1) is used in order to eliminate the NO 2 and SO 2 VCD in the reference.However, this method is based on the assumption that no strong spatial variations exist within the distance represented by the MAX-DOAS measurement (typically several kilometers).The variability of the mobile DOAS VCDs in the form of standard deviation of its values along such a typical distance is shown in Fig. 3a and b for SO 2 and NO 2 , respectively.The mobile DOAS VCDs are averaged for about twenty values from twenty spectra in Fig. 3a and b.Since the integration time of an individual spectrum is 10 s and the average van speed is 70 km h −1 , this means the van moves about 4 km in the time these twenty spectra are taken.In terms of these standard deviations, for most days no significant variations can be found.Exceptional large variations occur for SO 2 on 29 November, and 2 and 4 December, and for NO 2 on 29 November, and 4 and 21 December.These maybe related to the variations of the emission sources themselves.In general, however, the above assumption of generally small spatial variations appears reasonable to us.
The difference between the MAX-DOAS VCD and the mobile DOAS differential VCD is considered as the "cleanair" background (this difference is called "offset").The mobile DOAS differential VCDs are then converted into absolute columns by adding this "offset" to all values along the path.
The retrieval of vertical columns from the MAX-DOAS observations is described by Xu et al. (2010).Differential slant column densities are converted into tropospheric VCDs using tropospheric differential air mass factors (DAMFs) ac-cording to the following formula: The radiative transfer model SCIATRAN (Rozanov et al., 2002) is used to calculate DAMFs under the assumption that aerosol and trace gas profiles are given by constant values below the boundary layer height and exponential profiles above.The height of the boundary layer was on average about 1.5 km during our measurements from the lidar observations.For the radiative transfer calculations, aerosol extinction coefficients within the layer are taken from the lidar, while the NO 2 and SO 2 mixing ratios are obtained from the long-path DOAS measurements.Aerosol and trace gas profiles above the boundary layer are taken from the LOWTRAN database (Kneizys et al., 1988)  MAX-DOAS is a potential source of error.We estimated the effect on our retrieval to be less than 5 % by varying the setting of NO 2 and SO 2 loading (2.5 × 10 11 molec cm −3 , 6.25 × 10 11 molec cm −3 and 1.25 × 10 12 molec cm −3 ) within the boundary layer.Another uncertainty lies in the AOD (Aerosol Optical Depth, AOD) obtained from the lidar boundary layer height and extinction coefficients.A sensitivity study for different AODs (0.3, 0.5, 0.8 and 1.1), fixed boundary layer height of 1.5 km and constant NO 2 and SO 2 concentrations of 1 × 10 12 molec cm −3 shows a resulting change for the NO 2 and SO 2 AMFs at 20 • elevation of less than 10 %.

Emission calculation
The mobile DOAS measurement allows emission from pollutant sources to be further quantified for known wind direction, wind speed, and car speed.The total SO 2 and NO 2 emissions from a source are determined using the following Eq.( 2) (Johansson et al., 2009;Ibrahim et al., 2010): where VCD gas is the vertical column density, V m is the car speed, V w ⊥ is the component of wind speed perpendicular to the driving direction, and t is the time for acquiring one spectrum (i.e., integration time, about 9 s in this study).In this equation, the sum refers to all the measurement spectra along the path.Assuming additional factors for the partitioning between NO and NO 2 and the finite NO x lifetime, mobile DOAS measurements allow furthermore determining the total NO x (= NO + NO 2 ) emission (Ibrahim et al., 2010;Shaiganfar et al., 2011): (3) Here, R is ratio of NO x and NO 2 .This value is considered to be about 1.38 during daytime with an uncertainty of about 10 % according to the measurement results of Su et al. (2008) in the PRD.The factor c τ accounts for the destruction of originally emitted NO x on the way to measurement location.It depends on the wind speed (W ) and measurements distance (D) in the following way: The NO x lifetime t depends, amongst others, on the ozone and OH concentrations, as well as meteorological conditions.We here take a value of 5 h (Lin et al., 2010).The resulting c τ is in the range of 1.10-1.35assuming an average distance of 9 km from the emission source to the measurement location.
We estimate the uncertainty of c τ due to the uncertainties of the NO x lifetime, wind speed and distance to be about 10-15 %.
Wind direction and speed are taken from the miniature mobile weather station installed on the van.The car speed, location, and direction of each segment are from the GPS.The main errors in the estimation of SO 2 emission are uncertainties in the VCD retrieval (including uncertainties in the DOAS fit of SCDs and errors due to the conversion from SCDs to VCDs) and wind field (including the uncertainties of wind direction and wind speed).Errors of NO x emission calculations include the uncertainties in R and c τ .We estimate the error caused by approximating VCDs by SCDs with radiative transfer calculations of the zenith AMF to be about 10-15 %.Taking all the above error estimates into account results in an error for our VCD retrieval of less than 20 % and 25 % for NO 2 and SO 2 , respectively.The wind field is the largest error source not only due to its variation with time and location within the encircled area, but also because of uncertainties of its vertical distribution.The error caused by the wind field is discussed in detail in Sect.3.2.1.

Results and discussion
3.1 Distribution of SO 2 and NO 2 around GEA

Identification of emission sources
In this section, the distributions of SO 2 and NO 2 vertical columns along the measurement path are analyzed following the results of the mobile DOAS data described in Sect. 2. The possible emission sources are discussed further using these distributions for different wind fields.
Figure 3 shows the resulting MAX-DOAS VCDs during the measurement period.Figure 3a and b also illustrate the mobile DOAS VCDs and the offsets between the two instruments.The average offsets of SO 2 and NO 2 are 7.83 ± 1.42 × 10 15 molec cm −2 and 7.02 ± 2.41 × 10 15 molec cm −2 , respectively.An example result of SO 2 after SO 2 offset correction on 25 November is shown in Fig. 3c.Three peaks are also found in Fig. 3a  and b.The peaks of SO 2 (peak 1, peak 2, peak 3) are related to strong emissions of pollutant sources under the northeasterly wind.The peaks of NO 2 (peak 1 and peak 2) on 11 and 29 November are presumably caused by the same pollutant source, but the peak on 21 December (peak 3) may be related to the transport from other regions by the northwesterly wind.
Figure 4a shows the wind direction and wind speed during the whole experimental period.The wind came mostly from the north and northeast with the exception of 21 and 27 November, 5 and 13 December (southeast direction), 30 November and 21 December (northwest), and 4 and 23 December (east).Average wind speed at ground level ranged from 1 m s −1 to 5 m s −1 during the measurement period.According to the distributions of SO 2 and NO 2 around the path, four types of wind fields are distinguished in our study: north or northeast, southeast, east, and northwest.Typical spatial distributions of SO 2 and NO 2 vertical columns around GEA for these four different wind fields are shown in Fig. 5.The maps for SO 2 in Fig. 5 show a peak in the north irrespective of wind direction, suggesting an additional emission source outside the GEA and north from it.The comparison for wind direction from north/northeast and east versus west further suggests that it is located in the northeast rather than the northwest.No such peak exists for NO 2 in the north if the wind is coming from the north, so that we infer a SO 2 source (no dominant NO 2 emission) lies in the northeast of GEA (S1 in Fig. 1).
For the southeast wind, downwind peaks are found simultaneously for SO 2 and NO 2 in the northwest corner of our route (Fig. 5b).The corresponding downwind peaks for the reverse wind direction from the northwest are less pronounced, and thus sources of NO 2 and SO 2 within the encircled measurement area, but closer to the northern measurement route, can be identified (S2 in Fig. 1).The location of peaks in the downwind region for the north and northeast wind (Fig. 5a and c) is consistent with this conjecture.
Increased values of NO 2 appeared for all wind directions in the southeast/south of the measurement route, especially on 29 November and 4 December.These high column densities can most likely be attributed to traffic emissions from Humen Bridge in the south of GEA (S3 in Fig. 1), the main channel from Shenzhen and Dongguan to Guangzhou.During the Asian Games, the Guangzhou government set up a security checkpoint here, which occasionally caused traffic jams that enhanced high NO 2 concentration.

Comparison with OMI NO 2
We compare the results from mobile DOAS with OMI observations in this section in two ways.First, the NO 2 VCDs of the mobile DOAS are averaged for all mobile DOAS observations of a given day within one specific ground pixel (113.50 • E-113.75 • E, 22.75 • N-23.00 • N) and compared to OMI values within this pixel.Second, in order to explore spatial patterns, all NO 2 VCDs along the measurement path of the mobile DOAS are compared to OMI values on the 0.1 • × 0.1 • grid, as specified in Sect.2.3.Mobile DOAS values were averaged for all measurement days within the pixels of this higher resolved grid.
The comparison of NO 2 VCDs between the two instruments for the above-ground pixel is shown in Fig. 6.Our mobile DOAS values were higher than the OMI values most of the time, especially for the measurements from 27 November to 7 December.Aside from the averaged value of OMI for large areas and its insensitivity to the near-surface pollutant sources, the influence of clouds was also considered and treated as a dominant factor for the OMI and mobile DOAS observations.
Cloud fractions for our 14 days of measurement (Table 1) varied considerably during the measurement period.A cloud fraction of 0.4 was used as threshold to filter both data for comparison.This fraction was exceeded on 21, 27, and 29 November, as well as 7 and 9 December.Figure 7 shows the correlation between the two measurements for cloud fractions smaller than 0.4 for the ground pixel (113.50 • E-113.75 • E, 22.75 • N-23.00 • N).The correlation coefficient (R 2 ) is 0.88, indicating that relative changes observed by OMI and mobile DOAS were coinciding.
Although both instruments agree in general, high NO 2 VCDs were likely underestimated by OMI.These discrepancies, however, cannot be simply attributed to clouds.For example, the 22 and 23 December cloud coverage was lower, yet the discrepancies were higher compared to those of the 11, 19, and 25 November with higher cloud coverage.The disagreement between both data sets could also be caused by different assumptions on the calculation of their air mass factors, including different aerosol and trace gas profiles, ground albedo, etc.Furthermore, mobile and satellite measurements refer to slightly different times.Our mobile DOAS measurements were carried out from 10:00 to 13:00 (LT) and from 11:00 to 12:00 (LT) for the area of the satellite pixel (113.50  1 1 3 . 4 5 1 1 3 . 5 0 1 1 3 . 5 5 1 1 3 . 6 0 1 1 3 . 6 5 1 1 3 . 7 0   2     observations of NO 2 in general show little variation between 11:00 and 13:00 LT, see Fig. 8 (a similar observation was made for Hong Kong in Chan et al., 2012), we conclude that for our case the effect of the daily NO 2 cycle between 11:00 and 13:00 LT is of minor importance for this comparison of mobile and satellite measurements.
We now compare spatial patterns as described at the beginning of this section.Figure 9 shows that both mobile DOAS and OMI captured the high NO 2 VCDs in the northern and southern part of the mobile DOAS route (in the northern part due to industrial emissions, in the southern part due to vehicle emissions from the Humen Bridge), as well as low VCDs  in the western part (white circles in Fig. 9).The correlation analysis for these three areas (indicated by white circles in Fig. 9) is shown in Fig. 10.The correlation coefficient (R 2 ) is 0.72, suggesting that both observations indeed agree reasonably well.However, no such agreement can be found when all grid cells are taken into account.One reason for the lack of correlation might be that data are compared for the same location, but not the same time.Comparing two values for times when the diurnal variation cannot be neglected inevitably leads to mismatch between both data sets.According to Fig.   tions were carried out in the eastern part of the GEA.Lacking further data from longer measurements or models, we do not attempt a correction of this mismatch, but notice that it might have significant influence even for the relatively limited time of our mobile measurements.

Emission of SO 2 and NO x from GEA
As a key component in the estimation of SO 2 and NO x emissions, wind fields can be the source of large errors.Wind fields are generally used to estimate emissions about 400 m above ground level (according to the stack height of a power plant and the height of an elevated plume as discussed by Mellqvist et al., 2007).A wind profile radar could provide highly accurate data, but was not available for our measurements.Wind data from our mobile weather station are used instead.To minimize the error caused by the influence of surface processes on the wind field, measurement days with relatively constant wind field from ground to 400 m height are selected to estimate emissions.Figure 4b compares the wind direction at 400 m height to the surface from 9 November to 22 December.The wind field at 400 m is taken from observations at Kings Park (station number: 45004, 114.16  altitudes generally agree, with a relative deviation of ∼ 30 • .Based on this variation, we estimate the error of our emissions due to change of wind direction with height to be 20 %.While the wind direction in the higher atmosphere might vary little over a relatively small area, the wind speed can change significantly for different locations.We do not take into account differences in the wind speed profiles between different locations.Instead, the ratio of wind speeds at 400 m to that at ground level was estimated to be 1.2-1.3according to the national standard (GB/T 3840-91) from the environmental protection agency of China (Ministry of Environmental Protection of the People's Republic of China, 1992).So we simply use the wind speed at ground to estimate emission errors caused by uncertainties in the wind speed profile and arrive at a value of about 20-30 %.
Total emissions of SO 2 and NO x from the encircled GEA area obtained from Eq. ( 2) are shown in Fig. 11 for 14 days with a fractional cloud cover lower than 0.4 and a constant wind field.The entire measurement period is divided into five phases: before the Asian Games (9 and 11 November, phase 1), during the Asian Games (15, 17, 19, and 23 November, phase 2), the time between the Asian Games and the Asian Para Games (2, 3, and 4 December, phase 3), during the Asian Para Games (13, 17, and 19 December, phase 4), and after the games (21 and 22 December, phase 5).SO 2 emissions varied strongly between the different phases (by a factor of 5), where the lower and lowest emissions occurred during the Asian Games and Asian Para Games, respectively.A clear pattern for NO x emissions was not found, although these emissions decreased at the beginning of the games and increased again after their end.
The reduced emission of SO 2 during the Asian Games and Asian Para Games could be caused mainly by the pollution control strategy of the local environmental protection agency.Meteorological conditions also play an important role for emissions from GEA. Persistent rainfall from 10 to 12 December and again from 14 to 16 December helped to remove air pollutants, thereby decreasing emissions from GEA during the Asian Para Games.Table 2 summarizes the average SO 2 emission for the different phases of the measurement period.For times outside the games (phases 1, 3, and 5), SO 2 emission was estimated to be 9.50 tons per hour (approximately 83.2 × 10 3 tons per year), which is consistent with the value of 84.60 × 10 3 tons per year from the emission inventory (Guangzhou Municipality State of the Environment, 2010).On the other hand, the average NO x emission for this time was estimated to be 5.87 tons per hour.During the games (including phases 2 and 4) the emissions of SO 2 and NO x from GEA were reduced by 53.50 % and 43.50 %, respectively, compared with the time outside the games (including phases 1, 3, and 5).
Errors in the calculation of emissions are due to errors in the VCD retrieval, wind field, and car speed.The total error in the VCD retrieval was estimated to be less than 20 % for NO 2 and 25 % for SO 2 .The error of the car speed was about 1 % according to the GPS.The errors caused by wind direction and wind speed were about 20 % and 20-30 %, respectively, or about 30-35 % together.The total error for the estimate of SO 2 emissions amounts to about 35-40 %.Addi-tional uncertainties caused by R and c τ of 10 % and 10-15 %, result in a total error of NO x emissions of about 40-45 %.

Influence of SO 2 emissions from the GEA on the downwind region
Knowledge of emissions and transport from large area sources is crucial for the control and management of local environment problems.Guangzhou University Town (23.05 • N, 113.37 • E) hosted numerous events during the Asian Games, which raised issues on the air quality of the town.According to the geographical relationship between GEA and Guangzhou University Town from Fig. 1, the pollutants from GEA can diffuse to Guangzhou University Town when the wind direction is southeast.The SO 2 concentration in Guangzhou University Town is expected to increase due to the contribution of GEA emission.SO 2 is selected as the tracer gas because it originates from industrial emissions, unlike NO 2 which is still affected by local vehicle emissions.To investigate how emissions from GEA affect Guangzhou University Town for southeasterly winds, a long-path DOAS instrument is set up there to monitor possible downwind transport of SO 2 .The paths of air masses for southeasterly wind on 21 and 27 November, and 5 and 13 December are shown in Fig. 12.On these days, the wind traversed GEA and the SO 2 concentrations measured downwind at the University Town monitoring site by our long-path instrument were significantly higher, as illustrated in Fig. 13 (shaded box).Similar phenomena can be observed on 24 and 28 November and 10 and 12 December in Fig. 13, when the wind was also coming from the southeast.The average SO 2 concentrations for southeasterly and non-southeasterly winds measured by the long-path DOAS at the downwind location are 72.55 ± 13.06 µg m −3 and 24.58 ± 11.72 µg m −3 , respectively.Daily averages of SO 2 are enhanced by a factor 3 when air masses traversed GEA, compared to days when the wind came from other directions.Therefore, GEA is concluded to be a major contributor to SO 2 pollution in the Guangzhou University Town area.

Conclusions
In this paper, mobile DOAS observations carried out in GEA during the Guangzhou Asian Games from November to December 2010 are reported.These measurements were used   varying traffic emissions.Possible emission sources were determined to explain these distributions using the information from different wind fields.Pollutant sources in the northeast of GEA, outside the closed measurement route and sources in the north, within the area encircled, were also identified.Our NO 2 vertical columns were compared with OMI data and were found to be similar in spatial patterns.The correlation coefficient (R 2 ) of the vertical columns after cloud filtering was 0.88 within a specific ground pixel, but the absolute values measured by our mobile DOAS were mostly higher than the OMI data.SO 2 and NO x emissions from the GEA during the Asian Games period were also calculated.Lower and lowest emissions of SO 2 were found to occur during the Asian Games and Asian Para Games, respectively.Outside the Asian Games period, the average emission of SO 2 was estimated to be 9.50 ± 0.90 tons per hour (83.2 thousand tons per year), which is consistent with the value of 84.6 thousand tons per year from a local emission inventory.In comparison, the average emission of NO x was estimated to be 5.87 ± 3.46 tons per hour.During the games, the emissions of SO 2 and NO x were reduced by 53.50 % and 43.50 %, respectively.The error of total emissions was estimated to be about 40-45 % for NO x and 35-40 % for SO 2 .Using LP-DOAS measurements at the Guangzhou University Town, emissions from GEA were found to have a distinct impact on air pollution in this area depending on the wind direction.SO 2 concentrations were found to be about three times larger during our measurements when air masses crossed GEA.

Fig. 1 .
Fig. 1.Location of the Guangzhou eastern area (GEA).(a) The mobile DOAS measurement path encircling this area is indicated by the closed red line.SL: Shilou in Guangzhou City, location of MAX-DOAS and lidar; GU: Guangzhou University Town, location of LP-DOAS.S1, S2, and S3 represent three different pollutant sources.The maps of SL and GU are shown in (b) and (c), respectively.The arrow indicates the viewing direction (north) of the MAX-DOAS in (b).The light-path of LP-DOAS is about 788 m in (c).

Fig. 2 Fig. 3
Fig. 2 Example fit for (a) NO 2 recorded at 10:42 (LT) on 27 November, 2010 and (b) SO 2 at10:35 (LT) on 29 November, 2010 by our mobile DOAS instrument.Black lines denote the differential optical densities (DODs) of processed measurement spectra.Red lines show the result of the fit.The number given by VCD m is the differential VCD with respect to the Frauenhofer reference spectrum as described in section.

Fig. 2 .
Fig. 2. Example fit for (a) NO 2 recorded at 10:42 (LT) on 27 November 2010 and (b) SO 2 at 10:35 (LT) on 29 November 2010 by our mobile DOAS instrument.Black lines denote the differential optical densities (DODs) of processed measurement spectra.Red lines show the result of the fit.The number given by VCD m is the differential VCD with respect to the Frauenhofer reference spectrum as described in Sect.2.4.1.

Fig. 3
Fig. 3 Offset between MAX-DOAS and mobile DOAS values defined as the 781

Fig. 3 .
Fig. 3. Offset between MAX-DOAS and mobile DOAS values defined as the difference between these values for the measurement period for SO 2 (a) and NO 2 (b).The mobile DOAS value represents the average over about twenty spectra data acquired when the mobile DOAS passed the MAX-DOAS location.The MAX-DOAS value is the VCD acquired at the time of the passing by.(c) shows the example for SO 2 on 25 November 2010.Red line: original value from mobile DOAS; green line: result after adding the offset.Error bars indicate the standard deviation for the mobile DOAS as described in the text.
Fig. 4 (a): Time series of average wind direction and wind speed from 10:00 to13:00 (LT) during the measurement period recorded by our mobile miniature weather station.(b): Comparison of wind direction at surface and 400 m height.The surface wind fields are from the mobile weather station.The wind field at 400 m is taken from observations at Kings Park (station number: 45004, 114.16N, 22.31E) in Hong Kong (http://weather.uwyo.edu/upperair/sounding.html).

Fig. 4 .
Fig. 4. (a) Time series of average wind direction and wind speed from 10:00 to 13:00 (LT) during the measurement period recorded by our mobile miniature weather station.(b) Comparison of wind direction at surface and 400 m height.The surface wind fields are from the mobile weather station.The wind field at 400 m is taken from observations at King's Park (station number: 45004, 114.16 • E, 22.31 • N) in Hong Kong (http://weather.uwyo.edu/upperair/sounding.html).

Fig. 6 .Fig. 7
Fig. 6.Time series of NO 2 VCDs measured by mobile DOAS (black symbols) and OMI (red symbols) for the satellite ground pixel given by the coordinates (113.50 • E-113.75 • E, 22.75 • N-23.00 • N).The measurement time of the mobile DOAS within this area is from about 11:00 to 12:00 (LT).The bars show the standard deviations of OMI and mobile DOAS as described in the text.

Fig. 8
Fig. 8 Mean daily NO 2 cycle of the LP-DOAS from 11 November to 25 December 817

Fig. 9
Fig. 9 Comparison of NO2 VCD patterns measured by mobile DOAS and OMI (ground pixel: 113.25E-113.75E,22.75N-23.25N)during the whole measurement period (9 November to 23 December, 2010).The color coded circle indicates the mobile DOAS observations.The white circles in the northern and southern part show particularly high NO2 values captured by both instruments.The white circle in the western part contains low NO2 values for both instruments.The grid resolution is 0.1° × 0.1°.

Fig. 9 .
Fig. 9. Comparison of NO 2 VCD patterns measured by mobile DOAS and OMI (ground pixel: 113.25 • E-113.75 • E, 22.75 • N-23.25 • N) during the whole measurement period (9 November to 23 December 2010).The color-coded circle indicates the mobile DOAS observations.The white circles in the northern and southern part show particularly high NO 2 values captured by both instruments.The white circle in the western part contains low NO 2 values for both instruments.The grid resolution is 0.1 • × 0.1 • .

Fig. 10 .
Fig. 10 Correlation analysis of OMI and mobile DOAS NO 2 VCDs for the three 832

Fig. 10
Fig. 10 Correlation analysis of OMI and mobile DOAS NO 2 VCDs for the three areas (white circles) in Fig. 9. Values are averaged over the whole measurement period for both instruments.

Fig. 11 .
Fig. 11.SO 2 and NO x emissions estimated from mobile DOAS measurements between 10:00 to 13:00 (LT).Green lines indicate the beginning and end of the Asian and Asian Para Games.

Fig. 12 .
Fig. 12. Backward wind trajectories for the 200 m (red), 400 m (blue), and 800 m (green) heights at the Guangzhou University Town monitoring site from the NOAA HYSPLIT model (http://www.arl.noaa.gov/index.php) on 21 (a) and 27 (b) November, and 5 (c) and 13 (d) December.The star indicates the Guangzhou University Town monitoring site, the black box the measurement area of the mobile DOAS measurement.

Fig. 13 .
Fig.13.Daily average SO 2 concentrations measured by long-path DOAS at the Guangzhou University Town site.The shaded days indicate days when the wind came from the southeast and mobile DOAS measurements were conducted.For days with similar wind situations on 24 and 28 November and 10 and 12 December, we could not conduct mobile DOAS measurements due to vehicle limitation.

Table 1 .
Cloud fraction for the 14 days of measurement.Italic text represents the cloud fraction < 0.4.Bold text represents the cloud fraction > 0.4.Cloud fractions are taken from the OMI data product.

Table 2 .
SO 2 and NO x emissions as measured by our mobile DOAS for the different phases described in the text.