Laboratory for Climate and Ocean-Atmosphere Studies, Department of Atmospheric and Oceanic Sciences, School of Physics, Peking University, Beijing, China
Corresponding author: C. Li, Laboratory for Climate and Ocean-Atmosphere Studies, Department of Atmospheric and Oceanic Sciences, School of Physics, Peking University, Beijing 100871, China. (email@example.com)
Corresponding author: C. Li, Laboratory for Climate and Ocean-Atmosphere Studies, Department of Atmospheric and Oceanic Sciences, School of Physics, Peking University, Beijing 100871, China. (firstname.lastname@example.org)
 Using a CIMEL Sun photometer, we conducted continuous observations over the urban area of Shanghai (31°14′N, 121°32′E) from 18 April 2007 to 31 January 2009. The aerosol optical depth (AOD), Angstrom wavelength exponent, single scattering albedo (ω0), and aerosol particle size distribution were derived from the observational data. The monthly mean AOD reached a maximum value of 1.20 in June and a minimum value of 0.43 in January. The monthly averaged Angstrom wavelength exponent reached a minimum value of 1.15 in April and a maximum value of 1.41 in October. The frequencies of the AOD and Angstrom wavelength exponent presented lognormal distributions. The averaged ω0 at 550 nm was 0.94 throughout the observation period, indicating that the aerosols over Shanghai are composed mainly of scattering particles. The concentrations of coarse mode and accumulation mode aerosols over Shanghai were highest in spring compared with other seasons, especially for those particles with radii between 1.0 and 2.0 μm. The median radius of monthly averaged accumulation mode aerosols increased with increasing AOD, and fine particles accounted for the majority of the aerosol volume concentration. The ratios of the monthly averaged volume concentration of accumulation mode and coarse mode aerosols (Vf/Vc) were over 0.6 for all months studied and reached up to 1.94 in August. The volumes of the two modes changed with AOD, but their correlations presented different sensitivities, that is, the volume concentration of accumulation mode aerosols was more sensitive to variations in AOD than that of coarse mode aerosols. The aerosol volume concentration decreased with increasing ω0, indicating that the higher the volume concentration of aerosols, the higher the absorption in particle extinction properties. The increase in absorption was caused primarily by secondary species coated on black carbon (BC) and primary organic carbon (POC) particles.
If you can't find a tool you're looking for, please click the link at the top of the page to go "Back to old version". We'll be adding more features regularly and your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Aerosols, which are solid or liquid particles suspended in the atmosphere, are important indicators of climate change [Intergovernmental Panel on Climate Change, 2007]. Comprehensive knowledge of the optical and microphysical properties of aerosols forms the foundation of research on the environmental and climatic effects of aerosols. Given that the physical and chemical properties of aerosols are characterized by high spatial and temporal variability, the study of their influence over the environmental and climatic conditions depends on the determination of their spatial and temporal distributions, as well as on the accurate estimation of their optical properties. The optical properties of aerosols and the characteristics of their formation into cloud condensation nuclei are governed by aerosol chemical composition, particle size distribution, and particle hygroscopic ability. Emission sources and emission mechanisms indicate that different types of aerosols have varied optical and radiation characteristics [Eck et al., 1999; Dubovik et al., 2002; Kim et al., 2004]. Kiehl and Briegleb  concluded that the average size and chemical characteristics of particles exert considerable effect on aerosol optical properties. To reduce uncertainty regarding such properties, adding to our knowledge of aerosols is essential. Increased data on aerosols is especially vital to enhancing our understanding of aerosol chemical compositions, including sulfate, nitrate, ammonium, dust, sea salt, secondary organic aerosol, and EC/OC, particle size distribution, and optical characteristics. Simulating the effect of aerosols on atmospheric radiation by solving the radiative transfer equation necessitates the determination of AOD (τ(λ)), single scattering albedo (ω0(λ)), and asymmetry parameter. The optical properties of natural and anthropogenic aerosols determine the different roles they play in radiative forcing, thereby resulting in varied contributions to climate change. Many studies indicate that ω0 governs the positive or negative effects of radiative forcing, that is, heating or cooling; this effect is also related to surface reflectance. AOD determines the magnitude of radiative forcing [Hansen et al., 1997]. In addition, most aerosol radiative models calculate the radiative properties (e.g., τ(λ) and ω0(λ)) of aerosols by parameterization of their physical and chemical characteristics (e.g., particle size, shape, and composition) [Koepke et al., 1997; Hess et al., 1998]. The uncertainty in aerosol models (e.g., particle size, refractive index, ω0, and sphericity) is one of the primary sources of error in the satellite remote sensing of atmospheric AOD [King et al., 1992; Kaufman and Sendra, 1988], for some prior aerosol models are usually used in the retrieval of AODs from satellite observations. Bergstrom and Russell  estimated that when 1δ uncertainty exists, the change in ω0 (at 0.55 μm) of aerosols over the North Atlantic reaches up to 0.07, which means that the corresponding uncertainty of solar fluxes at the top of atmosphere will be about 21%. Elucidating atmospheric aerosol optical properties and their effect on radiation balance necessitates ascertaining the composition, size distribution, and total concentration of atmospheric aerosols [Santer et al., 1996]. Clarifying the mechanisms of aerosol radiative forcing depends on a thorough understanding of aerosol optical properties, as well as their spatial and temporal variations in different regions [Ackerman and Toon, 1981; Remer et al., 1997; Jacobson, 2001; Satheesh and Moorthy, 2005].
 Numerous observational studies on the aerosol optical properties over China or East Asia have been conducted. Mao et al.  analyzed the characteristics and regularity of change in the AOD over Beijing, as well as its relationship with meteorological conditions. In recent years, several scientific experiments on the optical properties and radiative characteristics of aerosols have been conducted in East Asia; these include the East Asian Studies of Tropospheric Aerosols – An International Regional Experiment [Li et al., 2007a], Asian Atmospheric Particle Environmental Change Studies [Nakajima et al., 2003], and Asian Sky Radiation Observation Network Experiment [Kim et al., 2004]. The above mentioned initiatives focused on inland aerosols, which are closely related to dust aerosols and urban pollution in large cities. Significant progress has also been made with regard to ground-based remote sensing networks in China, such as the Aerosol Remote Sensing Network (CARSNET) developed by the China Meteorological Administration [Che et al., 2009], the Chinese Sun Hazemeter Network developed by the Chinese Academy of Sciences [Li et al., 2007a], and the Sun and Sky Radiometer Network development in China [Takamura et al., 2002]. However, research on aerosol optical properties over near coastal urban areas has yet to be conducted. The Yangtze River Delta is located in the middle of China's eastern coast. It is a sub-tropical and temperate transition zone characterized by both monsoon and oceanic climates. The formation of atmospheric aerosols is influenced by the combined effects of continents and oceans. This region is also a densely populated and well-developed core economic area. Due to rapid development in recent years, high aerosol loading over this region has led to significant effects on the local crop production, ecological environment, and regional climate.Luo et al.  computed and analyzed the variation characteristics of China's AOD using data drawn from nearly 30 years' historical measurements in 46 national solar radiation stations and found that the Yangtze River Delta (YRD) region has experienced a rapid increase in AOD. Xu et al.  observed the physical, chemical, and radiative properties of aerosols over the Yangtze River Delta. Their study showed that the aerosol concentration over the agricultural areas is comparatively equal to that over any highly polluted agricultural region. Wang et al.  found large ranges of daily variations and weak seasonal variations in AOD with heavy aerosol loading and steady aerosol modes in the YRD region. Chen et al.  studied aerosol optical properties and their spatial and temporal variability in four sites in the Hangzhou region. Xia et al.  retrieved and analyzed the characteristics of aerosols in the Taihu area based on Sun photometer and surface irradiance data. However, while all of these studies contribute to the understanding of aerosol optical properties over the YRD region, they mostly focus on either variations in AOD and Angstrom index in the YRD region or relatively few samples from several seasons.
 We used Sun photometer measurements of the automated tracking CE318 (CIMEL Company) and combined them with aerosol data that were simultaneously gathered using in situ Aethalometer and GRIMM 180 air sampling equipment for black carbon (BC) and PM10/PM2.5 measurements, respectively, to determine the long-term characteristics of aerosol optical properties over a typical coastal urban region. The observational data were also used to analyze the optical and microphysical properties of aerosols over Shanghai. Parameters, such as AOD, Angstrom wavelength exponent,ω0(λ) at 550 nm, and volume distribution were derived from Sun photometer observations.
 The rest of the paper is organized as follows: section 2 provides a description of the observation site and experimental equipment, section 3 describes the algorithm used in deriving the key optical and microphysical parameters of aerosols, section 4 discusses the mixed optical characteristics of urban and marine aerosols over Shanghai, and section 5 presents the conclusions.
2. Experimental Site and Instruments
 The observation equipment used in this research was set on the roof of the Shanghai Pudong Meteorological Bureau building (31°14′N, 121°32′E; elevation = 14 m). Figure 1 shows the location of the observation equipment and the geographic features of Shanghai. Shanghai City is situated at the eastern tip of the Yangtze River Delta and halfway along China's eastern coastline; that is, it is adjacent to the East China Sea. As one of the most developed regions in the country, Shanghai has a population of 18.58 million, as well as numerous industrial enterprises and vehicles (1 million, as reported by the 2007 Statistical Bulletin of Shanghai), which emit a substantial volume of pollutant gases and aerosol particles. Around Shanghai, there are couple of densely populated cities, including Nanjing and Hangzhou, each of which has a population greater than 2 million, Suzhou, Wuxi, Ningbo, and Changzhou, each of which has population between 1 and 2 million, Nantong, Yangzhou, and Zhenjiang, each of which is inhabited by 0.5 to 1 million people, and Shaoxing, Taizhou, Huzhou, Jiaxing, and Zhousan, each of which has a population between 0.2 and 0.5 million. These cities are almost contiguous, constituting the Yangtze Delta Megalopolis. A vegetative region of about 10 km2 is located to the east of the observation site and the rest of the area is surrounded by commercial and residential buildings from all other directions. The observation site is located near a city traffic road and is exemplified by typical urban surface characteristics.
 Observations of column aerosol optical properties were conducted with a CE318 Sun photometer during the daytime for the entire duration of the study period. The instrument is commonly used to measure direct solar radiation at 8 channels with wavelengths of 1020, 936, 870, 670, 500, 440, 380 and 340 nm. It is stable and easy to operate and can withstand a variety of severe weather conditions, requiring minimal maintenance under such situations. The Sun photometer mostly employs atmospheric window channels, except for the 936 nm channel, which measures the intensity at the strong water vapor absorption band. The 670 nm channel also measures the intensity in the atmospheric window channel but presents weak ozone absorption. The bandwidth of 6 channels from 440 nm to 1020 nm is 10 nm, and the other two on 380 nm and 340 nm is 2 nm. CE318 runs automatically in accordance with scheduled procedures. It initiates observations in the morning when the air mass is 6.0 (the solar elevation angle is about 9°), after which the air mass decreases with increasing solar elevation angle. It stops operating in the evening when the air mass reaches 6.0 again. Continuous observations were conducted from 18 April 2007 to 31 January 2009. The Sun photometer was first calibrated in 2006 at Izana Observatory, one of the remote stations of the European mainland, after which it was brought to Lin'an, Zhejiang, for Langley calibration experiments from 8 August to 17 August 2007 and then to Beijing from 1 January to 10 January 2008 by inter-comparison with the China Aerosol Remote Sensing Network (CARSNET) masters at the Chinese Academy of Meteorological Sciences; hence, the observations may have been interrupted. The master instrument was calibrated by Langley plot analysis at the Mauna Loa Observatory following the Aerosol Robotic Network (AERONET) calibration protocol. The AOD difference between the master and the recalibrated instrument ranged from −0.005 to 0.015. The calibration of diffuse radiance was done once a year with the standard laboratory integrating sphere, and its relative uncertainty was better than 4% to 5% [Holben et al., 1998; Eck et al., 1999]. The process of cloud filtering and data quality control for direct irradiance retrieval was based on Smirnov et al. , while that for diffuse radiance retrieval was based on Holben et al. . We used SKYRAD (version 4.2) developed by Nakajima et al.  and Dubovik and King  to retrieve some parameters, such as ω0(λ) and volume size distribution. SKYRAD was developed for the PREDE Sun photometer, which is different from the CE318 Sun photometer, so transformation of CE-318 raw data to SKYRAD inputs was necessary. ASTPWin software was used to transform K7 format files from CE318 to ASCII files, after which the AODs were calculated. The ALL and ALR files produced included left and right scans of the almucantar plane, respectively. After cloud filtering and data quality control of the ASCII files, the validated data were re-arrayed according to the format of SKYRAD raw data. Preprocessing of input data used the altitude of the station to calculate the atmospheric pressure and the monthly TOMS satellite climatology data to calculate the ozone optical depth [Che et al., 2009]. The ground albedo was assumed to be 0.1 during analysis and remained unchanged except when the surface of the measurement site was extraordinary. The solid view angle of the CE318 Sun photometer was difficult to estimate because it cannot scan a small domain round the sun like the PREDE mode does. Therefore, the ω0(λ) and volume size distribution were retrieve through the AERONET algorithm from CE318 measurement data during the Lin'an calibration experiments. Then the solid view angle in SKYRAD was changed by iterative procedure using the above retrieved parameters from the AERONET algorithm. Finally, we obtained the converged solid view angle value that was used to calculate whole data sets, which represent aerosol observations through spring, summer, autumn, and winter. Valid sample numbers in each month are listed in Table 1. The mechanisms behind temporal variations in the aerosol optical properties were also analyzed.
Table 1. Sun Photometer Observation Samples From 18 April 2007 to 31 January 2009
 The SKYRAD method allows the size distribution, phase function, and ω0(λ) of the aerosols to be obtained from measurements of the direct solar irradiance (F) and diffuse sky radiance (E). When measured at ground level, these components are given by:
where F0 is the extraterrestrial solar irradiance, ω is the single scattering albedo, τaer(λ) is the AOD, P(Θ) is the total phase function at scattering angle Θ, q(Θ) is a multiple scattering term, and m0 is the optical air mass that can be approximated as m0 = 1/cosθ0 as long as the zenith angle θ0 ≤ 75°. ΔΩ is the solid view angle of the sky radiometer.
Nakajima et al. developed an optimized radiative transfer code for a plane-parallel atmosphere called the REDuced Multiple scattering program. To work with the diffuse component, the ratio R(Θ) is defined:
where β(Θ) is the total scattering coefficient, which includes molecular and aerosol scattering.
 The two quantities τ(λ) and β(Θ) can be expressed as:
where Kext and K are kernel functions. x = 2πr/λ, is the size parameter, v(r) is the columnar volume spectrum, which is defined as the volume of aerosol for an air column of unit cross section within a unit of logarithmic radius interval: v(r) = dV/d ln r (in cubic centimeters per square centimeters), rm and rM are minimum and maximum aerosol radii, respectively, and is the aerosol complex refractive index.
 The behaviors of Kext and K approximately determine the radius interval of reliable information of aerosol optical and physical features. Inspection of several kernels at refractive indices typical of the atmospheric aerosol shows that this interval ranges from 0.03 to 3 μm and 0.06 to 10 μm for only extinction and scattering data, respectively. When measurements of normalized diffuse sky flux are used together with the usual measurements of extinction, the radius interval ranges from 0.03 to 10 μm.
 The idea of the method is to iteratively eliminate the multiple scattering term q(Θ) from the data R(Θ) to uncover the coefficient β(Θ). In each of the steps, the code obtains the size distribution v(r) by the inversion of β(Θ) and τ(λ). This distribution is used as input for the radiative transfer code for recalculating in turn R′(Θ), which is compared with the experimental data to evaluate the average square difference ε(R). The process is repeated until the deviation is less than 10%.
Ångström  provided a formula that expresses the spectral dependence of AOD on the wavelength of incident light:
where τaer(λ) is the AOD, β is the Angstrom turbidity coefficient, which can be used to describe the general haziness of the atmosphere, and α denotes the Angstrom wavelength exponent, which can be obtained from the AODs of three wavelengths (i.e., 440, 670, and 870 nm) using the second order polynomial fit [Eck et al., 1999; Holben et al., 2001].
 The ω0(λ) is retrieved at wavelengths of 440, 670, 870, and 1020 nm [He et al., 2010]. Therefore, the ω0(λ) at a wavelength of 550 nm could be determined by linear interpolation as follows:
 A distribution function must be introduced to quantitatively analyze the statistical characteristics of the retrieved size distribution of aerosol particles. Many studies show that the bimodal lognormal distribution function can accurately describe the actual size distribution of aerosol particles [Shettle and Fenn, 1979; Remer and Kaufman, 1998], and has been validated as being universal and available in the YRD region. One of the validating studies is that of Xia et al. :
where C is the volume concentration of the particles, r represents the median radius, σ denotes the standard deviation, and the subscripts f and c represent the accumulation and coarse mode aerosols, respectively.
 This function divides the size distribution of aerosol particles into coarse and accumulation mode aerosols so that six parameters can be used to describe size distribution. These parameters are calculated as follows: Volume mean radius (logarithm of the average radius)
Standard deviation of the volume mean radius
Volume concentration (μm3/μm2)
4. Results and Discussions
4.1. Aerosol Optical Depth and Particle Size
4.1.1. Seasonal Characteristics
 The AOD and Angstrom exponent are basic parameters that characterize aerosol optical properties. They can be used to calculate the atmospheric aerosol content, determine aerosol mode and size distribution, and validate satellite-retrieved data related to aerosol optical properties. They are also key factors that determine the radiative and climatic effects of atmospheric aerosols [Hess et al., 1998].
Figure 2 shows monthly averaged changes in the AOD and Angstrom exponent over the study period. The monthly averaged AOD was lowest in January (0.43) and its maximum value (1.20) was achieved in June, the only month when it exceeded 1.0. Pan et al.  analyzed the characteristics of the AOD and Angstrom exponent using CIMEL Sun photometer data from 2007 to 2008 in five sites located in the YRD region and found results similar to those in the current work: the AOD over Shanghai is largest in the summer and reaches a maximum monthly average in June. In the present observational study, the Angstrom exponent was lowest from March to May, when the AOD was less than the one in the summer. The weather during autumn was always sunny, with rare precipitation and clear atmospheres. The monthly averaged Angstrom exponent was larger in autumn than in summer. The winter weather is dry and cold with stronger wind. Heavy wind is conducive to the dispersion of aerosols and the hydroscopic increasing effect of aerosol particles is weak due to lower relative humidity, so the average AOD was low during the winter.
 The higher AOD and lower Angstrom exponent in spring compared with those in autumn and winter may be partly caused by dusty weather due to the long-distance transport of dust from the surrounding and remote areas. The YRD region could be affected by dust storms from North China, as well as local pollution sources in spring, when precipitation is relatively low [Gong et al., 2003; Cui et al., 2009]. As shown in Figure 3, the days of dust and PM10/PM2.5 observed during the spring months significantly increased, especially in March, during which the location of interest experienced 3 days of dust weather, according to the weather record. In fact, while the main aerosols originated from local direct emissions (primary particles) and in situ nucleation (secondary particles), they mostly concentrated in the boundary layer [Yu et al., 2012]. The impact of dust particles extends to the middle troposphere, impacting the columnar averaged aerosol size and reducing the Angstrom exponent.
 The extreme variability of Shanghai's column aerosol optical properties in summer was caused mainly by regional static meteorological conditions and the regional transport of anthropogenic fine pollution particles from inland areas. Figure 4shows the spatial distribution of 850 hPa seasonal average wind vectors and geopotential height over East Asia. This distribution was derived from NCEP/NCAR reanalysis data from June 2007 to January 2009. In summer, the atmosphere of Shanghai was dominated by continental air masses from the southwest and the clean sea air had no effect on the city, such that a large quantity of aerosols emitted from the local YRD region accumulated under static atmospheric conditions. The presence of intense solar radiation in the summer promoted the gas-particle transformation effect in the atmosphere, thereby producing more fine aerosol particles.Buzorius et al. analyzed the mechanisms of secondary aerosol formation over the Asian continent by airborne, shipborne, and ground-based observations in the ACE-Asia experiment. From the ground-based observation of aerosol size distribution, the authors found that Aitken nuclei particles increase with the rise in nucleation mode aerosols. They also found that newly formed particles are composed primarily of ammonium sulfate. These airborne observations indicate that the increasing concentration of aerosols is often accompanied by increasing concentrations of SO2. Yamaji et al.  found that from May to June, the O3concentration of the boundary layer over East Asia is at its maximum, 30% to 60% of which originates from the regional anthropogenic emissions produced by photochemical effects. In addition, the relative humidity of the lower atmosphere of Shanghai in June is high, and the growth of hydrophilic aerosol particles increases the AOD. Researchers found that the atmosphere over East Asia contains a substantial amount of sulfate and ammonium, both of which are highly sensitive to gas-particle transformations and hygroscopic growth. Aerosols in Asia exhibit more pronounced hygroscopic growth effects than those in Europe and the eastern United States. Frequent biomass burning in eastern China in June also contributes extensively to the increase in AOD.Tian et al.  analyzed the consumption of biomass fuels in 1990s in rural China and found that straw is the main source of biomass emission. Cao et al. studied the emission factor of biomass consumption in Chinese provinces and cities in 2000 and found that the biomass consumption in eastern China is mainly induced by the open burning of straw. The optical properties of the aerosols produced by biomass burning are related to the age of the particles. Coagulation and gas-to-particle transformation increase the aerosol size such that the particles are transformed into accumulation mode aerosols with peaks near 0.2μm [Eck et al., 2003].
 In summary, the AOD over the Shanghai area during the spring and summer is large, indicating that the atmosphere throughout the two seasons is turbid, and this turbidity may be attributed to different factors. In spring, atmospheric aerosols include a mixture of dust brought about by northerly or northwesterly winds and local emissions, whereas aerosols in the summer are comprised of local urban/industrial aerosols when heavy wind days do not occur frequently, which could cause the accumulation of local pollutants due to stable weather [Duan and Mao, 2007]. High aerosol loading in summer is typically caused by hygroscopic growth under high humidity conditions, gas-particle transformations under high temperature, and high levels of radiation. This result is consistent with that ofZhou and Xiang , who viewed considerable urban air pollution as the direct cause of the increase in atmospheric turbidity.
4.1.2. Frequency Characteristics of AOD and Particle Size
 During the observation period, the AOD ranged from 0.0 to 3.0. This range was divided by an interval of 0.2, after which the frequency of the AOD was computed at each interval. The Angstrom exponent ranged from 0.0 to 2.0. The frequency of the Angstrom exponent at each 0.2 interval was calculated using the same method employed in the AOD calculations. The frequencies of the two aerosol parameters are plotted in Figure 5. The two frequencies are of lognormal distribution, that is, the average value of the data sets can represent the most likely value at each observation. The distribution of the AOD was positively skewed, whereas that of the Angstrom exponent was negatively skewed. The solid and dashed lines in Figure 5 represent the fitting curves of the frequency of the Angstrom exponent and AOD, respectively. F, x, and σ represent the peak, mean and standard deviation of lognormal distribution, respectively. The AODs were mostly concentrated at 0.2 to 1.0, accounting for 80% of the total value. Low cases of AOD (AOD <= 0.6) accounted for 47% of the total value during the study period, moderate cases of AOD (0.6 < AOD < 1.0) accounted for 33%, and high cases (AOD > 1.0) accounted for 20%. AODs greater than 1.6 rarely occurred, accounting for 8% of the total value. The most common AODs were those that ranged from 0.4 to 0.6, accounting for 32% of the total value.
 The frequency distribution of those samples with the Angstrom exponent (α) ranging from 1.4 to 1.6 accounting for 38% of the total samples. Those samples with α > 1.8 and α < 0.6 accounted for 0.3% and 1% of the total, respectively. The total samples are consistent with the lognormal distribution (mean, 1.29; median, 1.34), which is consistent with the results of a previous study on the characteristics of aerosols located in the YRD region by Pan et al. . This asymmetry is caused by the occasional occurrence of large-particle aerosols that affect Shanghai. Basically, aerosols over Shanghai are fine particles that originate from urban/industrial and biomass burning aerosols. On the basis of their average value, we deduce that aerosols over Shanghai are mainly continental aerosols. Such types of aerosols occur because locally or regionally transported dust particles, hygroscopic growth, and the combined growth of fine particles under heavy pollution affect the balance of the total number of samples.
4.2. Aerosol Single Scattering Albedo
 In the present study, the column average ω0 at 550 nm, with values ranging from 0.81 to 0.97 (Figure 6), was derived using two-year Sun photometer observations as bases. The observedω0imposed a wide range of variations in the atmosphere over Shanghai and is potentially related to the burning of materials in the city, the type of aerosol formation, long-distance transport, and environmental and meteorological conditions, among others. Thus far, no measurement or retrieval ofω0has been conducted in the Shanghai area; however, ground-based observation data gathered using the Aethalometer indicate that changes in BC concentration exhibit a trend opposite that of theω0recorded by the Sun photometer. Transportation near the observation site is the largest source of pollution. The exhaust gas from numerous vehicles is one of the main factors that influence aerosol formation. The BC from burned oil causes the aerosols over Shanghai to be particularly absorptive. The long-distance transport of dust from the surrounding and remote areas also affects the properties of aerosols over the city. During the observation period, the averageω0 was 0.94 ± 0.02, close to the value (0.93) observed in Lin'an [Xu et al., 2002], which is a rural site and about 200 km away from Shanghai, but far from the value (0.90 in 440 nm and decreasing slightly with wavelength) observed in Beijing [Xia et al., 2006]. Specifically, the ω0 of aerosols over Shanghai was higher than that in Beijing, indicating that the aerosol source and formation mechanism in these areas are somewhat different. Beijing is located in the north. The burning of coal during the winter for warmth produces large amounts of BC, further compounded by the presence of heavy industries in the surrounding areas of the city. As well, the terrain of Beijing is conducive to aerosol accumulation. In contrast, Shanghai is located near the sea. Marine aerosols are more strongly diluted by air pollution, and the absorption of these marine aerosols is very weak. Therefore, the aerosols over Shanghai have less absorption capacity than those in the northern cities of China. The aerosols over Shanghai are mainly generated from combustion processes induced by human activities. These processes include oil combustion, biomass burning, and a mixture of various industrial processes. In recent years, the implementation of industrial transformation, motor restrictions, and other measures has been beneficial to the reduction of aerosol absorption. Levy et al. classified aerosols as follows: urban/industrial aerosols are non-absorbing aerosols, with a representativeω0 value of 0.95; general, forest smoke, and aerosol emissions in developing countries are classified as moderately absorbing aerosols, with a representative ω0value of 0.90; and smoke aerosols from prairie fires are high-absorbing aerosols, with a representativeω0value of 0.85. According to this classification standard, they considered aerosols over Southeast Asia as consisting primarily of non-absorbing aerosols. Observations of Shanghai aerosols in the current work confirm this conclusion. In addition, a recent study byEck et al. verified that the aerosols over Southeast Asia are essentially non-absorbent.
Figure 7 plots the monthly average ω0 (550 nm) derived from the Sun photometer. The figure shows that the minimum (0.92) of monthly averaged values was reached in March; the minimum of all individual observations for a given year was also observed in the same month. A low ω0 generally represents the absorption properties of dust aerosols and biomass burning aerosols [Torres et al., 2005]. Zhang et al.  found that the ω0in East Asia during the spring significantly decreases under the influence of frequent dust storms and long-distance dust transmission. Using TOMS to measure the global UV-band column average ofω0, Hu et al.  also found that the ω0 in East Asia during the spring is at its lowest compared with other seasons. For visible bands, however, dust aerosols tend to be more scattering and less absorptive. The ω0 in Beijing during dust periods was observed to increase significantly from 0.90 to 0.93 compared with anthropogenic air pollution episodes [Xia et al., 2005]. We suppose that the dust transported to Shanghai from remote deserts in Northwest China traveled a long distance over many industrial and urban areas of China. The process may cause the dust particles were severely polluted and contaminated with BC [Jacobson, 2001]. Seinfeld et al.  also demonstrated that aerosols in mixed states (i.e., with addition of BC and other aerosols to the mineral particles) can change dust aerosol radiative effects in many ways. In this study, the maximum monthly averaged ω0 (0.95) was achieved in August. A high ω0generally corresponds to water-soluble aerosols under relatively high humidity conditions [Shettle and Fenn, 1979; Koepke et al., 1997], but many studies have found that the ω0 of urban/industrial aerosols continues to exhibit low values under relative humidity (RH > 90%). Hess et al.  obtained ω0 (550 nm) = 0.82, while Shettle and Fenn  obtained ω0 (550 nm) = 0.84. The lowest ω0 value in August in the present study was 0.85. This monthly mean value in the summer is consistent with observations of Taihu of the YRD region [Xia et al., 2007]. However, compared with the ω0 of other urban/industrial aerosols presented in the references, such as those from Mexico City with ω0 (550 nm) = 0.90 [Hess et al., 1998], Beijing, and Xianghe [Qiu et al., 2004; Xia et al., 2006; Li et al., 2007b], our results are higher. Shettle and Fenn  reported that ω0 (550 nm) = 0.79 under the same humidity conditions. When the relative humidity drops, ω0 decreases. Thus, for the relative humidity of Shanghai, which is generally high, the ω0 should not be very low. BC emissions in China, as observed by Streets et al. , exhibit a seasonal trend similar to that of the ω0 in the present work.
Table 2 shows the seasonal average of ω0 at different wavelengths inverted from the Sun photometer measurements. The average ω0 at each band was consistently higher than 0.91 and showed a decreasing trend with increasing wavelength in the visible bands. The scattering extinctions of winter aerosols at each band were almost the same. The range of ω0 tended to be larger with increasing wavelength in all other seasons except spring, during which dust outbreaks play a critical role in the formation of more coarse mode particles. For each band, the scattering capacity of aerosol particles was weakest in the spring but their absorption capacity during the same season was strongest compared with all other seasons. In contrast, the scattering capacity of aerosol particles was strongest in summer, but their absorption capacity during the same season was weakest compared with all other seasons.
Table 2. Average, Maximum, and Minimum ω0 at Four Bands for the Different Seasons
Figure 8 illustrates the frequency distribution of ω0 for different seasons. As shown in the figure, the ω0 with the highest frequency in spring, autumn, and winter was typically about 0.94; in summer, it was usually 0.96, accounting for 58% of the total number of samples. These results indicate that summer aerosol particles have the strongest scattering capacity. Spring and autumn exhibited a wide distribution range, but this result was especially pronounced in spring. The two seasons are alternate periods of warm and cold and more varied airflow activities can bring about even more aerosol sources, hence a higher probability of change in the scattering and absorption properties of aerosols exists in these seasons. Some samples for the two seasons had a ω0 less than 0.85. These absorption aerosols would impose a significant heating effect on the regional climate. The ω0 values of the summer and winter aerosols were concentrated at 0.86 to 0.96, and the values of 96% of these samples were concentrated at 0.92 to 0.96, indicating that the aerosol properties of the two seasons in Shanghai are stable and that their sources are relatively simple.
4.3. Aerosol Size Distribution
 The retrieved volume size distributions during the study period presented various modes in different samples, including monomodal, bimodal and trimodal. However, most aerosol spectra can be distinguished in the bimodal distribution, and volume concentrations usually reached minima when particle radii ranged between 0.2 and 5 μm. Inversion results for the volume distribution showed that the shape of each mode was relatively close to the lognormal distribution. However, some volume distributions deviated from the lognormal distribution. Research shows that this kind of bias yields only a minimal effect on radiation. In calculating the distribution of two aerosol modes, therefore, minimum volume concentrations between 0.2 μm and 5 μm were selected as the cut-off point for fine mode and coarse mode particles.Figure 9 shows the inversion of the average volume distribution for the four seasons. The form dV(r)/d ln r (the particle volume concentration per unit area per log radius for a vertical column of air) was adopted to denote the aerosol particle size distribution and facilitate graphical representation. The median radius and standard deviation of the particle volume concentrations for each mode of volume size distribution were calculated. Particles produced from coal burning are usually concentrated in areas smaller than 1 μm, whereas particles generated from vehicle exhaust are concentrated in areas smaller than 2 μm. Some observational data revealed that vehicle emissions are the main source of Aitken aerosol pollution (diameter = <0.1 μm). The peak of vehicle exhaust and photochemical smog pollution particles in urban areas is 0.002 μm, and the size of soil dust particles is generally greater than 2 μm [Zhao and Wang, 1991]. We can thus conclude that the first peak is related primarily to human activities, such as fuel combustion generated from coal burning and vehicle exhaust. Tang et al. considered the combustion process as the main source of sub-micron particles and believed that their representative fine particles originate primarily from human activities. The second peak is composed of soil dust particles from transportation, building construction, and sand transportation. Ocean waves can also produce a large number of sea salt particles (typical Angstrom exponent from 1.1 to 1.8), while the long-distance transmission of soil particles from arid and semi-arid areas (such as Northwest China and Inner and Outer Mongolia) can produce numerous coarse particles.
 The aerosol size distribution significantly differed with the changing of seasons. Factors that affect the particle size distribution include the combustion state, fuel properties, the living time of the particles, burning intensity, environmental temperature, and relative humidity. Spring volumes of large and small particles at each aerosol logarithmic radius interval were the largest compared with other seasons. The volume of large particles with sizes ranging from 1 μm to 2 μm was especially high. The concentration of accumulation mode aerosols over Shanghai was lowest in autumn, and that of the coarse mode aerosols was almost equal to that in winter. These results show that the particles generated from gas-particle conversion in autumn significantly decrease whereas large particles increase mainly because of the change in structures of aerosol sources. Particles of 0.1μm in size usually originate from collision and coagulation; such particles simultaneously develop from secondary aerosols. These particles also exhibit very small Brownian motion coagulation almost independent of gravity sedimentation. This effect stabilizes the particles and extends their lifetime, hence they tend to accumulate in the atmosphere. The photochemical reaction of the gas loses energy when the autumn light is weakened, which reduces the number of secondary aerosol particles. The reduction in precipitation affects the removal of large particles, which increases the proportion of coarse mode particles. The median radius of the two aerosol types is smallest in spring and largest in summer. The relative humidity in summer is high; thus, the particle size expands because of hygroscopic growth. Particles with sizes comparable to or greater than the wavelength of visible light determine the optical properties of aerosols and have the greatest contribution to the scattering of visible light, and directly affect atmospheric visibility. These particles are accumulation mode aerosols [Seinfeld and Pandis, 1998], the distributions of which were concentrated within a relatively small range over all of the seasons, among which particles with radii ranging from 0.1 μm to 0.5 μm accounted for 40.6% of the total volume in spring and 48.3% of the total volume in summer. Studying Los Angeles smog aerosols, Whitby and Liu  found that the mass spectrum exhibits a bimodal size distribution, the first and second peaks of which were found at diameters of 0.3 and 5 μm to 15 μm, respectively. From aerosol observations in Shanghai in the present study, the peak center of the aerosol volume distribution, which featured first and second peaks at diameters of 0.2 and 5.0 μm, respectively, is similar to that of Los Angeles smog aerosols.
Table 3 shows the fitting parameters of the monthly averaged bimodal size spectrum distribution and corresponding AOD, precipitation, and relative humidity. The monthly averaged median radius of the accumulation mode aerosols varied from 0.13 μm to 0.20 μm. In general, the median radius of the accumulation mode aerosols increased with increasing relative humidity (correlation coefficient ∼0.48), indicating the hygroscopic growth mechanism of fine particles in high humidity conditions [Kotchenruther et al., 1999]. The monthly averaged relative humidity was relatively higher at increased AOD values. Numerous observations have confirmed this finding. For example, Goddard Space Flight Center (GSFC) observations showed that AOD and columnar water vapor are highly correlated [Smirnov et al., 2000; Holben et al., 2001]. The Tropospheric Aerosol Radiative Forcing, Observational Experiment (TARFOX) experimental observations in the eastern coast of the United States indicate that water vapor adheres to the particle surface and contributes significantly to the increase in AOD [Hegg et al., 1997]. The relative humidity obtained from LIDAR measurements and the effective radius of aerosol particles derived from airborne remote sensing equipment in the TARFOX experiment are also highly correlated [Ferrare et al., 2000]. In addition to the effect of the hygroscopic growth of aerosol particles, the aggregate, touch, and gas-particle conversion processes can also increase the size of the particles [Reid et al., 1998; Jacobson, 2001]. In fact, observations of biomass burning aerosol particles in some cities (such as Mexico City) also showed that the size of fine mode aerosols increases with rising AOD but the relative humidity is not remarkably high in these areas. Using the Mie scattering theory as basis, Yang et al.  studied the effect of relative humidity on the columnar optical properties of aerosol particles. They obtained aerosol data and found that water vapor mainly affects the size distribution and refractive index of aerosol particles, especially those in the visible band. When the relative humidity is greater than 60%, the ability of the hygroscopic particles to absorb water vapor in the atmosphere is highly significant. A water film is gradually formed on the outer layer of dry aerosol particles, resulting in the increased size and change in refractive index of aerosol particles.
Table 3. Monthly Median Radius (r), Peak Volume Concentration (σ), and Volume Average Radius Standard Deviation (C) of the Accumulation Mode and Coarse Mode Aerosols Derived From the Aerosol Volume Size Distributiona
Also given is the ratio of the monthly average volume concentration of the accumulation mode and coarse mode aerosols (Vf/Vc), AOD(τ), as well as the precipitation and relative humidity.
Cumulative precipitation, rainfall within 24 h prior to each valid observation as one sample, sum of the rainfall for all valid observations within one month constitute the cumulative precipitation.
 The change in monthly averaged precipitation shows an imperceptible effect of precipitation on the AOD. On the basis of the size distribution, we found that the concentration of accumulation mode aerosols remains evident and that the peak concentration maintains high levels; in contrast, the concentration of coarse mode aerosols weakens as a result of wet deposition, with the peak concentration significantly dropping. The peak concentrations of coarse mode particles in July, August, and September dropped to about 0.1. Different removal efficiencies of fine and coarse mode particles may partly contribute to this phenomenon, that is, precipitation can dilute coarse particles but fine particles are not easily diluted. Some previous studies [e.g., Samara and Tsitouridou, 2000; He and Balasubramanian, 2008] on aerosol size distribution and scavenging ratios of major ionic species confirmed that precipitation can scavenge large sized particles more effectively than small sized particles. Fine mode particles mainly consist of secondary species that can form quickly after precipitation due to gas-particle conversion processes [Yu et al., 2012], while a longer process is necessary to recover the concentration of coarse mode particles from the sources with approximative constant emission. To distinguish the effects of these two mechanisms causing significant difference in concentration changes of coarse mode and fine mode after precipitation is challenging and beyond the scope of this study but will be paid close attention to in future studies employing other analytical techniques (such as chemistry and in situ measurement).
 By comparing the volumes of the two aerosol types, we found that the majority of the aerosol volume concentration consisted of fine particles. This phenomenon was particularly evident in the summer. In August, the volume of fine particles was almost double that of coarse particles (Vf/Vc= 1.94). In the spring months, the coarse particle volume began to dominate because of dust transport and local wind-blown soil particles. February showed the peak concentration (Vf/Vc = 0.63). By analyzing the aerosol observation data of the WMO background station at Lin'an, Yang et al.  found that the PM10/TSP values in 1991, 2002, and 2003 were all about 90%; the PM2.5/PM10 values for the same years were 46.52%, 69.33%, and 72.29%, respectively. These findings indicate that the aerosols over the YRD are mainly composed of fine particles. Observations from other cities also show similar results, in which the total volume of accumulation mode particles is larger than that of coarse mode particles [e.g., GSFC (Vf/Vc = 3), Creteil (Vf/Vc = 2.5)]. Overall, the results can be analyzed from the observation that the optical properties of aerosol over Shanghai are mainly determined by the scattering of fine particles.
 Many dynamic aerosol models calculate AOD from detailed aerosol microphysical information (such as aerosol concentration, mixing states, and compositions) [e.g., Yu et al., 2012], which can help facilitate the parameterization of aerosols in climate models and satellite remote sensing. Presenting aerosol optical properties as regressions with optical thickness helps outline the dynamics of the aerosol optical properties associated with the growth of the aerosol mass and aerosol processes (i.e., aging, particle size, and composition transformations, etc.) stimulated by the accumulation of aerosols in the atmosphere. Figure 10 shows the relationship between AOD and the aerosol concentration over Shanghai during the observation period. The total aerosol volume concentration increased with rising AOD (correlation coefficient ∼0.78). The accumulation mode and coarse mode aerosols presented an increasing trend with AOD. However, when distinguishing the fitting lines of the two aerosol types, accumulation mode aerosols were more correlated with AOD than coarse mode aerosols (Vf∼AOD correlation coefficient, 0.85; Vc∼AOD correlation coefficient, 0.47). This means that changes in the optical properties of aerosols over Shanghai were caused primarily by the formation, development, and removal of accumulation mode particles. This phenomenon is associated with the characteristics of Shanghai aerosols, given that the main physical processes of these aerosols are accumulation, coagulation, and growth over time. Although the long-distance transmission of large-scale particles also contributes to the aerosol concentration over Shanghai, it does not play a major role in its extinction during highly turbid weather.
Figure 11 depicts the relationship between ω0 and aerosol volume concentration, in which we consider that ω0 is related not only to refraction (mainly the imaginary part of refraction) but also to particle size. The change in aerosol properties and their links with physical, geographical, and meteorological conditions can be detected through this relationship. The aerosol volume concentration decreased with increasing ω0 (correlation coefficient ∼−0.37), indicating that the higher the aerosol volume concentration, the higher the absorption in aerosol particle extinction properties. The slope of the accumulation mode fitting line and its correlation coefficient were close to 0, indicating that the changes in the volume concentration of the accumulation mode have little effect on the scattering capacity of aerosols. However, the volume concentration of the coarse mode aerosols decreased with increasing ω0. Their correlation coefficient was −0.46 (sample number is 1431). This relationship indicates that large particles cause higher absorption in particle extinction capacity. Therefore, when Shanghai is affected by particles with larger size, the absorption of the aerosols particles is significantly enhanced. The YRD is known as one of the world's fastest developing urbanized areas. Vehicular emissions, thermal power plants, and other industrial units are the major polluting sources for BC and primary organic carbon (POC). The secondary species coated on BC and POC particles can significantly increase the size and imaginary parts of the refractive index of these particles, which have important impacts on their optical properties. The imaginary part of the refractive index of mixed compounds that absorb strongly, also known as brown carbon, can reach up to 0.27 [Alexander et al., 2008]. Yu et al. used a global chemistry transport model (GEOS-Chem) incorporating an advanced particle microphysics model to simulate spatiotemporal variations in AOD and found that coated BC and POC particles make important contributions to AOD in East Asia.Figure 11 suggests that the increase in AOD associated with the formation of brown carbon characterized by larger sizes could substantially increase the overall aerosol absorption.
5. Conclusions and Discussions
 The mean column optical and microphysical properties of aerosols over Shanghai were analyzed using two-year observational data gathered from Sun photometer measurements. To the best of our knowledge, this study is the first to quantitatively analyze the long-time series characteristics of aerosol optical properties derived from Sun photometer measurements in the Yangtze River Delta region in China. Our results reflect both similarities in analytical results and inconsistencies with the findings of previous research. By analyzing the long-time series inversion data on aerosols, we found that the AOD in Shanghai decreased through summer, spring, autumn, and winter. The AOD reached a maximum value of 1.20 in June and a minimum value of 0.43 in January. The Angstrom wavelength exponent reached a minimum value of 1.15 in April and a maximum value of 1.41 in October, indicating that the mixture of dust and local industrial emissions in spring is the main cause of turbid air over Shanghai. In summer, the high AOD was caused mainly by regional static meteorological conditions and the regional transmission of anthropogenic fine particle aerosols from inland areas. In addition, the frequencies of the AOD and Angstrom wavelength exponent both presented lognormal distributions. During the observation period,ω0 decreased with increasing wavelength. The average ω0 at 550 nm was 0.94, indicating that the aerosols over Shanghai are dominated mainly by scattering particles. March exhibited the lowest ω0 average value (0.92), whereas August showed the highest (0.95). This result is similar to the observations in Lin'an, Mexico City, and the Maldives. However, the aerosols over Shanghai are less absorptive than those over Beijing, suggesting that the source and formation mechanism of aerosols in the two cities are different. The average ω0 is different from that generated by the aerosol absorption model adopted in the Moderate Resolution Imaging Spectroradiometer (MODIS) operational retrieval at NASA/GSFC for aerosol products over this area. Frequent dust weathers that occur in spring and the secondary species coated on BC and POC particles were the main causes of the significant reduction in ω0. The low ω0 may change the cloud properties and indirectly affect radiative forcing [Ackerman et al., 2000]. Numerous models [Koepke et al., 1997; Hess et al., 1998] typically adopt the conclusion that the ω0 of dust aerosols ranges from 0.63 to 0.87. However, the broadband spectrum ω0 from aerial surveys is high at 0.95 [Fouquart et al., 1987]. Such a large difference considerably influences climate simulations because a small change in ω0 in the radiative forcing calculation of dust aerosols may transform the calculation results from heating to cooling [Hansen et al., 1997]. The median value of ω0 at 550 nm was 0.94 in the spring, autumn, and winter and 0.96 in the summer. The volumes of the two modes changed with AOD but their correlations presented different sensitivities, that is, the volume concentration of accumulation mode aerosols is more sensitive to variations in AOD than that of coarse mode aerosols. The aerosol volume concentration decreased with increasing ω0, indicating that the higher the volume concentration, the higher the absorption in the particle extinction properties. The increase in absorption was caused by the larger size particles, which mainly consist of secondary species coated on BC and POC particles. Urbanization is one key feature in the YRD region, and the observation site is located in the typical urban surface. Therefore, observations of the properties of aerosols over Shanghai present significant implications in future research on aerosol satellite remote sensing and climate forcing in the YRD and East Asian region. In particular, the annual and seasonal characteristics of aerosol changes can be assimilated to the General Circulation Model (GCM). The data can also be used to validate the simulation results of the GCM.
 The authors are grateful to OpenCLASTR project for using SKYRAD package in this research. Thanks are also given to Research team of Atmospheric Chemistry in Shanghai Meteorological Bureau for their kind suggestions on this paper. The study is partially supported by the research grants from the National Natural Science Foundation of China (NSFC, grants 40705013, 40775002, and 41175020), the Shanghai Science and Technology Committee Research Special Funds (grants 10JC1401600, 10231203803 and 10231203900) and the China Meteorological Bureau Public Welfare Special Funds (grant GYHY201106023). We would like to thank the three anonymous reviewers, whose useful comments have improved the paper.