Aerosol Optical Characteristics in Fukuoka and Beijing Measured by Integrating Nephelometer and Aethalometer : Comparison of Source and Downstream Regions

Akihiro UCHIYAMA Center for Global Environmental Research, National Institute for Environmental Studies, Tsukuba, Japan Bin CHEN Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences, China Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, Nanjing University of Information Science & Technology, China Akihiro YAMAZAKI Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan Guangyu SHI Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences, China Rei KUDO Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan Chiharu NISHITA-HARA, Masahiko HAYASHI Fukuoka Institute for Atmospheric Environment and Health, Fukuoka University, Fukuoka, Japan Ammara HABIB Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences, China and Tsuneo MATSUNAGA Center for Global Environmental Research, National Institute for Environmental Studies, Tsukuba, Japan (Manuscript received 28 February 2017, in final form 19 January 2018)


Introduction
Aerosol characteristics are an important factor in Earth's radiation budget, which is influenced by radiatively active gases, aerosols, and clouds.Aerosols change the radiation budget directly by absorbing and scattering solar radiation and indirectly through their role as cloud condensation nuclei, thereby increasing cloud reflectivity and lifetime (e.g., Ramanathan et al. 2001;Lohmann and Feichter 2005).The variation in the observed surface solar radiation depends on the presence of clouds, aerosols, and radiatively active gases.Aerosols disturb the solar radiation that reaches the Earth's surface.Some aerosols scatter solar radiation and enhance the planetary albedo, whereas others absorb solar radiation and trap energy in the climate system.
These processes are controlled by the aerosol optical properties: the scattering, absorption, and extinction coefficients; the single-scattering albedo (SSA), which is the ratio of the scattering coefficient to the extinction coefficient; and the light-scattering phase function.Therefore, the aerosol optical properties are important factors.In the 1970s, the importance of the aerosol optical properties was recognized (Yamamoto and Tanaka 1972), and measurement programs were initiated in several locations, including the South Pole, Mauna Loa, and Point Barrow (McComiskey et al. 2004;Delene and Ogren 2002;Sheridan et al. 2001).Awareness of the effect of aerosols on climate radiative forcing led to an increase in the number of measured variables and measurement sites in the 1990s.
In this study, several aerosol optical characteristics were measured using an integrating nephelometer and an aethalometer in two East Asian cities, Beijing and Fukuoka, from 2010 to 2014.Beijing is a well-known megacity in China whose economic activity has continuously increased over the past 30 years, resulting in increases in the population and number of vehicles.Fukuoka is one of the largest cities in western Japan.In the mid-latitude region, synoptic disturbances move from west to east.This movement causes air masses to also move from west to east, and the observation sites in Japan are thus affected by air originating from the continental area.Therefore, the modification of the aerosol characteristics during the transport of aerosols can be investigated by comparing the aerosol characteristics of the source and downstream cities.Furthermore, the aerosol characteristics were better clarified by comparing these two cities.
The Institute of Atmospheric Physics (IAP) of the Chinese Academy of Sciences (CAS) and the Meteorological Research Institute (MRI) of the Japan Meteorological Agency (JMA) have been measuring the aerosol optical properties and the surface downward solar irradiance to investigate the effect of the aerosol optical properties on the surface radiation budget as part of a cooperative Chinese and Japanese science and technology program.In this research program, in situ ground-based measurements of the scattering and absorption coefficients have been performed in Beijing and Fukuoka using an integrating nephelometer and an aethalometer.
The objective of this study was to characterize the aerosol optical properties in Beijing and Fukuoka using these measurements.The aerosol optical characteristics can be well understood by comparing the measurements obtained in the two cities.Some previous measurements of the aerosol properties in Beijing have been made over week-to month-long periods of intensive measuring campaigns, but few long-term, season-crossing observations have been reported.A two-year measurement survey by He et al. (2009) is the only season-crossing observation reported thus far.In the present study, measurements were performed over a four-year period.Using these data, the aerosol characteristics and their frequency distributions could be reliably obtained.However, the trends of the optical properties were not investigated because the fouryear measurement period is insufficient for such an investigation.
Section 2 describes the data and methods used in this study, the location of the observation sites, and the calibration of the scattering coefficients.Section 3 gives the monthly means and frequency distributions of the investigated optical properties.In Section 4, the characteristics of the optical properties are classified based on their extinction and absorption Ångström exponents, which are indices of the size distribution and the absorption composition, respectively.Section 5 describes the optical characteristics observed during winter in Beijing and spring in Fukuoka.The results are summarized in Section 6.

Instruments and measurement period
The scattering and hemispheric backscattering coefficients were measured using an integrating nephelometer (Aurora 3000, Ecotech, Australia).Using LED light sources, the nephelometer simultaneously measures the scattering coefficients at 450 nm (blue), 525 nm (green), and 635 nm (red).The angle ranges of the light sources are 9° -170° for total scattering and 90° -170° for hemispheric backscattering.Generally speaking, the inlet temperature is higher than the ambient temperature.Therefore, the relative humidity in the inlet of the nephelometer is lower than that of the outside air.This makes it difficult to measure the scattering coefficient at the outside air temperature and humidity.The effect of hygroscopic growth was removed, and the scattering and hemispheric backscattering coefficients were measured under dry conditions.The inlet of the nephelometer has a processor-controlled automatic heater, and the relative humidity threshold was set to 30 %.It was confirmed that the relative humidity in the inlet was less than 30 %.The instrument was operated at a flow rate of approximately 5 L min −1 (nominal value).
The absorption coefficients were measured using an aethalometer (Model AE31, Magee Scientific, USA) at seven wavelengths: 370, 450, 520, 590, 660, 880, and 950 nm.The aethalometer measured the attenuation of a beam of light transmitted through the sample collected on a quartz fiber filter while the filter continuously collected samples.The instrument was operated at a flow rate of 1 L min −1 in Beijing and 4 L min −1 in Fukuoka.Since aerosol concentration in Beijing was high, we reduced the flow rate so that the aethalometer operated stably.The absorption coefficient can be accurately measured using the recently developed photoacoustic method (Arnott et al. 1999) or the photothermal interferometric method (Sedlacek and Lee 2007).However, filter-based instruments were used because of their stability and ease of operation.Most filter-based absorption coefficient techniques suffer from various systematic errors that require correction (Coen et al. 2010;Weingartner et al. 2003;Arnott et al. 2005;Schmid et al. 2006;Virkkula et al. 2007).All of the scattering and absorption coefficient data were recorded as 1-min averages, and 30-min averaged data were used for data analysis.
The scattering and absorption coefficients were observed over a period of four years in each location: from March 2010 to February 2014 in Beijing and from August 2010 to May 2014 in Fukuoka.In the period from March 2010 to September 2011 in Beijing, the nephelometer was used without hemispheric backscattering measurements.Although the details are not described here, the differences between the analyzed results with and without hemispheric backscattering measurements were small.
In June 2011, the light source of the nephelometer installed in Fukuoka was discovered to be malfunctioning.During the period from the middle of January to June 2011, the extinction Ångström exponents α ext were very large in comparison with those from other periods.On the basis of this unusual discrepancy, it was assumed that the light source began to malfunction in the middle of January.Therefore, the data from this period were not used, and the measurement was restarted in January 2012.During the period from July 2012 to February 2013 in Beijing, all instruments were stopped while the room where the instruments were installed underwent renovation.

Observation sites
The aerosol optical properties were measured in Beijing, China, and Fukuoka, Japan, the locations of which are shown in Fig. 1a.Beijing is located in the area bordering the North China Plain and the Inner Mongolia plateau and is surrounded by the Taihang Mountains to the west and the Yanshan Mountains to the north.Beijing is a megacity with a population of more than 21,500,000.The measurements were made at the IAP (116.38°E,39.97°N), which is located in the northern part of the urban area of Beijing.The IAP is surrounded by a number of research institutes and residential and business complexes, and there are no factories nearby.The instruments were installed in a room on the roof approximately 35 m from the ground and 92 m above sea level.Sample air from outside the building was drawn into the instrument through an electric conductive tube passed through a window.The length of the tube was approximately 1.5 m, and the tube was connected to an isokinetic inlet.This inlet is not size-selective.The sample air was branched and guided to each instrument.It was confirmed that the instruments did not interfere with each other.The room was not air-conditioned.
Fukuoka is located on the northern shore of the island of Kyushu, facing the Sea of Japan, and is surrounded by the Sefuri Mountains to the south and southwest.Fukuoka is Kyushu's largest city, with a population of approximately 1,500,000.The measurements for this study were conducted at Fukuoka University Campus (130.36°E,33.55°N), which is located in the western part of the urban area of Fukuoka approximately 6 km from the sea and 1.5 km from the mountains.The university is surrounded by a number of residential quarters, and there are no factories nearby.The instruments were installed in a room on the fourth floor, approximately 15 m from the ground and 23 m above sea level.Sample air from outside the building was drawn into the instrument through an electric conductive tube passed through a window.The length of the tube was approximately 1.5 m.As in Beijing, the tube was connected to an isokinetic inlet.This inlet is not size-selective.It was also confirmed that the instruments did not interfere with each other.The room was air-conditioned to maintain a temperature of 25°C.
During the observation periods, some construction was done near both observation sites, which may have affected the measurements.

Calibration of instruments
The nephelometer is able to regularly monitor the output of the instrument by measuring the calibration gases without changing the calibration coefficients.Filtered air and CO 2 gas were used for the zero check and span check operations, respectively.The calibration check of the nephelometer was performed once per week at midnight.Correction coefficients were calculated from the calibration check data after each calibration check, and a quadratic function of time was fit to these coefficients using the method of least squares.
The aethalometer did not require a special calibration because it measures transmittance, which is a relative value.An important factor that influences measurement precision is the flow rate because it determines the sampling volume.The flow rate was measured using a precision soap film flow meter and compared with the recorded values; these measured flow rates were within 0.5 % of the recorded values.

Data processing method
Filter-based instruments are widely used at ground sites.However, most filter-based absorption coefficient techniques suffer from various systematic errors that require correction (Liousse et al. 1993;Petzold et al. 1997;Bond et al. 1999;Weingartner et al. 2003;Arnott et al. 2005;Schmid et al. 2006;Virkkula et al. 2007;Coen et al. 2010).In this study, the method by Coen et al. (2010) was used.Their correction scheme is based on four previously published methods that account for the optical properties of the aerosol particles embedded in the filter (Weingartner et al. 2003;Arnott et al. 2005;Schmid et al. 2006;Virkkula et al. 2007).
Because multi-wavelength scattering coefficient data can be used, the performance of the correction method developed by Coen et al. (2010) is expected to be very good.
Integrating nephelometers are widely used to measure aerosol scattering coefficients; however, they cannot measure light scattered in extreme forward or backward directions (scattering angles near 0° and 180°; Heintzenberg and Charlson 1996;Anderson et al. 1996;Anderson and Ogren 1998;Müller et al. 2009).To correct this truncation error, information on aerosol absorption properties and the particle size distribution is necessary (Bond et al. 2009).This study employed a recently developed method that uses multi-wavelength scattering and absorption coefficient data to correct the scattering coefficients (Uchiyama 2014, see Supplement 1).

Backward trajectory
To determine the characteristics of air masses during each season, backward trajectory analysis was conducted starting at a height of 500 m above the observation sites every 4 h from 2004 to 2013 in Fuku oka and from 2008 to 2013 in Beijing.These data were used to clarify the seasonal variation in the aerosol properties.When considering the seasonal variation of the aerosol characteristics, the resident time, which is the time an air mass spends within a given region before reaching the target city, was calculated from the backward trajectory data.The backward trajectory analysis was performed using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler 1999).

Aerosol properties
This section presents the time series, monthly and annual means, and frequency distributions of the aerosol properties.The extinction, scattering, absorption coefficients, the SSA, and the absorption Ångström exponent among aerosol properties are described in detail.The asymmetry factor, the extinction Ångström exponent, and the volume fractions of coarse and fine modes are briefly described and shown in detail in Supplement 2.

Extinction, scattering, and absorption coefficients
First, the measured extinction, scattering, and absorption coefficients C ext , C sca , and C abs are discussed.Figures 2a and 2b show the time series of the monthly mean scattering and absorption coefficients and their standard deviations.The monthly and annual mean values of the aerosol properties are given in Tables 1 and 2. As shown later, the frequency distributions of the aerosol properties deviate considerably from the normal distribution, but the mean value and the standard deviation are expressed in the form of mean ± standard deviation.
Some previous measurements of the aerosol optical properties in Beijing were made over week-to monthlong periods of intensive measuring campaigns (Bergin et al. 2001;Yan et al. 2008;Li et al. 2007;Garland et al. 2009), but few long-term, season-crossing observations have been reported (Table 3).He et al. (2009) studied the aerosol optical properties in Beijing using two-year data.According to He et al. (2009), the twoyear averages and standard deviations for C abs (532 nm) and C sca (525 nm) were 56 ± 49 and 288 ± 281 Mm −1 , respectively.The extinction coefficient with no wavelength correction was approximately C ext » C sca + C abs = 288 Mm −1 + 56 Mm −1 = 344 Mm −1 .The value of C abs obtained in the present study is smaller than that obtained in their study, and the values of C ext and C sca in the present study are larger than those obtained in their study.However, because the standard deviations are very large in Beijing, it is difficult to determine whether there is a significant difference between the present results and those obtained by He et al. (2009).Table 3 also gives measurement values that have been reported in other previous studies.Because these measurements were not conducted during the same time periods, it is difficult to compare the long-and short-period data.
Figures 2a and 2b and Tables 1 and 2 show the seasonal variation in C ext , C sca , and C abs ; in summer, these coefficients were small, whereas in winter, they were large.The seasonal variation in Fukuoka was more distinct than that in Beijing.
To investigate the characteristics of the air masses in every season, the resident time was calculated using the results of the backward trajectory analysis.The study area was divided into eight regions, and the area outside of East Asia was classified as the outside region (Fig. 1b).Tables 4 and 5 give the frequency with which each region had the longest resident time during the five days prior to the air mass arriving in the target city.The backward trajectory was analyzed monthly to determine these frequencies.
The results of the backward trajectory analysis also yielded the seasonal variation of the origins of the air masses arriving in the two target cities.Most air masses that reached Fukuoka in summer, winter, and spring were affected by the West Pacific Ocean region (region 8), the North Continent region (region 1), and the North Continent and Japan regions (regions 1 and 6), respectively.Most air masses that reached Beijing from November to April and May to October were affected by the North Continent region (region 1) and the East China region (region 3), respectively.The seasonal variation in C ext , C sca , and C abs in Beijing is unclear.The coefficient values in autumn and winter were large.He et al. (2009) found that C ext and C sca were largest in summer (Table 2 in their paper).However, according to the present results, the coefficients were not necessarily large in summer.The results of the sky radiometer analysis by Che et al. (2014) showed that the optical thickness was largest in summer and smallest in winter.Sky radiometer measurements cannot be made in very hazy conditions because the direct solar irradiances cannot be measured, making it difficult to distinguish between clouds and heavy haze.Therefore, the average value in winter was weighted by the results from days with light haze.As shown in Fig. 2b, C sca and C abs varied drastically in  (525 nm): extinction coefficient at a wavelength of 525 nm in units of Mm −1 C sca (525 nm): scattering coefficient at a wavelength of 525 nm in units of Mm −1 C abs (520 nm): absorption coefficient at a wavelength of 525 nm in units of Mm −1 ω 0 (525 nm): single-scattering albedo at a wavelength of 525 nm G (525 nm): asymmetry factor at a wavelength of 525 nm α ext : Ångström exponent for extinction coefficients α abs : Ångström exponent for absorption coefficients α abs_sw : Ångström exponent for absorption coefficients at wavelengths shorter than 520 nm α abs_lw : Ångström exponent for absorption coefficients at wavelengths longer than 520 nm V f : fine volume fraction (fraction of particles with radii less than 0.5 μm) V c : coarse volume fraction (fraction of particles with radii greater than 0.5 μm; Annual mean is the mean of monthly means.
winter in Beijing.Figures 2c, 2d, 2e, and 2f show the frequency distributions of C ext and C abs for each season, where spring, summer, autumn, and winter are defined as March-May, June-August, September-November, and December-February, respectively.In Fukuoka, most measured C ext values were less than 500 Mm −1 , and the most frequently recorded values of C ext were less than 100 Mm −1 .However, relatively large C ext values were observed in spring.In Beijing, values of C ext larger than 1000 Mm −1 were frequently observed.The most frequently recorded values of C ext were less than 100 Mm −1 , and values between 100 and 500 Mm −1 were recorded more frequently in spring and summer.
In Fukuoka, relatively large values of C abs and C ext were observed in spring, and larger values of C abs were observed less frequently in summer.In Beijing, the most frequently recorded value of C abs in spring was 25 Mm −1 , and a second smaller peak in the summer frequency distribution also occurred at C abs = 10 Mm −1 , as shown in Fig. 2f.In the other seasons, the most frequently recorded values were C abs < 10 Mm −1 .The percentage of recorded values between 10 and 60 Mm −1 was high in spring and summer, and this feature was particularly distinctive in summer.In autumn and winter, C abs exceeded 100 Mm −1 relatively frequently.

Single-scattering albedo
Figures 3a and 3b show the time series of the monthly mean SSAs (ω 0 ) and their standard deviations at a wavelength of 525 nm.Tables 1 and 2 give the monthly and annual means of ω 0 (525 nm).The scat- tering coefficients were properly corrected using the new method.
ω 0 (525 nm) in Fukuoka ranged from 0.75 to 0.95, and its annual mean value was 0.877 ± 0.053.As shown in Figs.3a and 3b, ω 0 (525 nm) in Fukuoka un- derwent seasonal variation; it was large in spring and small in autumn.ω 0 (525 nm) in Beijing ranged from No. of data 0.75 to 0.95, which is similar to the range in Fukuoka, and its annual mean was 0.868 ± 0.047.The variation of ω 0 (525 nm) in Beijing did not follow a clear sea- sonal trend, but ω 0 (525 nm) did change greatly from year to year.The average SSA obtained by He et al. (2009) was 0.80 ± 0.09, which is smaller than the average SSA measured in the present study.As shown in Supplement 2, the Ångström exponent for the extinction coefficient in Beijing was small throughout the year.This indicates that the measured aerosols included numerous large particles and that the scattering coefficients measured by the integrating nephelometer required a large correction.He et al. (2009) did not mention the correction of the scattering coefficient.The use of the corrected SSA provides the explanation of most of the difference between the present results and those obtained by He et al. (2009) (see Supplement 3).Che et al. (2014) analyzed sky radiometer (POM-02, Prede, Japan) data from Beijing and obtained seasonal average SSAs ranging from 0.93 to 0.96.These values are larger than the SSAs measured in the present study.The SSAs estimated from the sky radiometer measurements are column-averaged SSAs of aerosols under ambient conditions, which involve hygroscopic growth.The ground-based measurements in the present study were conducted under dry air conditions.Because the two measurements were conducted under different conditions, comparisons between the results of ground-based and sky radiometer measurements must be made carefully.
Figures 3c and 3d show the frequency distribution of ω 0 (525 nm) for every season.In Fukuoka, the most frequent value in spring was 0.915, and the width of the frequency distribution in spring was narrower than that in the other seasons.Most SSAs recorded in spring were more than 0.8.In the other seasons, a larger percentage of recorded SSAs were less than 0.8.In Fukuoka, the peak of the frequency distribution in summer was broad, and lower SSAs were observed.In summer, Fukuoka was mainly covered with air masses that originated in the West Pacific Ocean region (region 8), which are typically clean.Most light-absorbing aerosols present in Fukuoka were emitted from local sources, and their seasonal variation is small.Therefore, the relative contribution of C abs to C ext was high in summer, and lower SSAs were observed.
In Beijing, the most frequently recorded ω 0 (525 nm) values were 0.895 in spring, 0.905 in summer, and 0.885 in autumn and winter.The frequency distribution in winter was narrower than those in the other seasons.In summer, lower SSAs were observed more frequently than in the other seasons.The reason for this is that the most frequently recorded C abs in summer was 25 Mm −1 , which is larger than those in the other seasons.

Asymmetry factor, extinction Ångström exponent, and volume fraction
The results on the asymmetry factor G, extinction Ångström exponent α ext , and volume fractions of coarse and fine modes are briefly shown in this sec-tion.The details on these parameters are shown in Supplement 2.
Tables 1 and 2 give the monthly and annual means of these parameters and their standard deviations.
G (525 nm) in Fukuoka ranged from 0.5 to 0.7, and its annual mean was 0.599 ± 0.040.G (525 nm) in Beijing ranged from 0.6 to 0.75, and the annual mean was 0.656 ± 0.042.No clear seasonal variation was observed in both cities.G (525 nm) in Beijing was larger than that in Fukuoka.
The α ext values in Fukuoka ranged from 1.0 to 2.1.The annual mean was 1.555 ± 0.312.The α ext values in Beijing ranged from 0.2 to 1.5, and the annual mean  The aerosol volume was obtained by integrating the retrieved volume size distribution.The volume was then divided into two parts: fine-and coarsemode volumes (V f and V c ) with particle radii less and greater than 0.5 µm, respectively.V c in Beijing was larger than that in Fukuoka.In Fukuoka, V f was approximately 80 %.In Beijing, V c was approximately 60 % in autumn and winter, and both V f and V c were approximately 50 % in spring and summer.
The larger asymmetry factor in Beijing, the smaller α ext in Beijing, and the larger V c in Beijing than in Fukuoka are consistent.These results mean that, in Beijing, the particles present in the air were coarser than those in Fukuoka.

Absorption Ångström exponent
The wavelength dependence of the absorption coefficient can be approximated by an equation similar to that relating C ext and α ext : where α abs is the absorption Ångström exponent, which is dependent on the aerosol composition and aging stage (Russell et al. 2010;Clarke et al. 2007).The characteristics of α abs , which have not been investigated in previous studies, are discussed here.Furthermore, the relationships between α abs and other parameters are described in a later section.Figures 4a and 4b show the time series of the monthly means of α abs with their standard deviations.Tables 1 and 2 give the monthly and annual means of α abs .As shown in Figs.4a and 4b, α abs demonstrated a remarkable seasonal variation in both Fukuoka and Beijing; α abs was small in summer and large in winter.Most α abs values ranged from 0.6 to 1.5.The annual means of α abs were 1.106 ± 0.155 in Fukuoka and 0.977 ± 0.147 in Beijing.The values of α abs in Beijing were slightly smaller than those in Fukuoka.
Figures 4c and 4d show the frequency distributions of α abs during each season in Fukuoka and Beijing.The most frequently recorded values varied seasonally.The monthly mean α abs in summer in Fukuoka was approximately 1.0, which usually indicates that the absorbing aerosol is composed mainly of fresh black carbon.However, the frequency distribution in summer demonstrates that α abs values below 1.0 were observed frequently.In Beijing, the monthly mean α abs values in summer were less than 1.0, and the fre- quency distribution also demonstrates the existence of aerosols with α abs < 1.0.
Russell et al. ( 2010) conducted measurements using the Aerosol Robotic Network (AERONET) and found α abs values near 1 (the theoretical value for fresh black carbon) for aerosol columns dominated by urbanindustrial aerosols, larger α abs values for biomass burning aerosols, and the largest α abs values for Sahara dust aerosols.These are typical light-absorbing aerosols, which have α abs values greater than or equal to 1. Therefore, a simple external mixture of these aerosols cannot explain an α abs value of less than 1.Gyawali et al. (2009) observed biomass burning aerosols with α abs < 1.0 and demonstrated that such values of α abs can result from black carbon coated with either absorbing or non-absorbing material.Bergstrom et al. (2007) noted that the interesting observation of α abs < 1.0 may be the result of measurement uncer- tainties or somewhat large values of the imaginary part of the refractive index (ImRF) at longer wavelengths for certain particles.Additionally, very low values of α abs have been reported under different circumstances without explanation (Bergstrom et al. 2007;Clarke et al. 2007;Roden et al. 2006;Subramanian et al. 2007;Yang et al. 2009).Because α abs values below 1.0 were observed in both Fukuoka and Beijing, future studies must explain what conditions cause α abs to be less than 1.0.
Tables 1 and 2 give the monthly and annual means of the absorption Ångström exponents α abs_sw in the region of wavelengths shorter than 520 nm and α abs_lw in the region of wavelengths longer than 590 nm, which also showed a seasonal variation.It is known that brown carbon shows stronger absorption characteristics in the ultraviolet (UV) region than in the visible light region (Moosmüller et al. 2009).However, it is very difficult to interpret the α abs_sw and α abs_lw data, and thus, only the values are given in this study.
As shown in Tables 1 and 2, when α abs_sw and α abs_lw are similar to each other, α abs is smaller than both α abs_sw and α abs_lw .This indicates that the absorption coefficient is not a monotonically decreasing function of the wavelength; the absorption coefficients in the region of wavelengths between 520 and 590 nm are constant or have a small peak with respect to the wavelength.The cause of this wavelength dependence remains unclear; its determination would require further study of the absorption coefficient of aerosols.

Optical properties classified by extinction and absorption Ångström exponents
α ext is an index of the size distribution, and α abs is related to the aerosol components (Russell et al. 2010).Therefore, the data in this study were classified using these parameters, and the relationships between these parameters and the aerosol optical properties were investigated.The relationships between α abs and other parameters have not been investigated in previous studies.Classifications based on α ext and α abs have already been conducted by Russell et al. (2010) and Clarke et al. (2007).Russell et al. (2010) classified absorbing aerosols as desert dust, urbanindustrial, and biomass burning aerosols.Clarke et al. (2007) classified aerosols observed on aircraft as dust, biomass burning, and pollution plume aerosols.
Figure 5 shows scatter plots of α ext and α abs .The data used in these plots are one-day averages, and there are no distinct clusters.It appears to be difficult to classify these data based on the magnitudes of α ext and α abs .In Fukuoka and Beijing, the data were clustered around (α ext , α abs ) = (1.5, 1.1) and (1.0, 1.0), respectively.Both α ext and α abs in Fukuoka were slightly larger than those in Beijing.There was a weak positive correlation between α ext and α abs in both cities.

Absorption Ångström exponent and volume size distribution
To investigate the relationship between α abs and the volume size distribution, the data were classified according to α abs using the following bins: 0.2 -0.4,0.4 -0.6, 0.6 -0.8, 0.8 -1.0, 1.0 -1.2, 1.2 -1.4, and 1.4 -1.6.This relationship between α abs and the volume size distribution has not been discussed in previous studies.Figure 6 shows the volume size distribution classified by α abs .As indicated in the scatter plot of α abs and α ext , when α abs is small, the aerosol contains many large particles.This feature was observed in both Fukuoka and Beijing.The difference between the distributions in Fukuoka and Beijing was caused by differences in α ext ; α ext in Beijing was smaller than that in Fukuoka, and the aerosols in Beijing thus contained larger particles.

Extinction Ångström exponent and volume size distribution
To investigate the relationship between α ext and the volume size distribution, the data were classified according to α ext using the following bins: −0.5 -0.5, 0.5 -1.0, 1.0 -1.5, 1.5 -2.0, 2.0 -2.5, and 2.5 -3.0. Figure 7 shows the volume size distribution classified by α ext .For α ext < 1, the retrieved volume size distri- butions were bimodal with peaks at radii of approximately 0.1 and 2.0 µm, and for α ext > 1, the volume size distributions were monomodal with a peak at a radius of approximately 0.1 µm.These peaks at radii of approximately 0.1 and 2.0 µm correspond to the accumulation and coarse particle modes, respectively.Similar results were reported at Tsukuba by Uchiyama et al. (2014).

Absorption Ångström exponent and imaginary part of the refractive index
The ImRF was also determined using the analysis method developed by one of the present authors (Uchiyama 2014).The relationships between the ImRF and other parameters were thus investigated.
To investigate the relationship between α abs and the ImRF, the data were classified according to α abs .Because the dependence of the real part of the refractive index on α ext and α abs is small, the dependence of only the ImRF on α abs and α ext is discussed in this study.Figure 8 shows the wavelength dependence of the ImRF; dashed lines indicate few data points.In Fukuoka and Beijing, the ImRF shows a different tendency for the same α abs value.As shown in Fig. 6, the size distributions in the two cities at the same α abs value differed from each other.These differences in the size distribution caused the different tendencies in the ImRF.Roughly speaking, when α abs is small, the ImRF tends to be small.In Fukuoka, the ImRF increased with increasing wavelength.When α abs was large, the ImRF tended to be large.However, in Beijing, the ImRF decreased with increasing wavelength.
As mentioned in Section 3.4, Russell et al. ( 2010) obtained α abs values near 1 (the theoretical value for fresh black carbon) for aerosol columns dominated by urban-industrial aerosols, larger α abs values for biomass burning aerosols, and the largest α abs values for Sahara dust aerosols.In addition, according to the simulation results obtained by Gyawali et al. (2009) based on the coated sphere model, α abs is less than 1.0 for aerosols coated with light-absorbing or non-absorbing aerosols with relatively large cores and increases with increasing coating thickness for aerosols with relatively small cores.Furthermore, the figures in Gyawali et al. (2009) (Figs. 8, 9 in their paper) show that, if the light-absorbing aerosol coating is thick, α abs is greater than 1.0 regardless of the size of the core.This also indicates that α abs is greater than 1.0 for large light-absorbing aerosols.It was also  found that α abs is less than 1.0 for aerosols coated with non-light-absorbing aerosols with relatively large cores.When α abs is small, the ImRF is small, and the fraction of coarse particles is high, as shown in Fig. 6.Because the ImRF is small, the aerosol contains many non-absorbing components either through external or internal mixing.Sea salt particles, internally mixed particles rich in non-light-absorbing components, and aerosols that have undergone hygroscopic growth are conceivable as coarse and non-light-absorbing aerosols.According to the simulation results obtained by Gyawali et al. (2009), if the aerosol has a relatively large core and is coated with a non-light-absorbing aerosol, α abs is less than 1.0.This aerosol model is consistent with the present measurement results.Although the relative humidity in the nephelometer inlet was maintained at 30 % or less, it is also possible that the hygroscopically grown aerosols, which consisted of internally mixed light-absorbing aerosols, passed through the inlet of the nephelometer before they were sufficiently dried.
In Fukuoka, when α abs was large, the volume size distribution was monomodal (Fig. 6a), the fraction of fine particles was high, and α ext was large.In contrast, in Beijing, when α abs was large, the volume size dis- tribution was bimodal (Fig. 6b), the fraction of coarse particles was high, and α ext was small.As shown in the simulation results obtained by Gyawali et al. (2009), in Fukuoka, aerosols coated with light-absorbing or non-light-absorbing aerosols with relatively small cores may include fine light-absorbing aerosols, i.e., black carbon coated with secondary species like organic matter and nitrate or sulfate species from gasto-particle conversion.In Beijing, α abs was large, and α ext was small.The ImRF was large and decreased with increasing wavelength.These features are similar to those of mineral dust.

Extinction Ångström exponent and imaginary part of the refractive index
To investigate the relationship between α ext and the ImRF, the data were classified according to α ext .Figure 9 shows the wavelength dependence of the ImRF; dashed lines indicate few data points.When α ext was small (−0.5 £ α ext £ 0.5), the ImRF was small and decreased with increasing wavelength in both Fukuoka and Beijing.At medium values of α ext (1.0 £ α ext £ 2.0), the ImRF was large in both Fukuoka and Beijing.For large α ext (2.0 £ α ext £ 3.0), though few data points were available, the ImRF tended to be small in Beijing and large in Fukuoka.
When α ext is small, the fraction of coarse particles is high.The following two cases are considered as cases where α ext is small.One is the case of mineral dust aerosol.The other is a case where α abs is small.The mineral dust aerosol is characterized by small α ext and large α abs .Additionally, when α abs is small, the fraction of coarse particles is high, as stated in Section 4.1.The ImRF for mineral dust is large in the short visible wavelength region and decreases with increasing wavelength.The ImRF for aerosols with small α abs is small (see Section 4.3).Because the data are not distinguished by the size of α abs in Fig. 9, ImRF shows the average feature of aerosols with small and large α abs ; the ImRF was smaller than that of mineral dust aerosols and decreased with increasing wavelength.In Beijing, when α ext was large, the size distribution was monomodal, as shown in Fig. 7b.Therefore, the aerosols did not include mineral dust particles.The ImRF in the shorter wavelength region was large and decreased with increasing wavelength.Brown carbon has such characteristics, but because few data points were obtained, it was difficult to make a definitive conclusion.

Single-scattering albedo and extinction and absorption Ångström exponents
To investigate the relationships among ω 0 , α ext , and α abs , the data were roughly divided into the following bins according to the value of C ext : 1 -25, 25 -100, 100 -1000, and 1000 -5000 Mm −1 .The data were then classified according to α abs and α ext .Tables 6 and 7 give ω 0 (525 nm), and the cells are colored according to the value of ω 0 (525 nm); blue and red cells corre- spond to high and low values, respectively, and cells with few data points are not colored.Roughly speaking, as C ext increased, ω 0 (525 nm) tended to increase.At large α ext and small α abs (upper right of Tables 6, 7) and at small α ext and large α abs (lower left of Tables 6,  7), ω 0 (525 nm) tended to be large.Large α ext values indicate that the fraction of small particles is high.Therefore, the former case corresponds to newly produced and grown particles, including weakly absorbing or non-absorbing aerosols such as sulfate particles.Small α ext values indicate that the fraction of large particles is high.Therefore, the latter case corresponds to mineral dust.When both α ext and α abs were large, ω 0 (550 nm) was small.This may correspond to newly produced and grown particles, including absorbing secondary organic aerosols such as brown carbon.However, this region contains few data points (fewer than five).Therefore, strong conclusions cannot be drawn regarding cases with large values of both α ext and α abs .

Case studies
To demonstrate the usefulness of the data obtained in this study, two characteristic cases in Beijing (winter) and Fukuoka (spring) were preliminarily analyzed.

Optical properties during winter in Beijing
As discussed in Section 3, winter in Beijing is characterized by very large variations in C ext and C abs ; both very clean and very hazy conditions were observed.Although plots of the time series of C ext and C abs are not shown here, after air masses in the North Continent region (region 1) reached Beijing, the air became very clean, resulting in low C ext and C abs values.Then, the air gradually became turbid with daily variation, causing the C ext and C abs values to gradually increase.As C ext and C abs increased, the characteristics of the aerosols changed.
To investigate the changes in the optical properties as the conditions changed from clean to hazy, the data were divided into the following bins according to C ext : 1 -25, 25 -50, 50 -100, 100 -200, 200 -400, 400 -800,  800 -1600, and 1600 -5000 Mm −1 .Figure  10 shows the relationship between the aerosol properties and C ext .
As shown in Fig. 10a, when C ext was very small (C ext < 25 Mm −1 ), the wavelength dependence of the SSA was large, and in the middle range of C ext (25 Mm −1 £ C ext £ 200 Mm −1 ), the wavelength dependence was small.As C ext increased, the SSAs increased, and the wavelength dependence decreased.As shown in Fig. 10b, G decreased with increasing C ext and was smallest in the range of 200 Mm −1 £ C ext £ 400 Mm −1 .Then, as C ext increased beyond 400 Mm −1 , the asymmetry factors increased again.As shown in Fig. 10c, α ext was smallest when C ext was small.As C ext increased, α ext was maximized in the range of 200 Mm −1 £ C ext £ 400 Mm −1 and then decreased.α ext depends on V f and V c , and its change is consistent with the change in G.
As shown in Fig. 10d, α abs was also smallest when C ext was small.As C ext increased, α abs was maximized in the range of 200 Mm −1 £ C ext £ 800 Mm −1 and then decreased.The maximum α abs value was ap- proximately 1.2.As shown in Fig. 10d, when C ext was small, α abs_lw and α abs_sw were approximately 1.0.α abs_sw was maximized in the range of 200 Mm −1 £ C ext £ 400 Mm −1 and was approximately 1.35.α abs_lw was maximized in the range of 400 Mm −1 £ C ext £ 800 Mm −1 and was approximately 1.15 in this range.The absorption characteristics of brown carbon tend to be stronger in the UV region; α abs_sw was large in the shorter wavelength region (Moosmüller et al. 2009).The observed features in the range of 200 Mm −1 £ C ext £ 800 Mm −1 showed characteristics similar to those of brown carbon.As demonstrated by the variation in α ext and G, V f and V c were approximately 50 % in the middle range of C ext (200 Mm −1 £ C ext £ 400 Mm −1 ), and when C ext was smaller or larger, V f was low, and    V c was high, as shown in Fig. 10e.These changes in the aerosol characteristics due to changes in the aerosol amount indicate that, as air masses from the North Continent region (region 1) reached Beijing, the air became clean, C ext gradually increased, and the aerosol characteristics changed because of the local formation and emission of anthropogenic aerosols and their aging.As the number of pollution particles increased, there was a period when the number of smaller particles increased and α ext became large.Following this period, as the amount of air pollution increased, the number of larger particles increased, and α ext became small.In the former period, new particle formation and condensation likely dominated, and in the latter period, coagulation may have occurred.α abs in the former period was larger than that in the latter period.
The analysis in this study is limited because only the optical properties were considered.To better understand processes related to aerosols, it is necessary to make comprehensive measurements in future works, including measurements of precursor gases, the aerosol composition, the mixing state, and the size distribution.

Optical properties during spring in Fukuoka
As discussed in Section 3, spring in Fukuoka was characterized by relatively large C ext and C abs values.Although plots of the time series of C ext and C abs are not shown here, C ext (525 nm) and C abs (520 nm) changed periodically with the passage of a synoptic-scale disturbance.As with the case of winter in Beijing, the data were divided into bins according to C ext to investigate the dependence of the optical properties on C ext .The bins were the same as those used for Beijing, but no data with C ext > 800 Mm −1 were observed: 1 -25, 25 -50, 50 -100, 100 -200, 200 -400, and 400 -800 Mm −1 .
Figure 11 shows the relationship between the aerosol properties and C ext .Very few data had C ext > 400 Mm −1 .The dependence of the optical properties on C ext differed from that during winter in Beijing.When C ext was very small (C ext < 25 Mm −1 ), the wavelength dependence of the SSA was large.As shown in Fig. 11a, when C ext increased, the SSAs increased monotonically, and the wavelength dependence decreased.As shown in Fig. 11b, G  As C ext increased, α abs increased, reached a max- imum in the range of 50 Mm −1 £ C ext £ 100 Mm −1 , and then decreased.The maximum value of α abs was approximately 1.2 (Fig. 11d).Although the value of C ext at which α abs was maximized in Fukuoka differed from that in Beijing, the maximum values of α abs in Fukuoka and Beijing were the same.As shown in Fig. 11d, α abs_sw was always lower than α abs_lw .As demon- strated by the change in α ext and G, V f decreased and V c increased in the range of C ext > 200 Mm −1 in Fig. 11e.
During spring in Fukuoka, relatively large C ext values in the range of 50 to 200 Mm −1 were frequently observed.α ext in this range was approximately 1.7.Therefore, the fraction of small particles was high.These large C ext values may have been caused by air masses that did not include large particles passing over the polluted area in the continent.According to the trajectory analysis (Table 4), in spring, the inflow of air masses passing over Japan was also high.The insolation also rapidly increased in spring, resulting in a high aerosol production rate.This indicates that the large C ext values during spring in Fukuoka were partially caused by aerosols emitted and produced in Japan.
In the range of C ext > 200 Mm −1 , α ext was low.In Beijing, V c was always high, and α ext was low.There- fore, when C ext values exceeding 200 Mm −1 were observed during spring in Fukuoka, the air masses were assumed to have arrived from the heavily polluted continental area.However, the wavelength dependence of α abs in Fukuoka was different from that in Beijing.Therefore, the aerosol content was modified as the air masses moved from Beijing to Fukuoka and was mixed with locally emitted aerosols.

Summary and conclusion
The IAP (CAS) and MRI (JMA) have been measuring aerosol optical properties as part of a cooperative Chinese and Japanese science and technology program.From 2010 to 2014, the aerosol optical characteristics in two cities (Beijing and Fukuoka) located in East Asia were measured using an integrating nephelometer and an aethalometer, and long-term season-crossing data were obtained.Using a method developed by one of the present authors, scattering coefficients measured by the nephelometer were corrected more accurately than those in previous studies, and more reliable and accurate values of optical properties and their frequency distributions were obtained.The size distribution and complex index of refraction were  Research (Nos. 19340139 and 22310015).The authors would also like to thank Dr. Hara and Dr. Shiraishi of Fukuoka University for their support in obtaining measurements and maintaining instruments and three anonymous reviewers for their useful comments.Furthermore, we would like to express our appreciation to Dr. Kuji, the editor in charge of this paper.His detailed comments aided in the significant improvement of this paper.

Fig. 2 .
Fig. 2. Time series of monthly mean scattering and absorption coefficients with standard deviations in (a) Fukuoka and (b) Beijing.Normalized frequency distributions of extinction coefficients for every season in (c) Fukuoka and (d) Beijing.Normalized frequency distributions of absorption coefficients for every season in (e) Fukuoka and (f) Beijing.Winter, spring, summer, and autumn are defined as December-February, March-May, June-August, and September-November, respectively.The frequency distributions shown here are normalized to 1.
855 ± 0.347.α ext in Beijing was smaller than that in Fukuoka by approximately 0.7.

Fig. 5 .
Fig. 5. Scatter plot of the extinction and absorption Ångström exponents α ext and α abs in (a) Fukuoka and (b)Beijing.

Fig. 6 .
Fig. 6.Volume size distributions for α abs bins in (a) Fukuoka and (b) Beijing.A dashed line indicates that fewer than 25 data points were obtained for that bin.The total numbers of data points in Fukuoka and Beijing are approximately 44000 and 36000, respectively.

Fig. 7 .
Fig. 7. Volume size distributions for α ext bins in (a) Fukuoka and (b) Beijing.Dashed lines indicate that fewer than 25 data points were obtained for that bin.

Fig. 8 .
Fig. 8. Relationship between α abs and ImRF in (a) Fukuoka and (b) Beijing.Dashed lines indicate bins with fewer than 25 data points.

Fig. 9 .
Fig. 9. Relationship between α ext and ImRF in (a) Fukuoka and (b) Beijing.Dashed lines indicate bins with fewer than 25 data points.
Fig. 10.Relationships between aerosol characteristics and the extinction coefficient during winter in Beijing.(a) SSA at wavelengths of 450, 525, and 635 nm.(b) Asymmetry factor.(c) Extinction Ångström exponent.(d) Absorption Ångström exponents for all wavelengths, wavelengths shorter than 520 nm, and wavelengths longer than 590 nm.(e) Fine-and coarse-mode volume fractions.
was somewhat low in the range of 50 Mm −1 £ C ext £ 100 Mm −1 .For C ext > 200 Mm −1 , G was high.Additionally, as shown in Fig. 11c, α ext was somewhat high in the range of 50 Mm −1 £ C ext £ 100 Mm −1 and decreased in the range of C ext > 200 Mm −1 .The dependence of α ext on C ext was consistent with that of G.

Table 1 .
Monthly and annual means of aerosol properties in Fukuoka

Table 2 .
Same as Table 1 for Beijing.

Table 3 .
Aerosol optical properties measured in Beijing

Table 4 .
Frequency with which each region had the longest resident time for air masses reaching Fukuoka.The resident time is defined as the time an air mass stays in a given region within the five days prior to it arriving in the target city.Frequencies were calculated each month after backward trajectory analysis.

Table 5 .
Same as Table 4 for Beijing.

Table 6 .
Relationships among ω 0 , α ext , and α abs in Fukuoka.Cells where fewer than 10 data points were collected are not colored.−99 indicates no data.The number in parentheses is the number of data.The total number of data points is approximately 44,000.

Table 7 .
Same as Table6for Beijing.Cells where fewer than 10 data points were collected are not colored.−99 indicates no data.The number in parentheses is the number of data.The total number of data points is approximately 36,000.