Strong particle light absorption over the Pearl River Delta (south China) and Beijing (north China) determined from combined Raman lidar and Sun photometer observations



[1] Particle size and absorption properties have been determined from combined one-wavelength Raman lidar and Sun photometer measurements in the Pearl River Delta (South China) in October 2004, and in Beijing (North China) in January 2005. Particle effective radius varied around 0.24 μm in the south of China. Aerosols were strongly light absorbing throughout the continental haze layers. The mean imaginary part was around 0.02i and the single-scattering albedo was as low as 0.75 at 532 nm wavelength. Similar aerosol properties were found over Beijing.

1. Introduction

[2] Nearly 50% of the world's population lives in China, India and neighboring countries. Growth of population and economic rise in that region are causing severe environmental problems [Ramanathan et al., 2005], which will intensify in the coming decades. Unlike in Europe and North America the Chinese aerosols contain a much higher amount of light absorbing black carbon [Bergin et al., 2001] from industry and traffic exhausts, and the combustion of biomass and fossil fuel.

[3] Knowledge on particle light absorption is needed for a realistic estimation of the climatic impact of present and future aerosol pollution in Eastern Asia. Respective observations, however, are limited. Cao et al. [2003, 2004] investigated the relative contribution of carbonaceous aerosols to particulate matter near the surface in the south of China. Qiu et al. [2004] report on a systematic decrease of the single-scattering albedo over Beijing from 1996–2001.

[4] There is particularly little known on particle light absorption features in the atmospheric column. Therefore, combined Raman lidar and Sun photometer observations were carried out near Xinken (22.6° N, 113.6° E) in South China from 2–31 October 2004. Xinken lies on the banks of the Pearl River, about 80 km northwest of the center of Hong Kong. Approximately 250 million people live in an area of 60,000 km2. That makes it the most densely populated and largest urban area in the world.

[5] An overview on some important optical properties for the complete measurement period has been given by Ansmann et al. [2005]. Particle optical depth at about 532-nm wavelength ranged from 0.3–1.7. Ångström exponents were low with values between 0.65–1.35 (for the wavelength range from 380–502 nm) and between 0.75–1.6 (for the wavelength range from 502–1044 nm).

[6] This contribution extends the results on the optical aerosol properties toward important microphysical particle properties, i.e., particle mean (effective) radius, complex refractive index, and single-scattering albedo. For that purpose we applied for the first time a modified inversion scheme [Pahlow et al., 2006] to data collected with combined Raman lidar/Sun photometer observations. The results are complemented by observations made in Beijing in North China from 7–24 January 2005.

2. Methodology

2.1. Instruments

[7] A one-wavelength Raman lidar was used for vertically resolved measurements of the particle backscatter and extinction coefficients and of the backscatter-to-extinction (lidar) ratio at 532-nm wavelength. The data analysis procedure is described by Ansmann et al. [2005]. The accuracy of the backscatter and extinction coefficients is approximately 5%. The lidar ratio is accurate to ∼10%.

[8] Particle optical depth was determined with a Sun photometer at several wavelengths between 381–1044 nm. Uncertainties of optical depth are on the order of 1%–3% [Ansmann et al., 2005].

2.2. Inversion Algorithm

[9] Particle effective radius and complex refractive index were derived with an inversion algorithm [Müller et al., 1999]. The microphysical parameters are used to calculate the single-scattering albedo with a Mie-scattering code [Bohren and Huffman, 1983] at 532-nm wavelength. Originally the algorithm has been developed for the inversion of a combination of backscatter coefficients measured at 3–6 wavelengths and extinction coefficients measured at 2 wavelengths with the institute's Raman lidars.

[10] Because of the limited optical information, we had to combine the data from the vertically resolved night time lidar observations and the column-integrated daytime Sun photometer measurements. Another challenge follows from the fact that the number of measurement channels of particle extinction is larger than the number of particle backscatter coefficients.

[11] The code has been adapted accordingly [Pahlow et al., 2006]. Sensitivity analysis carried out with synthetic data sets showed that microphysical particle properties can be derived with acceptable accuracy for such data combinations. Results for particle size parameters are presented by Pahlow et al. [2006].

[12] The optical input data were generated by (1) calculation of haze-layer mean backscatter and extinction coefficients from the profile information measured with lidar, (2) conversion of optical depth, which is measured with Sun photometer, into layer-mean extinction coefficients, and (3) the additional calculation of backscatter coefficients at 381 nm, because sensitivity studies show that backscatter coefficients at two wavelengths are required for an accurate data inversion [Pahlow et al., 2006]. These steps will be described in the context of a characteristic measurement example of the continental haze layer over the Pearl River Delta (PRD). An extended discussion of data analysis is given by Tesche [2006].

3. Case Study From 19 October 2004

[13] Figure 1 shows the measurement from 19 October 2004. The particle backscatter coefficient shows that the main part of the haze layer extends to almost 2-km height. Extinction coefficients are as high as 600 Mm−1. The lidar ratio is around 60 sr, indicating anthropogenic pollution. For comparison extinction coefficients averaged less than 250 Mm−1 during the Indian Ocean Experiment (INDOEX) [Franke et al., 2003], where we observed pollution outbreaks from the Indian subcontinent and Southeast Asia.

Figure 1.

(top) Backscatter and extinction coefficients and the lidar ratio at 532 nm measured with Raman lidar in the Pearl River Delta from 10–12 UTC on 19 October 2004. Error bars denote one standard deviation. The extinction profile below 500-m height was obtained from the backscatter profile and an assumption for the lidar ratio. (bottom) Optical depth and Ångström exponents measured with Sun photometer on 19 October 2004. The uncertainty of optical depth is <5% for each data point. The uncertainty of the Ångström exponents is ∼20%.

[14] Optical depth was 0.96 at 532 nm, which is considerably larger than the peak value of 0.6 measured with Raman lidar during INDOEX [Franke et al., 2003]. The mean optical depth for the whole measurement period is 0.9 ± 0.3 [Ansmann et al., 2005], which also is considerably larger than the mean of 0.31 ± 0.08 measured during INDOEX in 1999/2000 [Franke et al., 2003].

[15] Figure 1 also shows the time series of optical depth and the Ångström exponent for the wavelength pair at 381/502 nm. Ångström exponents of 1 ± 0.2 were found close to sunset. The Ångström exponents over the PRD are quite similar to Ångström exponents of Indian particulate pollution (0.9–1.2) [Franke et al., 2003].

[16] We selected this example because we could combine the vertically resolved one-wavelength Raman lidar and the column-integrated Sun photometer observations to retrieve haze-layer mean properties. The backscatter coefficient remains rather constant within the haze layer. Profiles of relative humidity and potential temperature, which were derived from radiosondes launched at Hong Kong, showed a well mixed haze layer on that day. The lidar measured continuously. The observations do not indicate significant changes of the aerosol conditions between the final hour of Sun photometer measurements, which were used in the present analysis, and the first two hours of lidar observations.

[17] The inversion was carried out with a combination of backscatter coefficients derived at two wavelenghts and extinction coefficients derived at 6 wavelengths. Haze-layer mean backscatter (10 Mm−1sr−1) and extinction coefficients (557 Mm−1) were calculated from the lidar profiles. The haze-layer top height was ∼1.63-km. A definition of that height is given by Ansmann et al. [2005]. Tesche [2006] presents a sensitivity study on the influence of uncertainty of that height on the inversion results. Optical depths measured with Sun photometer were converted to mean extinction coefficients on the basis of the haze layer height.

[18] Pahlow et al. [2006] show that the inversion algorithm remains stable if extinction coefficients at ≥4 wavelengths are used. We picked the measurement channels at 381, 440, 780, 870, and 1020 nm. These wavelengths are not precisely those used by Pahlow et al. [2006], but they are close enough to rule out that significant errors are caused by these differences. Mean extinction values were 788, 695, 345, 314, and 240 Mm−1 at the five channels, respectively. We put 10% statistical noise on all optical data except those at 381 nm (see below). We repeated the inversion ten times, and averaged the individual solution sets.

[19] The lidar measured particle backscatter coefficients at one wavelength only. We obtained a second backscatter coefficient at 381-nm wavelength on the basis of the lidar ratio at 532 nm, and the Ångström exponent measured in the wavelength range from 381–502 nm in the last hour of the Sun photometer observations. The mean backscatter coefficient derived in that way is 14 Mm−1sr−1.

[20] We carried out a sensitivity analysis to determine the inversion error introduced by this artificial selection of the particle backscatter coefficient. The mean backscatter coefficient and the mean extinction coefficient at 381 nm were shifted toward ±20% larger/lower values, respectively. The inversion was then carried out with an additional 10% statistical noise put on these biased data sets. In most of the measurement cases we did not obtain reasonable inversion results if the lidar ratio at 381 nm was shifted from the mean value by ±20%. We conclude that the lidar ratio at 381 nm must be similarly high as the lidar ratio at 532 nm.

[21] The extinction information in the overlap region of the lidar was determined from the profile of the measured particle backscatter profile and assuming a particle lidar ratio of 55 sr which is the mean value from 500–1000 m height in Figure 1.

[22] Table 1 shows the results for effective particle radius (reff), real (mreal) and imaginary part (mimag) of the complex refractive index, and the single-scattering albedo (ω). The latter parameter was calculated from the retrieved particle size distributions and the complex refractive index. The uncertainty follows from the uncertainties of the microphysical parameters used in the Mie-scattering calculations.

Table 1. Microphysical Particle Parameters of Haze Measured in the Pearl River Delta on 19 October 2004a
  • a

    The symbols are explained in the text.

0.24 ± 0.06 μm1.59 ± 0.10.026 ± 0.0150.74 ± 0.12

4. Discussion

[23] Particles of South Chinese pollution are considerably larger than particles of anthropogenic pollution from North America and Europe, for which we usually find effective radii <0.2μm. Relative humidity was on average 50%–60% up to 2-km height throughout the measurement period [Ansmann et al., 2005]. Hygroscopic growth thus cannot explain the large particles. Single-scattering albedo varied around 0.77 (at 532 nm), which is at the lower end of numbers found for free-tropospheric north Indian pollution [Müller et al., 2003].

[24] Figure 2 shows the single-scattering albedo on 7 out of the 30 measurement days in the PRD. These cases were characterized by a well-mixed continental haze layer. The mean value of the single-scattering albedo varied between 0.75–0.8. In situ observations of particle size distribution and chemical properties were carried out at the lidar site [Eichler, 2006]. Single-scattering albedo was derived from Mie-scattering calculations. Two particle models were applied [Eichler, 2006]. Model 1 treats the particles as black carbon (BC) cores surrounded by a shell of non-absorbing material and water. In model 2 the particles are treated as a mixture of homogeneous spheres of BC, non-absorbing material, and water interdispersed with externally mixed BC. We estimate the uncertainty of the single-scattering albedo to at least 0.05.

Figure 2.

Single-scattering albedo at 532 nm determined from Raman lidar/Sun photometer observations (square), and derived from in situ observations with Mie-scattering model 1 (circle) and model 2 (triangle). Results are shown (a) for the PRD and (b) for Beijing.

[25] We have data from both instrument set-ups for six out of the seven measurement cases. We obtained the single-scattering albedo with model 2 only for day of year 297. Before that day we do not have the information on the chemical composition needed for the calculations. Agreement is good, but the numbers from the in situ observations are always in the upper range of values derived from the remote sensing instrumentation. One possible reason for the higher single-scattering albedo near the surface may be the influence of non light-absorbing sea salt from the South China Sea, which is close to the field site, and less light-absorbing dust generated by agricultural activity and traffic.

[26] Figure 2 also shows a time series of single-scattering albedo for Beijing. Only four days of lidar measurements could be used for the retrieval. A laser failure prevented observations after 17 January 2005. We cannot directly compare our lidar/Sun photometer results to those from the in situ instrumentation because respective observations are only available for the period from 17–24 January 2005. The time series of single-scattering albedo obtained with model 1 and model 2 is at the lower and upper end of values, respectively, determined from the remote sensing instruments. That result points at the large uncertainty the model calculations are affected with.

[27] Table 2 summarizes the results for effective radius, complex refractive index, and single-scattering albedo derived from the two instrument set-ups. Particle size and single-scattering albedo are rather similar at the PRD site. We conclude that the haze was strongly light absorbing from bottom to top of the haze layers. In contrast to the subtropical site of the PRD with high relative humidity, Beijing was characterized by continental climate with low relative humidity and low temperatures. Yet, also there we find comparably large particles and strong particle light absorption. According to Qiu et al. [2004] single-scattering albedo decreased from ∼0.85 to < 0.83 at Beijing between 1996–2001. We find a mean value of 0.78. Despite the large uncertainty of our results for single-scattering albedo there still seems to be a downward trend since 2001.

Table 2. Mean Microphysical Particle Parameters of Haze Layersa
ParameterPearl River DeltaBeijing
Lidar/SPMbIn SituLidar/SPMbIn Situ
  • a

    The results are derived from the measurement days indicated in Figure 2.

  • b

    SPM denotes Sun photometer.

  • c

    Mie-scattering model 1.

  • d

    Mie-scattering model 2.

reff0.24 ± 0.070.24 ± 0.010.23 ± 0.060.23 ± 0.01
mreal1.57 ± 0.111.62 ± 0.11
mimag0.022 ± 0.0150.019 ± 0.012
ω0.77 ± 0.120.84c0.78 ± 0.110.69c –0.89d

[28] Table 3 shows that our results on light-absorbing pollution are comparable to findings by other authors. Single-scattering albedo varied from 0.75–0.89 at several north Chinese sites between 1993–2001 [Qiu et al., 2004]. Single-scattering albedo at the continental-urban sites in general is lower than numbers describing continental-rural sites (Linan) and ocean sites (Yellow Sea, Gosan). In the latter case the impact of marine particles may be responsible for the lower light absorption. Higher single-scattering albedos were also obtained from Raman lidar observations in the Maldives during INDOEX [Müller et al., 2003]. Airmasses were advected from Southeast Asia, including South China.

Table 3. Single-Scattering Albedo Measured in Other Areas of South- and East Asia
SiteMeasurement PeriodωWavelength, nmReference
Beijing, ChinaJune 19990.81 ± 0.08550Bergin et al. [2001]
BeijingSpring 20010.88 ± 0.04500Xia et al. [2005]
BeijingJune 20030.88 ± 0.01550Eck et al. [2005]
BeijingWinter 1993–20010.82–0.87550Qiu et al. [2004]
Harbin, ChinaWinter 1993–20010.78–0.89550Qiu et al. [2004]
Zhengzou, ChinaWinter 1993–20010.81–0.87550Qiu et al. [2004]
Shenyang, ChinaWinter 20010.75–0.84550Qiu et al. [2004]
Linan, ChinaNovember 19990.95 ± 0.03500Xu et al. [2003]
Gosan, South KoreaNovember 20010.88 ± 0.02550Kim et al. [2005]
Yellow SeaSpring 20010.88 ± 0.03550Anderson et al. [2003]
Maldives (air from Southeast Asia)February 19990.93 ± 0.05532Müller et al. [2003]

[29] Our new approach to combine Raman lidar and Sun photometer observations offers a new way to investigate important properties of atmospheric pollution. One important restriction of that technique however is that the pollution has to be well mixed within one atmospheric layer. Our results show the importance of sounding the vertical structure of light absorption over China, which is still missing. One should keep in mind that the Chinese mainland acts as source region for long-range pollution transport to the Pacific Ocean and the Indian Ocean. The pollution layers, which are lifted above the marine boundary layer become thermodynamically decoupled, and can thus be transported over intercontinental distances.


[30] This work was supported by China National Basic Research and Development Program (2002CB410801 and 2002CB211605). We thank the China Meteorological Administration and the Hong Kong Observatory for providing radiosonde data.