Assessment of satellite-based aerosol optical depth using continuous lidar observation

Atmospheric Environment 140 (2016) 273e282

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Assessment of satellite-based aerosol optical depth using continuous lidar observation C.Q. Lin a, C.C. Li b, *, A.K.H. Lau a, c, d, Z.B. Yuan e, X.C. Lu c, K.T. Tse a, f, J.C.H. Fung c, d, g, Y. Li c, T. Yao c, L. Su h, Z.Y. Li c, Y.Q. Zhang c a

Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong, China Department of Atmospheric and Oceanic Sciences, School of Physics, Peking University, Beijing, 100871, China Division of Environment, The Hong Kong University of Science and Technology, Hong Kong, China d Institute for the Environment, The Hong Kong University of Science and Technology, Hong Kong, China e School of Environment and Energy, South China University of Technology, Guangzhou, Hong Kong, China f The CLP Wind/Wave Tunnel Facility, The Hong Kong University of Science and Technology, Hong Kong, China g Department of Mathematics, The Hong Kong University of Science and Technology, Hong Kong, China h Environmental Science Programs, The Hong Kong University of Science and Technology, Hong Kong, China b c

h i g h l i g h t s
5-year continuous lidar measurements were used to assess the satellite-based AOD.
Different seasonal variation in AOD is observed by MODIS and lidar over Hong Kong.
Aerosols in the upper mixing layer largely increase AOD at midday during summer.
Long-term average of AOD, estimated from satellite observation, may has large error.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 9 March 2016 Received in revised form 26 May 2016 Accepted 6 June 2016 Available online 7 June 2016

Due to a reliance on solar radiation, the aerosol optical depth (AOD) is observed only during the day by passive satellite-based instruments such as the MODerate resolution Imaging Spectroradiometer (MODIS). Research on urban air quality, atmospheric turbidity, and evolution of aerosols in the atmospheric boundary layer, however, requires 24-h measurement of aerosols. A lidar system is capable of detecting the vertical distribution of the aerosol extinction coefficient and calculating the AOD throughout the day, but routinely lidar observation is still quite limited and the results from MODIS and lidar sometimes are contradictory in China. In this study, long-term lidar observations from 2005 to 2009 over Hong Kong were analyzed with a focus on identification of the reasons for different seasonal variation in the AOD data obtained from MODIS and lidar. The lidar-retrieved AOD shows the lowest average level, but has the most significant diurnal variation during the summer. When considering only a 5-h period between 10:00 a.m. and 3:00 p.m. local time to match satellite passages, the average of the lidar-retrieved AOD doubles during the summer and exceeds that during the winter. This finding is consistent with the MODIS observation of a higher AOD during the summer and a lower AOD during the winter. The increase in the aerosol extinction coefficient in the upper level of the mixing layer makes the greatest contribution to the increase in the AOD at midday during the summer. These assessments suggest that large over-estimation may occur when long-term averages of AOD are estimated from passive satellite observations. © 2016 Elsevier Ltd. All rights reserved.

Keywords: MODIS Lidar Aerosol optical depth Aerosol extinction coefficient Hong Kong

1. Introduction

* Corresponding author. E-mail address: [email protected] (C.C. Li). http://dx.doi.org/10.1016/j.atmosenv.2016.06.012 1352-2310/© 2016 Elsevier Ltd. All rights reserved.

Given the large spatiotemporal variabilities and complex physical and chemical properties of aerosols, estimates of their corresponding earth’s radiative forcing patterns are still complex and

274

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

highly-uncertain (Min et al., 2009; Satheesh and Krishna Moorthy, 2005). Reduction of these uncertainties requires routine observation of aerosol distributions and properties at high spatiotemporal resolutions (Groß et al., 2013). Satellite sensors routinely observe the aerosol optical depth (AOD), a column integral of light extinction coefficients of aerosols in the atmosphere, at high spatial coverages (Jerrett et al., 2004; Kaufman et al., 2002). Observations of the AOD from the MODerate resolution Imaging Spectroradiometer (MODIS) instruments aboard NASA’s Earth observing system (EOS) polar-orbiting satellites, Terra and Aqua, have been extensively used to estimate the spatial distributions of aerosols (Chu et al., 2003; Li et al., 2003, 2015; Lin et al., 2015). Due to a reliance on solar radiation, AOD is observed only during the day from these passive instruments. Research on urban air quality, atmospheric turbidity, and evolution of aerosols in the atmospheric boundary layer, however, requires 24-h measurement of aerosols
rez-Ramírez et al., 2011). In addition, reliable (Berkoff et al., 2011; Pe aerosol measurements during nighttime would improve our understanding of the complete effect of aerosols in the climate feedback system (Johnson et al., 2013; Zhang et al., 2008). Unfortunately, ground-level measurement of the columnintegrated values for AOD from the aerosol robotic network (AERONET) (Holben et al., 1998) is also restricted to during the day, and is thus unable to provide information on the evolution of aerosols in the atmospheric boundary layer. A few new sensors that take advantage of moonlight (Berkoff et al., 2011; Esposito et al., 1998) or
rez-Ramírez et al., 2011) have been starlight (Herber et al., 2002; Pe developed to measure AOD during nighttime. These sensors, however, are either limited by the inconsistent availability of the lunar source and non-Lambertian reflectance properties of the moon (Kieffer, 1997), or expense and complex to collect sufficient starlight (Berkoff et al., 2011), leading to the fact that lunar and stellar measurements are still limited in use (Berkoff et al., 2011). The measurement of nighttime AOD from space is in its infancy and deserves further evaluation, but nevertheless is a subject of ongoing interest (Johnson et al., 2013). Lidar is an active remote sensing system that is capable of detecting the vertical distribution of the aerosol extinction coefficient and calculating the AOD throughout the day at high spatiotemporal resolution (Spinhirne, 1993). Long-term variation in the AOD and aerosol vertical distribution has been intensively investigated with the use of long-running lidar systems around the world. By analyzing observations from MODIS and lidar in eastern Asia, Kim et al. (2007) noted that seasonal variation in the AOD from the two data sets shared a similar feature, with a maximum in the spring when Asian dust storms frequently occur. On the basis of lidar measurements over M’Bour in Senegal from 2006 to 2008,
on et al. (2009) observed a maximum of dust activity and AOD Le (above 0.5) during summer. Raman lidar observations from Greece from 2001 to 2004 showed an average AOD of about 0.63, of which free tropospheric particles accounted for about 30% (Amiridis et al., 2005). By analyzing 1-year of lidar measurements in India, Sinha et al. (2013) showed that aerosols within the atmospheric boundary layer and in the free troposphere contributed about 0.37 (77.7%) and 0.12 (22.3%), respectively, to the total AOD. Therefore, lidar observation of the aerosol vertical distribution and AOD is valuable for the assessment of the long-term variation in the MODIS AOD. Although satellite-based lidar sensors are also able to observe aerosol vertical distribution during daytime and nighttime, the estimates are still suffer from coarse spatiotemporal resolution, signal attenuation, and large bias resulted from the assumed lidar ratio (Campbell et al., 2013). Long-term groundbased lidar observation, however, is still quite limited especially in China (Yang et al., 2013). Results from MODIS and lidar sometimes are contradictory. He et al. (2008) analyzed the characteristics of

the seasonal averages of aerosol profiles and observed the lowest AOD during summer with the use of 24-h lidar measurements from May 2003 to June 2004 over Hong Kong. In contrast, a higher level of AOD was observed during summer than during winter by analysis of the MODIS observations (Li et al., 2003). It is therefore important to identify the reasons for the different seasonal variation in the AOD data from MODIS and the 24-h lidar observation. In this study, long-term lidar observations from a 5-year period from 2005 to 2009 over Hong Kong were analyzed. The long-term running lidar provides crucial information for characterization of the long-term variation in the aerosol vertical distribution and AOD in this region. Diurnal, monthly, and seasonal variation in the aerosol vertical distribution and AOD were derived and analyzed with a focus on characterizing the different seasonal variation in AOD data from MODIS and the lidar. This paper is organized as follows. In the Data and Measurement section, the measurements from MODIS, lidar, a visibility sensor, and radiosondes are introduced. In the Results section, the diurnal, monthly, and seasonal variation in the aerosol vertical distribution and AOD from MODIS or the lidar are obtained and analyzed. Finally, the reasons for the different seasonal variation in the AOD data from MODIS and the lidar are further discussed.

2. Data and measurements 2.1. Lidar 2.1.1. Instrument The micropulse lidar (MPL) system was installed at Yuen Long (114.02 E 22.44 N, as shown in Fig. 1), a residential town in the northwest of Hong Kong. Hong Kong is located in the southeast part of the Pearl River Delta (PRD) region, which is one of the most highly-populated and rapidly-developing regions in China (Cao et al., 2003). The weather in this region is influenced by the southerly or southwesterly East Asia monsoon during the summer and by the northerly or northeasterly monsoon during the winter (Yuan et al., 2013). Rapid urbanization and industrialization have caused heavy air pollution problems in this region (Li et al., 2015;

Fig. 1. Map of Hong Kong and the Pearl River Delta (PRD) in China.

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

Zhong et al., 2013). The MPL system, produced by SESI Corporation (Science & Engineering Services, Inc.), operated continuously from 2005 to 2009 on a rooftop 28 m above ground level. The laser of the MPL system was emitted at an angle of 75 from the horizontal to avoid direct radiation from the sun. The bin time of the MPL receiver was set at 200 ns for a vertical resolution of about 29 m (30 m  sin75 ). The laser’s wavelength was 523 nm (Nd:YLF). The pulse repetition rate was 2500 Hz, which indicated a maximum detection range of 60 km. The per-pulse energy was restricted below 10 mj for eye safety. The blind zone caused by an incomplete after-pulse and overlap correction was about 130 m. Backscattering signals were recorded every 15 s. More details about the MPL system can be found in other studies (Campbell et al., 2002; He et al., 2006; Yang et al., 2013). To improve the signal-to-noise ratio, the MPL signals are averaged into data sets with intervals of 1 h. Because lidar signal attenuates quickly in the cloud and AOD is observed in the cloudfree condition by the passive satellite instruments, hourly MPL data that were significantly affected by clouds were removed by means of threshold values (Yang et al., 2013). Table 1 shows the data sample sizes of the hourly MPL measurements in cloud-free days during each of the seasons and years studied. A total of 10,707 hourly MPL measurements were obtained. Generally, data were less available in spring and summer due to greater influence from clouds. 2.1.2. Retrieval of aerosol extinction coefficient The method for retrieval of the vertical profile of aerosol extinction coefficient sa has been well given in a multiple of studies (Campbell et al., 2002; Chen et al., 2009; Chiang et al., 2007; Fernald, 1984; Klett, 1985; Sinha et al., 2013; Welton et al., 2002). The lidar solution is integrated by starting from the far end of the measurement range. A column-averaged extinction-to-backscatter ratio (so-called lidar ratio, sa) is usually assumed in lidar inversion. Independent observations of the AOD are typically used for estimate of sa (Cavalieri et al., 2010; Chen et al., 2009; Perrone et al., 2014; Wu et al., 2012, 2014). Because of a lack of the groundbased AOD observation at lidar site, we seek to find an alternative observation of AOD from satellites for estimate of sa. In this study, average of the MODIS AOD, as shown in Section 2.2, within an area with a 20-km radius at the lidar site is used to estimate sa. More details of the estimates of sa with the use of the MODIS AOD data can be found in He et al. (2006). The estimated sa is assumed to be constant for each day. Linear interpolation is then performed between these discrete values of sa throughout the entire study period.

275

Collection 6 (C6) data product suite has been released. In this study, the MODIS C6 level-2 AOD data from 2005 to 2009 were acquired from NASA’s Goddard Earth Sciences Distributed Active Archive Center (DAAC, https://ladsweb.nascom.nasa.gov/data/search.html). The MODIS AOD data were retrieved using the dark target algorithm over land and archived at a spatial resolution of 10 km  10 km (at nadir). Derived at 550 nm, the wavelength of the MODIS AOD product is in line with that of the lidar observation. Validation processes are under continuous development to ensure the high quality of the MODIS AOD product (Levy et al., 2013; Sayer et al., 2014; Tao et al., 2015). The accuracy of the MODIS AOD retrievals from the dark target algorithm over land has been determined on a global average of ±(0.05 þ 0.15ta), where ta is the AERONET-observed AOD (Levy et al., 2013). However, few evaluations of MODIS C6 AOD data have been conducted in China (Tao et al., 2015). Because there is no AERONET station collocated directly with the lidar system, the accuracy of the MODIS AOD data for this region are evaluated by AERONET level-2 AOD at a nearby station (HKPolyU, 121.500 E 25.030 N, as shown by a blue square in Fig. 1). With a distance of less than 20 km between the two stations, the accuracy of the MODIS AOD at HKPolyU station approximates that of the MODIS AOD at Yuen Long station. The AERONET-observed AOD is interpolated to 550 nm and averaged within ±30 min of the overpass time of the satellites. The MODIS AOD (from Terra and Aqua) is averaged within an area with a 20-km radius at the AERONET station. Fig. 2 shows a comparison of the AOD at the AERONET station at times for which the MODIS and AERONET-observed AOD data are both available from 2005 to 2009. Good agreement with a high correlation coefficient of 0.93 (N ¼ 452) is seen between the MODIS and AERONET-observed AOD data, depicting the reasonable accuracy of MODIS AOD in this region. 2.3. Visibility sensor The surface aerosol extinction coefficient sa,0, measured by a Belfort Model 6000 visibility sensor, was used to evaluate the lidar-

2.2. MODIS NASA’s EOS polar-orbiting satellites, Terra and Aqua, pass the equator at about 10:30 a.m. and 1:30 p.m. local time, respectively. The MODIS instruments aboard Terra and Aqua measure spectral radiation in 36 bands in 2330-km swaths. They provide daily AOD data that cover almost the entire globe. Recently, the MODIS

Table 1 Numbers of hourly lidar measurements available for cloud-free days during different seasons and years. Time

2005

2006

2007

2008

2009

Total

Spring Summer Autumn Winter Total

251 356 727 475 1809

215 392 753 562 1922

393 436 847 691 2367

481 512 781 446 2220

334 575 746 734 2389

1674 2271 3854 2908 10,707

Fig. 2. Comparison between the MODIS and AERONET-observed AOD data at a nearby station (HKPolyU) from 2005 to 2009.

276

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

retrieved aerosol extinction coefficient. The visibility sensor is collocated with the MPL system. Visibility is measured by means of the widely accepted principle of forward scattering. A high-output infrared light-emitting diode transmitter projects light into a sample volume. The light scattered at a forward direction of 42 is then detected by a receiver. The light source is modulated to provide excellent rejection of background noise and natural variation in the background light intensity. The visibility sensor is calibrated once every 6 months by a calibration kit supported by the Belfort Instrument Company. The sensor’s analog output signal is proportional to visibility. A total extinction coefficient at 550 nm is derived from the visibility and is directly included in a minute digital output (Tan et al., 2010). The hourly data are derived by averaging the minute-by-minute data within those hours that have more than 30 min-by-minute measurements. Ignoring extinction from atmospheric molecules, and the difference between the wavelengths of the visibility sensor and lidar, the hourly sa,0 observed by the visibility sensor is used to validate the lidar-retrieved result. 2.4. Radiosonde In this region, upper-air meteorological parameters are measured at King’s Park station (the World Meteorological Organization 45,004 weather station, 114.17 E 22.32 N, as shown by a red triangle in Fig. 1) by the Hong Kong Observatory (HKO) (Yim et al., 2007). The measurements have been made at 12-h temporal resolutions (8:00 a.m. and 8:00 p.m. local time) since 1921 with the use of pilot balloons. With a distance of less than 20 km between King’s Park and Yuen Long station, the vertical profiles of relative humidity (RH), observed using radiosondes at King’s Park station, are used for the analyses of the lidar observations. 3. Results 3.1. Vertical distribution of RH Some components of aerosols are highly hygroscopic and can therefore scatter light more efficiently under condition with higher RH (Li et al., 2013; Lin et al., 2015). Distribution of RH is therefore required to interpret the distribution of aerosol extinction coefficient. Fig. 3 shows the vertical distributions of the seasonal-mean RH observed by the radiosondes at King’s Park station from 2005 to 2009. In this study, March, June, September, and December are treated as the first month of spring, summer, autumn, and winter, respectively. Black-dashed lines represent the vertical distributions of the overall-mean RH during the entire study period. The

Fig. 3. Vertical distributions of the seasonal-mean RH observed by radiosondes at King’s Park station from 2005 to 2009 (black-solid lines). Black-dashed lines represent vertical distributions of the overall-mean RH during the entire study period. Individual observations at 8:00 a.m. or 8:00 p.m. local time in each day are plotted as gray lines.

individual observations at 8:00 a.m. or 8:00 p.m. local time each day are plotted with gray lines. Generally, the seasonal-mean RH is the highest during the summer (82.3± 8.5% near surface), followed by those in spring (78.3± 11.6%), autumn (71.6± 12.9%), and winter (69.3± 17.4%). The vertical profile of RH could be used to approximate the mixing layer height, at which the atmospheric moisture is significantly reduced and the vertical gradient of RH is at a minimum (Seidel et al., 2010; Wang and Wang, 2014). In this study, the seasonal mean of the mixing layer height can be approximated below 1 km during the early morning and late afternoon. In addition, the vertical profiles of the seasonal-mean RH have peaks in upper levels of the mixing layer. The maxima of the seasonal-mean RH are about 85%, 90%, 80%, and 75% in spring, summer, autumn, and winter, respectively. The data indicate that the RH value typically increases with the increasing height within the mixing layer in this region. Because the light extinction of aerosols grows exponentially as the RH increases, the high RH value is likely to lead to significant aerosol hygroscopic growth and enhance the aerosol extinction coefficient in the upper level of the mixing layer. 3.2. Evaluation of lidar-retrieved sa In situ measurement of the surface aerosol extinction coefficient

sa,0 from the visibility sensor is used to evaluate the lidar-retrieved aerosol extinction coefficient. Given the blind zone in the lidar observation, the near-surface (i.e., 130 m above ground level) aerosol extinction coefficient from lidar, also noted as sa,0, is compared with the visibility-derived value for sa,0. Because sa estimated during daytime is applied throughout each day, evaluations are conducted during daytime and nighttime to identify the difference in uncertainty between these two periods. A comparison between the hourly means of surface sa,0 from lidar and the visibility sensor during daytime (6:00 a.m. to 6:00 p.m.) and nighttime (6:00 p.m. to 6:00 a.m. in next day) is presented in Fig. 4. Each data point is associated with an hourly mean of the surface sa,0 from the lidar and the visibility sensor. The points in different colors represent data from different seasons. Correlation coefficient is comparable during daytime (R ¼ 0.86) and nighttime (R ¼ 0.85), suggesting that it is reasonable to apply sa throughout each day. Correlation for all four seasons is significant, and ranges from 0.75 to 0.88. Uncertainties of the lidar-retrieved sa profiles are likely to result from error in the overlap function, uncertainty of the lidar ratio profile, missing data in the blind zone, noise in the lidar-received signal, uncertainty in the reference signal, the effect from multiple scattering and so on. The overlap problem is solved experimentally using the technique demonstrated by Campbell et al. (2002). The uncertainty, resulted from the overlap problem, has been discussed by Welton et al. (2002) and estimated at less than 10% for this lidar data set (He et al., 2006). Lidar ratio is another main source of the systematic error for the lidar-retrieved sa. To reduce this error, lidar ratio was calibrated using independent observation of AOD from MODIS. An uncertainty of within ±3 sr for lidar ratio could result from the uncertainty of the MODIS measurement (He et al., 2006). To estimate AOD from the lidar retrieval, aerosol extinction coefficient is assumed uniform within the blind zone (i.e., 130 m near surface). In most case, the error resulted from this assumption is not large because the boundary layer is usually well-mixed especially during the daytime (He et al., 2008). The effect of multiple scattering, which has not been taken into account, also introduces errors. This error is typically lower than 10% and never higher than 20% based on numerical simulation results (Ackermann et al., 1999). The overall error of the lidar-retrieved sa is estimated at a range of 20%e30% of the lidar signal (He et al., 2006).

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

277

Fig. 4. Comparison between the hourly means of the surface sa,0 from lidar and the visibility sensor (a) during daytime (6:00 a.m. to 6:00 p.m.) and (b) nighttime (6:00 p.m. to 6:00 a.m. in next day). The points in blue, green, cyan, and red represent data from spring, summer, autumn, and winter, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.3. Diurnal variation in sa Fig. 5 shows the diurnal variation in the vertical distribution of the lidar-retrieved sa in (a) spring, (b) summer, (c) autumn, and (d) winter. The mixing layer heights from 8:00 a.m. to 6:00 p.m. local time, estimated by Yang et al. (2013) using the same lidar data set, are labeled with the black points. There are few steps for estimates of these mixing layer heights. First, the first derivative of Gaussian filter is applied to obtain the smooth vertical profile of gradient of the range-corrected lidar signal. Second, the height is identified, at which the gradient profile is at minimum. Third, false results, affected by the cloud within the mixing layer, are canceled. The seasonal mean of the mixing layer height is estimated at a higher level of 0.90 km in summer and at a lower level of 0.76 km in winter. Some characteristics of the lidar-retrieved sa are observed in four seasons. First, most aerosols accumulate within the mixing layer. Second, larger sa value, which is likely associated with higher RH and greater hygroscopic growth effect, is observed in the upper level of the mixing layer. The mean levels of sa, however, differ greatly in different seasons. Lower mixing layer, stronger regional transport of pollutants from inland, and less precipitation lead to higher level of sa in autumn and winter (Wu et al., 2013). In

contrast, the combined effect of greater vertical convection, prevalent cleaner air masses, and more frequent precipitation caused by the Asian summer monsoon lead to lower level of sa for most of the summer (He et al., 2008). It is also noted that a much greater diurnal variation in sa is observed during the summer. Although the mean level of sa is low, high level of sa is still observed within the mixing layer during the day in summer.

3.4. Monthly variation in AOD To depict the AOD at midday, the AOD observed between 10:00 a.m. and 3:00 p.m. local time by lidar are extracted for analysis. Fig. 6 shows monthly variation in the AOD among MODIS (blue bars) and the 5-h (green bars) and 24-h (red bars) lidar observation during the entire study period. Good agreement is found between the monthly-mean AOD from MODIS and the 5-h lidar observation. Both show bimodal variation, with two peaks from March to April and from August to October and two troughs from May to June and from November to January. An increase in the long-range transport of dust aerosols partly contributes to the higher AOD value in spring (He et al., 2008).

Fig. 5. Diurnal variation in the vertical distribution of the lidar-retrieved sa in (a) spring, (b) summer, (c) autumn, and (d) winter. The mixing layer heights from 8:00 a.m. to 6:00 p.m. in the different seasons are plotted with black dots.

278

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

Fig. 6. Monthly variation in the AOD from MODIS (blue bars), the 5-h lidar observation from 10:00 a.m. to 3:00 p.m. (green bars), and the 24-h lidar observation (red bars) during the study period. Error bars represent standard deviations. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.5. Seasonal variation in AOD and sa It is noted in Fig. 6 that the seasonal variation in the AOD from the 24-h and 5-h lidar data differ significantly, with the largest difference seen in summer. The seasonal variation in the AOD (bluesolid lines) and the near-surface sa,0 (green-solid lines) from the 24-h and 5-h lidar data are shown in Fig. 7(a) and (b), respectively. The seasonal variation in the MODIS AOD and the visibility-derived sa,0 are also plotted with blue and green-dashed lines, respectively, for validation. The details of these values are also presented in Table 2. The lidar-retrieved and the visibility-derived sa,0 from the 24-h and 5-h data have the same seasonal variation with the lowest value seen during summer and the highest value during winter. The AOD from the 24-h lidar data shows a lower average level of 0.33 ± 0.40 in summer and a higher average level of 0.48 ± 0.25 in winter. However, the AOD from MODIS or the 5-h lidar data has another type of seasonal variation, with a higher average level (0.67 ± 0.39 and 0.67 ± 0.58, respectively) in summer than that (0.53 ± 0.23 and 0.53 ± 0.24, respectively) in winter. The largest difference is observed in summer. The seasonal

mean of AOD from MODIS or the 5-h lidar data is double that from the 24-h lidar data in summer. Because the surface sa,0 is always lower during summer, significant extinction of aerosols in the upper level should make a large contribution to the increase in AOD that is seen at midday during summer. A weak difference (less than 20% of the AOD) is observed between the seasonal means of AOD from MODIS and the 5-h lidar data in spring. If MODIS AOD data are further filtered using lidar observation to eliminate the effect of cloud, the mean MODIS AOD in spring drops to 0.65 ± 0.31, which is much closer to 0.62 ± 0.40 observed by lidar. Despite this difference, the observation from the lidar is consistent with those from MODIS and the visibility sensor. The seasonal variation in the vertical distribution of sa from the 24-h lidar data is illustrated in Fig. 8(a), including standard deviation and overall mean. Generally, the majority of aerosols accumulate within the low layer. About 64%, 83%, and 91% of aerosol extinctions are below 1 km, 1.5 km, and 2 km, respectively. The vertical profiles of sa differ widely among the four seasons. The maximum of the seasonal-mean sa in summer is about 0.2 km1, which is much lower than that of the overall mean. In winter, the reduced solar radiation inhibits the transport of aerosols to high layers; sa are therefore lower than the overall mean in the high layers and higher than the overall mean in the low layers (He et al., 2008). Because of greater effect of the long-range transport of dust aerosols, higher sa are observed in the high layer during the spring, compared with other seasons (Xuan et al., 2000). The seasonalmean sa values in all four seasons from the 5-h lidar data, as shown in Fig. 8(b), are greater than those from the 24-h lidar data. The maximum of the overall-mean sa exceeds 0.4 km1 from the 5h lidar data. In particular, the greatest increase in sa is observed in summer. The maximum of the seasonal-mean sa reaches a value greater than 0.4 km1, which is about double that from the 24-h lidar data in summer. To identify the reasons for such difference between the values for sa from the 24-h and 5-h lidar data, the seasonal variation in the vertical distribution of the increase in sa (Dsa) is plotted in Fig. 9, including overall-mean Dsa. Seasonal means of the mixing layer height from 10:00 a.m. to 3:00 p.m. local time at about 0.87 km in spring, 0.94 km in summer, 0.94 km in autumn, and 0.77 km in winter are labeled. The vertical profile of the overall-mean Dsa has a maximum value of about 0.1 km1 at about 0.7 km above ground level, which is associated with the high-humidity layer. Because light extinction from aerosols grows exponentially as the RH increases, high RH in the upper level of the mixing layer is likely to lead to significant aerosol hygroscopic growth and enhance the

Fig. 7. Seasonal variation in AOD (blue-solid lines) and near-surface sa,0 (green-solid lines) from (a) the 24-h and (b) 5-h lidar measurement. Seasonal variation in the MODIS AOD (blue-dashed lines) and visibility-derived sa,0 (green-dashed lines) are also plotted for validation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

279

Table 2 Seasonal mean (±standard deviation) of AOD and near-surface sa,0 from the 24-h and 5-h lidar measurement, the MODIS AOD, and the visibility-derived sa,0. Values 24-h Lidar-retrieved AOD Lidar-retrieved sa,0 (km1) Visibility-derived sa,0 (km1) 5-h at midday Lidar-retrieved AOD MODIS AOD Lidar-retrieved sa,0 (km1) Visibility-derived sa,0 (km1)

Spring

Summer

Autumn

Winter

0.49 ± 0.35 0.23 ± 0.20 0.25 ± 0.20

0.33 ± 0.40 0.12 ± 0.16 0.15 ± 0.19

0.58 ± 0.37 0.24 ± 0.17 0.25 ± 0.19

0.48 ± 0.25 0.32 ± 0.20 0.32 ± 0.22

0.62 0.74 0.23 0.22

± ± ± ±

0.40 0.32 0.20 0.20

0.67 0.67 0.13 0.14

± ± ± ±

0.58 0.39 0.17 0.19

0.63 0.63 0.25 0.22

± ± ± ±

0.37 0.29 0.17 0.16

0.53 0.53 0.33 0.28

± ± ± ±

0.24 0.23 0.20 0.19

Fig. 8. Seasonal variation in the vertical distribution of sa from the (a) 24-h and (b) 5-h lidar data (black-solid lines). Horizontal error bars represent standard deviations of sa. Vertical distributions of the overall-mean sa during the entire study period are plotted with black-dashed lines.

aerosol extinction coefficient at midday, when more aerosols are lifted by solar heating to this level. In addition, the vertical profiles of the seasonal-mean Dsa differ greatly among the four seasons. Much larger value for Dsa, with a maximum exceeding 0.2 km1, is observed in summer, whereas much lower value for Dsa, with a maximum of about 0.05 km1, is observed in winter. These maxima are observed in the upper level of the mixing layer in all four seasons. In addition, the values for Dsa from high layers above 1 km are much higher during summer than those during winter, which indicates that the increase in the aerosol extinction coefficient at high layers makes a larger contribution to the increase in AOD at midday during summer than that during winter. Because of an incomplete overlap between the telescope fieldof-view and the emitted laser beam, systematic errors of lidar

signal are mainly observed in the low layer. To best eliminate the effect of the overlap, seasonal variation in the vertical distribution of the percentage of the increase in sa (Dsa/sa,24h) is plotted in Fig. 10, including overall-mean Dsa/sa,24h. Seasonal means of the mixing layer height from 10:00 a.m. to 3:00 p.m. local time are also labeled. High values in excess of 50% for Dsa/sa,24h are generally observed in the high layers above 3 km, which are likely due to the extreme low values for sa during nighttime in these layers. It is more important to note that Dsa/sa,24h still increases as height increases within the mixing layer. The vertical profile of the overallmean Dsa/sa,24h has a value of about 20%e30% at the upper level of the mixing layer. In addition, much larger value for Dsa/sa,24h, with a maximum exceeding 100%, is observed in the upper level of the mixing layer in summer, whereas much lower value for Dsa/sa,24h is

280

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

Fig. 9. Seasonal variation in the vertical distribution of Dsa (black-solid lines). Dsa is the increase in sa between the 24-h and 5-h lidar data sets. Vertical profiles of the overall-mean Dsa during the entire study period are shown with black-dashed lines. Black dots are used to label the seasonal means of the mixing layer height from 10:00 a.m. to 3:00 p.m. local time.

Fig. 10. Seasonal variation in the vertical distribution of Dsa/sa,24h (black-solid lines). Dsa/sa,24h is the percentage of the increase in sa between the 24-h and 5-h lidar data sets. Vertical profiles of overall-mean Dsa/sa,24h during the entire study period are shown with black-dashed lines. Black dots are used to label the seasonal means of the mixing layer height from 10:00 a.m. to 3:00 p.m. local time.

observed in winter. The strong increase in the Dsa/sa,24h by height within the mixing layer underscores the fact that aerosols in the upper level of the mixing layer make a substantial contribution to the increase in AOD at midday during summer. 4. Discussion and conclusions The MODIS-observed AOD data have been extensively applied in atmospheric and air-quality studies. The observation of AOD from MODIS, however, is restricted to during midday. Assessment of a representative AOD by MODIS for a long-term average requires 24h measurement of AOD. Due to a reliance on solar radiation, traditional ground measurements of AOD from AERONET are also restricted to during the day. A lidar system is capable of detecting AOD and profiling the aerosol extinction coefficient throughout the day at high spatiotemporal resolutions. Studies of long-term variation in AOD and aerosol vertical distribution, however, are still quite limited and the results from MODIS and the lidar sometimes are contradictory in China. In this study, long-term lidar observations from 2005 to 2009 over Hong Kong are analyzed, with a focus on identification of the reasons for the different seasonal variation in the AOD data obtained from MODIS and lidar. The MODIS C6 AOD data are used to estimate the lidar ratio, a prerequisite parameter in the lidar inversion.

The lidar-retrieved sa shows the lowest average level but the most significant diurnal variation during summer. By extracting a 5-h period at midday from 10:00 a.m. to 3:00 p.m. local time to match with the satellite passages, the average of the lidar-retrieved AOD doubles during summer and exceeds that level during winter. This finding is consistent with the MODIS observation of a higher AOD in summer and a lower AOD in winter. The maximum increase in the seasonal-mean sa between the 24-h and 5-h lidar data exceeds 0.2 km1 in the upper level of the mixing layer (~0.7 km above ground level) during summer, whereas a much smaller increase is observed during winter. Much greater solar heating could lift more aerosols to the upper level of the mixing layer and suspend them there for a longer period of time at midday in summer (He et al., 2008). Furthermore, because light extinction from aerosols grows exponentially as the RH increases, an RH that exceeds 90% in the upper level of the mixing layer is likely to lead to significant aerosol hygroscopic growth and substantially enhance the aerosol extinction coefficient at midday in summer. These findings underscore the fact that the increase in the aerosol extinction coefficient in the upper level of the mixing layer makes the largest contribution to the increase in AOD at midday in summer. Besides effect of humidity, the diurnal variation in aerosol extinction coefficient can be also related to other factors such as local emission and photochemical reaction (Peng et al., 2011; Zhang and Cao, 2015). On the basis of the ground-level measurement, PM2.5 concentration during the daytime is higher than that during the nighttime in Hong Kong, mainly due to heavier emissions of pollutants during the daytime (Shi et al., 2012). During spring and summer, the enhanced photochemical formation of secondary particles leads to higher PM2.5 concentration during daytime than that during nighttime in Guangzhou (Zhang and Cao, 2015). Therefore, these effects may also partially contribute to the variation in aerosol extinction coefficient. This study also underscores the importance of using continuous lidar observation for the assessment of long-term variation in satellite-observed AOD data. Due to significant diurnal variation in AOD, considerable error may be introduced when long-term averages of AOD are estimated from satellite observations. Vertical distribution of aerosol extinction coefficient is one of the most important impact factors in the satellite remote sensing of ground-level PM2.5 concentration from AOD (Lin et al., 2016). Some studies of satellite remote sensing of PM2.5 concentration have employed lidar observation to capture the vertical distribution of aerosols at specific locations (Chu et al., 2013; Schaap et al., 2009). A handful of pilot studies have been conducted to estimate groundlevel PM2.5 concentration from space during nighttime (Wang et al., 2016). These studies are still challenged by the lack of continuous measurement of aerosol vertical profile. Therefore, the analysis of continuous vertical profile of aerosols in this study also provides fundamental information for satellite remote sensing of PM2.5 concentration.

Acknowledgments The study is partially supported by the research grants from the “Strategic Priority Research Program” of the Chinese Academy of Sciences (Grant No. XDA05040000), the National High Technology Research and Development Program (863 Major Project, Grant No.: SQ2010AA1221583001) of China, the National Natural Science Foundation of China (NSFC, Grant No.: 41175020 and 41375008), NSFC/RGC Grant N HKUST631/05, and the Fok Ying Tung Graduate School (NRC06/07.SC01). The authors declare no competing financial interests.

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

References € lger, P., Wiegner, M., 1999. Significance of multiple scattering from Ackermann, J., Vo tropospheric aerosols for ground-based backscatter lidar measurements. Appl. Opt. 38, 5195e5201. Amiridis, V., Balis, D.S., Kazadzis, S., Bais, A., Giannakaki, E., Papayannis, A., Zerefos, C., 2005. Four-year aerosol observations with a Raman lidar at Thessaloniki, Greece, in the framework of European aerosol research lidar network (EARLINET). J. Geophys. Res. Atmos. 110, D21203. Berkoff, T.A., Sorokin, M., Stone, T., Eck, T.F., Hoff, R., Welton, E., Holben, B., 2011. Nocturnal aerosol optical depth measurements with a small-aperture automated photometer using the moon as a light source. J. Atmos. Ocean. Technol. 28, 1297e1306. Campbell, J.R., Hlavka, D.L., Welton, E.J., Spinhirne, J.D., Scott, V.S., Hwang, I.H., 2002. Full-time, eye-safe cloud and aerosol lidar observation at atmospheric radiation measurement program sites: instruments and data analysis. J. Atmos. Ocean. Technol 431e442. Campbell, J.R., Reid, J.S., Westphal, D.L., Zhang, J., Tackett, J.L., Chew, B.N., Welton, E.J., Shimizu, A., Sugimoto, N., Aoki, K., et al., 2013. Characterizing the vertical profile of aerosol particle extinction and linear depolarization over Southeast Asia and the Maritime Continent: the 2007e2009 view from CALIOP. Atmos. Res. 122, 520e543. Cao, J.J., Lee, S.C., Ho, K.F., Zhang, X.Y., Zou, S.C., Fung, K., Chow, J.C., Watson, J.G., 2003. Characteristics of carbonaceous aerosol in Pearl River delta region, China during 2001 winter period. Atmos. Environ. 37, 1451e1460. Cavalieri, O., Cairo, F., Fierli, F., Di Donfrancesco, G., Snels, M., Viterbini, M., Cardillo, F., Chatenet, B., Formenti, P., Marticorena, B., et al., 2010. Variability of aerosol vertical distribution in the Sahel. Atmos. Chem. Phys. 10, 12005e12023. Chen, W.-N., Chen, Y.-W., Chou, C.C.K., Chang, S.-Y., Lin, P.-H., Chen, J.-P., 2009. Columnar optical properties of tropospheric aerosol by combined lidar and sunphotometer measurements at Taipei. Taiwan. Atmos. Environ. 43, 2700e2708. Chiang, C.-W., Chen, W.-N., Liang, W.-A., Das, S.K., Nee, J.-B., 2007. Optical properties of tropospheric aerosols based on measurements of lidar, sun-photometer, and visibility at Chung-Li (25 N, 121 E). Atmos. Environ. 41, 4128e4137. Chu, D.A., Kaufman, Y.J., Zibordi, G., Chern, J.D., Mao, J., Li, C., Holben, B.N., 2003. Global monitoring of air pollution over land from the earth observing systemterra moderate resolution imaging spectroradiometer (MODIS). J. Geophys. Res. Atmos. 108. Chu, D.A., Tsai, T., Chen, J., Chang, S.-C., Jeng, Y., Chiang, W., Lin, N., 2013. Interpreting aerosol lidar profiles to better estimate surface PM2.5 for columnar AOD measurements. Atmos. Environ. 79, 172e187. Esposito, F., Serio, C., Pavese, G., Auriemma, G., Satriano, C., 1998. Measurements of nighttime atmospheric optical depth preliminary data from a mountain site in southern Italy. J. Aerosol Sci. 29, 1213e1218. Fernald, F.G., 1984. Analysis of atmospheric lidar observations: some comments. Appl. Opt. 23, 652e653. Groß, S., Esselborn, M., Weinzierl, B., Wirth, M., Fix, A., Petzold, A., 2013. Aerosol classification by airborne high spectral resolution lidar observations. Atmos. Chem. Phys. 13, 2487e2505. He, Q., Li, C., Mao, J., Lau, A.K.-H., Chu, D.A., 2008. Analysis of aerosol vertical distribution and variability in Hong Kong. J. Geophys. Res. Atmos. 113, D14211. He, Q.S., Li, C.C., Mao, J.T., Lau, A.K.H., Li, P.R., 2006. A study on the aerosol extinction-to-backscatter ratio with combination of micro-pulse LIDAR and MODIS over Hong Kong. Atmos. Chem. Phys. 6, 3243e3256. Herber, A., Thomason, L.W., Gernandt, H., Leiterer, U., Nagel, D., Schulz, K.-H., Kaptur, J., Albrecht, T., Notholt, J., 2002. Continuous day and night aerosol optical depth observations in the Arctic between 1991 and 1999. J. Geophys. Res. Atmos. 107. AAC 6e1. Holben, B.N., Eck, T.F., Slutsker, I., Tanre, D., Buis, J.P., Setzer, A., Vermote, E., Reagan, J.A., Kaufman, Y.J., Nakajima, T., 1998. AERONETda federated instrument network and data archive for aerosol characterization. Remote Sens. Environ. 66, 1e16. Jerrett, M., Arain, A., Kanaroglou, P., Beckerman, B., Potoglou, D., Sahsuvaroglu, T., Morrison, J., Giovis, C., 2004. A review and evaluation of intraurban air pollution exposure models. J. Expo. Sci. Environ. Epidemiol. 15, 185e204. Johnson, R.S., Zhang, J., Hyer, E.J., Miller, S.D., Reid, J.S., 2013. Preliminary investigations toward nighttime aerosol optical depth retrievals from the VIIRS Day/Night Band. Atmos. Meas. Tech. 6, 1245e1255.
, D., Boucher, O., 2002. A satellite view of aerosols in the climate Kaufman, Y.J., Tanre system. Nature 419, 215e223. Kieffer, H.H., 1997. Photometric stability of the lunar surface. Icarus 130, 323e327. Kim, S.-W., Yoon, S.-C., Kim, J., Kim, S.-Y., 2007. Seasonal and monthly variations of columnar aerosol optical properties over east Asia determined from multi-year MODIS, LIDAR, and AERONET Sun/sky radiometer measurements. Atmos. Environ. 41, 1634e1651. Klett, J.D., 1985. Lidar inversion with variable backscatter/extinction ratios. Appl. Opt. 24, 1638e1643.
on, J.-F., Derimian, Y., Chiapello, I., Tanre
, D., Podvin, T., Chatenet, B., Diallo, A., Le Deroo, C., 2009. Aerosol vertical distribution and optical properties over M’Bour (16.96 W; 14.39 N), Senegal from 2006 to 2008. Atmos. Chem. Phys. 9, 9249e9261. Levy, R.C., Mattoo, S., Munchak, L.A., Remer, L.A., Sayer, A.M., Patadia, F., Hsu, N.C., 2013. The Collection 6 MODIS aerosol products over land and ocean. Atmos.

281

Meas. Tech. 6, 2989e3034. Li, C., Mao, J., Lau, K.-H.A., Chen, J.-C., Yuan, Z., Liu, X., Zhu, A., Liu, G., 2003. Characteristics of distribution and seasonal variation of aerosol optical depth in eastern China with MODIS products. Chin. Sci. Bull. 48, 2488e2495. Li, C., He, X., Deng, Z., Lau, A.Kai-Hon, Li, Y., 2013. Dependence of mixed aerosol light scattering extinction on relative humidity in Beijing and Hong Kong. Atmos. Ocean. Sci. Lett. 6, 117e121. Li, Y., Lin, C., Lau, A.K.H., Liao, C., Zhang, Y., Zeng, W., Li, C., Fung, J.C.H., Tse, T.K.T., 2015. Assessing long-term trend of particulate matter pollution in the Pearl River delta region using satellite remote sensing. Environ. Sci. Technol. 49 (19), 11670e11678. Lin, C., Li, Y., Yuan, Z., Lau, A.K.H., Li, C., Fung, J.C.H., 2015. Using satellite remote sensing data to estimate the high-resolution distribution of ground-level PM2.5. Remote Sens. Environ. 156, 117e128. Lin, C., Li, Y., Lau, A.K.H., Deng, X., Tse, T.K.T., Fung, J.C.H., Li, C., Li, Z., Lu, X., Zhang, X., et al., 2016. Estimation of long-term population exposure to PM2.5 for dense urban areas using 1-km MODIS data. Remote Sens. Environ. 179, 13e22. Min, Q.-L., Li, R., Lin, B., Joseph, E., Wang, S., Hu, Y., Morris, V., Chang, F., 2009. Evidence of mineral dust altering cloud microphysics and precipitation. Atmos. Chem. Phys. 9, 3223e3231. Peng, G., Wang, X., Wu, Z., Wang, Z., Yang, L., Zhong, L., Chen, D., 2011. Characteristics of particulate matter pollution in the Pearl River Delta region, China: an observational-based analysis of two monitoring sites. J. Environ. Monit. Jem. 13, 1927e1934.
rez-Ramírez, D., Lyamani, H., Olmo, F.J., Alados-Arboledas, L., 2011. Improvements Pe in star photometry for aerosol characterizations. J. Aerosol Sci. 42, 737e745. Perrone, M.R., De Tomasi, F., Gobbi, G.P., 2014. Vertically resolved aerosol properties by multi-wavelength lidar measurements. Atmos. Chem. Phys. 14, 1185e1204. Satheesh, S.K., Krishna Moorthy, K., 2005. Radiative effects of natural aerosols: a review. Atmos. Environ. 39, 2089e2110. Sayer, A.M., Munchak, L.A., Hsu, N.C., Levy, R.C., Bettenhausen, C., Jeong, M.-J., 2014. MODIS collection 6 aerosol products: comparison between Aqua’s e-deep blue, dark target, and “merged” data sets, and usage recommendations. J. Geophys. Res. Atmos. 119, 2014JD022453. Schaap, M., Apituley, A., Timmermans, R.M.A., Koelemeijer, R.B.A., de Leeuw, G., 2009. Exploring the relation between aerosol optical depth and PM2.5 at Cabauw, the Netherlands. Atmos. Chem. Phys. 9, 909e925. Seidel, D.J., Ao, C.O., Li, K., 2010. Estimating climatological planetary boundary layer heights from radiosonde observations: comparison of methods and uncertainty analysis. J. Geophys. Res. Atmos. 115, D16113. Shi, W., Wong, M.S., Wang, J., Zhao, Y., 2012. Analysis of airborne particulate matter (PM2.5) over Hong Kong using remote sensing and GIS. Sensors 12, 6825e6836. Sinha, P.R., Manchanda, R.K., Kaskaoutis, D.G., Kumar, Y.B., Sreenivasan, S., 2013. Seasonal variation of surface and vertical profile of aerosol properties over a tropical urban station Hyderabad, India: VERTICAL AEROSOL CHARACTERISTICS. J. Geophys. Res. Atmos. 118, 749e768. Spinhirne, J.D., 1993. Micro pulse lidar. IEEE Trans. Geosci. Remote Sens. 31, 48e55. Tan, H., Chen, H., Wu, D., Deng, X., Deng, T., Li, F., Zhao, X., Bi, X., 2010. The performance evaluation and data correction of the forward scattering visibility sensor. J. Trop. Meteorol. 26, 687e693. Tao, M., Chen, L., Wang, Z., Tao, J., Che, H., Wang, X., Wang, Y., 2015. Comparison and evaluation of the MODIS collection 6 aerosol data in China. J. Geophys. Res. Atmos. 120, 2015JD023360. Wang, J., Aegerter, C., Xu, X., Szykman, J.J., 2016. Potential application of VIIRS Day/ Night Band for monitoring nighttime surface PM2.5 air quality from space. Atmos. Environ. 124 (Part A), 55e63. Wang, X.Y., Wang, K.C., 2014. Estimation of atmospheric mixing layer height from radiosonde data. Atmos. Meas. Tech. 7, 1701e1709. Welton, Ellsworth J., Voss, Kenneth J., Quinn, Patricia K., Flatau, Piotr J., Markowicz, Krzysztof, Campbell, James R., Spinhirne, James D., Gordon, Howard R., Johnson, J.E., 2002. Measurements of aerosol vertical profiles and optical properties during INDOEX 1999 using micropulse lidars. J. Geophys. Res. Atmos. 107, 8019. Wu, D., Fung, J.C.H., Yao, T., Lau, A.K.H., 2013. A study of control policy in the Pearl River Delta region by using the particulate matter source apportionment method. Atmos. Environ. 76, 147e161. Wu, Y., Guo, J., Zhang, X., Tian, X., Zhang, J., Wang, Y., Duan, J., Li, X., 2012. Synergy of satellite and ground based observations in estimation of particulate matter in eastern China. Sci. Total Environ. 433, 20e30. Wu, Y., Cordero, L., Gross, B., Moshary, F., Ahmed, S., 2014. Assessment of CALIPSO attenuated backscatter and aerosol retrievals with a combined ground-based multi-wavelength lidar and sunphotometer measurement. Atmos. Environ. 84, 44e53. Xuan, J., Liu, G., Du, K., 2000. Dust emission inventory in Northern China. Atmos. Environ. 34, 4565e4570. Yang, D., Li, C., Lau, A.K.-H., Li, Y., 2013. Long-term measurement of daytime atmospheric mixing layer height over Hong Kong. J. Geophys. Res. Atmos. 118, 2422e2433. Yim, S.H.L., Fung, J.C.H., Lau, A.K.H., Kot, S.C., 2007. Developing a high-resolution wind map for a complex terrain with a coupled MM5/CALMET system. J. Geophys. Res. Atmos. 112, D05106. Yuan, Z., Yadav, V., Turner, J.R., Louie, P.K.K., Lau, A.K.H., 2013. Long-term trends of ambient particulate matter emission source contributions and the accountability of control strategies in Hong Kong over 1998e2008. Atmos. Environ. 76, 21e31.

282

C.Q. Lin et al. / Atmospheric Environment 140 (2016) 273e282

Zhang, J., Reid, J.S., Miller, S.D., Turk, F.J., 2008. Strategy for studying nocturnal aerosol optical depth using artificial lights. Int. J. Remote Sens. 29, 4599e4613. Zhang, Y.L., Cao, F., 2015. Fine particulate matter (PM2.5) in China at a city level. Sci. Rep. 5, 14884.

Zhong, L., Louie, P.K.K., Zheng, J., Yuan, Z., Yue, D., Ho, J.W.K., Lau, A.K.H., 2013. Scienceepolicy interplay: air quality management in the Pearl River Delta region and Hong Kong. Atmos. Environ. 76, 3e10.