Accurate lidar ratio (LR) and better understanding of its
variation characteristics can not only improve the retrieval accuracy of
parameters from elastic lidar, but also play an important role in assessing
the impacts of aerosols on climate. Using the observational data of a Raman
lidar in Shanghai from 2017 to 2019, LRs at 355 nm were retrieved and their
variations and influence factors were analyzed. Within the height range of
0.5–5 km, about 90 % of the LRs were distributed in 10–80 sr with
an average value of 41.0
Aerosols in the atmosphere can affect the earth's climate by absorbing and
scattering solar radiation (direct effect of aerosols) (Huang et al., 2014;
Wang et al., 2013) or acting as cloud condensation nuclei, which can affect
cloud physical properties and precipitation (indirect effect of aerosols)
(Huang et al., 2006; Liu et al., 2019a, b; Yan and Wang, 2020). In
general, the vertical distribution information of aerosols is required to
improve our understanding of aerosol climate effects (Ferrare et al., 2001;
Sicard et al., 2011). For example, Wang et al. (2020b) found that
dust-forced radiative heating decreased significantly as the transport
height of dust aerosols decreased. A study by Lu et al. (2020) showed that
anomalous elevated aerosol layers above 2 km led to warming in the upper
atmosphere (
As an active remote-sensing instrument, the elastic scattering lidar can obtain vertical distribution information of aerosols; however, it is necessary to assume an aerosol extinction-to-backscattering ratio (i.e., lidar ratio, LR) in the retrieval process (Fernald, 1984; Welton et al., 2001), which can result in significant errors for the extinction coefficient, followed by aerosol optical depth (AOD). To our knowledge, the LRs at 355, 532, and 1064 nm are usually assumed to be 50 sr in China (Fan et al., 2018; Gong et al., 2015; Lv et al., 2020; Ma et al., 2019). In addition, CALIOP (Cloud-Aerosol Lidar with Orthogonal Polarization) onboard CALIPSO (Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations) can observe the vertical distribution of global aerosol optical properties, and its data products have been widely used around the world (Kim et al., 2018). The CALIOP algorithm first determines the type of aerosol according to the aerosol classification algorithm and then uses lookup table of multiple types of aerosol to determine the LR (Kim et al., 2018; Omar et al., 2009). Therefore, the quality of the CALIOP aerosol products depends on the accuracy of aerosol type identification and the consistency between actual LR and that in the lookup table (Painemal et al., 2019). LR is a complicated function of time and space, which depends on aerosol size distribution and particle composition (Reagan et al., 1988). LR is also affected by meteorological elements (Salemink et al., 1984), such as relative humidity (RH), which can change the aerosol particle size distribution and refractive index (Young et al., 1993). Therefore, good knowledge of accurate LR and its variation characteristics can not only improve the retrieval accuracy of parameters from elastic lidar, but also obtain information on aerosol types to trace the source of pollutants (Franke et al., 2001).
According to the definition of LR in Müller's (2003) study,
LR can be obtained by a variety of methods, such as high-spectrum-resolution lidar (HSRL), Raman lidar, and joint retrieval using sun photometer and elastic lidar (Zhao et al., 2018). Raman lidar can independently retrieve the extinction coefficient and backscatter coefficient of aerosols and obtain LR by combining elastic backscatter and Raman backscatter signals (Ansmann et al., 1992), which is the most widely used independent measurement method at present. Moreover, the LR measured by Raman lidar is a useful index to study the variations in aerosol physical properties (Ferrare et al., 2001).
A large number of observations and analysis of LR have been carried out based on Raman lidar all over the world. Since the establishment of the European Aerosol Research Lidar Network (EARLINET) in 2000, long time series observation data of vertical distribution and LR for various types of aerosol have been obtained on the European continent (Müller et al., 2007; Wandinger et al., 2016). In South Korea and Japan, the LR of Asian dust and biomass burning aerosols has also been studied based on Raman lidar (Murayama et al., 2004; Noh et al., 2007, 2008). The LR observed around the world usually shows different values due to different types of aerosols. However, long-term observations and research of LR in China are limited (Wang et al., 2016) due to the limitation of observation instruments. In particular, the observations and studies of LR are still rare in east China; however, range-resolved LR profiles based on independent measurement on a regional scale are very important. On one hand, the range-resolved LR obtained from ground-based Raman lidar can not only be used for comparison with 355 nm LR obtained from ATLID (Atmospheric LIDar) on EarthCARE (Earth Clouds and Radiation Explorer) planned to be launched by ESA (European Space Agency) (Liu et al., 2020a; Nicolae et al., 2018), but can also provide a reliable basis for the inversion hypothesis of elastic lidar in Shanghai and surrounding areas and improve product reliability for elastic lidar networks such as the Asian dust and aerosol lidar observation network. On the other hand, vertical distribution of aerosol absorption properties reflected by LR can be used as an input parameter for regional climate models (Mehta et al., 2018), which can further improve the calculation accuracy of radiative forcing.
In addition, studying the influence factors of LR in Shanghai can be conducive to understanding the LR variation characteristics and determining the source of pollutants. With these motivations, the vertical and temporal variations in LR and its influence factors in Shanghai were analyzed using the results retrieved from Raman lidar, which laid a solid foundation for the quantitative study of pollution and its causes in the future.
The Raman depolarization lidar (Raymetrics S.A., Athens, Greece, model
LR231-D300) used in this study is deployed on the roof of a building (31.1916
HYSPLIT-4 (Hybrid Single Particle Lagrangian Integrated Trajectory Model, Version 4) is a professional model jointly developed by the National Oceanic and Atmospheric Administration (NOAA) and the Australian Bureau of Meteorology for calculating and analyzing transport and diffusion trajectories of atmospheric pollutants and has been widely used in many studies around the world (Huang et al., 2012; Noh et al., 2007). It supports the input of a variety of meteorological data, and meteorological reanalysis data provided by NOAA were used in this study.
MERRA-2 (The Modern-Era Retrospective Analysis for Research and
Applications, Version 2) is an atmospheric reanalysis dataset provided by
the National Aeronautics and Space Administration (NASA) and the Global
Modeling and Assimilation Office (GMAO) (Gelaro et al., 2017). The aerosol
optical property data used in this study were derived from the 1 h average
product of the MERRA-2 tavg1_2d_aer_Nx dataset and CO column concentration from the MERRA-2
tavg1_2d_chm_Nx. The spatial
resolution of the two datasets was
0.625
ERA5 is a global atmospheric reanalysis dataset provided by the European
Centre for Medium-range Weather Forecasting (ECMWF) (Zhao et al., 2020). In
recent years, some studies have evaluated the accuracy of reanalysis data
provided by the ECMWF based on radiosonde data. For example, Luo et al. (2020)
found that the average RH discrepancy between the ERA-Interim and radiosonde was
within 10 % below 500 hPa. Song et al. (2020) found that the root mean
square error (RMSE) of ERA5 RH was 3.85 % compared with the RH profile of the radiosonde. The above results show that RH from reanalysis data has good
accuracy, and it has been widely used in various research fields (Sajadi et
al., 2020; Tzanis et al., 2019; Xiao et al., 2020). The RH data of ERA5 used
in this paper were divided into 37 layers vertically (1–1000 hPa). The
temporal resolution was 1 h and the spatial resolution was 0.125
Original signals need to be pre-processed before retrieval, including background subtraction, photon-counting signal dead-time correction, signal gluing, and overlap correction (D'Amico et al., 2016). The calculation of the glue coefficients in this study used the methods proposed by Newsom et al. (2009). In order to reduce the influences of lidar incompletely overlapping detection areas on retrieved results, only signals in the complete overlap area were used for retrieval. In addition, affected by the location altitude of the Raman lidar and the least square method used in the retrieval process, the lowest height of LR obtained by the Raman method was 569.5 m (a.s.l.). Since the Raman lidar used in this study can detect the Raman scattering signal of 387 nm nitrogen and signal-to-noise ratios of Raman signals in daytime are much lower than that in the nighttime, the 355 nm LR at night can be obtained through retrieval. The retrieval results of raw signals were counted by hour, and the hours with more than 15 minutes of retrieval results were regarded as effective observation hours. The retrieval results within the effective observation hours were averaged to obtain hourly average data. During the observation period, data of 667 effective observation hours were obtained through retrieval and statistics. The monthly distribution is shown in Fig. 1.
Effective observation hours per month from 2017 to 2019.
Previous studies indicated that there are some sources of errors in the retrieval. The relative errors of particle extinction coefficients caused by assumed air density profile are 1.5 % (Masonis, 2002), and the relative errors of particle backscatter coefficients caused by reference height can be 10 % (Ansmann et al., 1992). The mean deviations of particle extinction coefficients caused by signal detection are within 15 % in the 350–2000 m height range and within 20 % in the 3000–4000 m height range (Pappalardo et al., 2004). The difference is caused by different signal-to-noise ratios at low altitude and high altitude. Due to the low signal-to-noise ratio, there were usually more missing values at high altitudes.
Figure 2a shows the averaged LR profile for 667 h. Because of the variability of aerosol particle size and microphysical properties with
height (Singh et al., 2005), the averaged LR was characterized by large
variability, ranging from 17 to 82 sr. LR reached a maximum at the
height of 600 m and decreased with the increase in height. The averaged LR
in the height range of 0.5–5 km was 41.0
General variation in LR.
In order to investigate the variations in LR at different altitude ranges,
Fig. 2b presents the averaged LR for different altitude ranges. The
averaged LR from 0.5 to 1 km was 68.2
Figure 3 shows the frequency distribution of LR for different altitude
ranges. Overall, LRs were widely distributed in the altitude range of 0.5–5 km. In most cases (about 90 %) LR ranged from 10 to 80 sr with the
highest frequency of 17.3 % between 40 and 50 sr. It should be noted
that the number of observations trailed off at larger LR and the frequency
of abnormally large LR (
LR frequency distribution.
Figure 4a presents seasonal variation in LR over Shanghai during the observation period. The seasonal average LR was the largest in autumn with 47.6
LR temporal variations.
The statistics of the averaged LR for a different height range in each month were
shown in Fig. 4b. LR for all months decreased
with the increase in altitude. The averaged LR below 2 km was the largest in
October, which was attributed to smoke aerosols produced by biomass burning
in the surrounding cities and rural areas during the harvest season (Nie et al.,
2015). In view of LR vertical variations in different months, aerosols with
LR
Liu et al. (2012) reported that vast majority of aerosol particles in the
Yangtze Delta region (including Shanghai) were below 2 km. In order to
precisely analyze the variation characteristics of the LR in Shanghai, Fig. 5
shows LRs of 667 effective observation hours below 2 km. The abnormally large
LRs (
The averaged LR of effective observation hours. The white areas indicate invalid values.
From Eq. (1), we found that LR was negatively correlated with
Effect of
It is worth noting that LR responding to large
In order to further understand the influences of wind directions on LR and
its vertical distribution, a cluster analysis of back trajectories was used to
study the transport of atmospheric aerosols. Based on the HYSPLIT-4 model
(Franke et al., 2001; Noh et al., 2007), the 72 h backward
trajectories at the height of 1000 m were shown in Fig. 7a. The cluster
analysis resulted in four main air mass directions (Hänel et al., 2012;
Pietruczuk and Podgorski, 2009). Backward trajectory cluster analysis based
on the HYSPLIT model is widely used in atmospheric aerosol research (Wang et
al., 2020a; Xu et al., 2018; Zhang et al., 2020). We performed a
significance test on the cluster analysis results, and the one-way ANOVA
showed that
The averaged LR and
The averaged LR affected by the aerosols brought by air mass 2 was
approximately equivalent to the LR affected by air mass 1, with an
averaged value of 39.4
The averaged LR of air mass 3 was 44.2
The averaged LR affected by aerosols from air mass 4 was 42.6
In summary, the variations and vertical distributions of LR and
As mentioned previously, the vertical variations in absorbing aerosols and their influence factors played an important role in evaluating the aerosol radiation effect and studying the cause of pollution (Mishchenko et al., 2004). The LR vertical variation under different atmospheric turbidity has rarely been discussed; however, previous studies have analyzed the vertical profiles of LR in different pollution degree cases and their main concern was the averaged LR of the aerosol layer (Chen et al., 2014; Wang et al., 2016). AOD is an important parameter to characterize aerosol optical properties, which can reflect aerosol content in the atmosphere and is also an important index to evaluate atmospheric quality and visibility (Cheng et al., 2015; Hess et al., 1998; Qi et al., 2013). Previous studies have shown a positive correlation between AOD and LR by analyzing averaged LR for different AOD ranges (Ferrare et al., 2001; He et al., 2006) due to the increase in aerosol absorption and extinction caused by the increase in small particles (Takamura et al., 1994).
AOD was obtained by integrating 355 nm extinction coefficients in the range of 0.5–2 km. The averaged profile of LR below 2 km in different AOD ranges was drawn as shown in Fig. 8. Under clean conditions, LR decreased more dramatically with the increase in height. By contrast, the lack of significant vertical variability of LR under high atmospheric-turbidity conditions reflected the homogeneous vertical distribution of absorbing aerosols. The result that the vertical slope of LR presented a decreasing trend with increasing atmospheric turbidity can be explained by aerosol radiative effects on thermal structure and atmospheric stability. Under high atmospheric-turbidity conditions, aerosol particles that absorb a large amount of solar radiation during the day radiatively warm the surface at night but radiatively cool the air above the surface (Jacobson and Kaufman, 2006; Ramanathan et al., 2005). The decrease in the atmosphere stability due to the temperature difference increases vertical turbulence and results in the homogeneous vertical distribution of aerosols. By contrast, in the clear and pollution-free nights, the surface radiation cooling results in temperature inversion near the ground. The stable atmosphere is not conducive to the lifting of absorbing aerosols, resulting in a significant vertical variation in LR.
LR profiles in different AOD intervals.
Spatial distribution of AOD, AAOD, CAOD, and CO column concentrations in five cases. From left to right, different cases are represented, and from top to bottom, different tracers are represented. The blue star is the location of the Raman lidar.
As shown in Fig. 5, abnormally large LR occurred occasionally in relatively
high locations approximately above the top of PBL in spite of a usual decay trend in LR with height. To investigate the reasons, we selected 5 d with LR
Biomass burning is one of the important sources of PM, organic carbon (OC)
and BC in the atmosphere (Wu et al., 2020). It also emits
pollutant gases such as CO, SO
RH and LR profiles in three cases:
Although the abnormally large LR above 1 km was mainly relevant for the advection of biomass burning aerosols, it should be noted that the increasing aerosol extinction caused by the increase in RH could also result in large LR (Salemink et al., 1984). For example, Ackerman (1998) found that the LR of continental aerosols increased from 40 to 80 sr with RH. Figure 10 presents LR and RH profiles for three cases and shows that LR was a function of RH. The abnormally large LRs above 1 km had a good corresponding relationship with high RH, which demonstrated that the abnormally larger LRs were also related to high RH.
For the first time, a long-term (2017–2019) observation based on Raman lidar
was carried out in Shanghai. The aerosol 355 nm LR was retrieved, and the
variations in LR and their influence factors were analyzed. In the height
range of 0.5–5 km, about 90 % of LRs were distributed in 10–80 sr,
with an average of 41.0
LR and
We analyzed the spatial distribution of 500 nm AOD, AAOD, CAOD, and CO column
concentrations of five cases with LRs
The data presented in this paper are available from the corresponding authors upon request.
TL retrieved the data and wrote the paper. QH and YC formulated the project goals and edited and reviewed the paper. JL, QL, and WG downloaded and analyzed the reanalysis data. GH, WS, and XY revised the paper.
The authors declare that they have no conflict of interest.
We are grateful to NASA for providing MERRA-2 data and the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model. We also gratefully acknowledge the ECMWF for the provision of the ERA5 dataset.
This research has been supported by the National Key R&D Program of China (grant no. 2016YFC0201900), the National Natural Science Foundation of China (grant nos. 41975029, 91637101, and 91644211), the Science Research Project of Shanghai Meteorological Service (grant no. MS202016), the Chinese Ministry of Science and Technology (grant no. 2018YFC1506305), and the Fundamental Research Funds for the Central Universities (grant no. 2232019D3-27).
This paper was edited by Jianping Huang and reviewed by two anonymous referees.