Dependence of Backscattering Coefficients of Atmospheric Particles on their Concentration and Constitution under Dry and Humid Conditions at Southwestern Japanese Coast in Spring

The dependence of atmospheric particles’ backscattering coefficients (BSCs) on their concentration and constitution in the near surface air at southwestern Japanese coast in the spring of 2011–2015 was investigated. The BSCs were measured with a ceilometer at the wavelength 905 ± 5 nm. The size-segregated concentrations of particles larger than 0.3 μm were measured with optical particle counters. The constitution of particles in each of eight episodes, i.e., the portions of mineral dust, sea salt, sulfate and soot particles, was obtained with individual particle analysis by using an electron microscope. There was a close correlation (R = 0.76) between the BSCs and the volume concentrations of the particles when the relative humidity (RH) was lower than ~70%, regardless of the difference in particle constitutions. In contrast, the BSCs normalized with aerosol concentrations differed largely even at similar concentrations of particles when the RH was larger than 70%. The BSCs of particles dominated by marine origin (sea salt) could be several times larger than those dominated by land origin (dust and soot) at 90% RH, as the latter might be slightly larger than the BSCs under the dry conditions. On the other hand, the difference of volume-size distributions of particles likely also made the BSCs largely different from each other under the humid conditions. These results indicate that the concentration of atmospheric particles was the key parameter to determine the BSCs under dry conditions, while the BSCs under humid conditions were closely dependent on the content of deliquescent components and the size distributions of the particles.


INTRODUCTION
The scattering and absorbing of solar radiation by atmospheric particles are the major pathways for the particles to directly influence the distribution of energy in the atmosphere (Hansen et al., 1997).Understanding the dependence of the scattering on particles' physical and chemical properties, i.e., their size, chemical composition and deliquescent state, is crucial for an accurate description of aerosol radiative forcing in the examination of the relation between aerosol and radiation transfer (Houghton et al., 2001).In particular, deliquescent and hygroscopic particles can be enlarged via water vapor condensation, which may largely alter their performance in scattering and absorbing atmospheric radiation (Seinfeld and Pandis, 1997;Tang et al., 1997) .
The scattering of atmospheric particles has frequently been investigated in model studies and field observations.In numerical models, the extinction of light is usually quantified with the integration of scattering and absorbing effects of the particles (e.g., Takemura et al., 2002;Goto et al., 2011;Dai et al., 2015).Major aerosol types included in the models are carbonaceous, sulfate, soil dust, and sea salt (e.g., Takemura et al., 2002).In the consideration of hygroscopic growth of the particles, the extinction coefficient is usually estimated with a function of relative humidity.The chemical composition of atmospheric particles is usually considered a homogeneous mixture of multiple components.Different types of particles are included in models as being externally or internally mixed in the air.In fact, the morphology of a particle is also crucial for its optical properties.For example, the core-shell structure of aged soot particles, caused by the condensation of water vapor and the coating of secondary species, can cause large differences in light scattering and absorption in comparison with fresh soot particles (Chandra et al., 2004;Moffet and Prather, 2009).How to describe the contribution of such structures of particles was also challenged in model studies (e.g., Zhang and Thompson, 2014).
Field observations with lidars, sky radiometers, sun photometers, nephelometers and etc., on the ground or onboard flying objects, have been conducted in various areas (e.g., Hayasaka et al., 2007;Markowicz et al., 2008;Chen et al., 2014;Han et al., 2014;Kitakoga et al., 2014).The approaches applied in the observations have made the scattering and absorption of atmospheric particles available and provided quantitative data for the assessment of the warming/cooling effects of the particles.However, the correspondence of scattering and absorption to aerosol compositions was usually investigated using the composition of integrated aerosol samples (e.g., Carrico, 2003;Liu et al., 2008).Only the average characteristics of total collected particles during the periods of particle collection are available by integrated samples and this limitation has hindered a more accurate elucidation of the correlation between the optical and physiochemical properties of particles.In consequence, this kind of data has not allowed a parameterization suitable for an accurate description of the radiative forcing by atmospheric particles according to their constitution, and their changes that are caused by processes such as water vapor condensation.
A better understanding of the dependence of the optical properties on the shape, size and compositions of atmospheric particles will largely benefit the assessment of aerosols' role in the energy transfer in the atmosphere.Kandler et al. (2007) tried linking the characteristics of atmospheric dust particles quantified by individual particle analysis, with the optical properties of the particles.They encountered a very large difference between the observed values and the theoretical estimations, which was attributed to the diversity of the composition and morphology of the particles.Since then, little progress has been archived, until Denjean et al. (2015) recently reported that simply considering African dust particles, which had traveled across the Atlantic, as spherical particles might largely overestimate the radiative effect of the particles.
In this study, the backscattering coefficients (BSCs) and the concentrations of particles in the Asian continent outflow were continuously observed at a site on the southwestern Japanese coast.Single particle analysis using a scanning electron microscope equipped with an energy dispersive X-ray spectrometer was applied in order to quantify the size, shape and elemental composition of collected particles.The dependence of BSC of the particles on their concentrations and constitutions under different weather conditions was investigated.

Measurements of Backscattering and Number Concentration
The observations were carried out at Amakusa Environmental Research Unit (AERU: 32°19'N, 129°59'E, 35 m a.s.l.) of the Prefectural University of Kumamoto in the spring (March, April, May) of 2011-2015.AERU is located on the southwestern coast of Japan and faces the East China Sea with ocean to its northwest, west, and south (Fig. 1).There are a few agriculture fields to its east.Local influences are not expected in cases when air masses come from the south, southwest, west, or northwest directions.This ensures the site is suitable to the observations of air mass from the Asian continent after undergoing the marine environment.A ceilometer (Vaisala, CT25K) was used to measure the backscattering signals from surface to about 3 km at a time interval of 2 minutes.The ceilometer is an instrument to observe the altitudes of cloud bases.It can also be applied to observe vertical profiles of atmospheric particles by receiving the backscattering signals from the particles (Münkel et al., 2004).The wavelength of its laser is 905 ± 5 nm and the vertical resolution is 30 m.The photometry response per unit of aerosol mass concentration reaches its maximum when the particle diameter is close to 905 nm and the response for particles smaller than 0.3 µm and larger than 10.0 µm is less than 50% of the maximum (Gebhart, 2001).The backscattering coefficients (BSCs) in the lowest layer, from surface to 30 m, were used to show backscattering intensities of particles in the near surface air.It has been noticed that backscattering signals of ceilometers can be largely attenuated by the extinction of aerosols and by the absorption of water vapor in cases when the transmitted laser pulses travel long distances in the air.In case of 1.0 g cm -2 precipitable water along the laser path, the transmission is approximately 0.91 (Markowicz et al., 2008).In the lowest layer of this study, the travelling distance was about 60 m.The precipitable water is approximately 0.06 g cm -2 if the temperature is 15°C and the humidity is 80%.Therefore, the influence of water vapor is expected to be very small.Beam overlap correction was not applied to calibrate the BSCs for accurate values.The variability of this correction could be up to 20% near the surface (Markowicz et al., 2008).However, the absence of the correction does not influence the conclusions of this study because we used the data only in the lowest layer, and focused on the variation of the BSCs corresponding to particles, rather than the accurate values of the BSCs (Welton and Campbell, 2002).
Size-segregated number concentrations of particles in the range of diameter > 0.3, > 0.5, > 1.0, > 2.0 and > 5.0 µm were observed with optical particle counters (RION, KC-01D in 2011-2014, and KC-01E in 2015) at a time interval of 15 minutes.The optical particle counters (KC-01D and KC-01E) were calibrated before the observation campaign every year by the manufacturer, and a zero calibration was carried out once a week during the observation periods.To compare the backscattering intensity and the concentration of the particles, the size-segregated concentrations of particles in number were converted to concentrations in volume based on the fact that the scattering by a particle at the wavelength close to the particle size is approximately proportional to the volume of the particle (Pinnick et al., 1980).The integrated particle volume (IPV) of particles for each number-size distribution was used.The diameter used for a size bin was the medium size of the bin measured by the optical particle counters with a maximum size of 10.0 µm.According to Sakai et al. (2008), the IPV might be overestimated by 5-8% because of the use of the number concentration from optical particle counters.Meteorological conditions including surface pressure, temperature, relative humidity, wind speed, wind direction, and precipitation were monitored with a weather transmitter (Vaisala, WXT520).
Here and also in this study, coarse particles refers to particles of diameter larger than 1 µm.
One-hour averages of the BSCs, the size-segregated number concentrations of the particles, and consequently the IPVs were applied.The data during rain were excluded in order to avoid the influence of rain.In total, 7,132 sets of one-hour averaged BSCs, IPVs, and meteorological factors were screened for the investigation.

Particle Collection and Analysis
Observations were carried out at AERU in eight periods under different weather conditions (Table 1).Five samples, marked by D1-D5, were collected under relatively dry conditions (RH < 70%) and three samples, marked by W1-W3, under humid conditions (RH ~90%).D1, D2, D3 and D4 were collected at different stages of the passage of a dust-loading cyclone.In each time of sample collection, particles were collected on three Ti grids by using a single stage of PIXI cascade impactor (PIXI international).The flow rate of the sampling air of the impactor was 1 L min -1 .The 50% cutoff diameter of the impactor is 0.25 µm if the density of particles is 1.0 g cm -3 .The sampling time for one grid was between 0.5 and 3.0 minutes, depending on the status of air pollution.
Collected particles were analyzed with a scanning electron microscope, which is equipped with an energy dispersive X-ray analyzer.In the analysis, areas along a line crossing the particle impaction center were selected.Particles in the selected areas were photographed.The size of a particle was obtained from its micrograph with image analysis, which was the equivalent diameter of the particle projection area.
According to the elemental composition and the shape, the particles were classified into the following categories.Mineral particles were those containing silicon (Si) and aluminum (Al), but did not contain sodium (Na), and had irregular shapes.Sea salt particles were those containing Na and could be largely damaged by the electron beams in EDX analysis.Mixture particles were the mixture of minerals and sea salt, i.e., those containing Al, Si and Na and showed the above natures.Sulfate particles were those containing S and evaporated under the electron beams.Fly ash particles were those containing Al, Si, or iron (Fe) and had a spherical shape but did not evaporated under the electron beams.Soot particles were those containing remarkable carbon (C), had a fractal structure or showed a shape of aggregates or shrinkage of chain-like particles.Particles that could not be categorized into the above groups were summarized as others.

BSC vs. Aerosol Concentration
The BSCs and the IPVs were compared according to the extreme, intermedium, and low aerosol episodes under relatively dry (RH < 70%: In fact, the critical RH was around 70%.In the following descriptions, we simply use RH < 70% or RH > 70%) and humid (RH > 70%) conditions (Fig. 2).The BSCs were present in wide ranges.There was a close correlation between the two factors for extreme episodes under relatively dry conditions (Fig. 2(a): R 2 = 0.76).The correlation for intermedium episodes under relatively dry conditions was weaker than that for the extreme episodes (Fig. 2(b): R 2 = 0.55), and there was no correlation for low episodes under relatively dry conditions (Fig. 2(c): R 2 = 0.34).On the other hand, no clear correlations can be confirmed for the episodes under humid conditions (Figs. 2(d)-2(f)), except that there was likely a weak correlation for the extreme episodes under humid conditions (Fig. 2(d)).A general understanding is that large concentrations of particles usually result in large BSCs.
The BSC per unit concentration of particles was estimated with the ratios of BSC to IPV.The ratios were further normalized with the average of ratios when the relative humidity was less than 40%, which has been proved to be the critical relative humidity suitable for the investigation of aerosol particle scattering under dry conditions (Tang, 1996).Note that the BSCs linearly correlate with the IPVs if the normalized ratios are close to unit.Fig. 3 illustrates the normalized ratios versus the relative humidity.Under the conditions of RH < 70%, the ratios were frequently smaller than two and close to one.That means the BSCs were determined almost only by the volume concentrations of the particles.In contrast, the BSCs differed largely under humid conditions of RH > 70%, and no correlations between the ratio and RH can be confirmed.

Correspondence to Aerosol Constitutions under Relatively Dry Conditions
The relative humidity was smaller than 70% and the normalized ratios of BSCs to IPVs were close to one when samples D1-D5 were collected (Fig. 3 and Table 1).D1 and D2 were collected immediately after an episode of high aerosol particle loading.The concentrations were relatively low when D3 and D4 were collected.Along with the reduction of aerosol concentration, the BSCs decreased from 2.7 × 10 -4 -3.0 × 10 -4 sr -1 m -1 at the time of D1 and D2 collection to 1.9 × 10 -4 -2.4 × 10 -4 sr -1 m -1 at the time of D3 and D4 collection (Table 1).At the time of D5 collection,  the aerosol concentration was the highest among the eight samples and the BSC was also the largest among the samples D1-D5.We also checked the backward trajectories calculated by the NOAA HYSPLIT online model.The air parcels from which D1-D5 were collected arrived at AERU areas after rapid long-distance travel from northern China and Mongolia, where there are a number of desert areas (Fig. 1).Thus the air at the time of D1-D5 collection was typically postfrontal air (weather charts omitted).Similar postfrontal air has been frequently observed at the AERU areas in previous studies (e.g., Zhang et al., 2006).
The constitutions of D1-D5 particles were very different (Fig. 4).In D1, mineral particles and mixture particles together occupied approximately 40% (mineral, 12%; mixture, 25%) of the particles, and sea salt particles occupied 50%.When the dust became weak (D2), the fraction of mineral, mixture and sea salt particles decreased and sulfate particles constituted about 20%.There was also a large fraction of other particles.In D3 and D4, sea salt particles were the majority (D3: 76%; D4: 94%) and other particles including mineral particles occupied very small fractions.Different from the constitutions of D1-D3, soot particles constituted a fraction of 55% and mineral particles constituted a fraction of 18% in D5.
The above results indicate that the BSCs under the relatively dry conditions were independent from the composition and constituents of the particles.This was determined only by the concentrations of particles.The deliquesce relative humidity (DRH) of dust particles that have not been changed by atmospheric processes is larger than 94% because the dust particles contained less electrolytes and are very hydrophilic (Shi et al., 2008;Denjean et al., 2015).The DRH of sea salt particles is about 75% (Wexler and Seinfeld, 1991).Therefore, the DRH of particles mixed by mineral and sea salt components must be close to the DRH of sea salt particles.Particles of sea salt or sea saltmineral mixtures were the majorities in D1-D4, but the relative humidity was only about 50% (Table 1).Thus, the particles existed in a non-deliquescent state.Although the relative humidity was as large as 66% for D5, the particles were mainly soot particles and mineral particles, and considerable deliquescence of the particles was not expected.Note that atmospheric processed soot particles start to deliquesce at 63-76% (Li et al., 2014).Examples of a SEM image of particles in D1 is illustrated in Fig. 5(a), showing that most of the collected particles were in an irregular shape and did not have an apparent sign (usually a spherical shape) of particle's coagulation and dehydration.Therefore, the BSCs at the relatively dry conditions were the results of the scattering of naked (meaning the particles had no or less liquid solution coating) and non-deliquescent particles.A few studies investigated both the backscattering coefficients  and the volume or mass concentration of atmospheric particles under dry conditions (relative humidity smaller than 40%-60%), and reported an approximately proportional relation between the coefficient and the concentration (Lowenthal and Kumar, 2004;Münkel et al., 2004;Wang et al., 2012).

Correspondence to Aerosol Constitutions under Humid Conditions
Under the relatively humid conditions (RH > 70%), the normalized ratios of BSCs to IPVs were sporadically present in a range even when the concentrations of atmospheric particles were similar.The range became larger as the RH increased.In many episodes, the BSCs were much larger than those under relatively dry conditions, some even by up to 10 times.Samples of W1-W3 were collected under relatively humid conditions, 90-92% RH.The normalized ratio of BSC to IPV at the time of W3 collection was only slightly larger than those of D1-D5.In contrast, the ratios at the time of W1 and W2 collection were much larger than those of W3 and D1-D5 (Fig. 4).Individual particle analysis revealed that the difference of W2 in aerosol constitution from W1 and W3 was that W2 contained a large fraction (47%) of sea salt particles and mixture particles.In contrast, there were few sea salt-containing particles but a large fraction of sulfate particles in W1 and W3 samples (Fig. 4).
It could be expected that sea salt and sulfate particles, due to their hygroscopic and deliquescent properties, existed in the liquid phase and were present as wet particles under the humid (approximately 90%) but unsaturated conditions (Wexler and Seinfeld, 1991).According to experimental studies, the droplet growth factors of NaCl particles and sulfate particles could be more than 1.5 at the relative humidity 90% (Tang, 1997).Therefore, sea salt-containing particles (including sea salt particles and mixture particles) and sulfate particles must have been enlarged by the uptake of water vapor.In comparison with those in D1-D5 samples, the distinctive characteristics of particles in W2 were the apparent coating layers on many particles, which should be the consequence of particle deliquescence under the humid conditions (Figs. 5(a) and 5(c)).
The distinctive difference of W1 from W2 and W3 was its bi-mode volume-size distribution.In addition to the common mode at 2-5 µm of W1, W2 and W3, the distribution of W1 had a remarkable mode at 0.5-1.0µm (Fig. 4(b)).The volume concentrations of particles in the size range of 0.3-5.0µm for W1 (approximately 5.5 × 10 4 µm 3 L -1 ) and W3 (approximately 8.6 × 10 4 µm 3 L -1 ) were not very different.Therefore, it was very likely that the high concentration of W1 particles in the size range of 0.5-1.0µm, which was the size range of the optical particle counters covering the wavelength of the ceilometer laser, caused the relative large backscattering in comparison to the case of W3.
There were few sea salt-containing particles and the major hydrophilic particles were sulfate-containing ones in W1 and W3 samples.The sulfate particles were expected to be mainly composed of ammonium sulfate because only sulfur was the frequent detected element in the particles.Such particles are usually the coagulation of smaller particles and appear in accumulation mode size.The individual particle analysis revealed the spherical shape of sulfate particles in W1 and W3 and being frequently smaller than 1 µm, were apparently smaller than sea salt and mixture particles (Figs.5(b) and 5(d)).
These results indicate that the BSCs were closely correspondent to the constitution of atmospheric particles under the humid conditions and also could be different according to the size distribution.Large and hydroscopic particles, such as those abundant in sea salt, had a large ability of backscattering and dominated the scattering in the atmosphere under humid conditions, which was relevant to the particles' growth (Hale and Querry, 1973;Li, 1976).On the other hand, the presence of high concentration of particles at the size of the light wavelength might also be the reason for enhanced backscattering.
The results were further compared with the reported correlations between hygroscopic growth factors and backscattering coefficients for marine-, pollution-, and dustinfluenced air masses at the wave length 700 nm (Carrico et al., 2003).The correlations (the dot curves in Fig. 3) were suggested after a statistical investigation of aerosol radiative properties as the function of humidity without the consideration of aerosol size distributions during ACE-Asia.Regarding the relative humidity at D1-D5 and W1-W3 collection and the constitutions of the particles, the correlations between the BSCs and aerosols in the cases of D1-D5, W2, and W3 are quite consistent with the statistical correlations, although the wavelengths are different.However, for W1 which was a pollution case (Fig. 4(a)), the correlation was not consistent with the statistical results of Carrico et al. (2003).The consistence further supports the determination of backscattering by the concentration of particles under dry condition and indicates the critical role that sea salt played in the enhancement of the backscattering under humid conditions.The discrepancy under humid conditions suggests that other factors should be considered, on which the volume-size distribution was likely an important candidate according to the results of the present study.

SUMMARY
The backscattering coefficients and number concentration of atmospheric particles were observed at a site on the southwestern Japanese coast in spring.To verify the constitution of the particles, aerosol samples collected in eight periods under different weather conditions were analyzed with a SEM.Results show that, under the relatively dry conditions of relative humidity smaller than 70%, the backscattering coefficients were proportional to the aerosol concentration in spite of the difference of the aerosol constitution (R 2 = 0.76).Under relatively humid conditions, the backscattering coefficients were dependent on the constitutions of the particles, which was attributed to the deliquescence of sea salt-containing particles, and also likely correlated with the size distributions, i.e., mode sizes and concentrations.These results indicate that (1) large and hygroscopic particles corresponded to strong backscattering in terms of unit aerosols and concentrated aerosols in the size range close to the wavelength of light might significantly enhance the backscattering under humid conditions, while (2) the backscattering coefficients were independent from particle composition and constitution under dry conditions.

Fig. 1 .
Fig. 1.Location of observation site, Amakusa Environmental Research Unit (AERU), and the air-mass backward trajectories starting at 1000 m over the site (calculated with the NOAA HYSPLIT on-line model) when the particle samples of cases D1-D5 and W1-W3 were collected.

Fig. 2 .
Fig. 2. Scatter plots of BSCs vs. IPVs under relatively dry conditions [(a), (b), (c)] and humid conditions [(d), (e), (f)] with respect to the extreme (EX: left), intermedium (IN: middle), and low (LO: right) aerosol episodes.N is the number of data in each panel.Dash lines mark the linear regressions.

Fig. 3 .
Fig. 3. Scatter plot of normalized ratios of BSCs to IPVs vs. relative humidity.Diamonds mark the data when particle samples were collected.Dot curves show the fitting of the statistical results of Carrico et al. (2003) for "Marine", "Dust" and "Polluted" particles.

Fig. 4 .
Fig. 4. (a) Number fractions of different type particles in the eight episodes.Total numbers of particles analyzed for each episode are shown in the parentheses.(b) Volume size distributions at the time when W1, W2, and W3 were collected (one-hour average).

Table 1 .
Summary of sample collection time, relative humidity, temperature, number concentrations of particles of diameter > 5 µm (N c ) and of diameter 0.5-2.0µm (N a ), integrated backscattering coefficient (BSC).