Measurement report: Spectral and statistical analysis of aerosol hygroscopic growth from multi-wavelength lidar measurements in Barcelona, Spain

Abstract. This paper presents the estimation of the hygroscopic
growth parameter of atmospheric aerosols retrieved with a multi-wavelength
lidar, a micro-pulse lidar (MPL) and daily radiosoundings in the coastal region of
Barcelona, Spain. The hygroscopic growth parameter, γ, parameterizes
the magnitude of the scattering enhancement in terms of the backscatter
coefficient following Hänel parameterization. After searching for time-colocated lidar and radiosounding measurements (performed twice a day, all year round at
00:00 and 12:00 UTC), a strict criterion-based procedure
(limiting the variations of magnitudes such as water vapor mixing ratio (WMVR),
potential temperature, wind speed and direction) is applied to select only
cases of aerosol hygroscopic growth. A spectral analysis (at the wavelengths
of 355, 532 and 1064 nm) is performed with the multi-wavelength lidar, and a
climatological one, at the wavelength of 532 nm, with the database of both
lidars. The spectral analysis shows that below 2 km the regime of local
pollution and sea salt γ decreases with increasing wavelengths.
Since the 355 nm wavelength is sensitive to smaller aerosols, this behavior
could indicate slightly more hygroscopic aerosols present at smaller size
ranges. Above 2 km (the regime of regional pollution and residual sea salt) the
values of γ at 532 nm are nearly the same as those below 2 km, and its
spectral behavior is flat. This analysis and others from the literature are
put together in a table presenting, for the first time, a spectral analysis
of the hygroscopic growth parameter of a large variety of atmospheric
aerosol hygroscopicities ranging from low (pure mineral dust, γ
<0.2) to high (pure sea salt, γ > 1.0)
hygroscopicity. The climatological analysis shows that, at 532 nm, γ
is rather constant all year round and has a large monthly standard deviation,
suggesting the presence of aerosols with different hygroscopic properties
all year round. The annual γ is 0.55 ± 0.23. The height of the
layer where hygroscopic growth was calculated shows an annual cycle with a
maximum in summer and a minimum in winter. Former works describing the
presence of recirculation layers of pollutants injected at various heights
above the planetary boundary layer (PBL) may explain why γ, unlike the height of the layer
where hygroscopic growth was calculated, is not season-dependent. The
subcategorization of the whole database into No cloud and Below-cloud cases reveals a large
difference of γ in autumn between both categories (0.71 and 0.33,
respectively), possibly attributed to a depletion of inorganics at the point
of activation into cloud condensation nuclei (CCN) in the Below-cloud cases. Our work calls
for more in situ measurements to synergetically complete such studies based
on remote sensing.


close to saturation. In the literature the aerosol hygroscopic enhancement has been measured more often on the backscatter coefficient derived from lidar (Feingold and Morley, 2003;Fernández et al., 2015;Granados-Muñoz et al., 2015;Haarig et al., 2017;Lv et al., 2017;Navas-Guzmán et al., 2019;Chen et al., 2019;Pérez-Ramírez et al., 2021) and ceilometer (Bedoya-65 Velásquez et al., 2019) measurements than on the extinction coefficient derived from lidar measurements (Veselovskii et al., 2009;Dawson et al., 2020). Some intents to work on the attenuated backscatter coefficient derived from ceilometers were performed by Haeffelin et al. (2016) to help tracking the activation of aerosols into fog or low-cloud droplets. Others investigated the lidar ratio changes due to relative humidity and their effect on the classical elastic-backscatter lidar inversion technique (Zhao et al., 2017). 70 The present work takes advantage of observational capabilities quite unique at the site of Barcelona, NE Spain: a multiwavelength lidar system measuring at three elastic wavelengths since 2011 and a single wavelength micro pulse lidar working continuously 24/7 since 2015, as well as two radiosoundings launched everyday almost collocated to the lidars (the database starts in 2009). The paper deals with 1) the spectral analysis of the hygroscopic growth factor measured at three wavelengths, and 2) the climatological analysis of the hygroscopic growth measured at 532 nm in Barcelona. The spectral analysis is 75 motivated by conclusions from Dawson et al. (2020) who say that multispectral lidars are fundamental so as to "provide additional insight into the [hygroscopic enhancement factor] retrievals since the 355-nm wavelength is sensitive to smaller aerosols than the 532-nm wavelength". The climatological analysis is a partial answer to the call of several authors, e.g. like Bedoya-Velásquez et al. (2018), for further investigation extending the study periods to obtain results statistically more robust.
The structure of the paper is as follows: Section 2 describes the instrumentations and the methodology and Section 3 presents 80 the results of the spectral and climatological analysis. Conclusions are given in Section 4.

Lidars and radiosoundings in Barcelona
All measurements presented in this paper were performed at or close to the Barcelona lidar site at the Remote Sensing Laboratory of the Department of Signal Theory and Communications at the Universitat Politècnica de Catalunya (41.393ºN,85 2.120ºE, 115 m asl). Two lidar systems were used: the multi-wavelength (3β+2α+2δ+WV) ACTRIS/EARLINET lidar and the micro pulse lidar (MPL, 1β+1δ). The first system is run according to a regular weekly schedule and to monitor special aerosol events of interest. Aerosol optical properties from this system can be found in the ACTRIS database at https://actris.nilu.no/.
The system employs a Nd:YAG laser emitting pulses at 355, 532 and 1064 nm at a repetition frequency of 20 Hz. The measurements of the ACTRIS/EARLINET system are averaged over 30 or 60 minutes. The retrieved backscatter coefficients 90 at the three emitted wavelengths for the period 2010-2018 are used in this work. General details about the system can be found in Kumar et al. (2011). The MPL system runs continuously 24/7. It is part of the Micro-Pulse Lidar Network (MPLNET, https://mplnet.gsfc.nasa.gov/data?v=V3; Welton et al., 2001) since 2016. The system uses a pulsed solid-state laser emitting low-energy pulses (~6 μJ) at a high pulse rate (2500 Hz). All MPL measurements are averaged over 60 minutes. The MPL https://doi.org/10.5194/acp-2021-990 Preprint. Discussion started: 9 February 2022 c Author(s) 2022. CC BY 4.0 License. data used in this work are the backscatter coefficient profiles at 532 nm retrieved during the period 2015-2018. More technical 95 information about the system can be found in Campbell et al. (2002), Flynn et al. (2007) and Welton et al. (2018). All the MPL retrievals presented in this work were performed with in-house algorithms, and not with the MPLNET processing.
Radiosoundings measurements are launched twice a day (at 00:00 and 12:00 UTC) by the the Meteorological Service of Catalonia, Meteocat, at a distance of less than 1 km from the lidar site. The radiosoundings provide measurements of pressure, temperature, relative humidity and wind speed and direction. Data of the period 2010-2018 are used in the present work. At 100 this point, it is important to note the inherent spatial drift of radiosoundings and the long integration time of the lidar data (as long as 60 minutes) which may cause a loss of temporal and spatial coincidence between both retrievals. This effect can be enhanced during daytime when the atmosphere may change quickly. This has been demonstrated by a recent paper from Muñoz-Porcar et al. (2021) in which profiles of water vapor mixing ratio retrieved with lidar and radiosoundings were compared. The authors also highlighted the high variability of the profile of relative humidity in Barcelona due to the presence 105 of the sea coast, the mild temperatures of the Mediterranean climate inducing regularly land-to-sea and sea-to-land breeze regimes and the local orography.

Methodology
This paper deals with the enhancement factor of the particle backscatter coefficient, , as a function of relative humidity, , commonly noted in the literature. Since no other optical/microphysical property is considered here, the suffix is 110 omitted in the rest of the paper in order to alleviate the formulae. The wavelength dependency is indicated with a superscript . Finally, the backscatter coefficient at wavelength writes and the corresponding enhancement factor writes .
The methodology starts with the search of time co-located lidar and radiosoundings measurements within a difference smaller than ±120 minutes. The time co-located measurements are then analysed to look for vertical intervals ( , ) in which a monotonic increase of the particle backscatter coefficient and of the relative humidity simultaneously occurs. After fulfilling 115 the initial conditions, these vertical intervals are classified as hygroscopic growth cases following a strict criterion-based procedure including:  water vapor mixing ratio noted WVMR (maximum variation of 2 g kg -1 ),  potential temperature noted (maximum variation of 2 K),  wind speed (maximum variation of 2 m s -1 , 120  wind direction (maximum variation of 15 deg. . For all cases back trajectories were also calculated with HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Stein et al., 2015) in order to verify the aerosol origin inside the layers of the selected cases obtained by applying the criterion-based procedure. These criteria guarantee that the increase of is very likely due to an increase of the water vapor and not to a combined effect of the thermodynamics variables, and that the variations of are caused by an increase of the 125 https://doi.org/10.5194/acp-2021-990 Preprint. Discussion started: 9 February 2022 c Author(s) 2022. CC BY 4.0 License. aerosol size due to water uptake and not to changes in the aerosol composition or concentration in the analysed layer. Such criteria have been applied by other authors (Granados-Muñoz et al., 2015;Navas-Guzmán et al., 2019;among others).
In order to discard the cases including mineral dust which is known to be poorly hydrophilic, in case of doubts, the back trajectory analysis was completed with mineral dust forecast from the NMMB/BSC-Dust model (https://ess.bsc.es/bsc-dustdaily-forecast) and AERONET retrievals. 130 Figure 1 Vertical profiles of (a) at 3 wavelengths and ; (b) WVMR and . The horizontal dash lines indicate and obtained by applying the criterion-based procedure. The example is from 22 July, 2013, at 13:02 UTC.  For each case, each value of in the range [ , ] has a corresponding value of varying in the range [ , ]. For each case, we define the particle backscatter coefficient enhancement factor, , defined starting from 145 as: (1) quantifies the increase of when the relative humidity increases from to . A fitting of is performed with the so-called Hänel parametrization (Kasten, 1969;Sheridan et al., 2002) using the points available between and : 150 where is the hygroscopic growth parameter and parameterizes the magnitude of the scattering enhancement. This definition of the enhancement factor limits the range of relative humidity values for which it can be calculated, i.e. only for , which prevents direct comparisons when is not the same (Veselovskii et al., 2009 (Skupin et al., 2016;Titos et al., 2016;Haarig et al., 2017;Bedoya-Velásquez et al., 2018;Dawson et al., 2020). =40 % is a recommendation of World Meteorological Organization (2016) who demonstrates with in-situ measurements that the hygroscopicity growth effect on aerosols is minimized for values of the relative humidity below 40 %. The new scaled enhancement factor starting at =40 % is noted and expresses the increase of when 160 the relative humidity increases from to : Note that the term on the right-hand side multiplying is nothing else than the quotient of the intercepts of both enhancement factors ( and ) at =0 %. It is the first time a conversion of into is proposed. Figure 3 is an example showing the two retrieved enhancement factors and and their Hänel fit. One sees clearly the difference between restricted to [ , ] (spanning 21 % in the example presented) and spanning 50 % from 40 to 90%. In addition, to allow direct comparison of enhancement factors retrieved from different cases for different ranges, 170 this method has also the advantage of defining a common way for the calculation of the -value. The -value, also called the value (Titos et al., 2016), is defined as . It depends on both and . There is no consensus in the atmospheric community for the definition of the range of values which has a strong variability among studies as underlined by Titos et al. (2016). In this study, the -value is 85% with =40 %, and it applies for all cases.
It expresses the increase factor of the backscatter coefficient when the relative humidity increases from =40 % to 85 %. 175 Finally, to avoid outliers the cases with 85% greater than 10 with =40 % (which corresponds to 1.66) were not taken into account in the statistics. 180

Spectral analysis
In this section, only the ACTRIS/EARLINET lidar system, which has three elastic wavelengths, is considered. Among the backscatter profiles at the three wavelengths of 355, 532 and 1064 nm and the relative humidity profiles from radiosoundings available between 2010 and 2018, 32 potential cases of hygroscopic growth which fulfilled the selection criteria mentioned in 185 Section 2.2 were identified. From these 32 hygroscopic growth cases, the centre of the layer considered was below 2 km agl in 8 cases (25 %) and above 2 km agl in 24 cases (75 %). This reference height of 2 km was chosen based on Sicard et al. (2006Sicard et al. ( , 2011 and (Pandolfi et al., 2013) who showed that the separation between the surface mixed layers and possible decoupled residual/aloft layers in the Barcelona coastal area are more likely to occur around 2 km high. Interestingly Pérez-Ramírez et al. (2021) also considered altitude heights below 2 km (near surface and PBL) and above 2 km to show the temporal 190 evolution of and . Generally speaking, in the region of Barcelona, the aerosols present below 2 km, i.e. in the PBL, are representative of a coastal, urban background site and their chemical composition is dominated by anthropogenic, crustal and marine aerosols (Querol et al., 2001). Below 2 km, the aerosol type in Barcelona is defined as local pollution and marine aerosols. According to Pey et al. (2010), the mean annual urban contribution in Barcelona downtown of hydrophilic species such as SO , NO or NH is at least 53, 65 and 45 %, respectively, with respect to the regional background. For aerosol layers 195 above 2 km, Sicard et al. (2011) refer to "recirculation polluted air-masses" and Pandolfi et al. (2013) to either regional or Atlantic air-masses (African air-masses are discarded since the cases with the presence of mineral dust are not included in this study). Above 2 km, the aerosol type in Barcelona is defined as regional pollution and marine aerosols. It is important to mention that above 2 km, the height range reaches 5 km, the approximate height where the highest top of the layer is located.
To check the provenance of the aerosol layers below and above 2 km, we plot in Figure

Figure 4
Wind rose (a) below 2 km (8 cases) and (b) above 2 km (24 cases) from radiosoundings measurements in Barcelona. 205 The colorbar on the right applies to both plots.
For both wind roses, the most frequent wind speeds reach around 6 m s -1 designated as moderate breeze (World Meteorological Organization, 1992). This low wind speed maintains a good homogeneity of the atmospheric aerosols in the selected layers, i.e., a constant mixture of aerosols over time. Figure 5a and 5e show the resulting spectral Hänel fits at both height levels. Figure 5b-d and 5f-h show all enhancement factors 210 per wavelength below and above 2 km agl, respectively. In all those plots, the Hänel fits (solid lines) are calculated with the mean hygroscopic growth parameter ̅ of all individuals , and the variability associated to them (coloured shaded area) is calculated taking into account the standard deviation of all individual . The enhancement factors and Hänel fits in Figure   5 are those scaled in order to start at =40 %. The spectral values of 85% and ̅ at both height levels are reported in Table 1, which also includes the mean of the correlation coefficients, , of the individual pairs of ( , ), as 215 well as the average of the layer-mean Ångström exponents, , , between the pairs of wavelengths (355, 532 nm) and (532, 1064 nm). The first result to comment is that, independently of the height level, the correlation coefficients of the individual fits are high (>0.91) and present small fluctuation ( <0.06, except for =355 nm and above 2 km where =0.14). These high values of indicate the good correlation that exists between the profiles of the backscatter coefficients and the profiles of the relative humidity in the layer selected. 220 Below 2 km, the particle backscatter coefficient enhancement factor seems to have a clear spectral behaviour: 85% ( ) is 3.60 ± 2.47 (0.81 ± 0.41), 3.18 ± 2.07 (0.73 ± 0.40) and 2.68 ± 1.20 (0.65 ± 0.31) at 355, 532 and 1064 nm, respectively. This behaviour (decrease of 85% or with increasing wavelength) implies that the water uptake by the particles modifies the particle backscatter coefficient more strongly at shorter wavelengths than at larger wavelengths. Since the 355-nm wavelength is sensitive to smaller aerosols compared to larger wavelengths, larger fit 225 coefficients at 355 nm indicate slightly more hygroscopic aerosols present at smaller size ranges (Dawson et al., 2020).
, and , are 1.07 and 0.93, respectively. Although quite similar, the difference between both , also points out to a spectral sensitivity of the backscatter coefficient slightly larger at shorter wavelengths than at larger ones. Note en passant that our Ångström exponent values are in the range of column-averaged monthly values found by Sicard et al. (2011) and estimated from a long-term lidar database in Barcelona. Next, we aim at comparing our results with the literature. 230 The hygroscopic growth parameter depends on neither , nor , so the values of from the literature can be directly compared to ours. Contrarily, the -values depend strongly on and , so for the literature to be comparable with our values, the hygroscopic growth parameter from the literature is used to calculate 85% with =40 %. Also, we only considered works in which the enhancement factor was calculated for the backscatter coefficient measured with a lidar. Works in which the enhancement factor was calculated for the extinction coefficient 235 (Veselovskii et al., 2009;Dawson et al., 2020) or from in-situ data (Carrico et al., 2003;Titos et al., 2016;Skupin et al., 2016; among others) are not considered for comparison with our study. The literature results are summarized in Table 2 and represented in Figure 6, in terms of 85% and . The values in Table 2    Washington (Pérez-Ramírez et al., 2021). The presence of marine aerosols in Barcelona also explains why higher 275 85% are found compared to the rest of studies also dominated by pollution Granados-Muñoz et al., 2015;Chen et al., 2019) in which the presence of marine aerosols is not mentioned.
Above 2 km, the particle backscatter coefficient enhancement factor seems to have no spectral dependency: 85% ( ) is 2.96 ± 1.38 (0.71 ± 0.32), 2.88 ± 1.27 (0.70 ± 0.30) and 2.99 ± 1.38 (0.73 ± 0.30) at 355, 532 and 1064 nm, respectively. Such a flat spectral dependency has been observed only by (Navas-Guzmán et al., 2019) between 355 and 1064 280 nm for mineral dust. We believe that the spectral dependency of the total aerosol content is function of the chemical composition and of the concentration of each one of the hygroscopic components, and for this reason it is highly variable.
Although the hygroscopic parameter of relevant particle components has been reviewed by Liu et al. (2014), our conclusion calls for further chemical analysis and laboratory studies to determine the spectral behaviour of these relevant particles.

Climatological analysis at 532 nm
In this section we present for the first time a climatological analysis of the aerosol hygroscopic growth observed along a vertical range in ambient conditions by means of the hygroscopic growth parameter, , and the particle backscatter coefficient 295 enhancement factor at = 85 %, 85% . Data from both the multi-wavelength ACTRIS/EARLINET lidar (period 2010-2018) and the MPL (period 2015-2018) are considered. The common wavelength is = 532 nm. All data were screened according to the selection criteria mentioned in Section 2.2. Like in Section 3.1 all cases associated with a mineral dust intrusion are not included in this study. We find a total of 76 cases distributed along all months of the year. The monthly variation of 532 is represented in Figure 7. The mean-layer height ( ) and mean-layer relative humidity ( ) calculated as the 300 mean value in the hygroscopic layer of the height and relative humidity, respectively, is also plotted. The seasonal means of 532 , and are reported in Table 3. Winter includes the months of December, January and February; spring: March, April and May; summer: June, July and August; and autumn: September, October and November.
From the top plot of Figure 7, one sees that the aerosol hygroscopic growth parameter is in average rather constant all year round. However, for each single month, large standard deviations of 532 are observed (red shaded area in Figure 7), which 305 indicates the presence of aerosols with different hygroscopic properties all year round. The annual mean of 532 is 0.55.
While the seasonal deviations from that mean are small (≤ ± 0.03,  1.19 km). In regard of former works of (Sicard et al., 2006) establishing that the planetary boundary layer (PBL) in Barcelona is not significantly different between winter and summer seasons and that it is usually lower than 1.0 km, our findings suggest that hygroscopic layers are detected near the top or slightly above the PBL in autumn and winter, and clearly above the PBL in spring and summer. Although the hygroscopic aerosols are detected above the PBL in spring and summer, they might not be that different from the aerosols in the PBL. Indeed Pérez et al. (2004) showed that in Barcelona the 315 combined effects of strong insolation, weak synoptic forcing, sea breezes and mountain-induced winds create re-circulations of pollutants injected at various heights above the PBL and up to 4.0 km. Like 532 , the is also rather constant all year round which confirms that aerosol hygroscopic properties are not related to the level of humidity in the atmosphere. The growth parameters, some differences and similarities with our study are worth mentioning: 330  the annual mean of is 0.40, much lower than in our study (0.55) where the effect of sea salt is noticeable.
 the annual standard deviation of is 0.15, proportionally similar to = 0.23 found in our study.


is higher in winter (higher nitrate mass fraction) and lower in summer (higher organic mass fraction), whereas it is not season-dependent in our study.
An interesting result from Jefferson et al. (2017), complimentary to our analysis, is the retrieval of for sub-1 and sub-10 μm 335 particles which annual mean is 0.44 and 0.40, respectively. The lower sub-10 µm values of is attributed to the influence from soil dust. Our search for a special dependency of the hygroscopic growth with other factors (layer height, day/night, level of humidity, etc.) was rather unfruitful. However, by looking individually to all 76 cases, and in particular to the vertical profiles of the backscatter coefficient and relative humidity, we found a possible sub-categorization into 2 classes: cases with no cloud in the 345 vertical range examined (referred hereafter as No cloud) and cases where the hygroscopic behaviour was detected just below a cloud (referred hereafter as Below-cloud). From the whole dataset 55 cases were classified as No cloud and 21 as Belowcloud. We have checked that all cases were in sub-saturation humidity conditions ( < 100 %, the water is in vapour form and the aerosol cannot activate (yet) into a droplet). Figure 8 shows an example of both cases. The annual means of 532 , 85% , and for both cases are also reported in the bottom part of Table 3. On a yearly basis the Below- PBL where the formation of convective non-precipitating PBL clouds is frequent in coastal sites (Papayannis et al., 2017). In such cases, the relative humidity is higher than in the No cloud cases. The aerosol below the cloud starts to activate as cloud condensation nuclei, its size grows through adsorption of water vapour, its scattering properties increase, and as the aerosol 355 size grows, its potential to keep growing are reduced compared to a drier aerosol. All this is well illustrated in the two cases shown in Figure 8: while both the backscatter coefficient and the relative humidity are low ( < 55 %, < 1 Mm -1 sr -1 ) in the absence of clouds (Figure 8a), they are high ( > 80 %, > 2 Mm -1 sr -1 ) and increase strongly with height in the Below-cloud case (Figure 8b). So, although one would be tempted to visually attribute the strongest growth to the Below-cloud case, in practice the contrary occurs: 532 is higher for No cloud (0.87) than for Below-cloud (0.53). The same result is 360 reflected in the climatological data (Table 3): 532 = 0.58 for No cloud and 0.48 for Below-cloud. There are at least three reasons for that:  The first reason has been given above: the aerosol activation as cloud condensation nuclei in high humidity conditions reduces its potential to keep growing compared to the aerosol in drier conditions.  The second one is mathematical, definition-dependent and inherent to the aerosol composition. According to the 365 definition of the enhancement factor (Eq. 3), defined as a power law function normalized to a of 40 %, one sees that (bottom plots of Figure 8): 1) in cases with low , varies very little and a relatively high 532 is required to provoke departure of from unity; and 2) for high cases, varies steeply and a relatively low 532 is enough to provoke strong variations of .
 The third explanation is linked to the specific aerosol composition at the site. In Figure 9 we show box and whisker 370 plots of the percentiles of the seasonal values of 532 for all cases, and for the No cloud and Below-cloud cases.
When considering all cases, the results for the monthly statistics is reproduced for the seasonal one: 532 is not season-dependent. In spring and summer, 532 for No cloud is not significantly different from for all cases; for Below-cloud 532 is slightly larger but well within the seasonal standard deviation (see Table 3). The most important difference is in autumn when the mean 532 is 0.71 for No cloud and 0.33 for Below-cloud, i.e. 375 respectively well above and below the autumn mean of all cases (0.53).
This last paragraph presents a discussion on the possible explanations of the difference observed in autumn in Figure 9 which may rely on the concomitance of several factors. As far as the No cloud cases are concerned, in Barcelona hydrophilic inorganics such as nitrates (the most abundant) and ammonium are maximum in autumn and winter, while sulfates are minimum (Querol et al., 2001). The persistence of anticyclonic stagnating conditions during these seasons favours the 380 accumulation of the pollutants in the PBL (Querol et al., 2001;Pey et al., 2010). In such conditions more pollutants are prone to convective vertical motion and activation as cloud condensation nuclei if high humidity conditions are also present. Gunthe et al. (2011) showed that aged pollution particles in stagnant air (soluble inorganics dominate the mass fraction) are on average larger and more hygroscopic than fresh pollution particles (organics and elemental carbon dominate the mass fraction). Taken all together, these results support the higher values of 532 for the No cloud cases found in autumn (0.71) compared to the 385 https://doi.org/10.5194/acp-2021-990 Preprint. Discussion started: 9 February 2022 c Author(s) 2022. CC BY 4.0 License. rest of the year (0.52 -0.60). Two things in Figure 9 remain to be elucidated: why the hygroscopicity of the Below-cloud cases is lower in autumn than during the rest of the year, and why is it lower than that of the No cloud cases during the same season?
These questions are difficult to answer without complementary in-situ measurements, and at this point, only hypothesis can be formulated. The seasonal statistics available indicates that the mean-layer backscatter coefficient (not shown) in the hygroscopic layer of the Below-cloud cases are larger in autumn (1.71 Mm -1 sr -1 ) than in spring (0.73 Mm -1 sr -1 ) and summer 390 (1.35 Mm -1 sr -1 ); the same occurs for the but in a lesser extent (79.7 % in autumn vs. 74.3 and 76.5 % in spring and summer, respectively); and the opposite for the (0.60 km -i.e. within the PBL, see Sicard et al. (2006)-in autumn vs.
1.29 and 1.64 km -i.e. above the PBL-in spring and summer, respectively). Thus, in autumn the Below-cloud hygroscopic layers are within the PBL, hence the larger observed with respect to the spring and summer seasons. The lower autumn 532 could reflect a higher fraction of organics in the aerosol mixture in the PBL in autumn vs. a higher amount of inorganics 395 in the aerosol mixture above the PBL in spring and summer (possibly coming from the re-circulation of pollutants injected at various heights above the PBL, see Pérez et al. (2004). Note that this statement is a pure hypothesis. The literature emphasizes the complexity of the atmospheric aerosol hygroscopicity linked to their highly variable composition and chemical transformation. Cheung et al. (2020) suggest that the uptake of hydrophilic/hydrophobic species during particle growth and coagulation processes may influence the hygroscopicity of aerosols. Cruz and Pandis (2000) study the effect of organic mixing 400 and coating on the hygroscopic behaviour of inorganics and on NaCl particles in particular. Interestingly, they find that, depending on the organic mass fraction, the NaCl-organic mixtures could not only decrease (down to 40%), but also increase (up to 20%) the mixture hygroscopicity. More recently Ruehl and Wilson (2014) emphasize the new and complex relationship between the composition of an organic aerosol and its hygroscopicity and in the same field (Liu et al., 2018) study in the laboratory some of microphysical mechanisms involved in the hygroscopicity of secondary organic material. All studies call 405 for further laboratory and field research. The difference between the autumn Below-cloud (0.33) 532 and No cloud (0.71) is significant. The mean-layer backscatter coefficient in the hygroscopic layer of the Below-cloud (1.71 Mm -1 sr -1 ) is approximately twice larger than the No cloud cases (0.80 Mm -1 sr -1 ). The autumn / are higher / lower for the Below-cloud (79.7 % / 0.60 km) than for the No cloud cases (59.5 % / 1.93 km). It is possible here again to hypothesize that the lower 532 for Below-cloud could reflect a higher fraction of organics in the aerosol mixture in the PBL with higher 410 humidity conditions vs. a higher amount of inorganics in the aerosol mixture in the free troposphere in the No cloud cases.
Note that this statement is again a pure hypothesis. In summary, the observations in autumn show that the Below-cloud aerosols are detected in the PBL, at high relative humidities, and have large backscatter coefficients and low hygroscopic growth parameters. We close this section with a question: may the activation into CCN at the base of the cloud be affecting predominantly inorganics salts and thus generating a depletion of them and leave room to an organic-rich layer below the

Conclusions
A spectral and climatological analysis of the aerosol hygrowscopic growth parameter obevered in the atmospheric vertical column combining lidar and radiosoundings measurements is presented. The hygroscopic cases have been selected by filtering time coincident lidar and radiosoundings measurements, by detecting coincident backscatter coefficient and relative humidity increases with increasing height, and by limiting the variations of variables used as indicators of well mixed conditions such 430 as water vapor mixing ratio, potential temperature, and wind speed and direction. Results are presented in terms of the hygrowscopic growth parameter, , result of fitting the particle backscatter coefficient enhancement factor with the Hänel parametrization and the particle backscatter coefficient enhancement factor, 85% , at =85 % with =40 %. For our results to be comparable with the literature giving enhancement factors for a large variety of and , a very simple conversion for any values of / is proposed and applied to the literature values to get 85% 435 with =40 %.
The spectral analysis performed at the wavelengths of 355, 532 and 1064 nm distinguishes aerosols in layers below 2 km (regime of local pollution and sea salt) and above 2 km (regime of regional pollution and residual sea salt). Below 2 km, decreases with increasing wavelengths ( =0.81, 0.73 and 0.65; 85% =3.60, 3.18 and 2.68). This behaviour could be attributed to the aerosol size: the smaller the aerosol, the more hygroscopic. This hypothesis is supported by the Ångström 440 exponents which are higher for the pair (355, 532) than for (532, 1064), which points out to a spectral sensitivity of the backscatter coefficient slightly larger at shorter wavelengths than at larger wavelengths. Above 2 km the values of (0.71, 0.70 and 0.73; 85% =2.96, 2.88 and 2.99) are comparable to those below 2 km, and their spectral behaviour is flat. This analysis and others from the literature are put together in a table presenting for the first time spectrally the hygroscopic growth parameter and enhancement factors of a large variety of atmospheric aerosol hygroscopicities going from low (pure 445 mineral dust, < 0.2; 85% < 1.3) to high (pure sea salt, > 1.0; 85% > 4.0). In this table, the highest values of 85% (> 3) all represent situations with a notable fraction of sea salt; values of 85% between 2 and 3 are representative of polluted situations with different mixings; near the value of 2 we find biomass burning; between 1.5 and 2 rural background with automobile traffic; and values of 85% close to 1 correspond to clean and mineral dust cases. 450 The climatological analysis shows that at 532 nm is rather constant all year round and has a large monthly standard deviation suggesting the presence of aerosols with different hygroscopic properties all year round. The annual is 0.55±0.23 ( 85% =2.26±0.72). The height of the hygroscopic layers shows an annual cycle with a maximum clearly above the PBL in summer and a minimum near the top of the PBL in winter. Although the hygroscopic aerosols are detected above the PBL in spring and summer, they might not be that different from the aerosols in the PBL. Former works describing the 455 presence of re-circulation layers of pollutants injected at various heights above the PBL may explain why , unlike the height of the hygroscopic layers, is not season-dependent. The sub-categorization of the whole database into No cloud and Below-https://doi.org/10.5194/acp-2021-990 Preprint. Discussion started: 9 February 2022 c Author(s) 2022. CC BY 4.0 License. cloud cases reveals a large difference of in autumn between both categories (0.71 and 0.33, respectively), possibly attributed to a depletion of inorganics at the point of activation into cloud condensation nuclei in the Below-cloud cases. Our work calls for more in-situ measurements to synergetically complete studies, like this one, based mostly on remote sensing measurements.