Radiative effects of long-range-transported Saharan air layers as determined from airborne lidar measurements

. The radiative effect of long-range-transported Saharan air layers is investigated on the basis of simultaneous airborne high spectral resolution and differential absorption lidar measurements in the vicinity of Barbados. Within the observed Saharan air layers increased water vapor concentrations compared to the dry trade wind atmosphere are found. The measured proﬁles of aerosol optical properties and water vapor mixing ratios are used to characterize the atmospheric composition in radiative transfer calculations, to calculate radiative effects of moist Saharan air layers and to determine radiative heating rate proﬁles. An 5 analysis based on three case studies reveals that the observed enhanced amounts of water vapor within Saharan air layers have a much stronger impact on heating rate calculations than mineral dust aerosol. Maximum mineral dust short-wave heating and long-wave cooling rates are found in altitudes of highest dust concentration (short-wave: +0.5 Kd − 1 , long-wave: -0.2 Kd − 1 , net: +0.3 Kd − 1 ). However, when considering both aerosol concentrations and measured water vapor mixing ratios in radiative transfer calculations the maximum heating/cooling rates shift to the top of the dust layer (short-wave: +2.2 Kd − 1 , long-wave: 10 -6.0 to -7.0 Kd − 1 , net: -5.0 to -4.0 Kd − 1 ). Additionally, the net-heating rates decrease with height - indicating a destabilizing effect in the dust layers. Long-wave counter radiation of Saharan air layers is found to reduce cooling at the top of the subjacent marine boundary layers and might lead to less convective mixing in these layers. The overall short-wave radiative effect of mineral dust particles in Saharan air layers indicates a maximum magnitude of -40 Wm both characterized by a long-range-transported SAL. Measured vertical proﬁles of both R 532 and δ p(532) are used for the detection of the SAL-outlines. From enhanced values of R 532 and typical values of δ p(532) for mineral dust the vertical extent of SALs can be determined. All the three selected cross sections are of approximately 50 km length and described in the following: in these altitudes. Proﬁles of aerosol mass concentration highlight a pure dust regime ( c m(dust) ≈ 100 µ gm − 3 ; aerosol extinction coefﬁcients around 20 0.06 km − 1 ) at altitudes ranging from approximately 1.5-5.0 km altitude transitioning to a mixed marine and dust aerosol regime at lower atmospheric levels (0-1.5 km). The lidar measurements again do not indicate a pure marine aerosol regime at low altitudes. Both marine and settling mineral dust aerosol is found in the MBL (0-1.5 km). SALs NARVAL-II of water compared dry free trade-wind r m and R (532) in the lidar proﬁles (b) and (c) 25 also show a distinct correlation The SALs show almost uniformly increased water vapor mixing ratios ranging from 3-5 gkg − 1 compared to the surrounding free-troposphere. (case (b): 2.8-4.2 km; case (c): 2.5-4.8 km). Case (a) however, indicates that no distinct correlation of enhanced r m and R (532) could be observed in a SAL-free troposphere. The proﬁle shows a drop of r m to values smaller 1 gkg − 1 at altitudes greater 3 km, indicating the transition from the MBL to the dry free troposphere. Such a drop in humidity (which is coming along with a strong trade wind inversion caused by Headly cell 30 subsidence) was observed during most SAL-free periods in NARVAL-II, and is discussed by Gutleben et al. (2019b) in the framework of a detailed dropsonde analysis.


The radiative transfer model libRadtran
Calculations of both downward and upward irradiances as well as atmospheric heating rates are performed utilizing the radiative 5 transfer equation solver DISORT (Stamnes et al., 1988) with an improved intensity correction (Buras et al., 2011). The solver is embedded in the Library Radiative Transfer model (libRadtran; Mayer and Kylling, 2005;Emde et al., 2016) and is applied with 16 streams in the short-wave (0.12-4.0 µm) and long-wave (2.5-100.0 µm) spectra. At lower tropospheric levels (0-10 km) the model grid is set to vertical resolutions of 0.1 km. To save computational time the grid setting is changed to coarser resolutions at higher altitudes (see Table 1). 10 Time-expensive line-by-line calculations of spectral molecular absorption in the short-wave and long-wave spectral ranges are avoided by employing the sufficiently accurate correlated k-distribution absorption band parametrizations (Kato et al., 1999;Fu and Liou, 1992). Irradiances are then calculated by integrating over the respective parametrized absorption bands and height resolved diurnally averaged heating rates in the short-wave and long-wave spectra are derived from the difference in calculated radiation flux at the particular height intervals solving, at any vertical level z. Here, c p is the specific heat capacity of air at constant pressure, ρ(z) is the altitude dependent air density and F net (z)/δz represents the vertical change in net radiative flux at altitude z. The model temperature is parametrized using colocated dropsonde measurements. Reference profiles described by Anderson et al. (1986) are used to parametrize the trace gas concentrations from 0-120 km altitude. However, water vapor profiles and any information on the atmospheric aerosol 20 composition underneath the aircraft are taken from WALES lidar measurements which are interpolated accordingly to fit the model grid.
To minimize uncertainties in surface albedo (Claquin et al., 1998;Liao and Seinfeld, 1998) the bidirectional reflectance distribution function (BRDF; Cox and Munk, 1954a, b;Bellouin et al., 2004) is used. The BRDF derives sea surface albedo iv https://doi.org/10.5194/acp-2020-420 Preprint. Discussion started: 8 June 2020 c Author(s) 2020. CC BY 4.0 License. from 10 m-wind speeds measured by dropsondes and sea swell. Based on measurements by the MODIS-Aqua/Terra satellite during the field campaign, sea surface temperature is set to a fixed value of 302 K.

Aerosol optical properties from lidar measurements
In this study, the characterization of aerosol and water vapor profiles in libRadtran is performed using WALES DIAL and depolarization lidar measurements. Therefore, a method to identify profile-regions of different aerosol species and aerosol 5 concentrations using lidar measurements of α p(532) , R 532 , δ p(532) and r m was developed and is discussed in the following.

Aerosol classification
WALES lidar profiles of partilce linear depolarization ratio δ p(532) can be used to detect and identify Saharan dust marine aerosols in vertical atmospheric columns (Burton et al., 2012;Groß et al., 2013). δ p(532) for Saharan dust near source regions fluctuates around 0.3 (Freudenthaler et al., 2009;Tesche et al., 2009b;Groß et al., 2011b) and recent studies showed that this 10 ratio remains unchanged after long-range transport across the subtropical North Atlantic Ocean (Burton et al., 2012;Groß et al., 2015). Marine aerosol is composed of sea salt and water-soluble parts and is weakly depolarizing in a moist environment. Dry and stronger depolarizing marine aerosol (δ p(532) > 0.04; Murayama et al., 1999;Sakai et al., 2010) is therefore not expected since relative humidity inside the MBL was found to be always greater than 80 %. As a result, δ p(532) is a good proxy for the differentiation of mineral dust and less depolarizing marine aerosol (Sakai et al., 2010;Burton et al., 2012;Groß et al., 2013) 15 in NARVAL-II WALES lidar profiles. In this way three aerosol regimes can be determined in the dataset: I. pure mineral dust regime: δ p(532) ≥ 0.26, II. pure marine aerosol regime: δ p(532) ≤ 0.04, III. mixed regime -marine aerosol mixed with mineral dust: 0.04 < δ p(532) < 0.26.
Clear and aerosol-free regions are detected via filtering for no evident particle-backscattering (R 532 < 1.2). Cross sections of 20 an aerosol mask along the HALO flight tracks for libRadtran aerosol input are generated using these criteria.
In addition to the application of the detection scheme, the Saharan origin of detected mineral dust layers is verified utilizing the HYbrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT; Stein et al., 2015) with NCEP GDAS (National Centers for Environmental Prediction Global Data Assimilation System) input data (shown in Gutleben et al., 2019b). Starting locations and times for the backward trajectory calculations were chosen to match the center of detected mineral dust layers in 25 lidar data.

Conversion of aerosol extinction coefficients to aerosol mass concentrations
To run UVSPEC with aerosol input, particle mass concentrations in the classified aerosol regimes have to be determined.
These values are adopted for this study. Mineral dust and marine aerosol mass concentrations (c m,dust , c m,marine ) are then calculated by multiplying the derived aerosol volume concentrations with typical particle densities of ρ dust = 2.5 g cm −3 for 5 long-range-transported mineral dust and ρ marine = 2.2 g cm −3 for marine aerosol. Those particle densities are based on a study by Kaaden et al. (2009) who showed that SALs consist of a mixture of mineral dust particles together with sulfate particles.
The above equations allow the characterization of aerosol mass concentrations in the pure mineral dust regime (I) and marine aerosol regime (II). However, in mixed regimes (III) which appear frequently over the North Atlantic Ocean when SAL-mineral dust is settling to lower atmospheric levels, the mineral dust contribution to α p(532) of the total aerosol mixture 10 has to be determined before the application of the conversion coefficients. The aerosol extinction coefficient at 532 nm of a marine-mineral dust aerosol mixture α p(532),mix can be written as, with α p(532),marine and α p(532),dust being the marine aerosol and mineral dust particle extinction coefficient at 532 nm and x = α p(532),dust /α p(532),mix .

15
Using the known lidar ratios of marine and mineral dust aerosol at 532 nm (S p(532),marine 18 and S p(532),dust 47; Burton et al., 2012;Groß et al., 2013) and following the methods described in Tesche et al. (2009a) and Groß et al. (2011a) one can calculate the fraction x of dust contributing to the total particle extinction coefficient of the mixture, with the coefficients D dust and D marine : D marine := δ p(532),mix − δ p(532),marine S p(532),marine (1 + δ p(532),marine ) Finally, Eq. (3), ν dust(532) and ν marine(532) as well as ρ dust and ρ marine are used to calculate mineral dust and marine aerosol particle mass concentrations in mixed aerosol regimes (III).   Yi et al. (2011) showed that different representations of particle shapes result in a change of P(Θ) and can cause up to 30 % difference in the dust radiative forcing at top of the atmosphere (TOA). To minimize errors resulting from wavelength-interpolations Hess et al. (1998) established the readily available spectrally resolved OPAC database (Optical properties of Aerosols and Clouds) which includes modeled information on the above mentioned aerosol 5 optical properties for 61 wavelengths in the spectral range from 0.25-40 µm for various aerosol species. OPAC is a widely used data base in aerosol models and retrievals (e.g. Kim et al., 2004;Liu et al., 2004;Patadia et al., 2009) as well as general circulation and climate models for calculations in both the short-wave and the long-wave spectra. Thus, it is an appropriate tool to link lidar derived aerosol mass concentrations to aerosol optical properties in the classified aerosol regimes.
OPAC sea salt and water-soluble particle microphysical properties are modeled under the assumption of spherical particles 10 using Mie-Theory (Mie, 1908). The assumption of spherical particles is legitimate for radiative transfer simulations in the period of NARVAL-II since no dry and aspherical marine aerosol particles are expected at observed relative humidities of greater 80 % inside the derived marine aerosol regimes (Murayama et al., 1999;Sakai et al., 2010). Thus, a humidity dependent marine aerosol composition which refers to WALES measurements of water vapor mixing ratios together with dropsondederived temperature profiles is used in this study (see Table 2). 15 Mineral dust particles however, are characterized by highly irregular shapes (Falkovich et al., 2001;Kandler et al., 2011).
Hence, an assumption of spherical mineral dust particles in radiative transfer models using Mie-Theory is inappropriate and may lead to wrong results. Especially if particles are significantly larger than the transmitted wavelength (which is the case for most backscatter lidar systems) phase functions of aspherical particles are characterized by an increased amount of sideward but a reduced amount of backward scattering compared to spherical particles (Koepke and Hess, 1988;Gobbi et al., 2002;20 Nousiainen, 2009;Wiegner et al., 2009;Gasteiger et al., 2011;Sakai et al., 2014). For this reason mineral dust particle microphysical properties were updated in the latest OPAC version (v4.0; Gasteiger et al., 2011;Koepke et al., 2015) and are now calculated by means of the T-matrix method (Waterman, 1971) under assumptions of an aspect ratio distribution for prolate spheroids observed during the Saharan mineral dust experimentents SAMUM-I and SAMUM-II .
vii https://doi.org/10.5194/acp-2020-420 Preprint. Discussion started: 8 June 2020 c Author(s) 2020. CC BY 4.0 License. Several studies have shown that T-matrix theory substantially improves the agreement between measured and modeled aerosol optical properties of aspherical mineral dust particles (Mishchenko et al., 1997;Kahnert et al., 2005;Gasteiger et al., 2011) and are thus motivating its use in this study.
Results obtained from measurements during SALTRACE (Weinzierl et al., 2017) showed that the size distribution of longrange-transported mineral dust does not change significantly compared to the distributions measured at source regions. Grav-5 itational settling processes of large sub-micron particles in the course of the SAL-transatlantic transport are of a smaller magnitude than expected from Stokes gravitational settling calculations. Moreover, Denjean et al. (2015) found that the chemical composition and hygroscopy of mineral dust remains unchanged after long-range transport. Thus, a mixture proposed by Hess et al. (1998) which consists of four OPAC v4.0-components for desert aerosol optical properties is assumed in this paper (see Table 2). The second case represents a scenario with a detected elevated and long-range-transported SAL extending from 3-4 km altitude. The SAL shows increased backscatter ratios around 3.5 and high particle linear depolarization ratios >0.26.
δ p(532) and R (532) profiles feature sharp gradients to the above free-troposphere and to lower atmospheric levels. The

Saharan Air Layer Heating rates
Profiles of calculated short-wave, long-wave and net heating rates (24 h-averaged) for the three selected case studies are shown 5 in Figure 4. Since WALES is able to measure both water vapor mixing ratios and aerosol optical properties, total heating rate profiles and contributions of mineral dust to total heating rate profiles can be derived. The dust-contribution to the total heating rate is derived as the difference between heating rates that consider dust in the model and heating rates with no dust in the model atmosphere.
Observed SALs in case (b) and (c) are well mixed (constant potential temperature Θ around 315 K) and show enhanced 10 water vapor mass mixing ratios in the range from 2-5 g kg −1 compared to the surrounding dry free atmosphere. Both profiles have strong gradients of r m and Θ at the upper edge of the SAL (at the boundary to the above dry and aerosol-free trade wind atmosphere) indicating the two well-known SAL-related bounding inversions (Lilly, 1968). The MBL in all three cases is characterized by high relative humidities (r m : 10-16 g kg −1 ) and is capped by a temperature inversion (trade wind inversion) and a pronounced hydrolapse (r m drops from >15 to approximately 5 g kg −1 ). Due to water vapor absorption and emission the total heating and cooling rate profiles have a completely different shape.
Largest water vapor absorption of solar radiation takes place at the uppermost levels of the SAL leading to strong heating at these levels. Long-wave cooling due to emission of radiation towards space is also strongest at the top edge of the SAL since there is no heating from atmospheric counter radiation from higher atmospheric levels. This is why greatest total heating and 25 cooling rates are found at the upper edge of both observed SALs (short-wave: ∼2.2 K d −1 (both cases); long-wave: -6 K d −1 (case (b)) and -7 K d −1 (case (c)).
Adding short-wave and long-wave heating rates results in SAL-net heating rates that are entirely negative for both cases, taking values from -1.0 to -3.5 K d −1 (case (b)) and -0.5 to -5.5 K d −1 (case (c)). Moreover, an intensification of net radiative cooling with height towards the top of the SAL is evident. 30 Another increase in short-wave heating and long-wave cooling rates is found within the MBL which is characterized by a moist mixture of mineral dust and marine aerosols in all three cases. However, the shape of the MBL heating rate profile    pronounced MBL-inversion; cases (b) and (c)). Short-wave, long-wave as well as net heating rate profiles calculated for the dust-free case (a) show no distinct features above the MBL. In this case peak values of atmospheric heating and cooling correlate with regions of strongest gradients in r m (maximum long-wave cooling: -5 K d −1 ; maximum short-wave heating: 1.8 K d −1 ). This emphasizes the dominating effect of water vapor on atmospheric heating.

SAL radiative effects at surface level and top of the atmosphere 5
Saharan dust short-wave radiative effects at surface level and TOA ( Figure 5) are investigated by analyzing modeled solar zenith angle dependent short-wave irradiances for the three discussed scenarios. It is assumed that the observed profiles do not change and remain stationary within a 24 h time frame. Saharan dust short-wave radiative effects at surface level (RE SUR ) and top of the atmosphere (RE TOA ) are inferred as the difference between modeled irradiances considering mineral dust particles in the model atmosphere (E ↓tot(SUR) , E ↑tot(TOA) ) and irradiances calculated under assumption of no dust aerosol in the atmosphere 10 (E ↓nodust(SUR) , E ↑nodust(TOA) ), xiv https://doi.org/10.5194/acp-2020-420 Preprint. Discussion started: 8 June 2020 c Author(s) 2020. CC BY 4.0 License. and, Downward and upward irradiances are primarily determined by solar elevation, therefore having a symmetrical shape with maxima at noon (around 15:30 UTC). The longer the slant path of solar rays through SALs, the more Mie-and Rayleighscattering processes and the larger the fraction of backscattered light to space. As a consequence, RE SUR and RE TOA show

Discussion
In this study the effects of mineral dust particles and water vapor on radiative heating rate profiles in SAL-influenced regions were studied. It was found, that the enhanced water vapor concentrations within the SALs cause a decrease of radiative heating rates towards the top of the SALs. This negative gradient with height is in agreement with results found by Kim et al. (2004) who 20 focused on the effect of enhanced water vapor concentrations on atmospheric heating rates within Asian dust plumes. They also highlighted that derived atmospheric heating rates within the dust plumes are altered by enhanced water vapor concentrations and compared heating rate calculations including measured water vapor profiles to reference profiles. Calculated maximum short-wave heating and long-wave cooling rates were also found to be shifted from the center to the top of the dust layer when including the measured water vapor profile in their calculations. The strong negative trend of the heating rate profiles within 25 the SALs is supposed to decrease the static stability in the layers and to promote vertical mixing and convective development.
Vertical mixing in the SALs during their transport over the Atlantic Ocean was already proposed by Gasteiger et al. (2017) in an integrated study of active remote sensing, in-situ measurements and optical modeling. They suggested that vertical mixing within the SALs may counteract the Stokes gravitational settling during transport. Dropsonde measurements discussed in radiative effects in this study. However, results in this paper are of a slightly smaller magnitude due to thinning of the SAL during long-range transport (-40 W m −2 at the surface and -25 W m −2 at TOA). (Foltz and McPhaden, 2008) found that less 20 down-welling solar radiation in dust-laden regions may cause gradients in sea surface temperature and potentially impacts the evolution of clouds in the MBL.
During NARVAL-II, the majority of mineral dust particles was always located above the MBL and inside the SAL. However, during previous field campaigns it was observed that the vertical distribution of long-range-transported mineral dust can be highly variable (Reid, 2002). During the Puerto Rico Dust Experiment (PRIDE) in summer 2000, for example, the majority of 25 dust was in some cases observed to be located at lower atmospheric levels inside the MBL and in other cases it was observed to be located inside the SAL. A distinct seasonal pattern of Saharan dust transport towards the Atlantic Ocean was already found by Chiapello et al. (1995), who studied the vertical distribution of mineral dust particles at the beginning of long-range transport at the Cap Verde islands. They found that in contrast to the summer months, wintertime dust-transport towards the Atlantic Ocean is mainly taking place at lower atmospheric levels. Questions regarding the reasons for these variable vertical 30 distributions of mineral dust and whether there is a certain seasonal pattern in the vertical distribution not only at the beginning but also throughout the long-range dust transport can hopefully be answered in near future by analyzing data collected during the recent EUREC 4 A field campaign (ElUcidating the RolE of Clouds-Circulation Coupling in ClimAte; Bony et al., 2017) in January 2020 .
This study investigated the effects of dust and water vapor in long-range-transported SALs on atmospheric heating rates and radiative transport on the basis of airborne lidar measurements over the western subtropical North Atlantic Ocean. Simultaneously measured profiles of water vapor mass mixing ratios and aerosol optical properties were used to characterize both the vertically resolved aerosol and water vapor composition in radiative transfer simulations. 5 Lidar measurements in Saharan dust-laden regions indicated enhanced concentrations of water vapor in SAL-altitudes and radiative transfer simulations revealed that water vapor plays the dominant role for atmospheric heating rates in these heights.
Compared to water vapor, dust aerosol was identified to have a minor effect on total heating rates in SAL-altitudes showing small positive maximum heating rates of 0.3-0.5 K d −1 in the short-wave and slightly negative maximum cooling rates of -0.1 to -0.2 K d −1 in the long-wave spectrum at altitudes of highest aerosol mass concentration. Water vapor, however, was found 10 to contribute much stronger to total SAL-heating rates with maximum short-wave and long-wave heating of 1.8-2.2 K d −1 and -6 K d −1 to -7 K d −1 at the uppermost levels of the SAL. As a result, calculated net heating rates inside SALs are entirely negative and decrease with altitude.
SALs were also found to have a possible impact on cloud development in the MBL. Besides possible impacts on low-level circulations, SALs introduce additional atmospheric counter-radiation towards the top of the MBL. As a result, MBL tops in 15 dust-laden regions do not experience as much cooling as in SAL-free regions. This is also indicated by the heating rate profile in SAL-regions which is increasing with altitude and therefore counteracts the development of convection.
Last but not least, NARVAL-II lidar data were used to quantify the radiative effect of long-range-transported Saharan dust at surface level and top of the atmosphere. Maximum short-wave radiative effects of -40 W m −2 (surface) and 25 W m −2 (TOA) were found at intermediate zenith angles. 20 Summed up, radiative transfer calculations with NARVAL-II lidar data input highlighted the importance of correct representations of water vapor profiles in radiative transfer models and depicted the influence of mineral dust on the modification of solar irradiance throughout the atmosphere.
Data availability. The data used in this publication was collected during the NARVAL-II (Next-generation Aircraft Remote-Sensing for Validation Studies-II) campaign and is made available through the DLR Institute for Atmospheric Physics in the HALO database (German Author contributions. In the framework of the NARVAL-II field experiment MW and SG contributed to carry out all airborne lidar measurements used in this study. MW did the initial data processing. MG performed the analytic computations, analyzed the dataset and performed radiative transfer calculations with help BM and under supervision of SG. MG and SG took the lead in writing the manuscript. All authors discussed the results and contributed to the final manuscript. xviii https://doi.org/10.5194/acp-2020-420 Preprint. Discussion started: 8 June 2020 c Author(s) 2020. CC BY 4.0 License.