Cloud macro-physical properties in Saharan dust laden and dust free North Atlantic trade wind regimes: A lidar study

Abstract. Saharan dust is known to have an important impact on the atmospheric radiation budget, both directly and indirectly by changing cloud properties. However, up to now it is still an open question if elevated and long-range transported Saharan dust layers have an effect on subjacent marine trade wind cloud occurrence. Shallow trade wind clouds have a significant impact on the Earth's radiation budget and still introduce large uncertainties in climate sensitivity estimates, because of their poor representation in climate models. The Next-generation Aircraft Remote-Sensing for Validation studies (NARVAL) aimed at providing a better understanding of shallow marine trade wind clouds and their interplay with long-range transported elevated Saharan dust layers. Two airborne campaigns were conducted – the first one in December 2013 and the second one in August 2016; the latter one during the peak season of transatlantic Saharan dust transport. Airborne lidar measurements in the vicinity of Barbados performed during the second field campaign are used to investigate possible differences between shallow marine cloud macro-physical properties in dust-free regions and regions comprising elevated Saharan dust layers. The cloud top height distribution derived in dust-laden regions differs from the one derived in dust-free regions and indicates that clouds are shallower and convective development is suppressed. Furthermore, regions comprising elevated Saharan dust layers show a larger fraction of small clouds and larger cloud free regions, compared to dust-free regions. The cloud fraction in dusty regions is only 14 % compared to a fraction of 31 % in dust-free regions. Moreover, a decreasing trend of cloud fractions and cloud top heights with increasing dust layer vertical extent as well as aerosol optical depth is found.


combined active and passive remote sensing payload, including radar and lidar systems -probably the two most important instruments for vertically highly resolved measurements of aerosol and cloud properties. In addition, dropsondes were deployed to get information on the thermodynamic state of the atmosphere. This study is focused on the retrieval of horizontal and vertical distributions of both aerosols and clouds (i.e. cloud top height, cloud length) from lidar measurements performed in the course of NARVAL-II. 5 Chapter 2 gives an overview of the NARVAL-II field campaign and a description of the used lidar instrument. In Chapter 3 the general measurement situation during NARVAL-II is discussed and a detailed overview of the results is given. A short summary along with the conclusion of this paper is presented in Chapter 4.
2 Instruments and methods 2.1 NARVAL-II 10 In summer 2016 the Next-generation Aircraft Remote-Sensing for Validation Studies-II (NARVAL-II; Stevens et al. (2018, submitted)) were conducted to study the occurrence and formation of marine clouds during the subtropical North Atlantic wet season. As Saharan dust transportation over the Atlantic Ocean occurs quite frequently in northern hemispheric summer months, measurements were also dedicated to investigate the influence of the SAL on underlying shallow trade wind clouds.
During NARVAL-II, HALO was operated out of Barbados. The aircraft has a maximum range of more than 12 000 km and a 15 maximum cruising altitude of ∼15.5 km. It was equipped with a combined active and passive remote sensing payload including the lidar system WALES ), a 35.2 GHz cloud radar (Ewald et al., 2018, in discussion), microwave radiometers (Mech et al., 2014), a hyper spectral imager (Ewald et al., 2016), and the SMART instrument for radiation measurements  (Wendisch et al., 2001). Additionally a large number of dropsondes were deployed to get information on the atmospheric state.  Table 1 gives a detailed overview of all performed research flights including main research objectives. 5 This study focuses on the dust-laden RF2 to RF4 and RF6. A 22 h-lidar data set measured in the dust-free trades and a 16 hlidar data set measured in SAL trade wind regions is used to study differences in macro-physical cloud properties in the respective regions during NARVAL-II.

The WALES instrument
The WALES instrument ) is a combined airborne high spectral resolution (HSRL; Esselborn et al. (2008)) 10 and water vapor differential absorption lidar system (DIAL), built and operated by the Institute for Atmospheric Physics of the German Aerospace Center (DLR). The system provides highly resolved information on the vertical distribution of water vapor mixing ratio from measurements at four wavelengths around 935 nm. Additionally, it is capable of polarization sensitive measurements at 1064 nm and 532 nm wavelength. The 532 nm channel is also equipped with High Spectral Resolution Lidar (HSRL) capability, which allows to determine the extinction coefficient without assumption on scattering properties of aerosol 15 and cloud particles, hence enabling an enhanced characterization of them.
WALES measurements are performed in near nadir direction (2°-3°off-nadir angle) and provide vertical profiles of particle Depolarization data quality is ensured by frequent calibrations following to the ±45°method described by Freudenthaler et al. 5 (2009). Remaining relative uncertainties in aerosol depolarization measurements in the range from 10 to 16 % are primarily caused by the mechanical imprecision of the calibration setup (Esselborn et al., 2008) and possible atmospheric variations during the calibration. For backscatter and extinction measurements relative uncertainties of less than 5 % and 10 to 20 % have to be considered.

10
Based on the aerosol classification scheme described by Groß et al. (2013), WALES measurements can be used to identify and characterize layers of long-range transported Saharan dust. In this study the particle linear depolarization ratio at 532 nm (δ p532 ) is used as an indicator for non-spherical dust particles. Saharan dust δ p532 near source regions was found to take values around 30 % Tesche et al., 2009;Groß et al., 2011). This value does not change for long-range transported Saharan dust Burton et al., 2015;Groß et al., 2015;Haarig et al., 2017). Thus δ p532 is a 15 good proxy to distinguish long-range transported Saharan dust from less depolarizing marine boundary layer aerosols which typically take values around 3 % (Sakai et al., 2010;Burton et al., 2012;Groß et al., 2013). To reduce signal noise biases, an additional filter to flag mineral dust layers for regions with 532 nm-backscatter ratios (BSR 532 ) equal or higher 1.2 is applied (BSR 532 = 1 + β p532 /β m532 -where β p532 and β m532 are the particle and molecular backscatter coefficients). The origin of identified dust layers is further verified using calculated backward trajectories utilizing the HYbrid Single Particle Lagrangian Once verified as transported Saharan dust layer, the WALES HSRL particle extinction measurements at 532 nm are used to 25 calculate the aerosol optical depth of both the detected Saharan dust layers (τ SAL(532) ) and the atmospheric column ranging from the aircraft down to ground level (τ tot(532) ). Additionally, the Saharan dust layer's vertical extent ∆z SAL is defined as the sum of all dust-flagged 15 m-resolved height intervals within each vertical lidar profile.

Lidar-derived cloud macro-physical properties
Lidar derived cloud detection is usually performed using fixed signal thresholds (e.g. Medeiros et al. (2010); Nuijens et al. 30 (2009, 2014)) or by applying wavelet covariance methods for the detection of sharp gradients to the backscattered signal (Gamage and Hagelberg, 1993). During NARVAL-II it was found that BSR 532 in the cloud-free marine trade wind boundary layer as well as in the elevated SAL never exceeds a ratio of 10. Marine trade wind water-clouds are optically thick and thus  take much larger values. Based on these findings and to avoid potential miscategorizations of sharp aerosol gradients as cloud tops using wavelet transforms a fixed threshold of BSR 532 = 20 is used for the cloud/no-cloud decision.
To determine the cloud top height (CTH) in a vertical lidar profile the BSR 532 profile is scanned from flight level downwards and the first range bin where BSR 532 is greater or equal to the defined threshold is marked. Additionally, the whole profile is flagged as a 'cloud containing' profile. All 'cloud containing' profiles with cloud top heights in a certain altitude range are  neighboring cloud-free profiles. It should be mentioned that not the maximum cloud (gap) length of each individual cloud, but the along-track cloud (gap) length is derived. As a result, the amount of small clouds (gaps) in this study may be overestimated.

Dust measurements during NARVAL-II
In the following the measurement situation during the four HALO-flights used to characterize long-range transported Saharan dust layers (see Section 2.1) is summarized and their influence on subjacent marine trade wind clouds is investigated ( Figure   4).
During RF2 on 10 August a thin Saharan dust layer (∆z SAL <2 km) ranging from 2.5 to 5.0 km altitude was detected during  mean contribution of 51 % to τ tot(532) .
The following case study presents a detailed description of RF6 including an analysis of dropsonde-profiles in dust laden and dust free regions.   that their CTH is rarely higher than approximately 1 km. However, in dust free regions cloud top heights reach almost twice as high and up to 2 km.

Case study -19
Differences in meteorological parameters between SAL-regions and dust-free regions are discussed by looking at dropsonde measurements (D1, D2). Both dropsondes clearly indicate the so-called trade wind inversion (TWI) in the altitude range from 1.3 to 1.7 km height capping the moist MBL. The TWI is characterized by a rapid temperature increase of about 4 K within 5 400 m (not shown) and a strong hydrolapse (relative humidity (r) drops from >80 to ∼30 %). The MBL itself shows no significant variations in relative humidity in both the dust-free and the dust-laden regimes with mean values of 85 %. For a better visualisation of regions featuring high atmospheric stability the squared Brunt Väisälä frequency N 2 = g Θ dΘ dz , with g being the gravity of the Earth and Θ the potential temperature, is shown. N 2 shows regions of high atmospheric stability and thus strong restoring forces for a vertical air parcel displacement at the inversion altitudes. Enhanced atmopspheric stability is found at the 10 TWI for both analyzed dropsonde measurements. At higher altitudes the N 2 -profiles look different. In dust-laden regions the lower and upper boundary of the SAL are characterized by two additional well-known inversions Dunion and Velden, 2004;Ismail et al., 2010). Altogether, a total of three prominent inversion layers counteract convective development in dust-laden regions, whereas in dust-free regions only the trade wind inversion is present.

15
In a next step differences in macro-physical properties of shallow marine clouds between the identified dust-laden and dust-free flight segments are studied.

Cloud fraction and cloud top height
A fist indicator for differences in marine trade wind cloud occurrence is the cloud fraction CF. During NARVAL-II a total number of 3.2 × 10 4 one second resolved cloud tops were detected in trade wind regions (N CT (dust) = 8 × 10 3 ; N CT (nodust) = 2.4 × 10 4 ). 20 They contribute to an overall observed CF of 24 % within the measurement period. In dust-free regions a CF of 31 % was derived, while in SAL-regions CF was smaller by a factor of more than two (14 %). The next parameter to look for differences between the two cases is the CTH-distribution for both regions ( Figure 5). In the SAL-regions only a small fraction of clouds exceeds an altitude of 2 km and no cloud top is found at altitudes greater 2.5 km. The majority of cloud top heights (∼61 %) is found within the altitude range from 0.5 to 1.0 km. 26 % of all detected cloud top heights are located in the 1.0 to 1.5 km height 25 interval and only 11 % of that fraction contribute to the interval from 1.5 to 2.0 km altitude. Cloud tops in altitudes >2.5 km including deeper reaching convection with maximum top heights of 6 km are found in ∼16 % of all dust-free cloud profiles.
Underneath around 3 km altitude the CTH-distribution shows a two-modal structure with two local maxima ranging from 0.5 to 1.0 km (∼35 %) and 1.5 to 2.0 km altitude (∼20 %).
The statistical significance of observed differences in the distributions was checked by randomly resampling the respective 30 data-sets to smaller sub-sets and by comparing the shapes of the resulting distributions to the shape of the overall distributions.
The shapes of the resampled distributions showed no major differences compared to the overall distributions, thus it can be concluded that our NARVAL-II measurements indicate the presence of less and shallower clouds in Saharan dust laden trade wind regions compared to dust-free regions.

Cloud lengths and cloud gaps
As next step the cloud length and cloud gap length distributions of marine trade wind clouds in SAL-regions and mineral dust free regions are investigated (Figure 6, top). A total of 3688 and 2355 clouds were observed in dust free and dust-laden regions 5 during the NARVAL-II research flights. In both regions clouds with a horizontal extent of less than 0.5 km are by far the most prominent cloud type. Whereas 72 % of all clouds in SAL-regions are of this length, 65 % of all clouds detected in clear regions contribute to this length-interval. In both regions the frequency of cloud length occurrence decreases strongly with increasing cloud length. Relative frequency drops to ∼17 % (dust-laden) and ∼16 % (dust-free) in the length interval from 0.5 to 1.0 km.
Only 5 % of all clouds in dusty regions are observed to have a horizontal extent greater than 2 km. This fraction almost doubles 10 to 9 % in dust-free regions. The main contributor to this fraction are clouds with horizontal extents of more than 5 km (4 %).
Clouds of this length are basically only found outside dust-laden regions.
Another important parameter to highlight differences of cloudiness between SAL-regions and dust-free regions is the cloud gap length (Figure 6, bottom). Similar to the distribution of cloud lengths, also cloud gap frequencies decrease with increasing cloud gap length. In both regimes cloud gaps shorter 0.5 km are dominating. They contribute with 45 % and 35 % to the total 15 amount of observed cloud gaps in dust-free and dust-laden regions. A reverse picture emerges, when looking at the amount of cloud gaps greater than 5 km. A fraction of 17 % is found to be greater than 5 km underneath dust layers, whereas in dust-free regions these gap sizes contribute with 12 % to the distribution. Cloud gap fractions in range-bins from 1.5 to 4.5 km decrease  in both regions consistently with increasing cloud gap length.
The significance of the distribution-properties was again double-checked by the comparison to randomly resampled subdatasets. Overall, the cloud length and gap length distributions ( Figure 6) indicate, that the dust-laden trade wind regimes during NARVAL-II were characterized by a larger amount of small scale clouds and slightly greater cloud gaps, compared to the dust free regimes.

Connecting dust and cloud properties
As a further step the observed CTH and CF are related to the geometrical and optical depth (∆z SAL and τ SAL(532) ) of overlying mineral dust layers (Figure 7). Cloud fractions and heights in dust-flagged profiles of all four research flight are grouped together with respect to similar ∆z SAL (bin width: 0.2 km) and τ SAL(532) (bin width: 0.015). During NARVAL-II Saharan dust layers with maximum vertical extents of 4 km and maximum optical depths of 0.4 were observed ( Figure 2). However, 10 underneath optically thick dust layers (0.24 < τ SAL(532) ; 3.8 km < ∆z SAL ) not any cloud has been detected. to 0.24. Moreover, the variability of mean CTH in that range gets smaller, again indicating that higher-reaching convection is suppressed.
For the interpretation of these distributions the accumulated measurement-time in the respective intervals as well as the contribution of different research flights have to be taken into account. Mainly data collected in the course of RF3 contributes to SAL-measurements in the ranges 0.09< τ SAL(532) <0.24 and 2 km< ∆z SAL <4 km (Figure 2), thus being the main contribu-20 tor to observed increases of mean CTH and CF in regions of high τ SAL(532) and ∆z SAL . The remaining research flights (RF2, RF4 and RF6), were characterized by thinner dust layers that were rather decoupled from the MBL and contribute to regions of small τ SAL(532) and ∆z SAL .
Altogether, a decreasing trend of CTH and CF as a function of dust layer optical depth and vertical extent was detected during research flights over elevated and long-range transported Saharan dust layers. However, RF3 showed a predominant 25 and strongly pronounced transition layer that possibly altered the cloud layer resulting in an increased CF and CTH in the respective intervals of τ SAL(532) and ∆z SAL .

Summary and Conclusion
In this study airborne lidar measurements performed on-board the German high altitude and long-range research aircraft HALO during the NARVAL-II experiment over the North Atlantic trade wind region were used to investigate whether marine low regions implicate less, shallower and smaller clouds than dust-free regions. The overall derived cloud fraction in the dust-laden trades is 14 % and thus a factor of two smaller than the cloud fraction of 31 % derived from observations in the dust-free trades. These results are in good agreement with previous satellite remote sensing studies (Dunion and Velden, 2004) and model studies (Wong and Dessler, 2005;Stephens et al., 2004) which also suggest a convection-suppressing characteristic of the SAL with the main player being a dry anomaly in SAL-altitudes. During NARVAL-II long-range Saharan air layers were 5 not found to come along with dry anomalies, but were rather showing enhanced relative humidities in the range from 30 to 40 %. However, a suppressing characteristic of the SAL on subjacent marine clouds, is evident as well. Wong and Dessler (2005) also showed that the convection barrier increases with SAL-aerosol optical depth. To investigate a possible relation between SAL optical depth or layer vertical extent and marine trade wind CTH, the CTH and CF-distribution was analyzed as a function of SAL vertical extent and optical depth. It was found that mean CTH decreases with increasing 10 layer vertical extent for vertically thin layers (<1.5 km). Additionally, the mean CTH-variability for these layers is high, indicating the occurrence of higher-reaching clouds in those regions. There is no significant decrease of mean CTH for thicker dust layers, but a reduction of CTH-variability could be derived. Also a decrease in mean CTH-variability with increasing dust layer optical thickness starting at τ SAL(532) ≈ 1.2 could be detected. This indicates that optically and vertically thicker dust layers suppress the evolution of higher reaching convection. Moreover, a decrease in CF comes along with this reduction 15 in variability of the mean CTH. Underneath optically thick dust layers with 0.24 < τ SAL(532) not any cloud was detected.
Altogether, NARVAL-II lidar measurements indicate that there is a strong correlation between the presence of elevated and long-range transported Saharan dust layers and the occurrence and macro-physical properties of subjacent marine low clouds.
Further reaching questions regarding changes in radiation caused by the dust layer, changes in the general circulation patterns or the settling of dust particles into the cloud layer (Groß et al., 2016) could not be addressed within the present work and are 20 left to future studies.
Author contributions. In the framework of the NARVAL-II field experiment Martin Wirth and Silke Groß contributed to carry out all airborne lidar measurements used in this study. Martin Wirth did the initial data processing. Manuel Gutleben performed all analytic computations, statistically analyzed the data set and took the lead in writing the manuscript under consultation of Silke Groß. All authors discussed the results and contributed to the final manuscript.