2.1 NARVAL

In December 2013 and August 2016 the Next-generation Aircraft Remote-Sensing for Validation Studies (NARVAL; Klepp et al., 2014; Stevens et al., 2019) were conducted to study the occurrence and formation of marine clouds during the subtropical North Atlantic dry and wet seasons. 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 both NARVAL-I-South (here for simplicity referred to as NARVAL-I) and NARVAL-II, HALO was operated eastward of Barbados. The aircraft has a maximum range of more than 12 000 km and certified ceiling of 15.545 km altitude (max altitudes: NARVAL-I: ∼14 km; NARVAL-II: ∼15 km). During both campaigns it was equipped with a combined active and passive remote sensing payload including the lidar system WALES (Wirth et al., 2009), a 35.2 GHz cloud radar (Ewald et al., 2019), microwave radiometers (Mech et al., 2014), a hyper spectral imager (Ewald et al., 2016) and the Spectral Modular Airborne Radiation measurement System (SMART) instrument for radiation measurements (Wendisch et al., 2001). Additionally a large number of dropsondes were deployed to get information on the atmospheric state (NARVAL-I: 71; NARVAL-II: 218).

From 8 to 30 August 2016, 10 research flights (RF) comprising a total of 85 flight hours were conducted (Fig. 1). During four of those flights, flight patterns were specifically designed for an investigation of Saharan air layers and their impact on subjacent marine trade wind cloud regimes. Moreover, studying the large-scale atmospheric divergence was a main objective of the campaign (Bony and Stevens, 2019). This is why the flight patterns show many circles, i.e., during RF2, RF3, RF6–8 and RF10. Table 1 gives a detailed overview of all performed NARVAL-II research flights including the main research objectives.

This study focuses on the dust-laden research flights RF2, RF3, RF4 and RF6 of NARVAL-II. Cloud macrophysical properties measured during those flights are compared to properties observed during dust-free NARVAL-II flights. Datasets obtained during the NARVAL-II transfer flights from and to Germany (i.e., RF1 and RF10) are not included in the analysis because most measurements took place outside the trades and cirrus fields were present inside the trades. RF5 and RF7 are also excluded because cirrus fields covered most of the research area during RF5 and the objective of RF7 was to cross the Intertropical Convergence Zone (ITCZ) for several times. NARVAL-I lidar measurements inside the trades (10–20^∘ N) are used to compare obtained results from the 2016 summer season to the 2013 winter season.

In summary, 38 h of measurements during the summer season (22 h of lidar measurements during dust-free times, 16 h of lidar measurements with SAL present) and 44 h of measurements during the winter season are used to study differences in macro-physical cloud properties between the dust and non-dust times and different seasons.

Figure 1NARVAL research flight tracks: NARVAL-II dust-flights (color coded) and NARVAL-I and NARVAL-II dust-free flights (grey).

Table 1Overview of the conducted research flights during NARVAL-II in 2016 including dates, times of takeoff and landing, total duration of flights, research objectives, and flight hours in SAL regions (all times given in UTC; note Atlantic standard time is UTC−4; TBPB: Grantley Adams International Airport; EDMO: Airport Oberpfaffenhofen).

Download Print Version | Download XLSX

2.2 The WALES instrument

The WALES instrument (Wirth et al., 2009) is a combined airborne high spectral resolution (HSRL; Esselborn et al., 2008) 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 the 1064 and 532 nm wavelengths. 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 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 backscatter, linear depolarization and extinction from the aircraft down to the ground level. The vertical resolution of the WALES aerosol and cloud data is 15 m. The temporal resolution of the raw data is 5 Hz and is averaged to 1 Hz for a better signal-to-noise ratio. This results in a horizontal resolution of approximately 200 m at typical aircraft speed.

Depolarization data quality is ensured by frequent calibrations following the ±45^∘ method described by Freudenthaler et al. (2009). Remaining relative uncertainties in aerosol depolarization measurements are estimated to be in the range from 10 % to 16 % (Esselborn et al., 2008) and are primarily caused by atmospheric variations during the calibration. For backscatter and extinction measurements relative uncertainties of less than 5 % and 10 %–20 % have to be considered, respectively.

Figure 2The 10 d backward trajectories with starting points at the center of the respective Saharan air layers for the four NARVAL-II research flights leading over Saharan-dust-laden trade wind regions (RF2, RF3, RF4 and RF6).

2.3 Dust layer detection

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 nonspherical dust particles. Saharan dust δ_p532 near-source regions was found to take values around 30 % (Freudenthaler et al., 2009; Tesche et al., 2009; Groß et al., 2011). This value does not change for long-range-transported Saharan dust (Wiegner et al., 2011; Burton et al., 2015; Groß et al., 2015; Haarig et al., 2017). Thus δ_p532 is a 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 (BSR₅₃₂) equal to or higher than 1.2 is applied (BSR${}_{\mathrm{532}}=\mathrm{1}+{\mathit{\beta }}_{\mathrm{p}\mathrm{532}}/{\mathit{\beta }}_{\mathrm{m}\mathrm{532}}$ – 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 Integrated Trajectory model (HYSPLIT model; Stein et al., 2015) with NCEP GDAS (National Centers for Environmental Prediction Global Data Assimilation System) data input. Starting times and locations are chosen to match the center of detected mineral dust layers in the lidar profiles. The reliability of the backward trajectory calculations was checked by slightly modifying starting times and locations.

Once verified as transported Saharan dust layer, the WALES HSRL measurements are used to calculate the aerosol optical depth at 532 nm 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 Δz_SAL is defined as the sum of all dust-flagged 15 m resolved height intervals within each vertical lidar profile.

Figure 3Overview of the four NARVAL-II research flights leading over Saharan-dust-laden trade wind regions. (a) Cross sections of measured BSR at 532 nm and (b) applied mineral dust mask. (c) The 10 min boxcar average of the calculated dust layer vertical extent Δz_SAL. (d) The 10 min boxcar average of the derived total dust aerosol optical depth from aircraft to ground level τ_tot(532) (blue) and aerosol layer optical depth τ_SAL(532) (red). (e) Mean values and standard deviations of the measured 10 min averaged SAL particle linear depolarization ratio (δ_p532).

Download

2.4 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., 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 BSR₅₃₂ 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 BSR₅₃₂=20 is used for the cloud or no-cloud decision.

To determine the cloud top height (CTH), the BSR₅₃₂ profile is scanned from flight level downwards to 250 m altitude and the first range bin where BSR₅₃₂ is greater than 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 taken and divided by the total number of cloud-flagged profiles to obtain the CTH fraction in the respective bin of the overall CTH distribution. Similar to that the cloud fraction (CF) is defined as the number of all cloud-containing profiles divided by the total number of vertical lidar profiles.

For the calculation of cloud lengths along the flight path neighboring cloud-flagged vertical profiles are connected. The cloud length is then determined as a function of the respective geolocations (aircraft latitude and longitude) and CTH using the haversine formula. Cloud gaps are calculated analogously by connecting neighboring cloud-free profiles. Due to the instruments' maximum horizontal resolution of approximately 200 m, it is possible to resolve minimum cloud (gap) lengths of 200 m. 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.

3.1 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 Sect. 2.1) is summarized and their influence on subjacent marine trade wind clouds is investigated (Fig. 3). The Saharan origin of the observed dust layers is verified using 10 d backward trajectories with starting points at the center of the respective Saharan air layers (Fig. 2). All observed dust layers traveled for 5 to 10 d from the Adrar–Hoggar–Aïr region in central Africa to the measurement location over the western North Atlantic Ocean. In central Africa the SAL is formed by intense surface heating and dry convection, which mixes dust particles to altitudes of up to 6 km (Gamo, 1996).

During RF2 on 10 August a thin Saharan dust layer (Δz_SAL<2 km) ranging from 2.5 to 5.0 km altitude was detected during the whole flight. A mean δ_p532 of 30 % clearly classifies this elevated layer as a mineral dust layer. τ_SAL(532) took values around 0.15 – on average approximately 35 % of the total column aerosol optical depth during this RF. Unfortunately, bright and strongly reflecting clouds in the lidar field of view caused the safety circuit of the detector unit to shut down the device, causing some gaps in the continuous lidar dataset.

In contrast to RF2, a vertically and optically thick dust layer was observed during the whole RF3 on 12 August. δ_p532 of this layer ranged from 28 % to 30 %, thus confirming the presence of Saharan mineral dust. The layer had a maximum vertical extent of ∼4 km, showed aerosol optical depths around 0.2 and contributed on average with 60 % to the total column aerosol optical depth during that flight.

While RF2 and RF3 were designed for measurements solely in dust-laden regions, RF4 and RF6 on 15 and 19 August were planned for measurements in both dust-laden and dust-free regions within the very same research flight. Flight tracks were chosen to cross dust gradients frequently, resulting in multiple flight segments of dust and no dust along the flight track. Elevated aerosol layers showed mean δ_p532 of 30 % and could therefore be identified as SAL. While the SAL during RF4 ranged on average from 2.5 to 4.5 km, it reached higher to almost 6 km altitude during RF5. With τ_SAL(532)≈0.1 the dust layer during RF4 contributed on average 25 % to τ_tot(532). τ_SAL(532) during RF6 took higher values of up to 0.4 and showed a mean contribution of 51 % to τ_tot(532).

Figure 4Flight track of RF6 on 19 August 2016 on top of the Terra MODIS (Moderate Resolution Imaging Spectroradiometer) true color image (a) and the MODIS aerosol optical depth (AOD) product (b) at 13:40 UTC. Launched dropsondes are marked by colored dots (red dots: mineral dust laden regions, blue: dust free regions).

The following case study presents a detailed description of RF6 including an analysis of dropsonde profiles in dust-laden and dust-free regions.

Figure 5Mean vertical WALES lidar profiles of δ_p532 and mean vertical dropsonde profiles of relative humidity (r), potential temperature (θ), squared Brunt–Väisälä frequency $N^{\mathrm{2}}=\frac{g}{\mathrm{\Theta }}\frac{\mathrm{d}\mathrm{\Theta }}{\mathrm{d}z}$, and wind speed (u) and direction (wdir) in dust-free (I) and dust-laden (II) regions during RF6 on 19 August 2016 (horizontal bars indicate standard deviations). (III) Differences in water vapor mass mixing ratio (MR), potential temperature, relative humidity and wind speed between the two regions. Shaded regions mark the marine boundary layer (MBL, blue) and the Saharan air layer (SAL, orange).

Download

3.2 Case study – 19 August 2016

RF6 on 19 August 2016 took place in the area between 48 and 60^∘ W and 13 and 19^∘ N (Fig. 4). The ITCZ and associated deep convection were located 550 km south of the flight track at around 10^∘ N and are not expected to have an influence on our analysis. RF6 was planned to cross a sharp gradient between a dust-laden and a clear region at an altitude of approximately 8.25 km with about one-half of the measurement time in dust-laden and the other half in dust-free regions. The circular patterns of the flight track were flown for dropsonde-based divergence measurements. Whereas the first pair of circles was performed over a heavily dust-laden region in the southern part of the flight track, the second pair was performed in the northern part over an almost dust-free region. This is also seen in MODIS aerosol optical depth imagery at 13:40 UTC in Fig. 4b where the region around the southern circle shows a maximum aerosol optical depth greater than 0.4.

Measured cross sections of BSR₅₃₂ and the derived mineral dust mask (Fig. 3a, RF6 and b, RF6) show pronounced elevated mineral dust layers ranging from 2.5 to 5.0 km altitude, horizontally alternating with dust-free profile regions. Due to the conducted divergence measurements, dropsondes were launched frequently along the circular flight tracks and are used to compare vertical profiles of meteorological parameters in dust-laden regions to those in dust-free regions. For this purpose mean profiles of potential temperature θ, relative humidity and water vapor mass mixing ratio (r, MR) as well as wind speed and direction (u, wdir) of all dropsonde measurements in the respective dust-laden and dust-free regions are compared in Fig. 5 (panels Ib–d, IIb–d and IIIa–d). Additionally, lidar-derived δ_p532 is analyzed for both regimes (Fig. 5, Ia and IIa).

Inside the SAL-region a three-layer structure is present:

1. the marine boundary layer (MBL), reaching up to approximately 1.3 km height (δ_p532 smaller than 10 % indicates that marine aerosols are the dominant contributor to the aerosol composition of the MBL);

2. a transition or mixed layer extending from the MBL top to 2.8 km altitude with varying values of δ_p532 (10 $\mathit{\%}<{\mathit{\delta }}_{\mathrm{p}\mathrm{532}}<\mathrm{20}$ %); and

3. the elevated SAL, with typical δ_p532 for long-range-transported Saharan dust (δ_p532∼30 %) ranging from 2.8 to 3.8 km height.

The mean dust-free δ_p532 profile of the MBL and transition layer looks quite similar to the mean dust-laden δ_p532 profile. However, no SAL signature is detected. The low CF in the dust-laden region (southern part of the flight track) which is visible in the MODIS image (Fig. 4), is also evident in lidar measurements after the application of the described threshold method for cloud detection. Whereas a lot of cloud tops in heights ranging from 0.5 to 1.5 km altitude are detected in the northern part of the flight track (after about 16:45 UTC), almost no cloud is detected along the earlier southern flight path – with the exception of the transition region to the dust-free area (cloud top heights at ∼1.5 km altitude).

This is also evident in the calculated cloud fractions. CF is 20 % in dust-free regions. In the SAL region however, CF decreases to 11 % (including the clouds developing at the edges of the dust layer). Another characteristic of clouds in SAL regions is 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. Divergence measurements discussed by Bony and Stevens (2019) and Stevens et al. (2019) show that dynamical properties in the two regions are different as well. They found that MBL vertical velocity in the dust-free regime is directed upwards and could explain the observed increased cloud top heights.

For an investigation of the question why vertical wind speeds, cloud tops and cloud fractions are higher in the dust-free regime than in the SAL regime, differences in meteorological parameters between SAL regions and dust-free regions are analyzed by discussing mean profiles of all dropsonde measurements in the respective regions. Both the dust-laden and the dust-free regions clearly indicate the so-called trade wind inversion (TWI) in an altitude range from 1.5 to 1.8 km height capping the moist MBL. The TWI is characterized by a rapid temperature decrease of about 4 K within 400 m (not shown) and a strong hydrolapse (relative humidity (r) drops from >80 % to ∼30 %). In both regimes the MBL itself can be divided into a sub-cloud layer which extends from the ocean surface to 0.5–0.7 km and a cloud layer (Groß et al., 2016) which extends from the sub-cloud layer top to the TWI (0.5–1.8 km). Those two regions can be identified in profiles of Θ and humidity. Whereas the sub-cloud layer is well mixed (Θ = constant, MR = constant), the cloud layer shows a conditionally unstable lapse rate of 5–7 K km⁻¹ (saturated air parcels are unstable to vertical displacement). Overall, measurements of Θ and humidity show a stronger variation in the dust-free MBL than in the dust-laden one, suggesting the presence of more boundary layer clouds in dust-free regions.

Nuijens et al. (2009) and Nuijens and Stevens (2012) found that high wind speeds near the surface correspond to an increase in boundary layer humidity leading to a deepening of the cloud layer and increased area rainfall. Lonitz et al. (2015) used large-eddy simulations to show how higher relative humidities associated with observed dusty boundary layers change the evolution of the cloud layer. However, when comparing boundary layer wind speed and humidity in the two regimes no distinct differences can be observed, indicating that some other mechanism must be responsible for the observed differences in vertical wind speed, cloud fraction and cloud top height. The MBL of both regimes is dominated by northeasterly winds with speeds around 7 ms⁻¹. In dust-laden regions wind speeds in SAL altitudes are 4 ms⁻¹ lower than in the dust-free regions. This suggests that the SAL represents a decoupled layer, which penetrates into the trade-wind regime. Moreover, enhanced amounts of water vapor are observed inside the long-range-transported SAL. Relative humidity and water vapor mass mixing ratio show an increase of 2 g kg⁻¹ in SAL regions compared to the dust-free trade wind region. Such an increase has already been observed by Jung et al. (2013). From radiosonde measurements they found that the SAL transports moisture from Africa towards the Caribbean and gets moistened during transport by upwelling surface fluxes.

For a better visualization of atmospheric stability the squared Brunt–Väisälä frequency $N^{\mathrm{2}}=\frac{g}{\mathrm{\Theta }}\frac{\mathrm{d}\mathrm{\Theta }}{\mathrm{d}z}$, with g being the gravity of the Earth, is shown. N² shows regions of high atmospheric stability and thus strong restoring forces for a vertical air parcel displacement at the inversion altitudes. Enhanced atmospheric stability is found at the TWI for both regimes. At higher altitudes N² profiles look different. In dust-laden regions the lower and upper boundaries of the SAL are characterized by two additional well-known inversions (Carlson and Prospero, 1972; Dunion and Velden, 2004; Ismail et al., 2010). Inside the layer N² is almost zero – indicating a well mixed SAL regime. Furthermore, Θ points towards a neutral stratification in the interior of the SAL since it does not change with altitude. Altogether, a total of three prominent inversion layers counteract convective development in dust-laden regions, whereas in dust-free regions just the trade wind inversion is present.

In conclusion, one can suggest that the SAL potentially modifies radiative transfer, atmospheric stability and the evolution of vertical velocities, hence representing a proxy for reduced amounts of clouds and lower cloud top heights. To discuss this hypothesis, differences in cloud macro-physical properties are investigated for the whole NARVAL field campaign series.

Figure 6Histograms of detected cloud top height fractions during NARVAL-I and NARVAL-II with bins of 0.5 km size. Red bars illustrate the distribution of cloud top height fractions in SAL regions. Blue bars represent the derived cloud top height distribution from measurements in the dust-free trades during NARVAL-II. Grey bars show the derived cloud top height in the dust-free winter season during NARVAL-I.

Download

3.3 Differences in cloud macro-physical properties

3.3.1 Cloud fraction and cloud top height

A first indicator for differences in marine trade wind cloud occurrence is the cloud fraction CF. During NARVAL-II a total number of 3.2×10⁴ 1 s resolved cloud tops were detected in trade wind regions ($N_{\mathrm{CT}\left(\mathrm{dust}\right)}=\mathrm{8}\times {\mathrm{10}}^{\mathrm{3}}$; $N_{\mathrm{CT}\left(\mathrm{nodust}\right)}=\mathrm{2.4}\times {\mathrm{10}}^{\mathrm{4}}$). 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 2 (14 %). In winter (NARVAL-I) an almost 3 times higher CF of 37 % is derived. The next parameter to look for differences between the three regions is the CTH distribution (Fig. 6). 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 %) are found within the altitude range from 0.5 to 1.0 km in lower altitudes of the MBL-cloud layer. A total of 26 % of all detected cloud top heights are located in the 1.0–1.5 km height interval and only 11 % of that fraction contributes 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. Below 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 %). Several clouds were also detected in the lowermost 0.5 km of the atmosphere (∼1 %). Most likely those clouds are evolving or dissipating clouds at the bottom of the cloud layer.

In the dust-free winter season a shift of the distribution to higher altitudes is observed since most cloud tops were sampled in the interval from 2.0 to 2.5 km altitude (∼39 %). However, no cloud was observed at altitudes greater 3.5 km. This shift is caused by a slightly higher TWI in winter months, shown in Stevens et al. (2017), who compare mean dropsonde profiles of water vapor mixing ratio during NARVAL-I and NARVAL-II.

Figure 7Histograms of detected cloud lengths (a) and cloud gap lengths (b). Red bars illustrate the distribution of marine low cloud (gap) lengths located below Saharan dust layers. Blue bars represent the distribution derived from measurements in the dust-free trades during NARVAL-II. Grey bars show the derived distribution in the dust-free winter season during NARVAL-I.

Download

The statistical significance of observed differences in the distributions was checked by randomly resampling the respective datasets to smaller subsets 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.

Figure 8(a) Mean cloud top heights (middle) and cloud fraction (bottom) of clouds detected below Saharan dust layers as a function of Saharan dust layer vertical extent (Δz_SAL) – bin interval: 0.2 km. (b) Mean cloud top heights (middle) and cloud fraction (bottom) of clouds detected below Saharan dust layers as a function of Saharan dust layer optical depth (τ_SAL(532)) at 532 nm wavelength – bin interval: 0.015. Bars mark respective standard deviations of mean cloud top heights (1σ). The uppermost graphs in (a) and (b) illustrate summed measurement times in each interval.

Download