Nature and Extent of Shallow Marine Convection in Subtropical Regions: Detection with airborne and spaceborne Lidar-Systems over the tropical North Atlantic Ocean

Shallow marine cumulus convection over the Atlantic ocean near Barbados is studied with observations by airborne and spaceborne lidar instruments performed during the field campaign Next-generation Aircraft Remote Sensing for Validation Studies (NARVAL). For the first time airborne lidar measurements with the DLR high spectral resolution lidar system WALES on-board the German research aircraft HALO were conducted over the tropical North Atlantic Ocean. In the course of NARVAL several CALIPSO satellite underflights were performed, which allow comparisons of detected cloud top edges from the 5 two lidar instruments (i.e. WALES and CALIOP on-board CALIPSO). The study concentrates on the comparison and investigation of detected cloud top height distributions derived from measured WALES and CALIOP lidar profiles by use of a newly developed cloud detection algorithm. This allows to test the utilization of satellite based lidar systems for the observation of shallow marine convection. The distribution of cloud top heights during wintertime measurements shows a two-layer structure with maxima in ∼1000 m and ∼2500 m in both WALES and CALIOP measurements. Cloud top heights vary with latitude. 10 The analysed WALES and CALIOP data shows most frequent cloud tops in 10◦ to 20◦ N at heights from 1500 to 2500 m. A meridional decrease of detected cloud top heights over the subtropical North Atlantic Ocean, with lower values in the North, is observed. Approximately 36% of all clouds in the Atlantic trades are detected to have a horizontal extent of less than 1 km in the winter season. Cloud gaps shorter than 1 km dominate the Atlantic trades. They make up approximately 45% of all detected cloud gaps. 15


Introduction
In present day climate research clouds are one of the major contributors to uncertainties in estimates of the Earth's energy budget (Stephens, 2005).Clouds are highly variable in space and time and their occurrence cannot be predicted exactly.The quantification of cloud feedbacks in climate models is still one of the biggest challenges in present day climate science (Bony et al., 2015).Inadequate representation of clouds and moist convection in general circulation models is the main limitation in current representations of the climate system (Stevens and Bony, 2013).Large inter-model spread is found in low latitudes due to the dependence of cloud and precipitation responses on unresolved processes (Stevens and Bony, 2013).Differences in the representation of shallow marine convection in global climate models lead to large differences in climate sensitivity estimates (Bony and Dufresne (2005); Zelinka et al. (2012)).
These so called trade wind clouds form at the top of the subtropical marine boundary layer beneath the descending branches of the Hadley cell.They are often limited in altitude to at most 4000 m due to the presence of a prominent trade wind inversion, separating the moister marine boundary layer from the dry free troposphere (Stevens, 2005).Hence, they play an important role for the transport of moisture to the free atmosphere (Tiedtke, 1989).Shallow marine cumuli contribute about 60% to the net cloud radiative forcing and are one of the dominant contributors to global albedo (Hartmann et al., 1992).They cover about 12% of the sky over the Earth's oceans (Warren et al., 1986), but are extremely variable in spatial extent with time.The importance of shallow marine convection on climate in the trade wind regions was already identified in the mid-20 th century when campaigns started to explore the characteristics of shallow marine convection by studying meteorological processes and the structure of the moist boundary layer with the aid of airborne in-situ observations (Malkus, 1954(Malkus, , 1956(Malkus, , 1958)).Since then a large number of field experiments were conducted to better characterize shallow marine trade wind convection for numerical atmospheric models.
Hereby, macrophysical properties of cumulus clouds like their morphology, cloud size distributions or cloud top height (CTH) distributions play an important role.On the one hand these properties are used to evaluate cloud models (e.g.Siebesma and Cuijpers (1995)), on the other hand they serve as input parameters for model calculations to investigate the clouds radiative and dynamic effects on the environment (e.g.Zhao and Austin (2005a), Zhao and Austin (2005b)).Ground-based, shipborne or airborne measurements during field campaigns provide highly resolved observations of the macro-and microphysical cloud properties (e.g.Colòn-Robles et al. (2006); Siebert et al. (2013)), but are limited in space and time.Satellite measurements provide global coverage and long-term observations but the footprint of passive satellite observations mostly exceeds the smallscale structure of trade wind convection.In contrast, active remote sensing satellite measurements with the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) on-board the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) have a footprint diameter of 70 m and a footprint spacing of about 330 m (Winker et al., 2010).Thus, such measurements have frequently been used to study macrophysical parameters of trade wind cumuli and for model evaluation (e.g.Luo et al. (2016), Medeiros et al. (2010), Ahlgrimm andKöhler (2010), Tackett and Di Girolamo (2009)).However, up to now, no systematic evaluation of the applicability and constraints of CALIOP data for the use in studies of shallow marine convection was done.
The objective of the presented study is twofold.In a first step we want to evaluate if, and to which extent, spaceborne lidar measurements with the CALIOP lidar can resolve the small-scale character of trade wind convection in comparison to highly resolved airborne measurements.We therefore use measurements performed during the Next-generation Aircraft Remote Sensing for Validation Studies (NARVAL) mission (Klepp et al., 2014) on-board the German high-altitude and long-range research aircraft HALO (Krautstrunk and Giez, 2012) in December 2013 over the subtropical North Atlantic Ocean in combination with the spaceborne lidar system CALIOP.In a next step we want to use these airborne and spaceborne lidar measurements to investigate the size of shallow marine convection as well as their CTH distribution during NARVAL.Furthermore, we want to investigate the representativeness of the data collected during the NARVAL mission and therefore compare these data sets to longer time periods of CALIOP measurements and of different seasons.Exemplarily we use data of December, January and February (DJF) 2012/2013 and of summer (June, July, August -JJA) 2013.
A description of the used instruments and their measuring principles is given in Section 2. Section 3 presents the comparison of airborne and spaceborne lidar measurements and the statistical analysis of the NARVAL measurements with respect to the macrophysical properties of shallow marine convection.In Section 4 the results are discussed with respect to representativeness and comparisons to other studies.Section 5 concludes this work.

NARVAL
In December 2013 and January 2014, the Next-generation Aircraft Remote Sensing for Validation Studies (NARVAL) mission took place.NARVAL was designed as an airborne experiment using the German High Altitude and Long range research aircraft (HALO), which is a modified Gulfstream G550 business jet with a maximum range of more than 12000 km and a maximum cruising altitude of more than 15500 m (Krautstrunk and Giez, 2012).During NARVAL, the HALO aircraft was equipped with a set of remote sensing instruments.Besides the two main instrument packages, the water vapour and differential absorption lidar WALES (Water Vapour Lidar Experiment in Space, Wirth et al. (2009)) and the Halo Microwave Package (HAMP), a combination of a 36 GHz cloud radar and a set of microwave radiometers (Mech et al., 2014), the payload also included a miniDOAS system (Prados-Roman et al. (2011), Weidner et al. (2005)) and radiation measurements (Fricke et al., 2014).
In this study we only make use of the lidar data.The NARVAL mission consisted of two separate measurement periods.
In December 2013, measurements were performed out of Oberpfaffenhofen to Barbados as transfer flights as well as out of Barbados as local flights.The aim of the December flights was to study shallow marine convection and their environment.departing from Grantley Adams Airport (TBPB, Barbados).Six CALIPSO underflights were performed over the subtropical North Atlantic Ocean between 25 • W and 60 • W. Figure 1 shows the flight tracks of the conducted HALO flights during NARVAL.Furthermore, all CALIPSO ground tracks within a predefined area used for the analysis in this study during the NARVAL period are plotted.Altogether 12 daytime and 13 nighttime CALIPSO tracks crossed the defined measurement area during NARVAL.An overview of the conducted research flights during NARVAL including times of take-off and landing, and times of CALIPSO underflights is given in Table 1.

Lidar systems
For the this study, we use measurements of two different lidar systems -the airborne system WALES and CALIOP onboard the CALIPSO satellite.Both are briefly described in the following.

The WALES instrument
WALES is an airborne demonstrator built for a proposed, spaceborne mission to observe H 2 O concentrations in the atmosphere from space using the Differential Absorption Lidar (DIAL) technique (Bösenberg, 1998).In addition to the water vapour channels, WALES is equipped with High Spectral Resolution Lidar (HSRL) capability using an iodine filter (Esselborn et al., 2008), allowing simultaneous HSRL measurements at 532 nm and DIAL measurements in the absorption bands of water vapour between 935 and 936 nm.Furthermore, WALES performs polarization sensitive measurements at 532 and 1064 nm.
This study concentrates on the HSRL measurements at 532 nm.The raw data of WALES has a vertical resolution of 15 m and a horizontal resolution of 0.2 s (∼ 40 m at 200 ms −1 cruising speed).The data is post-processed to retrieve the aerosol backscatter ratio and the aerosol backscatter coefficients β part (Esselborn et al., 2008).For further technical information and description of the system see Wirth et al. (2009) and Esselborn et al. (2008).

The CALIOP instrument
The CALIPSO satellite has a sun synchroneous orbit in an altitude of about 705 km in nadir-pointing orientation.It crosses the equator at 13:30 (ascending node) and 01:30 (descending node) local solar time and has a 16-day repeat cycle.CALIOP is the spaceborne lidar instrument (Winker et al., 2007) on-board CALIPSO (Winker et al., 2010).It is a backscatter lidar performing simultaneous polarization sensitive measurements at 532 and 1064 nm.CALIOP data are provided in different data processing levels, from raw data to data products with large horizontal and vertical averaging lengths.For this study we use CALIOP Level 1B V4 532 nm data with a 30 m vertical and 330 m horizontal resolution expressed in terms of total attenuated backscatter coefficient β tot .

Data evaluation
For a consistent comparison of WALES and CALIOP instruments, data sets are converted into a common unit.In this study we use the ratio between total backscatter coefficient and molecular backscatter coefficient which is called backscatter ratio (BSR = β tot /β mol ).To calculate the molecular backscatter coefficient β mol , temperature and pressure fileds from Integrated Forecasting System model analysis (IFS) of the European Centre for Medium-range Weather Forecasts (ECMWF) are used.
The modelled fields are interpolated in space and time to match the flight paths of CALIPSO and HALO and their specific range resolutions.
To determine CTHs based on the BSR data, we define a BSR threshold for the cloud-/no-cloud decision.During NARVAL, aerosol was mainly located in the marine boundary layer.These aerosol layers showed BSR values between 1 and 10.BSR values for clouds were found to be much higher.For this study we define a threshold of BSR = 90 that marks the lower edge of the BSR range found for clouds during NARVAL.For the examination of shallow marine trade wind convection we only consider height ranges less than 4000 m altitude as shallow marine clouds usually do not advance to greater altitudes (Stevens, 2005).Furthermore, all determined profile heights less than 250 m (all altitudes are given above sea level) are excluded, as they are prone to surface echoes.The cloud detection algorithm scans profiles in the direction of beam propagation.If the threshold is exceeded once in a single lidar shot, the corresponding height is marked as CTH and the profile is marked as cloudy (Figure 2).We are aware that this detection algorithm does not capture lower cloud layers.However, we think that this approach gives a good upper estimate for the CTH distribution.Considering multiple cloud layers within one profile in the detection can even lead to false conclusions since the lidar signal is mostly saturated within clouds.
In a next step, all cloudy profiles are connected to determine cloud size distributions along the flight-paths.To detect beginnings and endings of a cloud, two cloudy profiles must be separated by at least one cloud-free profile.Otherwise, they will be attributed to the same cloud.
To compute lengths of clouds and cloudless areas, the Earth's shape is simplified to be a spheroid.The distance Ψ between two points A and B on a surface of a sphere is then calculated using, where Φ is the azimuthal angle and Θ is the polar angle.For distances of cloud and cloud gap points, Ψ is multiplied with the sum of the radius of the Earth (r ≈ 6370 km) and the mean CTH.

CALIOP / WALES comparison
To evaluate the potential of satellite based lidar measurements with the CALIOP lidar on-board CALIPSO we proceed as follows.First, we perform direct comparisons of the measured profiles during the CALIPSO underflights including the comparison of the derived CTH distribution.In a next step, we compare the macrophysical properties from airborne and spaceborne lidar for the whole NARVAL measurement period.

Case study -11 December 2013
During all NARVAL flights, the overall synoptic situation stayed constant.

Cloud top heights during NARVAL
In a next step we compare the overall distributions of detected CTHs during the whole period of NARVAL.We first analyse the CTH distribution of all conducted CALIPSO underflights and in a next step the CTH distribution of all conducted WALES and CALIOP measurements in the defined research area during NARVAL.Underflight profiles were found by manual filtering.
An overview of the number of used data sets and profile kilometres is given in Table 2.

Meridional distribution of detected cloud top heights
In the previous section we have shown that CALIOP measurements provide a good basis for studying the vertical distribution of CTHs.Consequential, we use the CALIOP measurements to study the meridional distribution of shallow marine convection CTHs to determine possible changes in their distribution with latitude.For this analysis we use CALIOP data over the subtropical North Atlantic Ocean between 60 • and 35 • W longitude and between 0 • and 30 • N latitude.Figure 6 shows the meridional

Cloud lengths and cloud gap lengths
For the analysis of cloud lengths, clouds with clearly detected boundaries are used (see Section 2.3). Figure 7 shows the two-dimensional frequency distribution of cloud lengths combined with mean CTHs.During NARVAL shallow marine clouds with a horizontal extent of less than 1 km were prevalent.They make up approximately 38% in CALIOP measurements.Also stratus clouds with extents larger than 100 km are detected in 7% of all CALIOP measurements.The maximum in mean CTH is located between 2000 and 2250 m.Most of the clouds have a size less than 1 km and CTHs between 2500 and 2750 m.They make up about 9% of all detected clouds.WALES data (not shown) and CALIOP are in good agreement, showing nearly the same distributions of relative cloud length frequency within this 10-day period.However, due to the better horizontal resolution of the WALES data, we were also able to resolve cloud sizes smaller than 1 km (up to 0.5 km -not shown).We found a uniform distribution of cloud sizes between 0 and 1 km.This resolution is not possible with CALIOP measurements as they have an effective resolution of about 330 m.In this study the reasons of the difference in CTH and meridional distributions between winter and summer measurements were not examined.Differences can be caused by variabilities in the general circulation pattern or due to the influence of aerosols.
Aerosols play an important role for the development and lifetime of clouds (Twomey, 1977) and may also modify the stability of the atmosphere (Gasteiger et al., 2016).Aerosol transport over the Atlantic Ocean and thus the aerosol distribution in the trade wind region is highly dependent on the general circulation and is subject to seasonal variations (Kaufman et al., 2005).
In August 2016, further airborne measurements in the vicinity of Barbados were conducted during the NARVAL-II mission to study shallow marine trade wind convection and exchange processes from shallow to deep convection.These measurements provide the opportunity (together with this study) to further examine the seasonality of trade wind clouds, the general circulation pattern as well as the effect of aerosols on cloud development.
Measurements were conducted in trade wind regions over the subtropical North Atlantic between 10 and 20 December 2013 at the beginning of the climatic dry season.In January 2014, HALO was operated out of Keflavik (Iceland) to study post-frontal convection and precipitation over the extra-tropical North Atlantic.In the following, the abbreviation NARVAL only refers to the flights performed during the first phase in December 2013 as this study focuses on the characterization of shallow marine trade wind convection.During NARVAL, eight research flights with almost 70 flight hours were conducted -four Caribbean transfer flights from or to Oberpfaffenhofen (EDMO, Germany) and four local flights over the subtropical North Atlantic Ocean

Figure 1 .
Figure 1.Flight-tracks of CALIPSO and HALO during NARVAL.White lines represent CALIPSO flightpaths in the investigation area ranging from 60 • W to 35 • W and 10 • to 20 • N (period: 10 to 20 December 2013, dashed: descending node, solid: ascending node).Red lines show the tracks of conducted HALO research flights.The green line indicates the profile discussed in Section 3.1.1.Black dots mark locations of dropped radiosondes.
Figure 3 gives an exemplary overview of the situation on 11 December 2013.Small irregularly scattered clouds dominate the area over Barbados and over the Atlantic Ocean.South of about 10 • N deep convective structures from the inter-tropical convergence zone (ITCZ) are present and north of about 30 • N cloud structures of extra-tropical weather regimes are visible.On 11 December 2013, we sampled the airmasses west of Barbados in several east-west flights at different latitudes form about 18 • N to about 10 • N. Along most parts of the flight path the situation was characterized by a moist marine sub-cloud layer below about 1 km topped by small scale cloud structures at ∼ 1 km and at ∼ 2.5 km.Weak aerosol structures were also visible between these two cloud layers.Above about

Figure 2 .
Figure 2. Exemplary visualisation of the developed algorithm applied on WALES data along the flight track (∼ 2 km).Colors represent measured backscatter ratios.Grey crosses indicate detected cloud tops.The horizontal solid black line marks the surface echo cut-off.The uppermost panel shows sequences of detected clouds (blue-filled polygons).

Figure 3 .
Figure 3. GOES -Satellite visible light image showing the large scale cloud situation around Barbados on 11 December 2013, 11 UTC.The red dot indicates the location of the island of Barbados (http://www.goes.noaa.gov).

Figure 4 .
Figure 4. Vertical profiles of determined BSR during a CALIPSO underflight on 11 December 2013 (WALES (a) and CALIOP profiles (b)).The cloud top detection algorithm is applied above 250 m (horizontal black solid line).The moment of vertical collinearity is marked by a black dashed line.Cloud top height frequencies are shown in (c) and (d) with 250 m vertical bin-size.

Figure 4
Figure 4 (c, d) shows the CTH distribution along the flight track of the underflight derived from WALES and CALIOP measurements.In general, both CTH distributions show a good agreement with a two-layer cloud structure.Both distributions have their maximum in detected CTH frequency at heights between 2250 to 2500 m due to the detected stratiform-like cloud layer.Over 47% of all CTHs derived from CALIOP measurements and 43% of all CTHs derived from WALES measurements are found in this height range.However, while the CTH distribution from the WALES measurements shows high values in the height bins between 2250-2500 m and 2500-2750 m with about 43% and about 32%, respectively, the CTH distribution derived from CALIOP measurements shows a contribution of almost 50% in the height bin between 2250-2500 m but a significantly lower value above.In contrast it shows about 10% more CTHs located in the height bin between 2000-2250 m compared to the CTH distribution derived from WALES measurements.Both distributions show local maxima in the order of 5% in height bins between 750 and 1250 m, representing the small scale convective clouds.No CTHs are detected above 3000 m.

Figure 5
Figure 5(a) presents the CTH distribution of all conducted underflights.Most of the detected cloud tops are found in heights between 2000 to 2500 m.More than 50% of all detected CTHs derived from measurements of both instruments are found in these heights.As already seen in the case study on 11 December 2013, the CTH distribution derived from WALES measurements shows larger values in the uppermost part of the height range 2000-2500 m while the CALIOP derived distributionshows a higher contribution in the lower part of this height range.Furthermore, the CTH distribution derived from both instruments have a local maximum in heights between 1000 and 1250 m, with WALES detecting almost twice as many cloud-tops as CALIOP in this height region (WALES: ∼ 15%; CALIOP: ∼ 7%).The differences in relative frequency of detected CTHs in each bin interval between the two compared data sets never exceed 10%.There are no overall systematic differences in the distributions, although WALES data shows a more pronounced two layer CTH structure.No cloud tops are found in heights between 2750 and 4000 m.

Figure 5
Figure 5(b)  shows the CTH distributions for the whole period of NARVAL in the defined area.As seen in the CTH distribution of the underflights the CTH distributions for all NARVAL measurement flights derived from WALES and CALIOP are in good agreement.Again, a two-layer structure is obvious with maxima in height ranges between 750-1500 and 1750-2500 m.The maximum derived from WALES measurements is found between 2250 and 2500m, while from CALIOP measurements it is found between 2000 and 2250 m.Altogether, about 60% of all CTH are located in height ranges between 1750 and 2750 m for both measurement systems.In the height range between 750 and 1500 m about 21% and 25% of all detected clouds were found from CALIOP and WALES measurements, respectively.

Table 2 .Figure 5 .
Figure 5. Relative frequency distributions of detected cloud top heights measured by WALES and CALIOP -(a) during all conducted CALIPSO underflights and (b) during the whole NARVAL period (spatial boundaries: 60 and 35 • W and 10 and 20 • N).The black lines indicate the differences between the compared profiles for each bin-interval with a bin-size of 250 m.

5
distribution of the CTHs derived from all CALIPSO overpasses in the defined area during the entire NARVAL period.One can see that low clouds are omnipresent over the entire meridional transect from 0 • to 30 • N. Most of the cloud tops are found at heights between 1500 and 2500 m.North of about 20 • N a lowering of CTHs compared to the regions between 10 • to 20 • N where the maximum of the cloud tops is found in higher altitudes, is clearly visible.Between 20 • and 30 • N the maximum in CTH frequency decreases from about 2500 m to about 1500 m.This can be explained by differential heating and therefore 10 deepening of the marine boundary layer towards the tropics.Furthermore, an abrupt transition from high CTH frequency to low CTH frequency is located at approximately 10 • N.This marks a transition to deep convective systems near the ITCZ.High Atmos.Meas.Tech.Discuss., doi:10.5194/amt-2016-333,2016 Manuscript under review for journal Atmos.Meas.Tech.Published: 26 October 2016 c Author(s) 2016.CC-BY 3.0 License.surface air temperatures and subsequent instability in the tropics are preconditions for air parcels to overcome a trade wind inversion and to build up deep convection.

Figure 8
Figure8illustrates the length distribution of cloudless areas along the CALIOP flight paths observed during NARVAL.Most gap lengths have a length of less than 1 km.They make up nearly 60% in analysed WALES data (not shown) and about 45%

Table 1 .
Overview of the conducted research flights during NARVAL (times given in UTC).