Life Cycle of Stratocumulus Clouds over one Year at the Coast of the Atacama Desert

. Marine stratocumulus clouds of the Eastern Pacific play an essential role in the Earth’s energy and radiation budget. Parts of these clouds off the west coast of South America form the major source of water to the hyper-arid Atacama Desert coastal region at the northern coast of Chile. For the first time, a full year of vertical structure observations of the coastal stratocumulus and their environment are presented and analysed. Installed at Iquique Airport in northern Chile in 2018/2019, three state-of-the-art remote sensing instruments provide vertical profiles of cloud macro- and micro-physical properties, wind, 5 turbulence and temperature, as well as integrated values of water vapor and liquid water. Distinct diurnal and seasonal patterns of the stratocumulus life-cycle are observed. Embedded in a land-sea circulation with a super-imposed southerly wind com-ponent, maximum cloud occurrence and vertical extent occurs at night, whereas minima during local noon. Night-time clouds are maintained by cloud-top cooling, whereas afternoon clouds re-appear within a convective boundary layer driven through local moisture advection from the Pacific. During the night, these clouds finally re-connect to the maritime clouds in the upper 10 branch of the land-sea circulation. The diurnal cycle is much more pronounced in austral winter with lower, thicker and more abundant (5x) clouds than in summer. This can be associated to different SST gradients in summer and winter, leading to a stable, respectively neutral stratification of the maritime boundary layer at the coast of the Atacama Desert in Iquique. vital moisture flora fauna Atacama Desert. The performed observations, for the first time, bring forth a full seasonal cycle of vertically resolved insights into the physical processes of the marine stratocumulus clouds interacting with topography of northwestern Chile. The observations resolve the cloud vertical structure, including cloud classification and microphysics, the turbulent structure of the ABL as well as the temperature. Additionally vertically integrated values of liquid water and water vapor have been recorded - all with a temporal resolution of seconds to minutes.

. Main parameters of the chirps used by the cloud radar. Hmin, Hmax and Hres are minimum, maximum and resolution of the height, vrange and vres are velocity range and resolution, respectively, and intL is the integration time over each chirp. RH. The differences between the four thresholds thus give a measure for the uncertainty for the special situation at the Iquique airport site.
To improve the accuracy of the temperature profiles, so-called boundary layer scans are performed every 15 minutes. These 155 scans are performed in the vertical plane at 70 • azimuth, and consist of 19 elevation angles with nine elevations from 4.2 • to 30 • , a zenith observation and symmetrical elevations on the opposite side. The scan ran thus from a direction to the Oyarbide measuring field towards the ocean. The idea was to eventually investigate differences in stratification towards the ocean and towards the mountains. Unfortunately, the lowest elevation angles were lower than the line of sight towards the cliff and the inland half of the scan could not be analysed meaningfully. Due to this reason, only the scans towards the ocean were used for 160 this enhanced temperature profiling. These boundary layer scans allows derivation of temperature profiles with low uncertainty (<0.5 K RMSE) in the lowest hundred meters increasing further up to the middle troposphere, (1.5 K RMSE, see Crewell and Löhnert, 2007). Vertical resolution of the temperature retrieval decreases also with height such that the inversion appears much broader and weaker than it is in reality. To estimate the height of the inversion we first identify the maximum temperature gradient, fit a second-order polynomial to the three gradient values around and determine the location of the maximum of this 165 polynomial as the height of the inversion.

Doppler Lidar
The Doppler wind lidar deployed is a Streamline XR from HALO Photonics (Worcestershire, Great Britain). It sends laser pulses at a wavelength of 1500 nm at a rate of 10 kHz into the atmosphere, measures the backscattered signal, integrates over one second and derives the along-beam mean Doppler velocity and backscatter coefficient at 30 m spatial resolution.

170
Maximum range is 15 km, but the Doppler signal detection needs sufficiently strong backscatter which typically limits the applicable range to the atmospheric boundary layer with higher aerosol contents. The instrument performed a conical scan at an elevation of 70 • every 15 minutes and at 24 azimuth angles (VAD scan) to derive a vertical profile of the horizontal wind vector based on the method described in Päschke et al. (2015). This scan was followed by a scan in the same vertical plane as the MWR (RHI scan). During the remaining time between scans, it measured vertical velocities at a temporal resolution of one 175 second to characterize the turbulence in the boundary layer. These vertical measurements sum up to 52min 44sec per hour and are used for an uncertainty estimate of the horizontal winds as well as for the boundary layer classification scheme described in Section 2.3.2. Vertical backscatter measurements are used to derive cloud base (Section 2.3.1).

Standard meteorology
Both the MWR and the CR are equipped with a standard weather station (WXT 530,Vaisala,Finland). The station measures 180 air pressure, wind speed and direction, temperature and relative humidity as well as precipitation, providing values every 1 sec (MWR) and 4 sec (CR), respectively. The two stations are mounted at 2.25 m and 1.25 m and thus give information about the near-surface stratification.

Sea surface temperature
We use two global Sea Surface Temperature (SST) datasets (GHRSST, 2008(GHRSST, , 2018   To gain more information about the observed clouds, we have applied the Cloudnet algorithm (Illingworth et al., 2007) which uses zenith observations only of the three remote sensors described above. This algorithm classifies the observed clouds into liquid, mixed phase or ice clouds as well as into a precipitating/non-precipitating class including drizzle detection.  gorithm is based on reflectivity from the CR, brightness temperatures from the MWR and backscatter profiles from the DL.
Additionally, data from the European Centre for Medium-Range Weather Forecasting Integrated Forecast System (ECMWF IFS) provides temperature and wind information throughout the whole atmosphere to be able to discriminate hydrometeor phase. The lidar provides mainly the lowest cloud-base, the MWR information about the presence of liquid water, and the radar about cloud-top and higher cloud boundaries. Doppler velocities from the CR allow identification of falling hydromete-200 ors, e.g. rain or snow. Together with the IFS temperature profiles, it is possible to identify regions of liquid water or ice clouds.
For the cloud frequency of occurrence (FOC) statistics, we focus on warm clouds with bases below 2 km asl. Only clouds with liquid droplets without ice or supercooled droplets are considered. In addition, cases of cloud liquid droplet detection with drizzle are included.

205
In addition to the standard Cloundet classification algorithm, we use the boundary layer classification scheme described by Manninen et al. (2018), which is a Cloudnet add-on product. This scheme can identify the regions in the atmospheric boundary layer below the cloud with significant turbulence and determine the origin of this turbulence. The classification is provided as a function of time and height at a resolution of 3 min and 30 m. It is based on calculations of the turbulent dissipation rate from the Doppler lidar vertical velocities, the skewness of the vertical velocity distribution, the derived horizontal wind speeds and 210 backscatter coefficients as well as the near-surface temperature gradient around 2 m height.
At every height and at every time, six classes of turbulence origin are provided: in cloud (if backscatter at a certain height is above a certain threshold), non-turbulent (turbulent dissipation rate below threshold), cloud-driven (skewness of vertical velocity distribution is negative), convective (unstable stratification at surface and turbulent conditions between the considered height and the surface), wind-shear (wind shear above a certain threshold). If a turbulent layer is neither 'cloud-driven', nor 'convective' nor 'wind-shear' it is assigned 'intermittent'. Additionally, the scheme provides information whether the surface layer is stable or unstable.

Satellite View of a typical Situation
A typical stratocumulus situation off the west coast of northern Chile is depicted in Fig. 2. The ocean on the left side of the 220 scenes is nearly completely covered by stratocumulus clouds while the Atacama, the Andes and the Altiplano to the right are nearly completely cloud free as they are above 1000 m above sea level and thus higher than the MBL.
The whole cloud deck is typically moving along the coast to the north, turning gradually to the northwest as it approaches the Peruvian coast. This movement can be depicted from the two open cell areas above the ocean which appear at different locations in the morning and noon image. From the displacement we can estimate for these scenes a velocity of around 3 m/s.

225
Typical values go up to 7.5 m/s.
In the morning scene, clouds reach inland at some places: to the northwest of Arica, around Antofagasta, especially at the Mejillones Peninsula at Antofagasta (ANF), and at the southern edge of the scene. All these locations lie at lower altitudes and the edge of the stratocumulus identifies where the landscape reaches cloud-top height. At the coast, in the morning, there are some gaps in the cloud deck, some of which can be related to valleys cutting through the coastal mountain range. At these 230 places dry desert air of the nocturnal flow from the Andes can probably reach the ocean and dissolve the clouds (see e.g. Schween et al., 2020). Additionally, to southeast of Iquique, a fog field can be seen inland in the central valley between the coastal mountain range and the slopes of the Andes (compare to Cereceda et al., 2008b;del Rio et al., 2021a;Böhm et al., 2021).
At noon the situation has changed: at some places between Antofagasta and Iquique, larger gaps in the stratocumulus cloud 235 deck opened and reach several tens of kilometers out over the ocean. This typical pattern can be observed nearly every day: between morning and noon these gaps form at the coast and during the day extend further over the ocean. In the afternoon, clouds form again at the coastal cliff and grow over the ocean. As a result the gaps in the cloud deck are closed in the evening or early night and a continuous stratocumulus cloud field extending from the open ocean to the coast has been reestablished.  tion at the site typically no clouds are present. Accordingly we can expect that during the summer months less clouds will be present.

Cloud Occurrence
The Cloudnet target classification scheme is used to investigate the frequency of occurrence (FOC) of warm clouds with cloudbase lower than 2 km asl (Fig. 5). Most clouds occur in winter and spring, when during night nearly 80% of the time clouds are 280 present, while during summer and autumn this reduces to 22% and 43%, respectively. During daytime FOC reduces by about 20%-points meaning nearly no clouds during summer afternoon and somewhat more clouds in autumn and spring. In winter there is nearly no diurnal variation in the FOC of clouds.
The vertical structure of the cloud varies as well (Fig. 6). In general clouds can be found at higher levels in summer than in the other seasons with lowest clouds in winter. But while in summer, and to some extent in autumn, there is no clear cluster 285 where we can expect clouds, during winter and spring they are clearly confined to a height range from 300 m to 1100 m in winter and 600 m to 1200 m in spring. The diurnal course reveals that the reduction of cloud occurrence between noon and afternoon occurs from below. The lower boundary of the region where to expect clouds, increases from morning hours to the afternoon by several hundreds of meters. This is in agreement with the observations of del Rio et al. (2021b) made some km to the south-east at the slopes of the coastal mountains in another year.

290
The same pattern can be observed if we investigate average cloud boundaries (Fig. 7). Over the whole day average cloud-top is at about the same height: 1000 m in autumn, 1100 m in spring and 900 m in winter. In contrast hereto, cloud-base is lowest during night and increases from sunrise until afternoon by about 150 m. An exception from this pattern occurs in summer when cloud-base as well as cloud thickness increases over the day which is more typical for shallow cumulus clouds. Nevertheless, cloud occurrence in summer is low and the variability of the cloud boundaries is larger than cloud thickness, such that these   The diurnal course with a constant cloud-top and a rising cloud-base during daytime indicates that processes within the MBL are the reason for the observed thinning of the cloud deck during daytime. A possible mechanism to explain this could be the absorption of solar radiation during daytime, so that long-wave radiative cooling at cloud-top is largely cancelled out and the formation of subsiding turbulent eddies at cloud-top is inhibited. As a result the sub-cloud layer is not well-mixed, 305 vertical moisture transport from the ocean is at least reduced, water loss by evaporation at cloud-top is not compensated anymore and the MBL becomes drier. A lower water vapor content means a higher cloud-base or even no cloud (Bretherton et al., 2004). However, this cannot explain why the cloud deck starts to dissolve at the coast. Another mechanism could be the

Liquid Water Path
As discussed above (sect. 2.2.2) we use four different retrievals based on different pre-processing of the radiosonde data to derive the liquid water path (LWP). The average difference between the resulting values is small (3 g/m 2 ) but shows a diurnal course with larger values during night (up to 18 g/m 2 ). Nevertheless, these differences are lower than the overall uncertainty of the applied LWP retrieval (20-30 g/m 2 ).

315
In general, average LWP is lowest in summer (average value 20 g/m 2 ) and highest in winter (70 g/m 2 ) (Figs. 9 and 10).
Diurnal courses show a recurring pattern with maximum values during night and a minimum in the early afternoon. Daily   (1050) lowest (950) drizzle occurence (%) lowest (1) highest (18) MBL stratification stable neutral turbulence mostly non-turbulent more turbulent mixing clouds already in the afternoon although temperature is then around its maximum. The upper branch of the sea breeze pushes these clouds over the ocean. Eventually the coastal cloud connects to the maritime stratocumulus deck and after sunrise the 460 process begins again.
We note here that the interaction of the land-sea-breeze with the 'Rutllant cell' (Rutllant et al., 2003) can not inferred from the wind lidar measurements as it is restricted to the aerosol loaded MBL. The Rutllant cell comprises strong winds inland at altitudes around 1000 m above sea level and moisture transport into the hyperarid Atacama desert (Schween et al., 2020). If the cell would extend over the ocean strong wind shear would occur at the top of the MBL and could provide moisture from 465 the MBL to the desert.

Conclusions
This paper presents ground-based remote sensing profiling observations of coastal stratocumulus clouds at the airport of Iquique at the northern coast of Chile. These clouds are a vital moisture supply for flora and fauna in the western part of the Atacama Desert.

470
The performed observations, for the first time, bring forth a full seasonal cycle of vertically resolved insights into the physical processes of the marine stratocumulus clouds interacting with topography of northwestern Chile. The observations resolve the cloud vertical structure, including cloud classification and microphysics, the turbulent structure of the ABL as well as the temperature. Additionally vertically integrated values of liquid water and water vapor have been recorded -all with a temporal resolution of seconds to minutes.

475
Desierto de Atacama, for maintenance and technical support of the instruments during their installation at Iquique airport. We acknowledge the ACTRIS Cloud Profiling Unit (CLU) for providing the dataset in this study, which was produced by the Finnish Meteorological Institute, and is available for download from https://cloudnet.fmi.fi/. We especially thank Ewan O'Connor from FMI to include the instruments in the Cloudnet operational processing and Tobias Marke for setting up the turbulence classification scheme. This work would not have been possible without our technicians Pavel Krobot and Rainer Haseneder-Lind and their efforts in preparation, packing and sending the instru-520 ments, setup and maintenance of the internet connection and support for the way back. We also thank Susanne Crewell and Thomas Rose for assisting during the instrument setup. Finally, the authors are very grateful to the efforts of Mario Mech, whose experienced support in preparing and carrying out the setup of he measurement station was indispensable for the success of the project.