Introduction
During the summer monsoon season in southern West Africa (SWA), stratiform
low-level clouds (LLC) frequently form during the night with a typical
cloud-base height (CBH) of several hundred meters above ground, and cover
extensive areas . Using multiyear (2006–2011)
surface synoptic observations and various satellite products,
presented a first climatology of the LLC
during the wet monsoon season (July–September) in SWA. They found that
shortly after sunset LLC frequently form along the coast of Guinea and spread
farther inland in the course of the night, while LLC are most frequent
upstream of elevated terrain. The LLC reach a maximum northward extent
between 09:00 and 10:00 UTC, with the maximum aerial coverage of
approximately 800 000 km2. Similar results were obtained by
and , who analyzed the
data collected during the African Monsoon Multidisciplinary Analysis (AMMA)
2006 special observing periods (May–October). Due to their persistence until
the early afternoon hours, LLC significantly influence the radiation budget
at the surface , and thus affect
the diurnal cycle of the atmospheric boundary layer (ABL) and regional
climate .
So far, only few studies focused on the analysis of mechanisms and factors
controlling the onset and maintenance of LLC in SWA.
used remote sensing observations at Nangatchori in
central Benin, performed regional simulations for
the whole 2006 monsoon season using the Weather Research and Forecasting
(WRF) model, while performed high-resolution
numerical simulations with the Consortium for Small-Scale Modeling (COSMO)
model for a case study and an area around the city of Savè (Benin).
found the formation of LLC to be related to the
onset of the nocturnal low-level jet (NLLJ) and the same conclusion is
obtained from the model simulations. proposed the
strong wind shear underneath the NLLJ, which leads to the destabilization of
the near-surface layer and increased turbulent upward mixing of cold air and
moisture, to be the major process for the cloud formation at Nangatchori. On
the other hand, modeling results of suggest that
turbulent processes related to the NLLJ are more dominant close to the coast,
while orographically forced lifting on the windward side of mountains is more
important farther inland. The importance of horizontal cold-air advection
with the southwesterly monsoon flow for the formation of LLC was found by
both modeling studies . Finally,
after LLC form, radiative cooling at the cloud top, as well as the vertical
mixing below the cloud, help to maintain the stratus deck. High-resolution
simulations suggest that additional processes could be important for LLC
formation, such as vertical cold-air advection, which is related to
orographically induced lifting as well as to gravity waves, and enhanced
convergence and upward motion upstream of existing clouds
.
While LLC and involved phenomena are the integral part of the West African
Monsoon system, climate models struggle to realistically represent them.
conducted a comprehensive analysis of global
climate models and found positive biases in NLLJ speed, negative biases in
LLC cover, and consequently large overestimation of solar radiation (of up to
90 W m-2). extended the analysis of
to the latest global climate model data sets.
While similar biases are found as in , the authors
have identified differences in subgrid cloud schemes as one of the possible
reasons why models struggle to realistically represent LLC.
Up to now, spatial and temporal investigations of LLC in this region have
been performed based mainly on satellite images, synoptic observations and
few modeling studies mentioned above, while high-quality observational data
sets were rare. Due to these limitations, processes that control the
formation and dissolution of LLC are still not fully understood. Moreover,
understanding of these processes has important practical implications, such
as improving operational forecast and predictions of the West African monsoon
in weather, seasonal and decadal climate simulations
. In order to fulfill this gap,
a comprehensive field campaign was conducted within the framework of the
Dynamics-aerosol-chemistry-cloud-interactions over West Africa (DACCIWA)
project in June and July 2016. The campaign
included ground-based measurements at three supersites in Ghana, Benin, and
Nigeria , radiosonde and aircraft measurements
. presented the
large-scale setting and synoptic and mesoscale weather systems, which
affected the region during the DACCIWA campaign, identifying different phases
of similar meteorological conditions, while gave
an overview of the diurnal cycle of the ABL conditions as well as of the
conditions related to nocturnal LLC at three ground-based supersites.
This study is conducted in concert with the analyses presented by
and . In this study we focus
on the description of the diurnal cycle of LLC and identification of physical
processes and factors that control the formation, maintenance, and
dissolution of LLC during one typical night with undisturbed monsoon
conditions and persistent LLC e.g.,. For this we use measurements performed at the Savè
supersite during intensive observation period (IOP) 8 (7–8 July 2016).
Specifically, we aim at identifying the main factors leading to the relative
humidity (RH) change and LLC formation and assessing the heat budget terms
for different phases during the life cycle of the LLC. Although we show only
one case study, the dominant processes are considered to be representative of
a major part of the DACCIWA campaign, especially for the post-onset phase of monsoon 22 June–20 July 2016,. This
is supported by the analysis of , who perform in a
consistent manner an analysis for 11 IOPs, and we find a good agreement
between these two studies. Additionally, investigate
LLC characteristics (vertical extent, coverage, onset time, and horizontal
distribution and evolution), as well as the intra-night variability of
boundary-layer conditions and processes relevant for LLC formation.
present a detailed statistical analysis of the
characteristics of the LLC and the dynamics in the lower atmosphere for a
41-day period.
This paper is organized as follows: in Sect. 2 a brief description of the
study site, data, and methods used is given. In Sect. 3, the evolution of LLC
is described, followed by the presentation of atmospheric dynamic and
thermodynamic conditions in Sect. 4. In this section we also analyze the RH
tendency and the heat budget. The discussion of results is presented in
Sect. 5, while the main findings are summarized in Sect. 6.
Data and methods
Meteorological measurements at the Savè supersite
In this study we analyze the data collected during the DACCIWA ground-based
measurement campaign, which took place between 14 June and 31 July 2016 at
the Savè (Benin) supersite (Fig. ). The comprehensive and unique
data set of the ground-based campaign consists of remote sensing and in situ
data , which enable the investigation of cloud
characteristics and dynamic and thermodynamic conditions at high temporal and
vertical resolutions . During the campaign, 15 IOPs
were conducted, during which, in addition to normal radiosondes launched at
standard synoptic times, frequent radiosondes were launched at regular
intervals in between the normal radiosondes. The Savè supersite
(8.00∘ N, 2.43∘ E; 166 m above sea level (a.s.l.)) is
located approximately 185 km inland from the coast in a moderately hilly
region which is favorable for LLC formation . The vegetation cover at the site is
characterized by grass and bushes (waist deep). A comprehensive measurement
setup at this supersite allows the detailed investigation of atmospheric
conditions and identification of atmospheric processes relevant to the
observed life cycle of LLC. Meteorological data used here comprise
near-surface meteorological parameters, 30 min averaged turbulence fluxes
and turbulence variables, which were calculated using the TK3.11 software
, and radiation fluxes .
(a) The black rectangle indicates the geographical location
of the DACCIWA area of interest. (b) Topographic map of the
investigation area. The three supersites, including the Savè supersite,
are indicated with black squares. The Accra coastal radiosonde station is
shown with a red circle.
The dynamic and thermodynamic conditions in the ABL were measured with
radiosondes (normal and frequent) and different continuously running active
and passive remote sensing instruments . The radiosondes were launched in
regular intervals of 1.5 h, starting at 17:00 UTC prior to the IOP day
(7 July) until 11:00 UTC on the IOP day. For Benin, local standard time
equals UTC plus 1 h. Additionally, the radiosounding measurements at the
coastal station Accra (Ghana, Fig. ) were performed as part of the
DACCIWA radiosonde campaign . High-resolution
information of flow conditions (wind speed and direction) is obtained from a
sodar (for lower part of the ABL) and an ultra-high frequency (UHF) wind
profiler (above 200 m a.g.l.) measurements. Additionally, Doppler lidar
azimuth scans at 15∘ elevation angle (plan-position indicator, PPI)
provided the information on the three-dimensional radial velocity field by
applying the velocity–azimuth display VAD;
technique. In addition to the retrieval of the mean horizontal wind field
with the VAD method, PPI scans can be used to quantify turbulence
e.g.,. We use PPI scans performed at a 15∘
elevation angle to estimate the variance of radial velocity for each range
ring and project them to the vertical axis. The procedure is as follows:
first we dismiss all range rings affected by clouds and remove outliers and
echos by hard targets. Then we estimate the mean wind speed and direction
using a simplified version of the VAD method . The
variance for each range ring is calculated by averaging the deviation from
the fitted curve and corrected for uncorrelated noise. We obtained the
information on liquid water path (LWP) and integrated water vapor (IWV) from
the microwave radiometer using the retrieval
algorithm provided by the University of Cologne . This algorithm was trained on a set of more than 12 000
radiosonde profiles measured between 1980 and 2014 in Abidjan (Ivory Coast).
These quantities are needed for the radiative transfer model, which we used
to obtain vertical profiles of radiative fluxes, and which is described in
Sect. .
Measurements of LLC characteristics
The cloud characteristics are documented by ceilometer (CBH), cloud radar
measurements (cloud top height, CTH), and infrared (IR) cloud camera (cloud
cover). The CHM15k ceilometer operates at a wavelength of 1064 nm and a
pulse rate of 5–7 kHz. It records the backscatter every 60 s up to
15 km a.g.l. and has a vertical resolution of 15 m. A 35.5 GHz cloud
radar was operated in the vertical stare mode to provide radial velocity and
reflectivity profiles from 150 m to 15 km a.g.l. at a vertical resolution
of 30 m and a temporal resolution of 10 s. The CBH is determined from the
attenuated backscatter coefficient profiles based on a threshold method
(manufacturer Lufft, personal communication, 2016), while the CTH is derived
from the measurements of radar reflectivity of hydrometeors using a threshold
of -35 dBz; i.e., reflectivities larger than -35 dBz are considered
clouds . In addition to this, information on sky
conditions is obtained with the IR cloud camera. This camera takes images of
the fraction of sky every 2 min with an opening angle of
43∘ × 32∘ (i.e., 166 m × 114 m covered
area at 200 m height) and a wavelength range of 7.5–13.5 µm. The
sky images are coded in three colors: red, green, and blue (RGB). The color
of the image depends on the emissivity of the sky and, consequently, on the
temperature. Thus, a red image indicates a relatively warm temperature, while
blue indicates colder temperatures. Based on this, it is possible to
distinguish cloud-free periods, periods with continuous cloud deck, as well
as periods with intermittent (stratus fractus) clouds.
The Spinning Enhanced Visible Infra-Red Imager (SEVIRI) data provided the
spatiotemporal characteristics of LLC in the larger area. The spatiotemporal
distribution and evolution of LLC in the study area is observed with the
SEVIRI sensor, mounted on the geostationary
Meteosat Second Generation satellite system. Information on LLC coverage is
inferred using three channels: the visible at 0.6 µm (VIS),
middle-infrared at 3.9 µm (MIR) and thermal-infrared at
10.8 µm (TIR) at their native resolution of 3×3 km2 (at
nadir) and a repeat rate of 15 min. At night, LLC are illustrated using the
brightness temperature difference of the TIR and the MIR channels, which is a
proxy for cloud droplet size and can thus be used to detect low clouds
(smaller droplets) . The underlying concept is based on the
assumption that the average droplet size within the LLC is smaller than in
higher-altitude clouds. As the emissivity difference of the TIR and the MIR
is dependent on cloud droplet size , the brightness
temperature difference between TIR and MIR has been used frequently to detect
low clouds from various satellite platforms e.g.,. However, during daytime, the channel at 3.9 µm
measures a mixture of outgoing thermal and reflected solar radiation, so this
method does not work after sunrise . Therefore, the
reflectance in the VIS channel is used to illustrate cloud coverage during
daytime. In order to easily distinguish between these two techniques,
different colormaps in Fig. are chosen for the different
techniques. For both, daytime and nighttime, higher-level clouds are masked
out by applying a TIR brightness temperature threshold at 283 K. Based on
observed temperature profiles, this approximates a cloud-top altitude of
2.7 km. It should be noted that edges of mid-level clouds might be missed by
the TIR filter. The satellite imagery shown in this study is meant to be a
purely qualitative representation of LLC occurrence and distribution during
the IOP.
The difference in brightness temperature for spectral channels 10.8
(TIR) and 3.9 µm (MIR) is shown in panels
(a)–(h) for the nighttime period. The range of the color
bar for the nighttime period (a–h) is between -1 and 3.5 K.
Purple color indicates LLC. Time is indicated in the top right corner in each
panel, with panel (a) showing 21:00 UTC on 7 July 2016 and
(l) showing 11:00 UTC on 8 July 2016. For the daytime period
(i–l) the reflectance in the visible channel (0.6 µm) is
shown and the range of the color bar is indicated in the legend. The gray
areas indicate the TIR brightness temperature threshold at 283 K, which
indicates higher-level clouds. The Savè supersite is indicated with a
black circle.
SBDART radiative transfer model
Vertical profiles of radiative fluxes are computed with the Santa Barbara
DISORT Atmospheric Radiative Transfer (SBDART) model. SBDART is a software
tool which computes plane-parallel radiative transfer in clear and cloudy
conditions within the Earth's atmosphere . Thanks to
the rich data set obtained within the field campaign, many of the input
parameters can be specified, allowing for a realistic modeling of radiative
fluxes. For example, based on the radiosonde measurements, vertical profiles
of pressure, temperature, and water vapor are used in the model, while the
microwave radiometer provides IWV values. The standard tropical ozone density
profile is used as input (linearly interpolated to model levels), since this
information is not available from measurements. In total, 65 vertical input
levels are specified, with 50 m resolution in the lowest 2.5 and 1 km
resolution between 3 and 16 km.
The spectral range of radiative flux calculations is selected to correspond
to the measurement range of the near-surface solar radiation instrument,
namely between 0.34 and 2.2 µm in the shortwave range and
4.5–42 µm in the longwave range with spectral resolutions of 0.01
and 0.1 µm, respectively. The solar illumination angles are
computed from the specified day of year, time, and geographic coordinates. A
spectrally uniform albedo equal to 0.2 is set for
the surface reflectance properties. The boundary-layer aerosol type is set to
typical rural, while the vertical optical depth of the boundary-layer aerosol
is defined as a mean daily aerosol optical depth measured on 7 July
(Level 1.5, http://aeronet.gsfc.nasa.gov/, last access: 15 March 2018)
and is equal to 0.36. In the case of cloud presence, cloud properties, such
as cloud-layer range and the optical thickness of the cloud layer, are
specified. The cloud optical thickness is determined from the LWP
measurements and the cloud droplet effective radius. We use the default value
of cloud droplet effective radius of 8 µm as this value is within
the range of aircraft measurements in the area . The phase function model used in cloud layers is Mie
scattering. All other input parameters correspond to the default ones.
Sensitivity tests were also performed in order to inspect how sensitive the
model output is to input parameters, and the configuration which gave the
best agreement of net longwave and shortwave radiation with the near-surface
observations was chosen.
Characteristics of the diurnal cycle of LLC
The SEVIRI-based information about the spatial distribution and temporal
evolution of LLC is shown in Fig. . In the early evening, some
patchy LLC are present in the investigation area and are confined to higher
terrain (Fig. ) of the Atakora Mountains range (Togo) and upstream
of the Oshogbo Hills (Nigeria) (Fig. a, b). After 22:00 UTC, the
first LLC formed southwest, i.e., upstream, of Savè and then extended to
the downstream side by 00:00 UTC (Fig. c, d). At the same time,
the area in the neighboring Nigeria covered with LLC was extending westwards
until the two areas merged into one large area around 01:00 UTC
(Fig. d, e). At about 02:00 UTC, LLC have already extended and
cover a substantial part of the domain (Fig. f), which continues
to grow in the course of the night (Fig. g), so that at 05:00 UTC
LLC cover the large part of the investigation area (Fig. h). After
sunrise at around 06:00 UTC (Fig. i), LLC start slowly to
dissipate (Fig. j, k); however, at 11:00 UTC (Fig. l)
their presence in the domain is still substantial.
(a) Time series of ceilometer backscatter (color) and CBH (black dots) derived from the backscatter profiles.
(b) The reflectivity of hydrometeors obtained by cloud radar
(color), the CBH (black dots), and CTH (open circles)
derived from the cloud radar using a threshold of -35 dBz.
(c) Time series of 30 min averaged LWP and IWV from microwave
radiometer. (d) The RGB coded image of sky conditions obtained by IR
cloud camera. The image shows relative contributions of red, green, and blue
in a given pixel, where the color of the pixel depends on the
emissivity of the sky area and its brightness temperature. The blue colors correspond to clear sky and the
red color indicates LLC.
The ceilometer backscatter measurements at Savè (Fig. a) show
some low- and mid-level clouds (up to 3 km, not shown) present between 18:00
and 22:00 UTC, followed by a cloud-free period. The LLC formed around
midnight, with CBH at around 300 m a.g.l., and the same was observed by
SEVIRI (Fig. d). These LLC are maintained during the rest of the
night and even after sunrise, with CBH at approximately 250 m a.g.l. After
around 08:00 UTC, the CBH rises approximately linearly with time. At first,
the CTH is observed roughly at 500 m a.g.l., indicating on average a 250 m
deep cloud layer, with a period between 01:00 and 03:00 UTC without a clear
cloud radar signal, therefore making it difficult to determine the CTH for
the whole period (Fig. b). After 03:00 UTC, LLC are persistent
until 08:00 UTC, with the CTH roughly constant at 650 m a.g.l. forming a
400 m deep cloud layer. After around 08:00 UTC, the CTH rises linearly as
well. The microwave radiometer measurements of the LWP reveal varying cloud
characteristics during the night (Fig. c). In the first couple of
hours LLC contain less liquid water, most likely due to the combination of a
shallow cloud layer and lower optical thickness, and according to the sky
conditions obtained with an IR cloud camera, they are intermittent as well
(Fig. d). After about 03:00 UTC, the LWP increases considerably
probably because the cloud deepens, while a continuous cloud cover is
observed. A minimum of IWV is observed just prior to the LLC onset and in the
course of the night only a slight increase is observed. The observed
differences in LLC characteristics suggest varying atmospheric conditions;
therefore, we inspect dynamic and thermodynamic conditions at Savè in the
next section.
Atmospheric conditions relevant for the diurnal cycle of LLC
Low-level jet and thermodynamic conditions
We start the investigation of atmospheric conditions during this IOP by
inspecting the horizontal wind field (Fig. ). The large-scale
conditions on this IOP are characterized by an about 1000 m deep monsoon
layer with southwesterly winds and the African easterly jet above
. Note that on this particular IOP the
observed monsoon depth is lower than for the whole DACCIWA investigation
period, which has a median depth of 2 km . The
minimum wind speed is found within the layer between about 1000 and
1500 m a.g.l., which corresponds to the transition layer between the
southwesterly monsoon flow and easterlies above. In the African easterly jet
layer, winds reach a maximum of 15 m s-1 at about 3500 m a.g.l.
see Fig. 4b in. During the afternoon and early
evening, a moderate northwesterly-to-southwesterly flow of 3 m s-1
prevails in the lowest 1500 m (Fig. a). The onset of NLLJ is
observed at 20:30 UTC, with an abrupt increase in wind speed up to a maximum
of 8 m s-1, at a height of 275 m a.g.l. At the time of the NLLJ
onset wind direction changes from westerly to southerly and
south-southwesterly. The height of the NLLJ maximum corresponds to the height
at which LLC form roughly 3.5 h later (Fig. b). Once the clouds
have formed, the NLLJ maximum shifts upwards to the height of around
450 m a.g.l., reaching the maximum speed of 10 m s-1
(Fig. a). A weakening of the NLLJ is seen after 04:00 UTC and the
axis is lifted to around 600–700 m a.g.l. With respect to wind speed close
to the surface, we observe a similar behavior to ,
with the wind speed only slightly increasing above 1.5 m s-1 after the
NLLJ onset (Fig. b). The wind direction becomes less variable
after the arrival of south-southwesterly NLLJ and southwesterly flow persists
in the course of the night and the following morning.
(a) Temporal evolution of the wind speed is shown in color,
while arrows show wind direction obtained from sodar and UHF. Black contours
show linearly interpolated potential temperature measured every 1.5 h by
radiosondes. The red dots show CBH and CTH as shown with open red circles.
(b) Time series of the near-surface 10 min averaged wind speed
(black) and wind direction (blue) measured by energy balance station. The
vertical black lines indicate the beginning of five different phases observed
during this IOP.
The potential temperature isolines show that after the sunset at 18:00 UTC,
a strong cooling of the layer close to the ground occurs, while coincident
with the NLLJ onset this layer becomes deeper and reaches approximately
750 m depth (Fig. a). The period between the onset of NLLJ and
the formation of LLC is characterized by the strongest decrease in
temperature. After the continuous LLC deck has formed around 02:30 UTC, the
temperature is roughly constant below the cloud base as well as within the
cloud layer. The increase in the temperature coincides with the increase in
the CBH after 08:00 UTC due to an evolving convective boundary layer (CBL).
So far we have seen that due to the observed varying atmospheric conditions,
the investigated period can be divided into different phases. The first phase
identified is the period between the sunset (18:00 UTC) and the onset of the
NLLJ (20:30 UTC), when the increasing static stability causes the decoupling
of the mixed layer from the stable surface layer, and this phase is denoted
the stable phase. This phase is followed by the jet phase,
a time period between the onset time of NLLJ and the formation of LLC
(00:00 UTC), which marks the beginning of the stratus phase. The
period of roughly 2.5 h after LLC formation is identified as stratus phase I, since inhomogeneous cloud cover and nonstationary atmospheric
conditions are observed. This is followed by stratus phase II, which
corresponds to the period between 02:30 and 06:30 UTC, with a persistent LLC
deck and quasi-stationary atmospheric conditions. The final
convective phase is associated with growing CBL and is characterized
by increased surface heating and lifting of the cloud base. Note that these
five phases do not occur only on this particular IOP.
found that the same phases can be distinguished for at least 10 other IOPs.
The wind shear underneath the NLLJ causes mechanical production of
turbulence, which is considered to be an important process leading to the LLC
formation .
Figure a shows the absolute values of the gradient Richardson
number (Ri), which we calculate from radiosonde profile measurements
for the 50 m averaged bins according to the following expression:
Ri=gθ∂θ/∂z∂U/∂z2, where g is the acceleration due to gravity,
θ is the potential temperature, and U is the horizontal wind speed.
Generally, turbulence is stronger as the Richardson number is smaller, while
Ri=0.21–0.25 is considered to be a critical Richardson number
below which the flow is fully turbulent. When Ri is above 1, the
flow is considered to be laminar e.g.,. Although two
Doppler lidars were deployed at Savè, we could not obtain reliable
measurements of vertical velocity fluctuations (σw) from the
vertical stare mode observations . However, for an
assessment of turbulence in the nocturnal cloud-free ABL, radial velocity
measurements during PPI scans performed with the scanning Doppler lidar at
Savè can be used too. The standard deviation of the radial velocity
(σrv) measured by Doppler lidar is shown in
Fig. b.
(a) The gradient Richardson number shown in color is
calculated from radiosonde measurements. Black isolines show the vertical
potential temperature gradient in K (100 m)-1 calculated from
radiosonde data. The white dots show CBH and open white circles denote CTH.
(b) Variance of the radial velocity obtained by Doppler lidar
measurements. The black contours show horizontal wind speed (in m s-1)
measured by radiosondes, while black dots show CBH and open black circles
denote CTH. (c) Time series of the near-surface 30 min averaged TKE
(black) and stability parameter (ζ, blue). (d) Time series of
the near-surface sensible heat flux (H0, black) and latent heat flux
(LE0, blue) measured by the energy balance station. The vertical black
lines indicate the beginning of five different phases observed during this
IOP.
The importance of the NLLJ for LLC formation was first reported by
for the SWA region, while
found similar importance for the nocturnal stratus in the Great Plains (USA).
The signature of the NLLJ in the near-surface measurements is expected to be
mostly seen in the TKE and not necessarily in the mean wind speed, as
suggested by in their AMMA study.
Figure a, b show that before the sunset, a CBL is still present at
17:00 UTC, while during the stable phase turbulence decays due to the lack
of mechanically generated mixing since the wind speed is rather low.
Additionally, due to the longwave radiative cooling of the surface after
sunset resulting in a negative sensible heat flux (Fig. d), the
stable ABL develops and the negative buoyancy suppresses vertical mixing in
the ABL, which is evident from low values of the TKE and a high stability
parameter (ζ, Fig. c). The stability parameter is defined as
the ratio of a height z and the Obukhov length
L=-u*3/(kgθ‾w′θ′‾), where u* is
the friction velocity, k=0.4 is the von Kármán constant, and
w′θ′‾ is the kinematic heat flux. This parameter is
traditionally used as a measure of stability in the surface layer. Its
magnitude is not directly related to static stability, but a positive sign
indicates statically stable conditions, and a negative one implies unstable
conditions e.g.,. Simultaneously with the NLLJ onset,
turbulence increases in the upper and lower shear zones of the jet. In the
jet phase, below the NLLJ maximum and at the surface, static stability
decreases, enabling stronger turbulent mixing and increase in TKE
(Fig. a, c), while simultaneously the sensible heat flux decreases
from -10 W m-2 to its maximum value of -16 W m-2 during the
night (Fig. d). This enhanced vertical mixing is likely dominated
by large coherent eddies and leads to the vertical
transport of cold air from the radiatively cooled surface. In the first
couple of hours after the LLC formed, mostly intermittent turbulence
(0.25<Ri<1) is present within the LLC in the shear zones of the
upper part of the jet (Fig. a), while increased turbulent mixing
is evident below the CBH (Fig. a, b), which results in a highly
turbulent ABL. We notice that there is quite good agreement in the
information about the turbulence intensity obtained from different
measurement systems. The profiles of radial velocity variance show higher
values in the lower and upper shear zones of the NLLJ maximum, as well as
below the CBH.
Temporal evolution of relative humidity (color) and potential
temperature in Kelvin (panel a, isolines) and specific humidity in
g kg-1 (panel b, isolines) in the lowest 1.5 km obtained from
radiosonde profiles performed every 1.5 h. The arrows show the horizontal
wind vector from radiosondes. The CBH is indicated with white dots and the
CTH with open white circles. Time series of 10 min averaged RH (c),
temperature, and specific humidity (d) measured by the energy
balance station 4 m a.g.l. The vertical red and black lines indicate the
beginning of five different phases observed during this IOP.
Figure a, b show temporal evolution of RH in combination with
potential temperature and specific humidity. In the stable phase the increase
in RH is confined to the lowest 100 m, with the simultaneous decrease in
temperature and increase in specific humidity. At the surface, temperature
decreases by 4 ∘C, while a small increase in specific humidity
occurs (∼1 g kg-1), which finally leads to an increase in RH
from 70 % to 85 % (Fig. c). We observe the drop in
temperature by 3 ∘C in the period between 20:00 and 00:00 UTC,
while simultaneously RH increases by 10 % in the layer below
700 m a.g.l. We note that simultaneously with the RH increase after
20:00 UTC, an increase in ceilometer backscatter is observed as well
(Fig. ). This is most likely related to the aerosol hygroscopic
growth, i.e., the size and composition of particle change due to their water
vapor uptake . At the beginning of
the jet phase a slight increase in specific humidity (∼0.7 g kg-1 between 20:00 and 21:30 UTC) in the layer up to
1100 m a.g.l. is recorded. This increase in specific humidity occurs
simultaneously with the increased turbulent mixing due to the NLLJ. In the
second part of the jet phase (21:30–23:00 UTC), specific humidity decreased
by 0.6 g kg-1 in the same layer. After saturation has been reached,
LLC form. The unsteady conditions during the subsequent roughly 2 to 3 h
(stratus phase I) are reflected in the RH measurements, with the sonde
released at 02:00 UTC not reaching saturation. At the same time, the
decrease in temperature is accompanied by the decrease in moisture; i.e.,
specific humidity decreased at a rate of 0.5 g kg-1 h-1 below
600 m a.g.l. This suggests that the air mass behind the NLLJ is drier than
the environment at Savè. After 03:00 UTC, the conditions below the cloud
base are quasi-stationary. The cooling of the near-surface layer weakens
after the cloud deck forms, which leads to a near-neutrally stratified
surface layer (ζ≈0, Fig. c) and contributes to
further vertical mixing (Fig. c). After 08:00 UTC, RH starts to
decrease with a simultaneous increase in temperature and specific humidity.
Based on radiosonde measurements, CBH and CTH can be determined by applying
different criteria to RH measurements, such as the criteria described in
. Their criteria detect cloud layers when RH is
larger than 99 %. Comparison of RH profiles with CBH and CTH shown in
Fig. a clearly shows that three radiosonde profiles would indicate
a deeper cloud layer than detected by the cloud radar. While on average there
is a good agreement between the CBH estimates from ceilometer and
radiosondes, RH measurements can suggest a too high CTH due to the
condensation on the sensor even after the sonde has risen above the cloud
top. This issue highlights the advantage of the DACCIWA ground campaign and
the multitude of instruments deployed allowing for the multiple estimates of
certain parameters and their cross-validation.
In the following sections we present the analysis of processes relevant for
the evolution of LLC and assess their relevance during different phases.
Relative humidity tendency
The observed changes in RH are a result of temperature (T) and/or specific
humidity (q) changes, i.e., RH increases due to increase in specific
humidity and/or decrease in temperature. In order to quantify whether the q
or T change has a stronger influence on the RH tendency and consequently on
LLC formation, we determine their respective contributions using consecutive
radiosonde measurements. The terms of the RH tendency equation are derived
from the time derivative of RH=e/es, where e is the
water vapor pressure and es is the saturation water vapor
pressure. In the next step, we incorporate the Clausius–Clapeyron relation
∂es∂T=LvesRvT2 and the definition of
water vapor pressure, e=q0.378q+0.622p, where T is the air
temperature in Kelvin, p is the air pressure in hPa, Lv is the
latent heat of vaporization (2.5×106 J kg-1), and
Rv=461.5 J kg-1 K-1 is the gas constant for water
vapor. Finally, the contribution of absolute values and tendencies of q and
T to RH tendency is calculated according to
∂RH∂t︸(I)=pes0.622(0.378q+0.622)2∂q∂t︸(II)-pesqLv(0.378q+0.622)RvT2∂T∂t.︸(III)
Terms of Eq. () are calculated directly from available
radiosonde measurements of RH, q, T and p. In
Eq. (), term (I) is the observed RH tendency, term (II)
represents the contribution from q change and term (III) from T change.
Term (III) includes the minus sign, which means that a positive value of this
term indicates cooling and vice versa. For the calculation of RH, q and T
tendencies we use soundings released at 18:30 and 20:00 UTC for the stable
phase, at 20:00 and 23:00 UTC for the jet phase, at 23:00 and 03:30 UTC for
the stable phase I, at 03:30 and 06:30 UTC for the stratus phase II and at
06:30 and 11:00 UTC for convective phase. Other quantities in
Eq. () are calculated as averages of all soundings within
each phase. The results are shown in Fig. . For each of the terms
in Eq. () the range of uncertainty is determined by
calculating uncertainty propagation from the known uncertainties of
measurements, which are σT=0.2 ∘C for T and
σRH=2 % for RH.
The observed RH tendency profiles (Term (I) in
Eq. ) obtained from radiosonde measurements for different
phases during IOP 8 are shown in blue. The contributions from the specific
humidity changes (Term (II) in Eq. ) and temperature
changes (Term (III) in Eq. ) are shown in green and red,
respectively. The shading indicates the range of uncertainty of each term
determined based on the uncertainty propagation calculations. The colored
circles show values of different terms of Eq. () obtained
from near-surface measurements. The black circles denote the median
horizontal wind speed (WS) profile for each phase. The mean cloud layer is
indicated in gray shading.
During the stable phase, RH increases in the layer below 300 m a.g.l., with
a maximum of about 4 % h-1 (Fig. a). Above this level up
to roughly 700 m a.g.l., RH is almost constant, while above RH decreases.
The decrease in the temperature is mostly responsible for the increase in RH
below 300 m, while on average there is a small positive moisture
contribution. The median wind speed profile in this phase is less than
3 m s-1 in the lowest 1 km. The large uncertainty for each of the
terms in stable phase is due to the fact that the tendencies are calculated
over a relatively short time period of just 1.5 h. During the next phase,
the layer with a significant increase in RH deepens to about 700 m a.g.l.,
with a maximum rate of 6 % h-1 (Fig. b). The main bulk
of this change is caused by cooling, while moisture change is negligible
during the jet phase. The layer of the maximum change corresponds to the
level of the NLLJ maximum. This is in agreement with results in
, who found that on average cooling is the main
process leading to the increase in RH and saturation, while moistening
contributes only little.
At the end of the jet phase the clouds form, but are intermittent during
stratus phase I, which is characterized by almost constant RH within the
cloud layer, while below the cloud base a small decrease in RH of
-1 % h-1 is recorded (Fig. c). Although we still
observe cooling in the lowest 1 km during this phase, the rate is much lower
compared to the stable and jet phases. At the same time a competing, stronger
negative contribution of the specific humidity change is observed. The median
wind speed in the jet layer has even increased to 9 m s-1, with the
NLLJ maximum shifting upwards to the cloud top. As the LLC deck grows and
becomes thicker, the NLLJ maximum shifts further upward towards the cloud top
during stratus phase II (Fig. d). As this phase lasts even after
sunrise, the daytime heating causes the weakening of the jet wind speed, and
we observe a stronger temperature increase above the CTH than below the CBH.
On average, a small positive RH tendency exists below the NLLJ maximum (about
0.5 % h-1), mostly due to a positive q tendency. In the
convective phase, the temperature continues to increase below 700 m a.g.l.,
and has a stronger contribution to negative RH tendency, while the q
tendency is small (Fig. e). The analysis of RH tendency shows that
on average temperature changes are more pronounced than moisture changes,
especially in the period prior to LLC formation as well as after sunrise.
Therefore, in the next step we investigate processes responsible for the
observed temperature changes.
Heat budget analysis
Since the contribution of the temperature change is more dominant, i.e.,
cooling is the dominant process for LLC formation, we investigate in more
detail the heat budget during this IOP. The conservation equation for the
mean potential temperature (θ), with the molecular term neglected, is
equal to e.g.,
∂θ‾∂t︸(I)=-u‾∂θ‾∂x+v‾∂θ‾∂y+w‾∂θ‾∂z︸(II)+1ρcp∂Qj*∂z︸(III)-LvEρcp︸(IV)-∂w′θ′‾∂z︸(V),
where cp=1004 J kg-1 K-1 is the specific heat at constant air
pressure, ρ is density of the air, Qj* is the net radiation flux,
E is the mass of water vapor per unit volume per unit time being created by
a phase change from liquid or solid to gaseous and w′θ′‾ is
the kinematic heat flux. We use radiosoundings in the same manner as in the
previous section to calculate the potential temperature tendency (term I).
The advection term (II) is considered here as a residual term since we can
not calculate this term for each phase, but is estimated for a specific
period in Sect. . The radiative flux divergence term
(III) is determined using the radiative transfer (SBDART) model
. The latent heat release term (IV) is relevant in
the case when clouds are present, and is determined from the LWP measurements
(Fig. c), assuming that the liquid water content is linearly
distributed over the cloud layer depth (h), therefore, the phase change
term equals LvEρcp=-Lvcpρh∂(LWP)∂t. Finally, the divergence of sensible heat flux
(V) is determined using the mean surface values of sensible heat flux
(H0, Fig. d) for the respective phase and assuming linear
decrease up to the top of the inversion layer (stable phase), to the NLLJ
maximum (jet phase) or to the CBH during the nighttime conditions (stratus
phase I and II). For the daytime conditions, we analyzed measurements of
turbulent fluxes obtained by unmanned arial system (UAS) ALADINA
in order to get the insight
into their characteristics. Since the flight times of the UAS do not
correspond to the time period of the convective phase, it is not possible to
include them in the analysis directly. However, the analysis of 20 flights
during the morning hours on 8 different days (not shown) indicates that it is
reasonable to assume that sensible heat flux decreases linearly with height
and equals -0.2H0 at the CTH e.g.,.
Vertical profiles of heat budget terms: potential temperature
tendency (blue), phase change term (green), radiation flux divergence term
(red), divergence of the sensible heat flux (orange), and residual term
(black) shown for different phases during IOP 8. The shaded gray area
indicates the mean cloud layer.
Figure shows the vertical profiles of heat budget terms for the
five phases. The strong cooling of the layer below 300 m a.g.l. during the
stable phase leads to the formation of the stably stratified nocturnal ABL. A
large part of the observed cooling is due to the surface longwave radiative
flux divergence with a maximum cooling rate of -0.22 K h-1. When
vertically averaged up to 275 m a.g.l., this term explains 29 % of the
observed temperature change, while contribution from sensible heat flux
divergence is only 7 %. The residual term is found to be the largest,
with 64 % contribution to the observed temperature change during this
period (Fig. a). The large residual term is most likely caused by
the cold pool outflow, which resulted from the early evening local rainfall
event which occurred approximately 15 km south of Savè and moved westward
during its life cycle (between 19:00 and 21:00 UTC), as revealed by X-band
radar data (not shown). These results are in general agreement with findings
by , who found that the strongest radiative flux
divergence is observed in the early evening under weak-wind and clear-sky
conditions, which may contribute even up to 48 % to the observed cooling
in the lowest 48 m a.g.l.
After the arrival of the NLLJ, the layer of the strongest cooling deepens to
700 m a.g.l., with the maximum cooling rate of -1.2 K h-1 at the
height of the NLLJ maximum (Fig. b). Below the NLLJ maximum, the
contribution of longwave radiation, which is still active during this
cloud-free period, and of sensible heat flux divergence to the observed
cooling, is approximately equal, i.e., 16 %. Based on Ri and
radial velocity variance values (Fig. a), there is evidence of
increased turbulent mixing below the NLLJ maximum, suggesting upward
turbulent transport of cold air leading to cooling and an increase in RH in
this layer (Fig. a, b). However, the contribution of longwave
radiation and sensible heat flux divergence (due to turbulent mixing) to the
observed cooling below the NLLJ maximum is substantially lower compared to
the 68 % contribution from the residual term. We assume that cooling due
to the horizontal cold-air advection, associated with the onset of NLLJ, is
most likely the reason for the strong decrease in temperature during the jet
phase and is considered in more detail in the next subsection. These results
are in general agreement with results obtained for 11 different IOPs by
; i.e., the average 22 % radiative flux divergence
contribution to the cooling during the stable and jet phases is in agreement
with their results. Note that direct comparison of magnitudes is not advised
since the methods applied differ slightly between the studies; therefore,
some differences in the contributions from sensible heat flux divergence and
horizontal advection to the observed cooling are obtained during these two
phases.
During stratus phase I, the observed cooling below the CBH is mostly due to
the strong vertical wind shear, which causes an increase in the sensible heat
flux divergence and contributes 48 % of the observed temperature change
(Fig. c). The contribution of the radiative cooling is similar to
the jet phase and is equal to 13 %. At the same time, the strong
radiative cooling of -1.3 K h-1 at the cloud top helps to maintain
the cloud layer, which consequently evolves into dense stratus clouds by the
end of this phase. The phase change term is positive within the cloud layer
due to condensational heating. During stratus phase II, the atmospheric
conditions below roughly 500 m a.g.l. are quasi-stationary; therefore, no
substantial difference between the different terms is observed. The most
pronounced feature is the strong radiative cooling of -3 K h-1 at
the cloud top (Fig. d). Normally, this strong radiative cooling at
the cloud top leads to entrainment of air and increased turbulent mixing
within the ABL (due to the density differences at the CTH) and, subsequently,
to the development of a strong capping inversion at the cloud top
e.g.,. However, we do not observe the expected
strong temperature inversion at the cloud top. Instead, in our case the
radiative cooling is counter-balanced by the large-scale advection of warmer
air in the layer up to 2 km, leading to the observed weak heating at the
cloud top. This horizontal warm-air advection is accompanied by a wind
direction change (Fig. ).
After sunrise, solar radiation heats the surface, causing positive sensible
heat flux (Fig. d) and evolving CBL. Turbulent mixing due to
buoyancy leads to upward transport of warm air from the surface and warming
of the layer below the CBH, which explains 52 % of the warming of the CBL
(Fig. e). However, due to the fact that sensible heat flux
increases rapidly during convective phase (the mean and standard deviation
are 93 and 45 W m-2, respectively), the estimation of the turbulence
term is associated with high uncertainty. Within the clouds, the phase change
term is negative due to the evaporative cooling.
Horizontal temperature advection
The large residual during the jet phase suggests that the horizontal cold-air
advection related to the NLLJ arrival has an important contribution to the
observed temperature change, which consequently led to the saturation and LLC
formation. In conditions of undisturbed southwesterly monsoon flow,
and observed a frequent
occurrence of a stationary front, which formed along the Guinean coast in the
afternoon and was located several tens of kilometers inland. This front was
reflected in a strong gradient between the relatively cool air mass over the
Gulf of Guinea and warm air over land. Northward propagation of the front
started after 16:00 UTC, i.e., after decay of turbulence in the CBL, and it
reached the Savè region around 21:00 UTC. A similar stationary frontal
structure was seen in the simulations by for the
coast of Mauritania. They related its stationary during the day to a balance
between horizontal advection within the onshore flow and turbulence in the
CBL over the land. Based on previous numerical simulations, as well as the
investigation of conditions along the coast (using radiosonde data) and at
Savè, similar processes are expected to occur along the Guinean coast
during the monsoon season. Specifically, we expect the horizontal cold
advection to be related to the Gulf of Guinea maritime inflow which reaches
Savè in the evening .
Comparison of vertical profiles of horizontal wind speed
(a) and potential temperature (b) in Accra (solid line) and
Savè (dashed line) at 17:00 (red) and 23:00 UTC (green). (c)
The mean horizontal advection estimated between the coast and Savè for
different front locations (50, 75, 100, and 125 km from the coast) during
the 6 h period (17:00–23:00 UTC) using radiosonde measurements in Accra
and Savè (blue), while the error bars show 1 standard deviation. The
potential temperature change observed at Savè during this period is shown
in gray.
Figure a, b show the vertical profiles of wind speed and potential
temperature from radiosoundings at the coast (Accra) and Savè. The
conditions at the coast at 17:00 UTC are characterized by a strong monsoon
flow of 8 m s-1, compared to low winds at Savè. On the other hand,
the conditions at Savè are much warmer, with a well-mixed CBL. Generally,
there are large differences in the conditions between the coast and Savè,
as the coastal station seems to be in the cold maritime air mass, while a
well-developed CBL dominates the conditions at Savè (Fig. a, b).
At 23:00 UTC strong winds in Accra are still present, while the potential
temperature decreased slightly. At the same time conditions at Savè have
changed substantially: the wind profile is now characterized by a pronounced
NLLJ up to 8 m s-1, which is the same as at the coast, and potential
temperature decreased considerably to about 299 K in the layer below
600 m a.g.l.; i.e., it has nearly the same value as at the coast during the
daytime. Based on these considerations we conclude that the front of the Gulf
of Guinea maritime inflow with maritime air mass already passed Savè at
this time. In the layer above 750 m a.g.l. conditions at Savè do not
change considerably during this period.
The estimation of the horizontal temperature advection is based on several
assumptions, which are described in detail in . These
include (i) the assumption of homogeneous temperature distribution along the
coast, (ii) neglecting the zonal wind component, (iii) gradual (linear)
increase in temperature in the south–north direction within the maritime
inflow air mass at a certain distance from the coast (due to the position of
the maritime air mass front), and (iv) constant temperature in the
continental ABL north of the front. Recent modeling studies
indicate that the maximum inland
penetration of the maritime air mass front in the afternoon hours is between
50 and 125 km inland from the coast. Therefore, we estimate the contribution
of horizontal advection to cooling at Savè during the stable and jet
phases, using radiosonde measurements at the coast (Accra) and Savè
(Fig. c). The meridional temperature difference for four different
front locations (50, 75, 100, and 125 km inland) is determined at
17:00 UTC, while the mean (meridional v) wind is the average of
measurements at 17:00 UTC in Accra and 23:00 UTC at Savè. The horizontal
advection (-vΔθΔy-1) estimate indicates a maximum
cooling rate of -2.7 K (6 h)-1 at the height of the NLLJ maximum.
When we compare estimated horizontal advection to the potential temperature
tendency at Savè for the time period 17:00 to 23:00 UTC, above the NLLJ
maximum there is an almost perfect fit (Fig. c), indicating that
cooling in this layer can be explained by horizontal advection. Below the
NLLJ maximum, the contribution of horizontal advection is 55 % to the
observed cooling, which explains the large part of the heat budget residual
term during the jet phase and is in good agreement with the estimated average
contribution from horizontal advection to cooling during the stable and jet
phases for the whole DACCIWA campaign . Since for
IOP 8 radiosonde data are available every 1.5 h, we could determine the
contribution of horizontal advection to the heat budget for the stable and
jet phases separately. Related to the jet phase, which is characterized by
the largest temperature change and the arrival of the NLLJ, the contribution
by cold-air advection accounts for almost 70 %, while its contribution
accounts for 55 % when relating it to the period from 17:00 to
23:00 UTC.
Discussion
Satellite images reveal that during this particular IOP, first LLC form east
of the Atakora Mountains in Togo and upstream of Oshogbo Hills in Nigeria.
The location of LLC confined to mountainous regions suggests that
orographically induced lifting is one important process relevant for their
formation as found in modeling studies of and
. In subsequent hours, the clouds in the region of the
Atakora Mountains extend towards the northeast, i.e., upstream of Savè. The
evolution of LLC at Savè suggests that they are not advected from
southwest, where they form first, but they most likely form due to favorable
atmospheric conditions, such as horizontal cold-air advection, longwave
radiative cooling, and sensible heat flux divergence. During the DACCIWA
campaign, different regions of LLC formation were recorded; i.e., in some
cases the first clouds formed over higher terrain, while for other nights
they seem to be independent of terrain features, and these are presented in
more detail in .
The heat budget analysis suggests that the most relevant process for the LLC
formation is cooling due to horizontal cold-air advection, which is
associated with the arrival of maritime inflow indicated by cold air mass and
southwesterly NLLJ. Although there are large uncertainties related to the
estimation of the horizontal temperature advection ,
we are certain that this process contributes the most to the LLC formation.
Our results suggest that the contribution from horizontal cold advection can
be up to 68 % during the jet phase when cooling is strongest. The
difference in the horizontal advection estimation between this study and
comes from the limitations in the temporal resolution
of radiosondes; i.e., can not perform the heat budget
analysis in a consistent manner since during some IOPs radiosondes between
17:00 and 21:00 UTC are not available. For this reason they need to include
periods when horizontal advection is not active, which results in a lower
contribution of 55 % during the longer time period (stable and jet
phases). The horizontal cold-air advection was also found to be an important
process in numerical simulation .
Our results highlight the importance of having a dense network of
measurements, as well as resolving the atmospheric conditions at high
temporal and spatial resolution in order to adequately quantify all processes
relevant for LLC. For example, quantified the
cooling rate due to horizontal cold-air advection to be
-5 K (12 h)-1. This rate is similar to the observed potential
temperature change between 17:00 and 23:00 UTC at Savè
(Fig. c). However, as seen in this study, the cooling due to the
cold-air advection at Savè occurs during a much shorter period (20:00 to
00:00 UTC, Fig. ), suggesting that our rate of -1 K h-1
during the jet phase gives a more reliable estimation of the horizontal
advection magnitude. This cold air mass behind the maritime inflow is also
drier compared to the continental air mass, thus giving an observational
confirmation of findings from numerical simulations
.
On the other hand, previous observational studies identified upward mixing of
moisture, due to the increased vertical wind shear related to the NLLJ, as
the main process for the cloud formation . Although
the vertical mixing of cold air from the surface layer aloft is found to have
a non-negligible contribution to the overall cooling, which leads to the
saturation and cloud formation, this is not the main process
cf..
In the simulations of and ,
a shift of the NLLJ maximum towards the cloud top was found, which is now
verified by the observations. This upward shift of the NLLJ maximum is found
to be due to the change in the stratification within the cloud layer,
compared to the cloud-free period (jet-to-stratus phase I change). The
wind-shear-induced mixing below the NLLJ maximum (at the height of
275 m a.g.l., Fig. a) reduces the gradient of the potential
temperature, resulting in a less stably stratified layer below the NLLJ
maximum than above. With the cloud formation, vertical gradients in wind
speed and temperature decrease further, causing the thickening of the mixed
layer and consequently the shift of the inversion layer upwards.
found the shift in stratification within the clouds
to be caused by the enhanced TKE, condensational heating, radiative cooling
at the cloud top, and upward motion in the stable ABL. All of the processes
are observed in our study as well, except for the latter, since due to the
uncertain measurements of vertical wind velocity by Doppler lidars we can not
estimate the contribution of upward motion.
Summary and conclusions
The data collected during a comprehensive DACCIWA ground-based field campaign
at the supersite in Savè (Benin) on 7–8 July 2016 are analyzed in order to
investigate the diurnal cycle of LLC and related atmospheric processes. This
particular time period is chosen since the conditions during this case study
reflect typical conditions and features related to undisturbed southwesterly
monsoon flow. These typical features include the onset of NLLJ and the
formation of LLC. The associated dynamic and thermodynamic conditions allow
the identification of five different nocturnal phases. These include the
stable phase indicating the period after the sunset and before the onset of
the NLLJ, when the wind speed in the ABL is low and increasing static
stability causes the decoupling of the mixed layer from the stable ABL. The
jet phase starts with the onset of NLLJ related to the arrival of Gulf of
Guinea maritime inflow. The formation of LLC marks the beginning of the
stratus phase, which is divided in stratus phase I, since the inhomogeneous
cloud cover and nonstationary atmospheric conditions are observed, and
stratus phase II, which corresponds to the period with a persistent stratus
deck and quasi-stationary atmospheric conditions. The convective phase starts
approximately 1 h after the sunrise and is associated with growing CBL and
lifting of the cloud base.
Shortly after the sunset, the stably stratified nocturnal boundary layer
developed due to the contributions of longwave radiative cooling of the
ground (29 %), sensible heat flux divergence (7 %), as well as some
local effects due to cold pool outflow from early evening convection, causing
the decoupling of the mixed layer from the stable surface layer (stable
phase). Within the stable ABL, RH increased with a maximum rate of about
4 % h-1 mostly due to strong cooling in this layer. The jet phase is
characterized by the southwesterly NLLJ, which causes an increase in wind
speed of up to 8 m s-1 and the NLLJ maximum height at about
275 m a.g.l. The strong wind shear below the NLLJ causes increased vertical
turbulent mixing and consequently the erosion of the surface inversion,
leading to the coupling of the residual and surface layer. The effect of
horizontal cold-air advection, related to the Gulf of Guinea maritime inflow,
which brings the cold maritime air mass and a prominent NLLJ wind profile, is
found to have the dominant role in the observed strong cooling during the jet
phase. The residual term of the heat budget is considered to correspond to
the horizontal temperature advection term, and we find that it can contribute
up to 68 % to the observed temperature change below the NLLJ maximum
during the jet phase. The contribution from both radiative cooling and
sensible heat flux divergence is 16 %, respectively. The cooling at a
rate of -1.2 K h-1 persists for approximately 3 h, causing the
continuous increase in RH at a rate of 6 % h-1, until finally the
saturation is reached and LLC form at the height corresponding to the NLLJ
maximum height. No significant contribution from the moistening is found
during this phase. The nonstationary conditions during stratus phase I are
observed due to competing influences of processes leading to cooling, namely
turbulent mixing and cold-air advection, and a dry air advection, related to
the drier maritime air mass behind the maritime inflow front, thus leading to
inhomogeneous and thin cloud cover. A combining effect of the vertical wind
shear below and above the NLLJ maximum and the presence of LLC leads to the
change in stratification, causing lower static stability in the sub-cloud
layer and higher at the cloud top, which in turn results in an upward shift
of the NLLJ maximum (from 275 to roughly 400 m a.g.l.). This shift of the
NLLJ maximum towards the layer of maximum static stability continues in
stratus phase II as quasi-stationary conditions are established. Turbulent
mixing is an important factor leading to the cooling below the cloud base,
while strong radiative cooling at the cloud top with a rate of approximately
-2 K h-1 helps to maintain thick stratus. In the morning, the
52 % contribution from sensible heat flux divergence to the observed
heating below the CBH is the largest and, consequently, leads to continuous
warming of the CBL, lifting of the CBH, and dissolution of LLC.
Overall, this study presents the first detailed observational analysis of the
complete diurnal cycle of LLC and processes leading to their formation,
maintenance, and dissolution over southern West Africa. This comprehensive
data set enabled the verification of the previous numerical results, as well
as revealed some new findings enabling better understanding of processes
related to the West African monsoon. This mostly concerns the role of the
maritime inflow and associated horizontal advection of cold but also drier
air for the LLC formation. Furthermore, this detailed analysis of the diurnal
cycle of LLC and related conditions and processes is expected to
substantially contribute to the development of the conceptual model of the
LLC life cycle.