Representing about 30 % of the Earth's total cloud cover, upper tropospheric clouds play a crucial role in the climate system by modulating the Earth's energy budget and heat transport. When originating from convection, they often form organized systems. The high spectral resolution of the Atmospheric Infrared Sounder (AIRS) allows reliable cirrus identification, both from day and nighttime observations. Tropical upper tropospheric cloud systems have been analyzed by using a spatial composite technique on the retrieved cloud pressure of AIRS data. Cloud emissivity is used to distinguish convective core, cirrus and thin cirrus anvil within these systems. A comparison with simultaneous precipitation data from the Advanced Microwave Scanning Radiometer – Earth Observing System (AMSR-E) shows that, for tropical upper tropospheric clouds, a cloud emissivity close to 1 is strongly linked to a high rain rate, leading to a proxy to identify convective cores. Combining AIRS cloud data with this cloud system approach, using physical variables, provides a new opportunity to relate the properties of the anvils, including also the thinner cirrus, to the convective cores. It also distinguishes convective cloud systems from isolated cirrus systems. Deep convective cloud systems, covering 15 % of the tropics, are further distinguished into single-core and multi-core systems. Though AIRS samples the tropics only twice per day, the evolution of longer-living convective systems can be still statistically captured, and we were able to select relatively mature single-core convective systems by using the fraction of convective core area within the cloud systems as a proxy for maturity. For these systems, we have demonstrated that the physical properties of the anvils are related to convective depth, indicated by the minimum retrieved cloud temperature within the convective core. Our analyses show that the size of the systems does in general increase with convective depth, though for similar convective depth oceanic convective cloud systems are slightly larger than continental ones, in agreement with other observations. In addition, our data reveal for the first time that the fraction of thin cirrus over the total anvil area increases with the convective depth similarly for oceanic and continental convective systems. This has implications for the radiative feedbacks of anvils on convection which will be more closely studied in the future.
High clouds cover about 30 % of the Earth (e.g.,
In the tropics, where these high clouds are most abundant, they are part of
large mesoscale systems of a characteristic size of tens of thousands km
Within the last decade, numerous studies focused on these mesoscale
convective cloud systems (MCSs). Their structure and life cycle were studied
by using composite techniques applied to satellite imagery and radar
(e.g.,
In addition, organized convection was studied by statistical analysis of
cloud regimes defined by similar cloud pressure and optical depth within grid
cells
In this article, we use infrared (IR) sounder data to study mesoscale deep
convective systems and, more specifically, their horizontal extent and IR
emissivity distribution. The high spectral resolution of IR sounders, in
particular the Atmospheric Infrared Sounder (AIRS) aboard Aqua since 2002 and
the Infrared Atmospheric Sounding Interferometers (IASIs) aboard MetOp since
2006, allows reliable cirrus identification, both from day and nighttime
observations (e.g.,
One of the World Climate Research Programme grand challenges is to determine
the role of convection in cloud feedbacks (
Proxies of convective intensity/strength or convective depth may be given by
vertical updraft (e.g.,
Details on the AIRS cloud retrieval and the construction of UT cloud systems
are given in Sect.
AIRS is a high spectral resolution infrared spectrometer, aboard the
polar-orbiting EOS Aqua satellite with an equatorial crossing at 01:30 and
13:30 LT
The Laboratoire de Météorologie Dynamique (LMD) cloud property retrieval is based on a weighted
Recently, we have developed a modular cloud retrieval code (CIRSs, clouds from
IR sounders;
In this analysis, to facilitate the reconstruction of the UT cloud systems
from the AIRS cloud properties, it is convenient to grid the data, keeping
the statistics and occurrence of the individual cloud types inside the grid,
while the physical parameters (
Before reconstructing the horizontal extent of the UT cloud systems, a
critical question has to be addressed: how do we define UT clouds? Most studies
on tropical MCSs' life cycle and structure used the IR brightness temperature
(
Median and quartiles of cloud IR brightness temperature (red) and
retrieved cloud temperature (blue) as a function of cloud emissivity for high
clouds (
The present cloud system approach, employing cloud altitude (temperature, pressure) and opacity (emissivity), has the advantage of a clear distinction between high and low clouds based on cloud pressure and of thin and thick cirrus, based on cloud emissivity. This is important since, as discussed in the introduction, this new UT cloud system approach aims to explore the horizontal structure of the UT cloud systems, including thin cirrus.
Since the AIRS initial spatial resolution is more adapted to study organized
convection rather than small-scale shallow convection, we revise the
definition of upper tropospheric clouds (i) towards slightly higher clouds
and (ii) by using a tropopause-dependent definition. Hereafter, UT clouds
will be considered as those being at most 250 hPa below the tropopause,
corresponding to a maximum cloud pressure of about 350 hPa and a height of
about 8 km in the tropics. It should be stressed that the standard high
cloud definition of
Average value of cloud emissivity for bins of
Typically, a convective system is composed of an opaque precipitating core
which detrains cirrus in the form of an anvil at the height of neutral
buoyancy (e.g.,
Median and quartiles of maximum (dashed black) and average (solid
black) rain rate from AMSR-E, and average vertical winds (solid red) from
ERA-Interim, as a function of cloud emissivity for high clouds (
To study the horizontal extent of cloud systems, a full spatial coverage is
required. However, in the tropical region where the cloud systems will be
explored, 30
Geographic map of AIRS cloud data for 1 July 2007 at 01:30 LT.
In the course of the study, several questions emerged, such as how many
neighbors to use and what should be the shape of the region for the
neighbors to be included in the interpolation. The reason we draw readers'
attention to these details is due to the irregular gap area shape and size
which vary with latitude. The optimal filling configuration was deduced by
statistically comparing the fractions of each of the UT cloud types in the
grid cells with real data and those with interpolated data, but also by
visually examining geographical maps of cloud types, such as that in the top panel of
Fig.
Once the gaps are filled, we apply a composite technique to reconstruct the
upper tropospheric cloud systems; adjacent grid cells containing UT clouds
and sharing a common side are grouped. The grid cells must contain more than
70 % of UT cloud types within all AIRS measurements in order to be
considered in the procedure. For interpolated grid cells, the threshold is set
slightly lower, to 65 %, as this 5 % difference corrects for an
observed bias in the UT cloud amount of the interpolated areas. To ensure the
spatial continuity of cloud systems, the average cloud pressure difference
between two adjacent grid cells must be lower than 50 hPa; this is a
legitimate value, as it is slightly above the uncertainty of retrieved
To identify opaque areas inside the built UT cloud systems, which potentially
enclose convective core(s), a second grouping is performed. The emissivity
limit for the opaque area definition is set to 0.9. The cloud system is then
considered as a “convective” one when containing at least one grid cell
with
The bottom panel of Fig.
Geographic maps of
We find that upper tropospheric cloud systems cover about 20 % (25 %)
of the tropical band, defined as 30
Fraction of occurrence, coverage and median size for isolated cirrus
systems, systems with one convective core and with multiple convective cores,
over the latitude band 30
Figure
As discussed in the introduction, the impact of UT cloud systems on the
Earth's energy budget depends on their horizontal extent, their emissivity
distribution and the temperature difference between the cloud and its
underlying surface (lower clouds or Earth surface). The latter has been
explored by
Hereafter, we are primarily interested in the horizontal cloud system emissivity structure, rather than in the total coverage over the tropical band; to keep the uncertainties low, we consider only convective cloud systems which are composed of more than 80 % of real data.
Figure
Median values and standard errors of fraction of convective core
(green), thick (magenta) and thin (cyan) anvil as a function of cloud system
size. In red are cloud system size density function distributions for
The composition of a convective system (convective part, thick and thin
anvils) depends on the system life-cycle stage (as illustrated in Fig. 9d
of
Figure
Convective core fraction kernel density estimate (solid line) and histogram for single- (red) and multi-core (blue) systems (AIRS data, 2003–2015).
By stratifying single-core convective systems according to their fraction of
convective area within the cloud system, we explore whether their physical
properties follow an evolution pattern which corresponds to different life-cycle stages. To do so, taking in account the convective fraction
distribution of single-core systems of Fig.
Single-core systems over land and ocean, the former having a fraction of land
convective grid cells over total convective grid cells above 0.5 and the
latter below 0.5, are further separated to early afternoon (PM) and night
(AM) since diurnal variations are expected. The statistics at each
“maturity step” are shown in Fig.
Number of single-core cloud systems in each maturity step separately over the ocean and over land and during the night (AM) and early afternoon (PM) (AIRS data, 2003–2015).
Median values and standard errors of physical properties of
single-core convective systems for the 11 maturity steps defined by
fraction of convective area (1, 0.78, 0.65, 0.55, 0.47, 0.40, 0.34, 0.29,
0.24, 0.19, 0.13, 0.01) separately over the ocean and over land, and during
the night (AM) and early afternoon (PM):
Figure
The minimum temperature of the convective core is the only variable which
does not have a clear behavior. When considering specific regions, like the
three land regions and three ocean regions discussed in
We are interested in studying the relationships between anvil properties and
convection when the systems are mature. Therefore, we are confident in
isolating these systems (according to Fig.
As discussed in the introduction, there are different proxies describing the
convective intensity/strength or convective depth, which might give
insight into different aspects of convection. The level of neutral buoyancy
(LNB), which can be computed from atmospheric soundings, describes the
convective environment and sets the potential vertical extent for convective
development
In the following, we will investigate convective depth only for mature
convective systems, defined according to Fig.
Median and standard error of maximum convective core rain rate as a function of minimum temperature within the convective core for mature single-core systems separately over land (red) and ocean (blue) (AIRS and AMSR-E data, 2003–2009).
The top panel of Fig.
Whereas it is straightforward to determine the minimum temperature within a
single-core convective system, it is more difficult to consider this proxy
for multi-core convective systems. The latter might be composed of several
convective subsystems in different phases of development. Nevertheless, we
build for those systems the average
Median and standard error of horizontal extent versus minimum
temperature within convective core for mature single-core
From Figs.
The top panel of Fig.
Median and standard error of thin cirrus over total anvil area for
mature systems as a function of minimum temperature within convective
core(s), for single-core systems, separately over land and
ocean
We have built upper tropospheric cloud systems, using cloud pressure and emissivity retrieved from 13 years of AIRS observations. These data have been used to investigate the properties of tropical UT cloud systems and, in particular, relationships between the convective depth given by cloud top temperature in the mature stage of a convective cloud system and the properties of the surrounding cirrus anvils.
Median and standard error emissivity within cloud system as a
function of the normalized distance to the convective core. Mature
single-core systems are considered for three classes of convective depth
represented by intervals in
The benefits of the present UT cloud system database compared to other data and methods are that (1) IR sounder data have a large instantaneous coverage and are sensitive to thin cirrus down to an emissivity of 0.1 (0.2 in visible optical depth) during day and night, and (2) our cloud retrieval provides the physical properties of altitude and emissivity decoupled, allowing us to reconstruct the horizontal extent of the UT cloud systems and then to distinguish between deep convective cloud systems and isolated systems and to resolve their emissivity structure, which is essential for determining the radiative feedback of the anvils on convection. For our investigation, we first needed to establish proxies to identify (1) convective cores, (2) mature deep convective systems and (3) the convective depth of a mature convective system. It was demonstrated, using rain rate and large-scale vertical winds, that in the tropics UT opaque clouds with an emissivity close to 1 have a large probability to stem from convection, even though they include probably a part of stratiform rain. Therefore, the cloud emissivity permits us to differentiate convective cores, cirrus and thin cirrus anvils as well as to identify single-core and multi-core convective systems. UT cloud systems cover about 20–25 % of the tropics. While the frequency strongly decreases from isolated cirrus towards multi-core convective systems, the latter's coverage is the largest. The fractional area of the convective core within a cloud system has already been proven to be a maturity stage proxy. Though considering only two measurements per day, the evolution of properties of single-core convective systems could still be statistically followed by using convective fraction within a cloud system as a proxy for maturity since our results are compatible with findings using a better temporal resolution. The size of the convective core reaches a plateau and then decreases during the stage of dissipation, guiding us to define mature convective systems as those with a convective core fraction between 0.1 and 0.3.
Several proxies of convective intensity/strength or depth exist, giving insight into different aspects of convection. With our data, we could probe mature convective cloud systems' characteristics with respect to the convective core minimum temperature, a variable indicative of the convective depth. It could be shown that colder convective systems (meaning those rising higher) have larger values of maximum rain rate within the convective core (a tendency more marked over land), as well as larger cirrus anvils (a tendency more marked over the ocean). Both findings are in agreement with previous studies. Compared to other methods, our approach provides the unique opportunity to also study the horizontal emissivity structure within the anvils. It was revealed that the fraction of thin cirrus over the total anvil area increases with increasing convective depth, similarly for oceanic and continental mature convective systems and both for single- and multi-core systems. Regional analyses, besides some observed amplitude variations over the same surface type, confirmed these tendencies. We also demonstrated that with increasing convective depth, the emissivity of the anvil decreases, in general, more sharply with increasing distance to the convective core. This might have important implications for the radiative effects of these systems, in relation to a convection intensity increase in a warming climate.
The above findings are very promising and the observed relationships might provide observational metrics for studying detrainment processes with cloud-resolving models or even climate models, if their spatial resolution is similar to the one of our database, and for constraining parameterizations related to convection and detrainment. Combined with variables derived from other data sets, such as vertical cloud structure and corresponding heating rates, atmospheric humidity, surface temperature, level of neutral buoyancy, vertical and horizontal winds, this database will be the basis to address questions on feedbacks between anvils and convection and on their modulation of the atmospheric circulation, in particular with respect to climate change. Furthermore, Lagrangian transport analysis could be used to indicate the origin of the isolated cirrus systems and to assess the link between convective sources and the air entering the stratosphere. Moreover, when meteorological reanalyses are available at higher spatial and temporal resolution, exploration of lag correlations between variables, such as vertical winds, size of convective core, rain rate and other atmospheric condition variables, could give a better understanding of convection mechanisms.
The AIRS L2 cloud data (2003–2015) which have been used
for constructing the UT cloud systems have been produced at LMD, using the
CIRS retrieval method (Feofilov and Stubenrauch, 2017). These data will be
made available at the French Data Centre AERIS by the end of 2017. An earlier
version of AIRS-LMD cloud data (2003–2009), which is very similar, is
available at
The authors declare that they have no conflict of interest.
This research was supported by the Centre National d'Etudes Spatiales (CNES), the Centre National de la Recherche Scientifique (CNRS) and the European Space Agency (ESA) within its framework of ESA Climate Change Initiative. The authors thank the Earth Observing System AIRS and AMSR-E teams for providing the data. The calculations have been performed at the ClimServ IPSL center. The authors also thank two anonymous reviewers who helped to improve this paper through their thoughtful comments. Edited by: G. Feingold Reviewed by: two anonymous referees