ESSDEarth System Science DataESSDEarth Syst. Sci. Data1866-3516Copernicus PublicationsGöttingen, Germany10.5194/essd-9-881-2017Cloud property datasets retrieved from AVHRR, MODIS, AATSR and MERIS in the framework of the Cloud_cci projectStengelMartinmartin.stengel@dwd.dehttps://orcid.org/0000-0001-5449-0701StapelbergStefanSusOliverSchlundtCorneliaPoulsenCarolineThomasGarethhttps://orcid.org/0000-0002-7341-1420ChristensenMatthewhttps://orcid.org/0000-0002-4273-6644Carbajal HenkenCintiahttps://orcid.org/0000-0002-3408-5925PreuskerReneFischerJürgenDevasthaleAbhayWillénUlrikaKarlssonKarl-GöranMcGarraghGregory R.ProudSimonPoveyAdam C.GraingerRoy G.MeirinkJan Fokkehttps://orcid.org/0000-0001-6682-5062FeofilovArtemhttps://orcid.org/0000-0001-9924-4846BennartzRalfBojanowskiJedrzej S.https://orcid.org/0000-0001-8460-4183HollmannRainerhttps://orcid.org/0000-0001-9222-4025Deutscher Wetterdienst, Frankfurter Str. 135, 63067 Offenbach, GermanyRutherford Appleton Laboratory, Didcot, Oxfordshire, UKInstitute for Space Sciences, Freie Universität Berlin, Carl-Heinrich-Becker-Weg 6–10, 12165 Berlin, GermanySwedish Meteorological and Hydrological Institute (SMHI), Norrköping, SwedenDepartment of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, UKNational Centre for Earth Observation, University of Oxford, Oxford, OX1 3PU, UKRoyal Netherlands Meteorological Institute (KNMI), De Bilt, the NetherlandsLaboratoire de météorologie dynamique (LMD), Paris, FranceUniversity of Wisconsin, Madison, Wisconsin, USAVanderbilt University, Nashville, Tennessee, USAMeteoSwiss, Zurich, SwitzerlandMartin Stengel (martin.stengel@dwd.de)23November2017928819046June201720June201726September20175October2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://essd.copernicus.org/articles/9/881/2017/essd-9-881-2017.htmlThe full text article is available as a PDF file from https://essd.copernicus.org/articles/9/881/2017/essd-9-881-2017.pdf
New cloud property datasets based on measurements from the passive
imaging satellite sensors AVHRR, MODIS, ATSR2, AATSR and MERIS are presented.
Two retrieval systems were developed that include components for cloud
detection and cloud typing followed by cloud property retrievals based on the
optimal estimation (OE) technique. The OE-based retrievals are applied to
simultaneously retrieve cloud-top pressure, cloud particle effective radius
and cloud optical thickness using measurements at visible, near-infrared and
thermal infrared wavelengths, which ensures spectral consistency. The
retrieved cloud properties are further processed to derive cloud-top height,
cloud-top temperature, cloud liquid water path, cloud ice water path and
spectral cloud albedo. The Cloud_cci products are pixel-based retrievals,
daily composites of those on a global equal-angle latitude–longitude grid,
and monthly cloud properties such as averages, standard deviations and
histograms, also on a global grid. All products include rigorous propagation
of the retrieval and sampling uncertainties. Grouping the orbital properties
of the sensor families, six datasets have been defined, which are named
AVHRR-AM, AVHRR-PM, MODIS-Terra, MODIS-Aqua, ATSR2-AATSR and MERIS+AATSR,
each comprising a specific subset of all available sensors. The individual
characteristics of the datasets are presented together with a summary of the
retrieval systems and measurement records on which the dataset generation
were based. Example validation results are given, based on comparisons to
well-established reference observations, which demonstrate the good quality
of the data. In particular the ensured spectral consistency and the rigorous
uncertainty propagation through all processing levels can be considered as
new features of the Cloud_cci datasets compared to existing datasets. In
addition, the consistency among the individual datasets allows for a
potential combination of them as well as facilitates studies on the impact of
temporal sampling and spatial resolution on cloud climatologies.
For
each dataset a digital object identifier has been issued:
Satellite-based datasets of geophysical variables are crucial for climate
research as they represent observations of the Earth's climate system, which
can be used for both the analysis of the climate and its variability as well
as guidance for atmospheric model developments. These datasets evolve
periodically by mainly two activities: (1) extending and improving the
underlying radiance record by adding new satellite recordings and applying
new inter-calibration to the entire record, and (2) the development and
application of more advanced retrieval systems and the utilization of
additional or more frequent auxiliary data which often undergo regular
updates themselves.
In the last few decades, activities to process and reprocess global,
high-quality cloud property datasets based on long-term satellite measurement
records have been undertaken with increased effort. The backbone of most of
the multi-decadal climate datasets of cloud properties has been the National
Oceanic and Atmospheric Administration (NOAA) Polar Operational Environmental
Satellites (POES) series. The Advanced Very High Resolution Radiometer
(AVHRR) has been on board the NOAA satellites since the end of the 1970s
(i.e. NOAA-5 and beyond). AVHRR is a passive imaging sensor, where the source
of measured radiation is not emitted by the instrument. Instead, the upwards
reflected solar and emitted thermal radiation is measured at the top of the
atmosphere (TOA). This is done in abutting pixels that assemble a seamless
image. With its four to six spectral channels, AVHRR allows the retrieval of
key cloud properties. AVHRR has been a significant contributor to many global
cloud climatologies, e.g. the International Satellite Cloud Climatology
Project ISCCP;, the
Pathfinder extended dataset PATMOS-x; and
the Climate Monitoring Satellite Application Facility's (CM SAF) cloud,
albedo and radiation dataset CLARA-A1/A2;.
Since the 1990s the National Aeronautics and Space Administration (NASA) and
the European Space Agency (ESA) have launched research satellite missions,
e.g. Terra, Aqua, the European Remote Sensing Satellite (ERS-1/2) and the
Environmental Satellite (Envisat), that carry AVHRR heritage sensors. These
are the Moderate Resolution Imaging Spectroradiometer (MODIS), the
Along-Track Scanning Radiometers (ATSR-1/2) and the Advanced Along-Track
Scanning Radiometer (AATSR), which provide an increased number of spectral
channels as well as higher spatial resolution (≤ 1 km footprint
size) than AVHRR. The cloud datasets derived from these measurement records
cover more than one decade and are thus becoming useful for climate studies.
Examples of related cloud property datasets are the Global Retrieval of ATSR
cloud Parameters and Evaluation GRAPE; for
ATSR/AATSR, the NASA MODIS Collection 5
and Collection 6
.
The MODIS and ATSR/AATSR sensors include the spectral channels of AVHRR but
have additional ones in the visible, near-infrared and, in the case of MODIS,
also in the thermal infrared. However, even when restricted to the AVHRR
heritage channels, their increased spatial resolution as well as their
contribution to increasing the observation frequency motivates their
consideration in climate research, in particular in conjunction with AVHRR.
Most of the aforementioned cloud property datasets have improved over the
years and have now reached quality levels that facilitate qualitative and
quantitative assessments of clouds in the Earth's climate system
e.g., including studies to understand cloud processes and
the evaluation of atmospheric models. However, there is still potential for
advancing such datasets.
A common shortcoming of existing datasets is the absence of uncertainty
information for pixel-level retrievals (Level-2 data) as well as for daily
and monthly averages (Level-3 data). These uncertainties should be derived
using a mathematically sound framework with uncertainty propagation. Another
improvement to cloud property datasets is to ensure that the properties
retrieved using mainly shortwave measurements are radiatively consistent with
those mainly based on thermal infrared measurements. This is known as
spectral consistency and is important to ensure that subsequent simulations
of TOA radiances using these retrieved cloud properties match the measured
radiances in all spectral bands. The same can be inferred for TOA broad-band
fluxes produced using the retrieved parameters. Spectral consistency is not
maintained in existing cloud retrievals e.g.
despite being of particular importance to, for example, studies investigating
the impact of cloud properties and their change on TOA broadband fluxes and
latent heating rates.
The ESA Cloud_cci project covers the cloud component within ESA's Climate
Change Initiative . The overarching aim of the ESA
Cloud_cci project has been the generation of state-of-the-art cloud property
datasets based on European and non-European satellite missions including the
investigation of their synergistic capabilities. This was achieved by
characterizing and advancing measurement records of passive sensors of ESA and non-ESA satellite missions
;
developing physical retrieval systems for cloud properties with spectral consistency over all utilized spectral bands (see above for
definition of spectral consistency and see Sect. 2.1 for the set of the spectral bands that have been utilized for each sensor
considered);
generating multi-decadal global cloud datasets, based on both single sensors and on a synergistic use of multiple sensors, including
uncertainty estimates which are propagated through all processing levels.
The retrieval systems presented in this paper are based on the optimal
estimation (OE) technique e.g. and are used to
derive a set of cloud variables simultaneously using the visible,
near-infrared and thermal infrared measurements. The retrieval systems were
used to generate cloud property datasets spanning the entire available
measurement record from 1982 until 2014. In the first phase of Cloud_cci
project, prototype versions of the datasets (version 1.0) were generated. In
this paper, version 2.0 of the Cloud_cci datasets is introduced by
presenting a concise overview of the most important technical and scientific
aspects. Section gives an overview of the
Cloud_cci datasets. This includes a description of the underlying
measurement records, the retrieval systems used, the cloud variables produced
at different processing levels, and the propagation of the Level-2
uncertainties. In Sect. selected examples of the
datasets are shown and discussed, and Sect. reports
the most important validation results. Section
summarizes the paper.
Composition of the Cloud_cci datasets
The following satellites and sensors were used in Cloud_cci:
AVHRR on board the NOAA POES satellites (NOAA-7, -9, -11, -12, -14, -15, -16, -17, -18, -19) and on board the European Organisation for the Exploitation
of Meteorological Satellites (EUMETSAT) Meteorological operational satellite Metop-A;
MODIS on board NASA's Aqua and Terra satellites;
ATSR-2 and AATSR on board ESA's research satellites ERS-2 and Envisat;
the Medium Resolution Imaging Spectrometer (MERIS), also on board
Envisat.
Considering imaging and orbital characteristics of the sensors processed, six
datasets were compiled as given in Table . For all
datasets digital object identifiers (DOIs) have been established (also given
in Table ). Figure reports
the local Equator-crossing times of all sensors considered.
Overview of all sensors processed in Cloud_cci and their duration
as a function of the daytime Equator-crossing time (AM: ante meridiem, before
noon; PM: post meridiem, after noon). Sensors belonging to the same dataset
are shown in the same colour.
List of Cloud_cci datasets together with the corresponding retrieval scheme, the sensor(s), satellite(s) used and the time period covered as well as the digital object identifiers (DOIs) issued.
AM: ante meridiem, before noon; PM: post meridiem, after noon; CC4CL: Community Cloud retrieval for CLimate; FAME-C: Freie Universität Berlin AATSR MERIS Cloud retrieval; N: NOAA.
The measurement records used
Measurements from the passive imaging sensors AVHRR, MODIS, ATSR2, AATSR and
MERIS sensors were used to produce the Cloud_cci datasets. Each sensor has
different imaging characteristics, such as differences in swath width, which
leads to varying observation frequency for any given position on Earth. All
sensors operate in a sun-synchronous polar orbit. Each individual orbit has
an
ascending and descending part, which roughly corresponds to either daytime or
night-time conditions, which are thus also referred to as daytime and night-time
node. For the MERIS+AATSR dataset, the night-time observations are ignored.
Individual AVHRR and MODIS sensors cover the globe nearly twice a day with
the daytime and night-time nodes of their orbits. Due to their narrow swath
width, ATSR2 and AATSR need 3 days to cover the full globe with daytime
and night-time observations (Fig. ). With
respect to the local Equator-crossing time of the daytime node, the
AVHRR-carrying satellites were separated into AM (a.m. – ante meridiem,
before noon) and PM (p.m. – post meridiem, after noon). In the following
sections further characteristics of the measurement data that form the basis
of the Cloud_cci datasets are summarized.
Daily Level-3U cloud mask composites for 22 June 2008 demonstrating
the spatial coverage of the daytime nodes of selected sensors within 24 h. See Table for definition of Level-3U.
AVHRR
The second and third generations of the AVHRR sensor (AVHRR/2 and AVHRR/3)
provide measurements in two visible, one near-infrared and two thermal
infrared channels with the following (approximate) central wavelengths: 0.6,
0.8, 3.7, 10.8 and 12.0 µm. Exceptions are daytime observations of
NOAA-16 in 2001–2003 and the entire records of NOAA-17 and Metop-A, for
which a 1.6 µm channel was switched on during the day and used
instead of the 3.7 µm channel. The AVHRR swath width is
2399 km, which facilitates full global coverage (daytime and
night-time) twice daily. More information about the AVHRR sensor can for
example be found at https://www.wmo-sat.info/oscar/instruments/view/62.
The Cloud_cci AVHRR-AM and AVHRR-PM datasets are based on
measurements from AVHRR/2 and AVHRR/3 on board prime polar-orbiting NOAA POES
and Metop satellites, i.e. AVHRR-PM: NOAA-7, NOAA-9, NOAA-11, NOAA-14, NOAA-16,
NOAA-18 and NOAA-19; AVHRR-AM: NOAA-12, NOAA-15, NOAA-17 and Metop-A. All
measurements used are global area coverage (GAC) data with a footprint size
of 1.1 km× 4.4 km and a sampling distance of
3.3 km along track and 5.5 km across track between the
centres of the GAC footprints. GAC data are derived from the originally
measured Local Area Coverage (LAC, footprint size:
1.1 km× 1.1 km) data by on-board averaging of four
neighbouring LAC pixels every third scan line. Only the GAC record is
available for AVHRR with global and nearly seamless temporal coverage since
the early NOAA satellites. The two visible channels and the
1.6 µm channel (if present) were intercalibrated using MODIS
Collection 6 measurements (Heidinger et al., 2017; an update of
). For the IR channels no further calibration was
performed beyond the on-board blackbody calibration.
MODIS
MODIS is a 36-channel passive imaging sensor with a swath width of
2330 km. More information on the MODIS sensor is available at
https://www.wmo-sat.info/oscar/instruments/view/296. Out of the
original 36 spectral channels, measurements from 5 of them (the AVHRR
heritage channels, MODIS channel numbers 1, 2, 20, 31, 32) were used to
retrieve cloud properties from MODIS-Terra (in 2000–2014) and MODIS-Aqua (in
2002–2014). The MODIS sensor is calibrated on board. The calibration
approach employs radiometric, spatial, and spectral calibrators and the moon
as reference . MODIS sensors are well known for their high
calibration accuracy. MODIS Level-1b data of the Collection 6 release were
obtained from NASA. The Collection 6 Level-1 data are expected to show
several improvements over Collection 5 data due to improved calibration
methodologies (see for example
http://mcst.gsfc.nasa.gov/calibration/collection_6_info). The footprint
size of the MODIS Level-1 data is 0.25 km× 0.25 km
to 1 km× 1 km (depending on channel); here we used
data at 1 km× 1 km resolution for all channels.
ATSR2 and AATSR
The passive imaging sensors ATSR2 and AATSR have seven spectral channels in
the solar, near-infrared and thermal infrared range with central wavelengths
between 0.55 and 12 µm, of which five (the AVHRR heritage
channels; ATSR2/AATSR channel numbers 2, 3, 5, 6, 7) were used. At nadir the
ATSR2 and AATSR pixel resolution is approximately
1 km× 1 km with a swath width of 500 km.
The sensor is designed to be self-calibrating. Two integrated thermally
controlled blackbody targets are used for calibration of the thermal
channels, while an opal visible calibration target illuminated by sunlight is
used for calibration of visible and near-infrared channels. More information
about the ATSR-2 and AATSR sensors is available at
https://www.wmo-sat.info/oscar/instruments/view/56 and
https://www.wmo-sat.info/oscar/instruments/view/2. Version 3.1 of
the ATSR2 and AATSR Level-1 data was used. The available ATSR2 Level-1 record
covers 1995–2002, while the AATSR Level-1 record covers 2002–2012.
MERIS
MERIS is a passive imaging sensor whose measurements were synergistically
combined with AATSR measurements in this study, making use of the fact that
both sensors are mounted on the same platform (Envisat) but have
complementary spectral characteristics. The pixels of both sensors were
spatially matched. MERIS measurements outside the AATSR swath width
(500 km, which is narrower than the MERIS swath width of
1150 km) were not used. The collocated synergy product has a swath
width of 493 pixels. This is related to collocating the curved AATSR grid
with the MERIS grid. More information on the matching procedure can be found
in . The spatial resolution of the MERIS reduced
resolution mode is 1.2 km× 1 km and thus very similar
to AATSR. More information about the MERIS sensor is available at
https://www.wmo-sat.info/oscar/instruments/view/277. MERIS Level-1 data
of the third reprocessing has been used
(https://earth.esa.int/documents/700255/707222/A879-NT-017-ACR_v1.0.pdf/6fa86bec-9945-4e39-808e-3801f2e3962b).
In addition to the above-mentioned AVHRR heritage channels of AATSR, MERIS
channel 11 (spectrally located in the oxygen-A absorption band around
761 nm) and channel 10 (a nearby window channel located around
753 nm) were used. An empirical stray-light correction was applied to
the reflectance of the MERIS oxygen-A absorption band channel
. For this correction, the spectral smile
effect in the MERIS measurements , which is the
variation in the channel centre wavelength along the field of view, as well
as the amount of stray light in the MERIS oxygen-A absorption band channel,
was determined.
Retrieval systems
Based on a comprehensive comparison to existing cloud property retrieval
systems applicable to passive imaging sensors , the
two Cloud_cci algorithms were further developed and then used to generate
the Cloud_cci climate records. For datasets derived from AVHRR and from the
AVHRR heritage channels of MODIS, ATSR-2 and AATSR, the Community Cloud
retrieval for CLimate CC4CL;
retrieval system was employed. For the MERIS+AATSR dataset, the Freie
Universität Berlin AATSR MERIS Cloud
FAME-C; retrieval system was employed.
Common to both systems is the OE technique, which uses a Levenberg–Marquardt
non-linear inversion method to iteratively fit simulated TOA radiances to the
measured TOA radiances. The ability to include a prior knowledge of the
retrieved quantities is built into the method. The OE technique supports
comprehensive error propagation, allowing measurement error, forward model
error (due to approximations and assumptions, which are made in the modelling
of TOA radiance) and uncertainties in a priori knowledge to be combined to
give a rigorous estimate of the uncertainty on retrieved values on a
pixel-by-pixel basis.
CC4CL and FAME-C were employed to primarily retrieve the following cloud
properties: cloud mask (CMA), cloud phase (CPH), cloud optical thickness
(COT), cloud effective radius (CER) and cloud-top pressure (CTP). From these
properties, cloud-top temperature (CTT), cloud-top height (CTH), liquid water
path (LWP), ice water path (IWP) and cloud albedo (CLA) were also determined.
A short description of all cloud properties is given in
Table . The next two subsections briefly summarize the
main characteristics of CC4CL and FAME-C with in-depth details to be found in
the references given therein.
CC4CL
The CC4CL retrieval system has three main components: cloud detection to
retrieve CMA; cloud typing to retrieve CPH; and an OE retrieval of COT, CER
and CTP. The three components are framed by a pre-processing (e.g.
spatio-temporal mapping of all data fields and clear-sky radiative transfer
simulations) and a post-processing (e.g. merging, consistency check, quality
control). The cloud detection is performed by applying an artificial neural
network (ANN) that uses the AVHRR heritage channel measurements,
illumination, scan angles, and auxiliary data as input. The ANN was trained
to mimic the COT of the Cloud-Aerosol Lidar with Orthogonal Polarization
CALIOP;, which is the main payload of the
Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO)
satellite. Cloudy pixels are identified where the ANN-estimated COT exceeds
pre-defined thresholds (with the remaining pixels classified as clear), thus
defining the CMA. Based on CALIOP data in the ANN training set, a cloud mask
uncertainty is determined. The cloud typing is based on a threshold decision
tree documented in and
. Various cloud types exist for either liquid or
ice phase, which allows the simplification to a binary CPH information. The
central part of CC4CL is an OE estimation of COT, CER and CTP, which is based
on earlier developments of the Oxford RAL retrieval of Aerosol and Cloud
retrieval ORAC; but has undergone
major updates since then. As mentioned above, a cloud model is iteratively
modified to fit the simulated radiances to the measurements. For this, a fast
forward radiative transfer model is included using scalar reflectance,
transmission and emission operators. These operators interact with (1) the
direct beam solar source and/or the diffuse thermal source from above and (2)
both direct and diffuse surface reflectance from a bidirectional reflectance
distribution function (BRDF) surface in the solar channels, as well as diffuse
atmospheric and surface emission in the thermal channels from below. The
operators are a function of the state vector elements COT and CER, and solar
and instrument geometry and compiled in look-up tables (LUTs) precalculated
using the DIScrete Ordinates Radiative Transfer
DISORT; solver. The simulations were done
separately for liquid and ice clouds. Liquid clouds are represented with a
modified gamma distribution and ice cloud single-scattering properties are taken from .
Clear-sky transmittance and radiance profiles are computed using the Radiative
Transfer for Television Infrared Observation Satellite Program (TIROS)
Operational Vertical Sounder (TOVS) RTTOV; model version 11.3 . For each
iteration the above-cloud and below-cloud clear-sky transmittances and
radiances are interpolated from the corresponding RTTOV profiles as a
function of CTP. From the derived state vector variables, the properties CTT,
CTH, CLA, LWP and IWP are inferred with LWP and IWP calculations being based
on for all cloudy conditions. A full list of
retrieved cloud variables is given in Table . The reader
is referred to and for
more details on CC4CL. The CC4CL system was used for the generation of the
Cloud_cci datasets AVHRR-PM, AVHRR-AM, MODIS-Aqua, MODIS-Terra and
ATSR2-AATSR.
FAME-C
FAME-C is an OE-based retrieval system for cloud properties using TOA
radiance measurements from AATSR and MERIS. As an initial step a cloud
detection is performed as described in . This is followed
by the cloud typing procedure of and
, which is additionally simplified to provide
binary liquid-ice information. For daytime pixels identified as cloudy and
assigned with a cloud type, an OE-based retrieval of COT and CER is
performed. The OE retrieval was initially based on developments documented in
. Required LUTs for mapping cloud properties
to visible and near-infrared reflectances were composed through radiative
transfer simulations utilizing the Matrix Operator Model
MOMO;. From the retrieved
COT and CER, LWP and IWP are computed using for
liquid clouds and optically thick ice clouds. For optically thin ice clouds,
the conversion of is applied. Using the retrieved
CPH and COT, a cloud-top temperature retrieval is conducted using the AATSR
thermal infrared channels. Radiative transfer simulations for AATSR are done
using RTTOV version 11.2. The CTT is further converted to CTH and CTP using
collocated numerical weather prediction (NWP) profiles of pressure, temperature and height. These CTP values
are provided as first guess to a second CTP retrieval based on MERIS
measurements in an oxygen-A absorption band channel and a nearby window
channel. The difference in sensitivity of both cloud height retrievals to
different kinds of cloudy situations was analysed in
.
The full list of retrieved cloud properties using the FAME-C system largely
overlaps with the CC4CL output and is thus also given in
Table . The reader is referred to
for more details on the FAME-C. The FAME-C
system was used for the generation of the Cloud_cci MERIS+AATSR dataset.
Product definitions
The full suite of Level-2 cloud properties derived from both retrieval
systems is CMA, CPH, CTP, CTH, CTT, COT, CER, CWP, and CLA (at two wavelengths), where CWP represents either LWP in pixels with liquid clouds or IWP
in pixels with ice clouds. Nearly all of these properties are accompanied by
uncertainty measures that are direct outputs of OE (or derived from them).
Exceptions are CC4CL CMA, for which the uncertainty is empirically estimated
from validation work . Furthermore, CMA from FAME-C and
CPH from both retrieval systems are not accompanied by uncertainty
information yet. Besides Level-2, two additional processing levels exist:
Level-3U and Level-3C, which are explained in Sect. .
List of generated cloud properties. See Sect. for more information on the processing levels Level-2, Level-3U and Level-3C.
VariableAbbreviationDefinitionCloud maskCloud fractional coverCMACFCA binary cloud mask per pixel (Level-2, Level-3U) Subsequently calculated monthly total cloud fractional cover (Level-3C); also separated into three vertical classes (high, mid-level, low clouds) following ISCCP classification of .Cloud phaseLiquid cloud fractionCPHThe thermodynamic phase of the cloud (binary: liquid or ice; in Level-2, Level-3U) The monthly liquid cloud fraction (Level-3C) using the binary cloud phase information.Cloud optical thicknessCOTThe line integral of the absorption and the scattering coefficients along the vertical in cloudy pixels.Cloud effective radiusCERThe projected-area-weighted mean radius of the cloud drop and crystal particles, respectively.Cloud-top pressureCloud-top heightCloud-top temperatureCTPCTHCTTThe air pressure at the top of the uppermost cloud layer – direct output of OE. Height of cloud top, inferred from CTP using ERA-Interim profiles. Air temperature at the cloud top, inferred from CTP using ERA-Interim profiles.Cloud water path (containing ice and liquid water path)CWP (LWP, IWP)The vertically integrated liquid/ice water content in a cloudy column; derived from CER and COT following .Joint cloud property histogramJCHA spatially resolved two-dimensional histogram of combinations of COT and CTP for each spatial grid cell (Level-3C only).Spectral cloud albedo at 0.6 µmSpectral cloud albedo at 0.8 µmCLA_vis006CLA_vis008*The black-sky albedo derived for channel 1 (0.67 µm) and 2 (0.87 µm*), respectively (experimental product).
* For FAME-C, the cloud albedo is derived at 1.6 µm instead of
0.87 µm.
Processing levels of Cloud_cci data products. The footprint refers to the area on the Earth's surface that is covered by one sensor pixel at nadir view.
Processing levelFootprint sizeDescriptionLevel-2 (pixel data)MODIS: 1 km AATSR: 1 km AVHRR: 5 km MERIS+AATSR: 1 kmRetrieved cloud properties at the same resolution and location as the native sensor measurement (Level-1)Level-3U (daily composites)Global equal-angle, latitude–longitude grid with 0.05∘ resolution (MODIS-Europe: 0.02∘)Cloud properties of Level-2 data granules sampled to a global grid without combining any observations from overlapping orbits. Only sampling is done. Common alternative notations for this processing level are Level-2B or Level-2G. Temporal coverage of this product is 1 day.Level-3C(monthly averagesand histograms)Global equal-angle, latitude–longitude grid with 0.5∘ resolutionCloud properties combined (averaged) from a single sensor into a global grid; sampled for the histograms. Temporal coverage of this product is 1 month.
Cloud_cci product portfolio, also featuring day/night and liquid/ice separation for some properties. All products listed exist for each dataset.
* For FAME-C, the cloud albedo is derived at 1.6 µm instead of 0.87 µm.
From pixel-based retrieval data (Level-2) to daily and monthly properties (Level-3)
Level-2 data were the input to the Level-3 processing and underwent a
spatio-temporal sampling and averaging. Level-3U products are global
composites, defined on a latitude–longitude grid at
0.05∘× 0.05∘ resolution. Level-3U fields hold
Level-2 data which were sampled into the Level-3U grid within a 24 h time
window. The most important aspects of this sampling procedure are as follows: (1) only
that Level-2 pixel that has the lowest satellite zenith angle is kept in each
Level-3U grid cell and (2) the actual footprint size of each pixel is
considered (which depends on the sensor and scan angle), which can lead to
more than one Level-3U grid cell being filled by one single Level-2 pixel
observations. The Level-3U composition was done for each day, keeping the
ascending and descending nodes of the orbits in separate fields, which
roughly corresponds to separating daytime and night-time observations. The
Level-3U products also hold a variety of ancillary data information apart
from the retrieved cloud properties. Taking advantage of the high spatial
resolution of the MODIS sensor, additional Level-3U products were produced
for MODIS for a 0.02∘× 0.02∘ grid covering the
European area within 15∘ W to 45∘ E and 35 to
75∘ N (not shown).
Level-3C products are defined on a latitude–longitude grid with
0.5∘× 0.5∘ resolution and hold monthly summaries of
the Level-2 data, such as averages and standard deviation. In addition,
monthly histograms were composed for CTP, CTT, CER, COT, CWP, CLA, each
separated into liquid and ice clouds, and for combinations of COT and CTP
(joint cloud property histogram, JCH). Table summarizes
most important characteristics of all processing levels. The binning of the
Level-3C histograms is given in Table .
Table reflects the available cloud properties for each
processing level.
Due to differences in spatial resolution and swath width between the
considered sensors, the spatio-temporal observation frequency is very
different. The effect of this on monthly scale is demonstrated in
Fig. by the number of daytime observations per
0.5∘ grid cell per month.
Number of daytime observations (pixels) per 0.5∘ grid cell
for June 2008 and (a) AVHRR-PM/NOAA-18, (b)
AVHRR-AM/Metop-A, (c) MODIS-Aqua, (d) MODIS-Terra,
(e) AATSR and (f) MERIS+AATSR. For the sake of
comparability only the daytime number is shown because no night-time
observations are included in MERIS+AATSR. Grey-shaded areas indicate regions
with no daylight and thus without daytime observations.
Propagating the uncertainties
Different metrics can be used to represent the uncertainty of monthly mean
Level-3C products. A simple and often used metric is the standard deviation
σSD (Eq. ) calculated over the same set
of retrieved pixels (xi) that is used for the calculation of the mean
(〈x〉):
σSD2=1N∑i=1N(xi-〈x〉)2,
with N being the number of pixels.
The OE approach provides pixel-based retrieval uncertainties (σi)
that are based on mathematically consistent propagation of the uncertainties
of the input data (e.g. auxiliary data, measurement data, background errors)
into the Level-2 product space (see for example ,
and ). For the Cloud_cci datasets, the Level-2
uncertainties were further propagated into Level-3C products by two measures:
the mean of the pixel-based uncertainties (〈σi〉, Eq. ) and the mean of the squares of the pixel-based
uncertainties (〈σi2〉, Eq. ).
〈σi〉=1N∑i=1N(σi)〈σi2〉=1N∑i=1N(σi2)
With these measures it is possible to include the OE-based Level-2
uncertainties when quantifying both the true, natural variability
(σtrue) of the observed geophysical variable
(Eq. ) and the uncertainty of the calculated mean
(σ〈x〉, Eq. ).
σtrue2=σSD2-(1-c)〈σi2〉σ〈x〉2=1Nσtrue2+c〈σi〉2+(1-c)1N〈σi2〉
These equations assume a bias-free Gaussian distribution for both the Level-2
uncertainties and the retrieved variables. This assumption is inaccurate for
many variables of the presented properties, which introduces some limitations
to the presented approach. Hence, the propagated uncertainties are meant to
be experimental for these dataset versions. Assuming Gaussian and bias-free
distributions, the estimated natural variability represents the standard
deviation (thus the inner 68 % percentile) of the distribution around the
true mean. Furthermore, the estimated uncertainty of the monthly mean
represents the 68 % confidence interval around the calculated monthly
mean.
The given framework was applied to all cloud properties and their OE-based
uncertainties in the generation of the Level-3C products. The results are
discussed using the example of COT from the AVHRR-PM dataset for June 2008.
Figure shows global maps of monthly mean
COT and the corresponding monthly standard deviation (panels a and b) as
calculated from the retrieved Level-2 values. The estimate of the true
variability is shown along with the estimated uncertainty of the calculated
mean (panels c and d) for an uncertainty correlation (c) of 0.1. Panels (e)
and (f) show the results for an uncertainty correlation of 1.0. The exact
correlation is not known and it is likely to have spatial and temporal
variability. Due to this, two very different values have been chosen to
illustrate the sensitivity. In this example, regions with high mean COT tend
to be characterized by high spatio-temporal variability in the underlying
data, which is apparent in the increased standard deviation. An exception is
the northern part of the Atlantic and Pacific oceans, where the standard
deviation is low while the mean COT is high. A noticeable feature is found in
the stratocumulus regions, which are located in the eastern parts of the
subtropical ocean regions. There, the COT is very stable and thus shows low
standard deviations. Another very dominant feature is a band of rather high
mean COT accompanied with high variability, in the storm track regions of the
Southern Ocean.
Now, assuming a rather low uncertainty correlation of 0.1, the estimated true
variability becomes very small (Fig. c).
This is due to the second term on the right side of
Eq. (), in which the mean of the squared retrieval
uncertainties is not significantly reduced when using a correlation of 0.1.
In other words, if the retrieval uncertainties are only slightly correlated,
they contribute to a broadening of the observed COT distribution, causing
only a minor systematic shift of the distribution. In this scenario, the
retrieval uncertainties can explain a large portion of the observed standard
deviation. In some regions the second term of Eq. () is
even larger than the first term (the observed variance), which results in
negative values of the estimated natural variability. Such negative values
are non-physical and indicate an improper correlation in corresponding
regions. They have been set to 0 in the plots. The uncertainty of the mean is
also relatively small for a correlation of 0.1
(Fig. d). This is due to all three terms of
Eq. () becoming small: the third term because of the
relatively large N, the second term because of the low
correlation and the first term because of the low estimated natural
variability, additionally minimized by the division by N. In
other words, having small systematic uncertainties (i.e. a low uncertainty
correlation) leads to a low uncertainty of the mean, which is dominated by
sampling uncertainties decreasing with increasing N.
As a second example, an uncertainty correlation of 1.0 (panels e and f of
Fig. ) is considered. The second term on
the right hand side of Eq. () vanishes in this case,
leading to the estimated true variability being equal to the observed
standard deviation. The uncertainty of the mean is also increased, which is
in contrast to the previous scenario with low correlation. A correlation of
1.0 eliminates the third term of Eq. (). The first term
decreases with larger N, although the natural variability is now
larger than the previous example, leaving the second term dominating the
uncertainty on the mean, which is the arithmetical mean of the retrieval
uncertainties. Since for cloud optical thickness the retrieval uncertainty is
usually highest for clouds with high COTs, the uncertainty of the mean is
highest in regions dominated by such clouds.
The Cloud_cci Level-3C products contain the individual uncertainty terms
σSD, 〈σi〉 and 〈σi2〉 for
each variable in addition to the mean. This allows posterior calculations of
σ〈x〉 and σtrue for any given
uncertainty correlation.
Monthly standard deviation (a) and monthly mean
(b) for cloud optical thickness (COT). Panels (c) and
(d) show the estimated natural variability and uncertainty of the
mean (d) for a correlation of 0.1. Panels (e) and (f)
are the same as panels (c) and (d) but for an uncertainty correlation of
1.0. All data are from AVHRR-PM in June 2008.
Product examples
In this section most Cloud_cci products are discussed using the example of
the AVHRR-PM dataset, i.e. (1) Level-3U data of NOAA-18/AVHRR for
22 June 2008 and (2) Level-3C data for the month of June 2008.
Figure shows CMA, CFC, CPH, liquid cloud fraction,
COT and CER. Figure shows CTP, LWP, IWP and CLA. In
Fig. the JCH is presented in two ways: (1) shown as
a global COT-CTP histogram aggregated over all grid cells and (2) the
relative fraction of a certain subset of clouds, in this case cumulus clouds
according to the ISCCP definition given in : clouds
with CTP larger than 680 hPa and with COT lower than 3.6.
In Fig. time series of selected cloud
properties are shown for monthly, latitude-weighted averages (within a
60∘ S–60∘ N latitude band) of Cloud_cci AVHRR-AM,
MODIS-Terra, ATSR2-AATSR and MERIS+AATSR datasets. All datasets are
relatively stable. However, the time series exhibit some systematic offsets
between the datasets. Though these offsets have not been investigated in
detail yet, it is currently assumed that they are caused by the following
three reasons. (1) Differences in the spectral characteristics of the AVHRR
heritage channels used, i.e. the position and shape of the spectral response
functions, which is only accounted for empirically in cloud detection and
cloud typing schemes. (2) The applied, but maybe still imperfect, calibration
of the measurement records. (3) Differences in the spatial resolution of
AVHRR (footprint size of 1 km× 4 km) compared with
MODIS and AATSR (1 km× 1 km footprint size) may
have a significant impact. Figure shows the
time series for Cloud_cci AVHRR-PM and MODIS-Aqua datasets. Considering the
time series of all datasets, some inhomogeneities are found for AVHRR-AM,
AVHRR-PM. These are mainly due to differences in local observation time of
the individual satellites. A significant portion of this is due to a growing
delay in local observation time with satellite lifetime caused by the drift
of the satellite orbit. For AVHRR-AM, there is an additional jump in local
observation time between the early morning orbits of NOAA-12 and NOAA15 and
the mid-morning orbits of NOAA-17 and Metop-A. Variability in local
observation time means that different parts of a diurnal cycle of clouds are
sampled. Also, the solar zenith angle and the relative azimuth angle between
satellite and sun change with local observation time and can also lead to
inhomogeneities in a time series. Statistical correction methods for
mentioned effects exist e.g. and should be
applied precedent to any trend analysis. The impact of spectral deviation
among the individual AVHRR sensors of a dataset is assumed to have a smaller
impact compared to the satellite drift effect. For ATSR2-AATSR a small jump
in the time series of some cloud properties is found at the sensor
transition. This is primarily driven by differences in the dynamic range of the
3.7 µm channel. The channel saturates more often for ATSR-2. This
is particularly evident for CFC, LWP and IWP.
Another feature in the AVHRR-PM CFC time series
(Fig. ) is worth mentioning. In 1982 (and
onwards) and in 1991 (and onwards) positive anomalies are found which are
related to the major eruptions of the El Chichón (Mexico) and Pinatubo
(Philippines) volcanoes. A first analysis revealed that heavy aerosol
loadings are sometimes detected as clouds. As this seems to be a general
feature of CC4CL and FAME-C, cloud detection and cloud-top properties of all
datasets should be used with caution in heavy aerosol conditions.
Level-3U (a, c, e, g) and Level-3C (b, d, f, h) of
cloud mask/fraction (a–b), cloud phase/liquid cloud fraction
(c–d), optical thickness (e–f) and effective radius
(g–h) for the AVHRR-PM dataset for June 2008. For the Level-3U
examples, the ascending nodes of the orbits are shown, which roughly
correspond to the daylight portions of the orbits of NOAA-18. COT, LWP, IWP
and CLA are only retrieved during daytime conditions. Areas with no valid
retrievals in this day/month are grey-shaded.
As Fig. but for cloud-top pressure
(a–b), liquid water path (c–d), ice water path
(e–f), and spectral cloud albedo at 0.6 µm(g–h)
for Level-3U (a, c, e, g) and Level-3C (b, d, f, h) products. Panels (c)
and (e) both show the Level-3U cloud water path, which represents
liquid water path in liquid cloud pixels and ice water path in ice cloud
pixels.
(a) Joint cloud property histogram, globally aggregated over all
grid cells. (b) Global map of relative occurrence of cumulus clouds
(according to ISCCP definition of with
CTP > 680 hPa and COT < 3.6) with respect to all clouds. Data shown are from AVHRR-PM in June 2008.
Time series of monthly mean cloud properties of Cloud_cci
datasets, with thin lines being (from top to bottom) time series of monthly
mean cloud fraction (CFC), liquid cloud fraction (CPH), cloud-top pressure
(CTP), cloud optical thickness (COT), cloud effective radius (CER), liquid
water path (LWP) and ice water path (IWP), overlaid with a running average
(thick lines). Shown are those Cloud_cci datasets that are based on
so-called morning satellites. The monthly means are calculated for
60∘ S–60∘ N (latitude-weighted). The time series shown
have not been corrected for satellite drift or diurnal cycle.
As Fig. but for Cloud_cci datasets
based on afternoon satellites. The time series shown have not been corrected
for satellite drift or diurnal cycle.
Validation summary
Cloud_cci datasets between 2006 and 2014 were collocated in space and time
with observations from the CALIOP instrument mounted on board the CALIPSO
satellite. The active measurements of CALIOP observations can be considered
as an accurate reference for CMA, CPH and CTH. However, it is important to note
that cloud properties from CALIOP are physically different to that given by
the Cloud_cci products. For CTH, for example, the active sensor detects where
the density of particles sharply increases – the physical cloud-top edge.
Passive sensors observe the entire atmospheric column simultaneously. The CTH
retrieved from passive sensors is an effective average through the cloud-top
layer. As such, it is expected that Cloud_cci CTHs are lower than those
observed by CALIOP, in particular for clouds containing fuzzy
(semi-transparent) cloud layers. To account for this, various metrics have
been considered for the validation of Cloud_cci cloud properties, where the
CALIOP properties have been adjusted with respect to the optical depth
profile.
Summary of cloud mask (CMA) validation results for Cloud_cci
datasets when compared against CALIOP. Validation measures are the
probabilities of detecting cloudy and clear scenes (Hit rate,
PODliquid, PODice; see Appendix
for the definition of these terms) and bias. Also given is the number of
collocated pixels. The scores are separated into two cloud optical thickness
thresholds (COTthres) representing above which CALIOP COT the
CALIOP pixel was classified cloudy.
* Time window used for collocations was ±15 min for
ATSR2+AATSR, MERIS+AATSR and MODIS-Terra. For all others a ±3 min
window was used.
Summary of cloud phase (CPH) validation results for Cloud_cci
datasets when compared against CALIOP. Validation measures are probability of detection (POD) and bias of liquid cloud occurrence.
Also given is the number of collocated pixels. The scores are separated into two cloud optical depth levels (CODlev) representing at which top-down COD into the cloud the reference CALIOP CPH was taken.
* Time window used for collocations was ±15 min for
ATSR2+AATSR, MERIS+AATSR and MODIS-Terra. For all others a ±3 min
window was used.
Summary of cloud-top height (CTH) validation results for Cloud_cci
datasets when compared against CALIOP. Validation measures are standard deviation (SD) and bias. Also given is the number of collocated pixels.
All scores are separated into liquid and ice clouds (both Cloud_cci dataset and CALIOP had to agree on phase) and into two cloud optical depth levels
(CODlev) representing at which top-down COD into the cloud the reference CALIOP CTH was taken.
* Time window used for collocations was ±15 min for
ATSR2+AATSR, MERIS+AATSR and MODIS-Terra. For all others a ±3 min
window was used.
Summary of liquid water path (LWP) validation results for Cloud_cci
datasets when compared against the passive, microwave-based (MW), monthly
mean LWP data of in the period 2003 to 2008. The
validation was performed for three oceanic stratocumulus regions for which
the frequency of ice cloud occurrence is very small: SAF, west of Africa
(10–20∘ S, 0–10∘ E); SAM, west of South America
(16–26∘ S, 76–86∘ W); NAM, west of California
(20–30∘ N, 120–130∘ W) – see text for details. Reported
are the mean LWP of the MW as well as the bias and standard deviation (SD)
for each Cloud_cci dataset with respect to the MW (all values in
g m-2). In addition, the values are given in relative terms in percent in
brackets. The correlation coefficient (r) is also given.
* The mean values given are for the reference data at 10:30
when compared against AVHRR-AM, MODIS-Terra, ATSR2-AATSR and MERIS+AATSR,
and at 13:30 when compared against AVHRR-PM and MODIS-Aqua.
The CAL_LID_L2_05kmCLay-Prov product has been used, which was downloaded
from ICARE Data and Service Center (http://www.icare.univ-lille1.fr).
All collocations are based on searching for the nearest neighbour in the
Cloud_cci Level-3U data to each CALIOP observation. Due to the
similar orbital characteristics of NOAA-18, NOAA-19 (both are part of
AVHRR-PM), Aqua and CALIPSO, a very large set of collocations between these
passive imagers and CALIOP was found with only short temporal mismatches. A
time window of ±3 min was used in these cases. The orbital
characteristics of NOAA-17, Metop-A (both are part of the AVHRR-AM dataset),
Envisat (part of the ATSR2-AATSR and MERIS+AATSR datasets) and Terra deviate
significantly from CALIPSO. Thus, for these satellites the collocation time
window was extended to ±15 min. These collocations are, however, still
limited to the very high latitudes around 70∘ north and south and are thus
occasionally affected by snow and ice as well as low solar-zenith-angle
conditions.
In Table the validation results for CMA are
presented, i.e. the probabilities of correctly detecting cloudy and clear-sky
scenes (Hit rate, PODcloudy, PODclear; see
Appendix for the definition of these terms), the bias
and the number of considered pixels. In a first set-up, the distinction of
clear-sky and cloudy scenes in CALIOP data was made without applying any
cloud optical depth threshold (upper part of
Table ). The hit rates (i.e. the fraction of pixels
that were correctly labelled clear or cloudy by the Cloud_cci CMA with
regard to
CALIOP) range from 73 to 91 %, with the highest values for MODIS-Terra.
The biases range from -13 to -1 % indicating a slight underestimation
of cloudiness, despite the fact that the PODcloudy is
significantly higher than PODclear for all datasets, which can be
explained by the higher frequency of cloudy scenes compared to clear-sky
scenes. Removing optically very thin clouds from the CALIOP data (i.e.
classifying all clouds with optical thickness lower than 0.15 as clear sky in
the CALIOP data) significantly improves the agreement of Cloud_cci data with
CALIOP (lower part of Table ). The hit rates are
mostly increased (except for MODIS-Terra) and the biases are reduced (except
MODIS-Terra and ATSR2-AATSR).
In Table the validation results for CPH are
presented, i.e. the probabilities of correctly detecting liquid and ice
clouds (Hit rate, PODliquid, PODice; see
Appendix for the definition of these terms), the bias
and the number of considered cloudy pixels. As for CMA, two validation
set-ups have been used: in the first the CALIOP cloud phase from the
uppermost reported cloud layer was used (upper part of
Table ), while in the second set-up the CALIOP cloud
phase was taken at that level in the cloud at which the top-down cloud
optical depth is 0.15 or higher (CODlev=0.15; lower part of
Table ). For the first set-up, the probabilities of
detecting the correct phase range from 72 to 78 %, with highest values for
AVHRR-PM and ATSR2-AATSR. All datasets except MERIS+AATSR and ATSR2-AATSR
show a clear liquid bias, meaning an overestimation of the occurrence of
liquid clouds at the cost of ice clouds. A liquid bias can be explained by a
lack of sensitivity of the passive imager retrievals to optically very thin
ice cloud layers above liquid cloud layers. This is supported by the lower
PODice values compared to PODliquid. The hit rates for
phase agreement between Cloud_cci datasets and CALIOP increase for
CODlev=0.15, which is mainly driven by a better detection of ice
phase (increased PODice) while at the same time the detection
efficiency for liquid phase (PODliq) decreases only slightly. The
biases change towards ice (reduced liquid bias or ice bias instead of liquid
bias). The results show that the correct cloud phase determination for
passive imager retrieval is very sensitive to phase changes in the uppermost
cloud layers (e.g. between the physical cloud top and 1 optical depth into
the cloud). In addition to the CPH comparisons presented, the scores were
again calculated including only conditions for which no phase change occurred
in the CALIOP data between the physical cloud top and 1 optical depth into
the cloud (not shown). The agreement between Cloud_cci datasets and CALIOP
data further improves significantly. Hit rates increase by 2 to 10 %,
mainly driven by a much better detection probability for ice clouds.
In Table the validation results for CTH are
presented, i.e. standard and mean deviations. The comparisons were limited to
those collocated pixels for which both CALIOP and the Cloud_cci dataset
report clouds and valid retrievals of cloud phase and cloud-top height. In
addition, the data were restricted to cases where the phase assignment between
CALIOP and the Cloud_cci dataset is in agreement. Removing this uncertainty
in the phase assignment gives a clearer picture of the actual cloud-top
height retrieval, since the first guess of COT, CER and CTP in the Cloud_cci
retrieval systems is a function of the prior determined cloud phase. The
comparisons were separated into liquid and ice cloud conditions and carried
out twice using different top-down cloud optical depth levels
(CODlev) at which the reference CTH was selected from CALIOP
profiles. For liquid clouds the standard deviation between Cloud_cci
datasets and CALIOP is around 1 km and the bias is generally below
0.2 km (except MERIS+AATSR with a bias of 0.79 km). These
values do not change significantly when selecting the reference CALIOP CTH at
CODlev=0.15 (bottom part of Table ),
which indicates that water clouds usually do not have small optical
thicknesses. For ice clouds the agreement with CALIOP is lower as expected.
Standard deviations range from 1.9 to 2.8 km and the bias is
generally negative (thus an underestimation of CTH for ice clouds in
Cloud_cci datasets) between -2.5 and -3.6 km. These negative
biases are reduced to -1.9 to -2.8 km when the CALIOP CTH is
taken at CODlev=0.15. Standard deviations are also reduced by
about 0.2 km for this setting. The agreement of the Cloud_cci ice
cloud CTH to CALIOP further improves with increasing CODlev (not
shown). For example, at CODlev=1.0 the biases for ice cloud CTHs
are decreased to -0.78 to -1.99 km.
LWP retrievals were validated against LWP data derived from satellite-based,
passive microwave (MW) data . Microwave radiation can
normally fully penetrate clouds. Thus, MW measurements can provide a direct
measurement of the total liquid cloud condensate amount. Shortcomings of the
MW data usually exist for scenes with low LWP and scenes with clouds that
also contain large solid (ice) and liquid (rain) particles. Because of this
and because of the different orbital characteristics of the MW-sensor
carrying satellites our validation focused on Level-3C (i.e. monthly
averages) in three stratocumulus regions for which ice cloud occurrence is
very low. The regions are the oceanic area west of Africa at
10–20∘ S, 0–10∘ E (SAF hereafter), the oceanic area west
of South America at 16–26∘ S, 76–86∘ W (SAM hereafter),
and the oceanic area west of California at 20–30∘ N,
120–130∘ W (NAM hereafter). The data have an
accuracy of 15–30 % and contain monthly mean diurnal cycle products,
from which the 10:30 and 13:30 values were taken to match the
Cloud_cci morning and afternoon sensors, respectively. For the validation
scores presented in Table , only the common overlap
period among all Cloud_cci datasets and the MW data were considered (2003 to
2008). The validation scores vary among the Cloud_cci datasets but also
among the three regions under consideration. Considering the correlation
coefficients, the SAF region exhibits the best agreement with the MW for all
Cloud_cci datasets, which might be due to the large seasonal cycle of LWP in
this region. Bias and bias-corrected root mean square errors (bc-RMSEs) do
not differ from the other regions, except for the afternoon satellite
datasets AVHRR-PM and MODIS-Aqua, which show best scores for SAF. The
ATSR2-AATSR and MERIS+AATSR datasets have the largest deviations to the MW
data for all three regions. For all other datasets very small positive or
moderately negative biases are found, and thus a slight underestimation of LWP
compared to MW. Considering the given uncertainty estimates for the reference
data (15–30 %), one can still conclude that there is agreement
between all Cloud_cci datasets and MW in nearly all regions. It is worth
mentioning that the agreement with MW reduces when considering the time period
before 2003 for AVHRR-AM, AVHRR-PM and ATSR2-AATSR. This is mainly due to
problems with the earlier satellites, which can also be seen in the LWP time
series plots of Figs. and
.
Beyond the limited validation results presented in this paper, a
comprehensive effort has been carried out to compare the Cloud_cci datasets
with other, well-established datasets such as PATMOS-x, CLARA-A2 and MODIS
Collection 6 . Their results prove the quality of the
Cloud_cci datasets.
All presented Cloud_cci datasets are freely available.
DOIs have been issued for all datasets (see Table ) with
each DOI landing page containing a brief summary of the corresponding dataset
and a link to the data access page
(http://www.esa-cloud-cci.org/?q=data_download).
Summary
In this paper cloud property datasets generated within the ESA Cloud_cci
project were presented. The datasets are based on passive imager measurements
on board polar-orbiting satellites. The measurement records have been
characterized carefully and, in the case of AVHRR, been re-calibrated. Two
retrieval systems were developed: (1) the Community Cloud retrieval for
CLimate (CC4CL) which was applied to AVHRR as well as to the AVHRR heritage
channels measured by MODIS, ATSR2 and AATSR, and (2) the Freie
Universität Berlin AATSR MERIS Cloud (FAME-C) which was applied to
combined MERIS+AATSR measurements.
Based on these new developments, global cloud climatologies were generated
for all mentioned sensors spanning their entire life time. The datasets are
named: AVHRR-AM, AVHRR-PM, MODIS-Terra, MODIS-Aqua, ATSR2-AATSR and
MERIS+AATSR. The cloud properties derived are cloud mask/fraction, cloud
phase, cloud-top pressure/height/temperature, cloud optical thickness, cloud
effective radius, liquid/ice water path and spectral cloud albedo. The data
is available as pixel-based retrievals (Level-2), globally gridded composites
(Level-3U) and monthly summaries of the cloud properties (Level-3C):
averages, standard deviations and histograms. The OE-based uncertainty
information per pixel (contained in Level-2 and Level-3U) was propagated
into Level-3C data using an introduced mathematical framework
(Eqs. and ). While the main
characteristics of all datasets are very similar to the AVHRR-PM examples
shown, it needs to be noted that some deviations exist. These are mainly
introduced by differences in spatio-temporal observation frequency, remaining
differences in spectral properties among the considered sensors and
differences in retrieval systems, i.e. for MERIS+AATSR dataset.
Level-2 validation of cloud mask, cloud phase and cloud-top height against
CALIOP revealed a probability of correct detection of cloudy and clear-sky
scenes between 70 and 90 % (hit rates), with a general
underestimation of cloud occurrence frequency when compared to all detected
clouds in CALIOP data, which reflects the detection limitation of passive
imagers. Neglecting optically very thin clouds in CALIOP data improves the
agreement in terms of probability of detection and biases. For cloud phase,
hit rates of 70–80 % are reached with a bias towards liquid clouds (except
the MERIS+AATSR dataset) when comparing to the uppermost cloud layer of
CALIOP data. When comparing against the CALIOP phase taken at a top-down
cloud optical depth of 0.15 into the cloud, hit rates increase by
approximately 5 % along with a reduction in the biases. Validating
cloud-top height gives generally small standard deviations and biases for
liquid clouds, while the agreement with CALIOP is lower for ice clouds. For ice
clouds, a strong dependence on the reference level from which the CALIOP
cloud-top height is taken is found. Biases reduce to -0.8 to
-1.99 km when the reference CALIOP cloud-top height is taken at a
top-down optical depth of 1 into the cloud top. Monthly mean liquid water
path was validated against passive, microwave-based satellite data. The mean
and standard deviations are relatively diverse and strongly dependent on the
dataset and region. However, for most Cloud_cci datasets and considered
regions agreement with the reference data within the reported uncertainty of
the reference data (15–30 %) was found.
The validation results presented here, as well as the very comprehensive
Cloud_cci validation report , have proven the
comparability of Cloud_cci datasets with already existing datasets of the
same kind. However, additionally ensuring spectral consistency and adding
rigorously propagated uncertainty measures make the Cloud_cci datasets
distinct from them. The evaluation process of the presented datasets has also
revealed some limitations, of which the most important ones are listed in
Appendix . In the near future, the Cloud_cci
retrieval systems will undergo a revision, e.g. improving the forward models
and LUTs and revising the BRDF for snow and ice surfaces. Based on these
developments the datasets will be reprocessed, also including more recent
time periods. Along with that, the product portfolio will be extended to
include broad-band radiation flux properties at the surface and TOA, which will
allow several new applications such as studying the cloud radiative effect.
Applying the same retrieval system to multiple sensors also facilitates a
combination of the individual datasets, ideally leading to higher quality.
This will be subject to future studies. In addition to such a combination of
the datasets, the consistency among them (i.e. by using the same retrieval
system) will also enable studies that investigate the impact of the imaging
characteristics of the different sensors on derived cloud climatologies.
These imaging characteristics are, for example, the spatial resolution
(1 km× 1 km for AATSR/MODIS vs. 1 km× 4 km for
AVHRR GAC) as well as the observation frequency driven by the sensor swath
width (2399 km for AVHRR vs. 500 km for AATSR).
The calculation of the probabilities of detection (PODs) and the hit rates of
binary events (e.g. clear sky/cloudy, liquid/ice phase) is based on the
entries of the contingency table (Table ).
The POD for a certain event (e.g. event 0) is determined according to
Eq. (). Thus, it is defined as the number of the agreements
with the reference for a certain event, divided by the total number of this
event in the reference data.
POD0=N00(N00+N10)
The calculation of the hit rate is given in Eq. ().
The hit rate is defined as the number of cases in which agreement with the
reference data was found, divided by the number of all cases.
Hitrate0=(N00+N11)(N01+N10+N00+N11)
Known limitations
The most significant limitations of the Cloud_cci datasets, as revealed
during the evaluation, are listed. As not all limitations apply to all
datasets, a mapping table is provided (Table ).
Cloud detection shortcomings (overestimating of CMA and CFC) in conditions of high aerosol loadings, e.g. severe volcano eruptions or (local) heavy desert dust outbreaks.
Cloud detection shortcomings during polar night due to missing visible information and very cold surface temperatures (mainly affecting CMA and CFC).
Shortcomings in cloud detection (affecting CMA and CFC) and optical property retrievals (CER, COT, LWP, IWP, CLA) in regions with high surface reflection of solar radiation,
e.g. in areas of sun glint over ocean or in land areas with snow, ice and desert soil surfaces (high albedo).
Instabilities in the time series (of all cloud variables) due to satellite drift and/or switch in local overpass time. Satellite drift or diurnal cycle correction is required
before using the datasets for trend analysis.
Instabilities in the time series due to switching of near-infrared channels (affecting mainly the retrieval of CER and thus LWP and IWP): (a) during 2 years of NOAA-16 the AVHRR 1.6 µm
channel is switched on during daytime, while for the rest of the AVHRR-PM time series the 3.7 µm channel is measured during daytime, and (b) in the AVHRR-AM time series,
NOAA-12 and NOAA-15 have the 3.7 µm channel measuring during daytime while NOAA-17 and Metop-A have 1.6 µm.
Significant overestimation of CER of ice clouds and IWP due to erroneous composition of radiative transfer look-up tables.
Overestimation of CER and COT for snow/ice surfaces and high solar zenith
angles.
The AVHRR on NOAA-12 and NOAA-15 satellites measure in near-twilight conditions, due to the early morning orbits of these two satellites, for which the retrieval of all cloud properties,
especially the optical properties, is very difficult.
Inconsistencies in the 3.7 µm channel between the ATSR-2 and AATSR affected CPH, CMA and CFC.
Additional errors introduced when converting cloud-top level properties (CTH, CTP and CTT) to each other using potentially incorrect model profiles. However, these errors are assumed to be
significantly smaller than the actual retrieval errors of CTP.
Larger errors in cloud property retrieval (all properties except CMA and CFC) in multi-layer cloud conditions in particular when a high, optically thin ice cloud overlays an optically thick,
lower liquid cloud. See for an attempt to capture cloud properties from multiple cloud layers.
Larger sampling error (affecting all cloud properties) accompanied by artificially increased observed variability due to low observation frequency for some sensors.
Mapping the listed limitations above to the Cloud_cci datasets they apply to.
The authors declare that they have no conflict of
interest.
Acknowledgements
This work was supported by the European Space Agency (ESA) through the
Cloud_cci project (contract no.: 4000109870/13/I-NB). The authors would like
to acknowledge the help of NASA Goddard Space Flight Center in providing
MODIS Collection 6 Level-1 data and the help of NOAA and the University of
Wisconsin for providing the AVHRR GAC Level-1 data and corresponding
intercalibration coefficients for the visible and near-infrared channels of
AVHRR. Edited by: Alexander
Kokhanovsky Reviewed by: two anonymous referees
References
Baran, A., Shcherbakov, V., Baker, B., Gayet, J., and Lawson, R.: On the
scattering phase-function of non-symmetric ice-crystals, Q. J.
Roy. Meteor. Soc., 131, 2609–2616, 2005.
Baum, B. A., Menzel, W. P., Frey, R. A., Tobin, D. C., Holz, R. E., Ackerman,
S. A., Heidinger, A. K., and Yang, P.: MODIS cloud-top property refinements
for collection 6, J. Appl. Meteor. Clim., 51,
1145–1163, 2012.Bourg, L., D'Alba, L., and Colagrande, P.: MERIS SMILE effect characterisation
and correction, ESA Technical Note, available at: https://earth.esa.int/documents/700255/707220/MERIS_Smile_Effect.pdf (last access: 21 November 2017), 2008.Carbajal Henken, C. K., Lindstrot, R., Preusker, R., and Fischer, J.: FAME-C:
cloud property retrieval using synergistic AATSR and MERIS observations,
Atmos. Meas. Tech., 7, 3873–3890,
10.5194/amt-7-3873-2014, 2014.Carbajal Henken, C. K., Doppler, L., Lindstrot, R., Preusker, R., and Fischer,
J.: Exploiting the sensitivity of two satellite cloud height retrievals to
cloud vertical distribution, Atmos. Meas. Tech., 8,
3419–3431, 10.5194/amt-8-3419-2015, 2015.Carbajal Henken, C. K., Preusker, R., Fischer, F., and Hollmann, R.: ESA
Cloud Climate Change Initiative (ESA Cloud_cci) data: Cloud_cci
MERIS+AATSR L3C/L3U CLD_PRODUCTS v2.0, Deutscher Wetterdienst (DWD) and
Freie Universität Berlin (Dataset Producer),
10.5676/DWD/ESA_Cloud_cci/MERIS+AATSR/V002, 2017.
Carro-Calvo, L., Hoose, C., Stengel, M., and Salcedo-Sanz, S.: Cloud Glaciation
Temperature Estimation from Passive Remote Sensing Data with Evolutionary
Computing, J. Geophys. Res.-Atmos., 121, 2169–8996, 2016.
Dee, D., Uppala, S., Simmons, A., Berrisford, P., Poli, P., Kobayashi, S.,
Andrae, U., Balmaseda, M., Balsamo, G., Bauer, P., Bechtold, P., Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N.,
Delsol, C., Dragani, R., Fuentes, M., Geer, A. J., Haimberger, L., Healy, S. B., Hersbach, H.,
Hólm, E. V., Isaksen, L., Kållberg, P., Köhler, M., Matricardi, M., McNally, A. P., Monge-Sanz, B. M.,
Morcrette, J.-J., Park, B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., and Vitart, F.: The ERA-Interim
reanalysis: Configuration and performance of the data assimilation system,
Q. J. Roy. Meteor. Soc., 137, 553–597, 2011.
Enriquez-Alonso, A., Sanchez-Lorenzo, A., Calbó, J., González, J.-A.,
and Norris, J. R.: Cloud cover climatologies in the Mediterranean obtained
from satellites, surface observations, reanalyses, and CMIP5 simulations:
validation and future scenarios, Clim. Dynam., 47, 249–269, 2015.
Eyre, J.: A fast radiative transfer model for satellite sounding systems,
European Centre for Medium-Range Weather Forecasts (ECMWF), Research Dept.
Tech. Memo. 176, available from ECMWF, 1991.
Fell, F. and Fischer, J.: Numerical simulation of the light field in the
atmosphere–ocean system using the matrix-operator method, J.
Quant. Spectrosc. Ra., 69, 351–388, 2001.
Fischer, J. and Grassl, H.: Radiative transfer in an atmosphere–ocean system:
an azimuthally dependent matrix-operator approach, Appl. Optics, 23,
1032–1039, 1984.
Foster, M. J. and Heidinger, A.: PATMOS-x: Results from a diurnally corrected
30-yr satellite cloud climatology, J. Clim., 26, 414–425, 2013.Gómez-Chova, L., Muñoz-Mari, J., Izquierdo-Verdiguier, E., Camps-Valls,
G., Calpe, J., and Moreno, J.: Cloud screening with combined MERIS and AATSR
images, in: 2009 IEEE International Geoscience and Remote Sensing Symposium,
vol. 4, 761–764, 10.1109/IGARSS.2009.5417488, 2009.
Ham, S.-H., Sohn, B.-J., Yang, P., and Baum, B. A.: Assessment of the quality
of MODIS cloud products from radiance simulations, J. Appl.
Meteor. Clim., 48, 1591–1612, 2009.
Hansen, J. E. and Travis, L. D.: Light scattering in planetary atmospheres,
Space Sci. Rev., 16, 527–610, 1974.
Heidinger, A. K., Straka III, W. C., Molling, C. C., Sullivan, J. T., and Wu,
X.: Deriving an inter-sensor consistent calibration for the AVHRR solar
reflectance data record, Int. J. Remote Sens., 31,
6493–6517, 2010.
Heidinger, A. K., Foster, M. J., Walther, A., and Zhao, X.: The pathfinder
atmospheres-extended AVHRR climate dataset, B. Am.
Meteorol. Soc., 95, 909–922, 2014.
Heymsfield, A. J., Matrosov, S., and Baum, B.: Ice water path-optical depth
relationships for cirrus and deep stratiform ice cloud layers, J.
Appl. Meteorol., 42, 1369–1390, 2003.
Hocking, J., Rayer, P., Rundle, D., Saunders, R., Matricardi, M., Geer, A.,
Brunel, P., and Vidot, J.: RTTOV v11 Users Guide; NWP-SAF Report, Met,
Office: Exeter, UK, 2014.
Hollmann, R., Merchant, C., Saunders, R., Downy, C., Buchwitz, M., Cazenave,
A., Chuvieco, E., Defourny, P., de Leeuw, G., Forsberg, R., Holzer-Popp, T., Paul, F., Sandven, S., Sathyendranath, S., van Roozendael, M., and Wagner, W.: The ESA
climate change initiative: Satellite data records for essential climate
variables, B. Am. Meteorol. Soc., 94, 1541–1552,
2013.Hollstein, A., Fischer, J., Carbajal Henken, C., and Preusker, R.: Bayesian
cloud detection for MERIS, AATSR, and their combination, Atmos.
Meas. Tech., 8, 1757–1771, 10.5194/amt-8-1757-2015, 2015.
Karlsson, K.-G. and Johansson, E.: Multi-Sensor calibration studies of
AVHRR-heritage channel radiances using the simultaneous nadir observation
approach, Remote Sens., 6, 1845–1862, 2014.Karlsson, K.-G., Riihelä, A., Müller, R., Meirink, J. F., Sedlar, J.,
Stengel, M., Lockhoff, M., Trentmann, J., Kaspar, F., Hollmann, R., and
Wolters, E.: CLARA-A1: a cloud, albedo, and radiation dataset from 28 yr of
global AVHRR data, Atmos. Chem. Phys., 13, 5351–5367,
10.5194/acp-13-5351-2013, 2013.Karlsson, K.-G., Anttila, K., Trentmann, J., Stengel, M., Fokke Meirink, J.,
Devasthale, A., Hanschmann, T., Kothe, S., Jääskeläinen, E.,
Sedlar, J., Benas, N., van Zadelhoff, G.-J., Schlundt, C., Stein, D.,
Finkensieper, S., Håkansson, N., and Hollmann, R.: CLARA-A2: the second
edition of the CM SAF cloud and radiation data record from 34 years of global
AVHRR data, Atmos. Chem. Phys., 17, 5809–5828, 10.5194/acp-17-5809-2017,
2017.
King, M. D., Menzel, W. P., Kaufman, Y. J., Tanré, D., Gao, B.-C.,
Platnick, S., Ackerman, S. A., Remer, L. A., Pincus, R., and Hubanks, P. A.:
Cloud and aerosol properties, precipitable water, and profiles of temperature
and water vapor from MODIS, IEEE T. Geosci. Remote, 41, 442–458, 2003.
Levenberg, K.: A method for the solution of certain non-linear problems in
least squares, Q. Appl. Math., 2, 164–168, 1944.
Lindstrot, R., Preusker, R., and Fischer, J.: Empirical correction of stray
light within the MERIS oxygen A-band channel, J. Atmos.
Ocean. Technol., 27, 1185–1194, 2010.Marchant, B., Platnick, S., Meyer, K., Arnold, G. T., and Riedi, J.: MODIS
Collection 6 shortwave-derived cloud phase classification algorithm and
comparisons with CALIOP, Atmos. Meas. Tech., 9, 1587–1599,
10.5194/amt-9-1587-2016, 2016.
Marquardt, D. W.: An algorithm for least-squares estimation of nonlinear
parameters, J. Soc. Ind. Appl. Math.,
11, 431–441, 1963.
McGarragh, G., Poulsen, C., Christensen, M., Thomas, G. Povey, A., Proud, S.,
and Grainger, R.: The ORAC multilayer cloud retrieval algorithm, Atmos. Meas.
Tech. Discuss., in
preparation, 2017a.McGarragh, G. R., Poulsen, C. A., Thomas, G. E., Povey, A. C., Sus, O.,
Stapelberg, S., Schlundt, C., Proud, S., Christensen, M. W., Stengel, M.,
Hollmann, R., and Grainger, R. G.: The Community Cloud retrieval for CLimate
(CC4CL), Part II: The optimal estimation approach, Atmos. Meas. Tech.
Discuss., 10.5194/amt-2017-333, in review, 2017.
Norris, J. R., Allen, R. J., Evan, A. T., Zelinka, M. D., O'Dell, C. W., and
Klein, S. A.: Evidence for climate change in the satellite cloud record,
Nature, 536, 72–75, 2016.
O'Dell, C. W., Wentz, F. J., and Bennartz, R.: Cloud liquid water path from
satellite-based passive microwave observations: A new climatology over the
global oceans, J. Clim., 21, 1721–1739, 2008.
Pavolonis, M. J. and Heidinger, A. K.: Daytime cloud overlap detection from
AVHRR and VIIRS, J. Appl. Meteorol., 43, 762–778, 2004.
Pavolonis, M. J., Heidinger, A. K., and Uttal, T.: Daytime global cloud typing
from AVHRR and VIIRS: Algorithm description, validation, and comparisons,
J. Appl. Meteorol., 44, 804–826, 2005.
Platnick, S., King, M. D., Ackerman, S. A., Menzel, W. P., Baum, B. A.,
Riédi, J. C., and Frey, R. A.: The MODIS cloud products: Algorithms and
examples from Terra, IEEE Trans. Geosci. Remote, 41,
459–473, 2003.Platnick, S., King, M. D., Meyer, K. G., Wind, G., Amarasinghe, N., Marchant,
B., Arnold, G. T., Zhang, Z., Hubanks, P. A., Ridgway, B., and Riédi, J.: MODIS Cloud Opti-
cal Properties: User Guide for the Collection 6 Level-2 MOD06/MYD06 Product
and Associated Level-3 Datasets, Ver- sion 1.0, available at:
https://modis-images.gsfc.nasa.gov/_docs/C6MOD06OPUserGuide.pdf (last
access: 31 May 2016), 2015.
Platnick, S., Meyer, K. G., King, M. D., Wind, G., Amarasinghe, N., Marchant,
B., Arnold, G. T., Zhang, Z., Hubanks, P. A., Holz, R. E., Yang, P.,
Ridgway, W. L., and Riedi, J.: The MODIS
Cloud Optical and Microphysical Products: Collection 6 Updates and Examples
From Terra and Aqua, IEEE Trans. Geosci. Remote, 55,
502–525, 2017.Poulsen, C. A., Siddans, R., Thomas, G. E., Sayer, A. M., Grainger, R. G.,
Campmany, E., Dean, S. M., Arnold, C., and Watts, P. D.: Cloud retrievals
from satellite data using optimal estimation: evaluation and application to
ATSR, Atmos. Meas. Tech., 5, 1889–1910, 10.5194/amt-5-1889-2012, 2012.Poulsen, C., McGarragh, G., Thomas, G., Christensen, M., Povey, A., Grainger,
D., Proud, S., and Hollmann, R.: ESA Cloud Climate Change Initiative (ESA
Cloud_cci) data: Cloud_cci ATSR2-AATSR L3C/L3U/L2 CLD_PRODUCTS v2.0,
Deutscher Wetterdienst (DWD) and Rutherford Appleton Laboratory (Dataset
Producer), 10.5676/DWD/ESA_Cloud_cci/ATSR2-AATSR/V002, 2017.
Rodgers, C. D.: Inverse methods for atmospheric sounding: theory and practice,
vol. 2, World scientific, 2000.
Rossow, W. B. and Schiffer, R. A.: Advances in understanding clouds from ISCCP,
Bull. Am. Meteorol. Soc., 80, 2261–2287, 1999.
Saunders, R., Matricardi, M., and Brunel, P.: An improved fast radiative
transfer model for assimilation of satellite radiance observations, Q.
J. Roy. Meteor. Soc., 125, 1407–1425, 1999.Sayer, A. M., Poulsen, C. A., Arnold, C., Campmany, E., Dean, S., Ewen, G. B.
L., Grainger, R. G., Lawrence, B. N., Siddans, R., Thomas, G. E., and Watts,
P. D.: Global retrieval of ATSR cloud parameters and evaluation (GRAPE):
dataset assessment, Atmos. Chem. Phys., 11, 3913–3936,
10.5194/acp-11-3913-2011, 2011.
Schiffer, R. and Rossow, W. B.: The International Satellite Cloud Climatology
Project(ISCCP)- The first project of the World Climate Research Programme,
American Meteorological Society, Bulletin, 64, 779–784, 1983.
Stamnes, K., Tsay, S.-C., Wiscombe, W., and Jayaweera, K.: Numerically stable
algorithm for discrete-ordinate-method radiative transfer in multiple
scattering and emitting layered media, Appl. Optics, 27, 2502–2509, 1988.Stapelberg, S., Stengel, M., Karlsson, K.-G., Meirink, J, F., Bojanowski, J.,
and Hollmann, R.: ESA Cloud_cci Product Validation and Intercomparison
Report (PVIR), issue 4, rev. 1, 19 april 2017 edn., available from:
http://www.esa-cloud-cci.org/?q=documentation, 2017.
Stengel, M., Mieruch, S., Jerg, M., Karlsson, K.-G., Scheirer, R., Maddux, B.,
Meirink, J., Poulsen, C., Siddans, R., Walther, A., and Hollmann, R.: The clouds
climate change initiative: Assessment of state-of-the-art cloud property
retrieval schemes applied to AVHRR heritage measurements, Remote Sens.
Environ., 162, 363–379, 2015.Stengel, M., Sus, O., Stapelberg, S., Schlundt, C., Poulsen, C., and
Hollmann, R.: ESA Cloud Climate Change Initiative (ESA Cloud_cci) data:
Cloud_cci MODIS-Aqua L3C/L3U CLD_PRODUCTS v2.0, Deutscher Wetterdienst
(DWD), 10.5676/DWD/ESA_Cloud_cci/MODIS-Aqua/V002, 2017.Stengel, M., Sus, O., Stapelberg, S., Schlundt, C., Poulsen, C., and
Hollmann, R.: ESA Cloud Climate Change Initiative (ESA Cloud_cci) data:
Cloud_cci AVHRR-PM L3C/L3U CLD_PRODUCTS v2.0, Deutscher Wetterdienst (DWD),
10.5676/DWD/ESA_Cloud_cci/AVHRR-PM/V002, 2017.Stengel, M., Sus, O., Stapelberg, S., Schlundt, C., Poulsen, C., and
Hollmann, R.: ESA Cloud Climate Change Initiative (ESA Cloud_cci) data:
Cloud_cci MODIS-Terra L3C/L3U CLD_PRODUCTS v2.0, Deutscher Wetterdienst
(DWD), 10.5676/DWD/ESA_Cloud_cci/MODIS-Terra/V002, 2017.Stengel, M., Sus, O., Stapelberg, S., Schlundt, C., Poulsen, C., and
Hollmann, R.: ESA Cloud Climate Change Initiative (ESA Cloud_cci) data:
Cloud_cci MODIS-Aqua L3C/L3U CLD_PRODUCTS v2.0, Deutscher Wetterdienst
(DWD), 10.5676/DWD/ESA_Cloud_cci/MODIS-Aqua/V002, 2017.Stephens, G.: Radiation profiles in extended water clouds, II:
Parameterization
schemes, J. Atmos. Sci., 35, 2123–2132, 1978.
Sun, B., Free, M., Yoo, H. L., Foster, M. J., Heidinger, A., and Karlsson,
K.-G.: Variability and trends in US cloud cover: ISCCP, PATMOS-x, and
CLARA-A1 compared to homogeneity-adjusted weather observations, J.
Clim., 28, 4373–4389, 2015.Sus, O., Stengel, M., Stapelberg, S., McGarragh, G., Poulsen, C., Povey, A.
C., Schlundt, C., Thomas, G., Christensen, M., Proud, S., Jerg, M., Grainger,
R., and Hollmann, R.: The Community Cloud retrieval for Climate (CC4CL), Part
I: A framework applied to multiple satellite imaging sensors, Atmos. Meas.
Tech. Discuss., 10.5194/amt-2017-334, in review, 2017.
Terai, C., Klein, S., and Zelinka, M.: Constraining the low-cloud optical depth
feedback at middle and high latitudes using satellite observations, J. Geophys. Res.-Atmos., 121, 9696–9716, 2016.
Walther, A. and Heidinger, A. K.: Implementation of the daytime cloud optical
and microphysical properties algorithm (DCOMP) in PATMOS-x, J.
Appl. Meteor. Clim., 51, 1371–1390, 2012.Watts, P. D., Mutlow, C. T., Baran, A. J., and Zavody, A. M.: Study on cloud
properties derived from Meteosat Second Generation Observations, EUMETSAT ITT 97/181 Final Rep., Eur. Organ. for the Exploit. of Meteorol. Satell., Darmstadt, Germany,
http://www.eumetsat.int/Home/index.htm (last access: 21 November 2017), 1998.
Winker, D. M., Vaughan, M. A., Omar, A., Hu, Y., Powell, K. A., Liu, Z., Hunt,
W. H., and Young, S. A.: Overview of the CALIPSO mission and CALIOP data
processing algorithms, J. Atmos. Ocean. Technol., 26,
2310–2323, 2009.Xiong, X. and Barnes, W.: An overview of MODIS radiometric calibration and
characterization, Adv. Atmos. Sci., 23, 69–79,
10.1007/s00376-006-0008-3, 2006.