METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
Algorithm Theoretical Basis Document
（二行） for Cloud Type/Phase Product
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
（二行）
IMAIToshiharu**,
Takahito* andSUZUE
YOSHIDA
Ryo**and YOSHIDA Ryo*
MOURI Kouki*, IZUMI
Hiroshi*
（三行）
Abstract
This paper describes the algorithm theoretical
basis for the Himawari-8 Cloud Mask Product
Abstract
(CMP)
developed
by
the
Meteorological
Satellite
Center.
CMP is part
ofproduct,
the Fundamental
Cloud
The cloud type/phase product is an element of the fundamental
cloud
which incorporates
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
cloud
mask, cloud
and cloud top height. The Meteorological Satellite Center (MSC) of Japan
operational
sincetype/phase
7 July 2015.
Meteorological
(JMA)
separated
type/phase
and cloud
height, which
were
The CMPAgency
algorithm
is based
on thecloud
cloudmask,
maskcloud
technique
of NWC-SAF
fortop
MSG/SEVIRI
and
NOAA/NESDIS
for GOES-R/ABI.
Most cloud
tests involve
based on
contained
in miscellaneous
level 2 products
in detection
the Himawari-7
era, threshold
and put methods
them together
as a
radiative
transfer
calculation
using
NWP
data
as
atmospheric
profiles.
The
thresholds
are
modified
fundamental cloud product. The cloud type/phase product provides information on cloud type
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
(opaque/semi-transparent/fractional) and the phase of cloud drop (water/ice/unknown or mix) at the cloud
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
top.algorithm
The analysis
based
the results
NWCSAF
(NoWCasting
Facility)
was method
applied istomainly
SEVIRI
data,onwith
showing
hit ratios Satellite
with theApplication
MODIS product
cloud
type
algorithm
because
the method is well established and relatively easy to understand. Cloud
were
around
85% for
all seasons.
phase comparison between the MSC product （三行）
and the Level 2 product of MODIS (MYD06L2) and
Calipso (CAL_LID_L2_01kmCLay-ValStage1-V3-30) was carried out from 20 July to 2 August 2015.
began
operation.
1. Introduction
The ice phase capture ratio was greater than 80% for both
MODIS
and Calipso.
（一行）
This paper describes the algorithm theoretical basis for
Himawari-8 is the new geostationary satellite of the
CMP.
Japan Meteorological Agency (JMA). It was launched on
（一行）
7 October 2014 and began operation on 7 July 2015. The
2. Algorithm
1. Introduction
Cloud type studies produce comparatively qualitative
satellite
carries the Advanced Himawari Imager (AHI),
information,
which is greatly improved over past imagers in terms of
2.1 Overviewenabling determination of clouds as opaque,
Cloud detection
the CMPinalgorithm
based type
on
its Himawari-8
number of bands
temporal/special
resolution
carriesand
theitsAdvanced
Himawari
Imager
semi-transparent
or in
fractional
the MSCiscloud
comparison
of observation
data difference
and clearis sky
(Bessho which
et al.has2016).
Using AHI bands
data, for
JMA’s
(AHI),
16 observational
the
product.
Brightness
temperature
the data
main
calculated
from
numerical
weather
prediction
(NWP)
Meteorological
Satellite
Center
(MSC)
developed
the
provision of a wealth of information on cloud
metric utilized to determine cloud type. The algorithm
data. If observation data differ from clear sky data, the
Himawari-8 Cloud Mask Product (CMP).
microphysics.
Meteorological
satellite
operating
theoretical basis document for the cloud product of
pixel is cloudy. To discriminate between cloudy and clear
Cloud mask is used to discriminate cloudy pixels
organizations around the world have made out cloud type
NWCSAF (Meteo France 2012) describes the cloud type
pixels, it is desirable for observation variables in clear
from clear ones in satellite data. Numerous satellite
and
cloudrequire
phase in
recognition
of their importance,
determination
that MSC to
usesdiffer
for reference.
GOES-R
and cloudy conditions
and for The
the physical
products
cloud
mask information.
By way and
of
the
Meteorological
Center(SST)
(MSC)
of Japan
cloud
theoretical
document
reason type
behindalgorithm
the difference
to bebasis
explicit
(e.g.,
example,
Sea SurfaceSatellite
Temperature
(Yasuda
and
temperature,2010)
reflectance
and with
emissivity
the but
top these
and
Shirakawa 1999),
Clear (JMA)
Sky Radiance
(CSR)
(Uesawa
Meteorological
Agency
also began
to provide
the
(Pavolonis
also deals
cloud of
types,
transmittance
relating
theInatmosphere).
2009) and Aerosol
Detection
(Okawara
et al. 2003)
fundamental
cloud product
when
Himawari-8
enteredare
its
are
cloud phases
such toas clouds
ice or and
water.
cloud phase
Rather
than
encompassing
retrieval
of
these offive
derived
in
clear
pixels.
Conversely,
Cloud
Grid
operative period. MSC simultaneously separates cloud
studies, the difference in the imaginary part
the
variables, most of the tests in the CMP algorithm involve
Information (Tokuno 2002) is processed in cloudy pixels.
mask, cloud type/phase and cloud top height from other
complex refractive index between two bands is used to
the use of brightness temperature or reflectance data with
There were no MSC cloud mask products before the
level 2 satellite products and puts them together into a
determine the ice or water particle phase for IR bands.
similar tendencies. Cloud mask tests are based on the
start of Himawari-8’s operation; each satellite product
single
calledmask
the fundamental
can
Strabala
et al.algorithm
(1994) proposed
cloud particle(NWC)
phase
cloud mask
of the NoWCasting
had itsoutput
own cloud
process. Ascloud
cloudproduct.
mask isItalso
be
used for
for most
otherHimawari-8
level 2 products
those for
discrimination
using 8-,Facility
11- and(SAF)
12-micron
brightness
Satellite Application
(Meteo-France,
required
products,such
MSCasdecided
to
2012) and the
Oceanic
develop CMP
as a resource
available
for Himawari-8
aerosols,
sea surface
temperature
and volcanic
ash, and
temperatures
and National
their difference
fromand
dataAtmospheric
collected by
Administration
Environmental
products
common.
CMP
a part of their
the Fundamental
users
noinlonger
need
to iscompute
own cloud
the
high-resolution(NOAA)/National
infrared sounder (HIRS)
on board
Satellite,
Data,
and
Information
Service
(NESDIS)
Cloud
Product
(FCP),
which
contains
cloud
phase,
type
information.
NOAA’s polar orbiters. For near-infrared bands, Baum et
(Heidinger and Straka III, 2012).
and top height for Himawari-8/AHI pixels (Mouri et al.
Radiative transfer calculation using NWP data for
2016a, 2016b). FCP has been produced since Himawari-8
* System Engineering Division, Data Processing Department, Meteorological Satellite Center
** Meteorological Satellite and Observation System Research Department, Meteorological Research Institute
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
August 31, Division,
2015, Accepted
NovemberDepartment,
20, 2015) Meteorological Satellite Center
**(Received
System Engineering
Data Processing
(Received September 1, 2015, Accepted 20 November 2015)
--19
1 --
－ 19 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
al. (2012) investigated cirrus clouds using MODIS data. （二行） tropopause level. Although effective emissivity differs
This document outlines the cloud type/phase product and
from the product of actual cloud emissivity and actual
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
the related algorithm, which is based on the NWCSAF
cloud amount, it has similar characteristics in that
（二行）
algorithm (Meteo France 2012). The product includes
observed radiance is related to the ratio of contribution
IMAI
Takahito
* and YOSHIDA
Ryo**top radiation and surface radiation. This
information on cloud type and cloud
phase.
The
from cloud
（三行）
document also describes the required input data, output
means that thick cloud has high effective emissivity
data and evaluation.
and thin cloud has low effective emissivity.
EUMETSAT:
Abstract
An
intergovernmental
European
for the Himawari-8
Maskthat
Product
1.1. Content This paper describes the algorithm theoretical basis
organization
founded Cloud
in 1986
operates the
(CMP) developed by the Meteorological Satellite Center.
CMP
is
part
of
the
Fundamental
Cloud
Meteosat
geostationary
satellite
and
the
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
This is theoperational
Algorithmsince
Theoretical
Basis Document for
low-earth-orbiting Metop and Jason satellites. The
7 July 2015.
the cloud type/phase
product.
It describes
organization
runs a satellite
application facility
(SAF)
The CMP
algorithm
is basedthe
on algorithms
the cloud mask technique
of NWC-SAF
for MSG/SEVIRI
and
NOAA/NESDIS
for GOES-R/ABI.
tests involve
threshold
methods based
used and covers
related practical
considerations.Most cloud detection
to support
research,
development
and onoperational
radiative transfer calculation using NWP data as atmospheric
profiles.
The
thresholds
are
modified
activities not carried out at EUMETSAT headquarters.
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
1.2. Definitions, acronyms and abbreviations
𝜑𝜑𝜑𝜑{𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠} : Angle between sun and satellite.
results showed that the CMP hit ratio derived from such
comparison was more than 85%. The CMP
JMA:
Japan hit
Meteorological
Agency.
algorithm was applied to SEVIRI data, with results
showing
ratios with the
MODIS product
LRC: Local radiative center as incorporated in GOES-R
AHI: Advanced
Himawari
(mounted on
were around
85% for Imager
all seasons.
（三行） Cloud Type and Cloud Height Algorithm Theoretical
Himawari-8 and -9 for observation with 16 bands;
central wavelengths: 0.47, 0.51, 0.64, 0.86, 1.6, 2.3, 3.9,
1. Introduction
6.2, 6.9, 7.3, 8.6, 9.6, 10.4, 11.2, 12.4, 13.3 microns) .
（一行）
Aqua: NASA’s low-earth-orbit
earth science satellite.
Himawari-8 is the new geostationary satellite of the
𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 : Surface albedo depending on surface type, sun
Japan
Meteorological Agency (JMA). It was launched on
zenith
angle,andatmospheric
conditions
7 October 2014
began operation
on 7 July and
2015. other
The
variables.
satellite
carries the Advanced Himawari Imager (AHI),
BRDF:isBi-directional
reflectance
distribution
which
greatly improved
over past
imagers function.
in terms of
its
numberNASA’s
of bandssatellite
and its for
temporal/special
resolution
Calipso:
analyzing clouds
and
(Bessho
al. 2016).
UsingandAHI
data,
aerosolsetin relation
to weather
climate.
TheJMA’s
unit is
Meteorological
Satellite
Center
(MSC)
developed
equipped with cloud-aerosol lidar for observationtheof
Himawari-8 Cloud Mask Product (CMP).
aerosols and cloud microphysics.
Cloud mask is used to discriminate cloudy pixels
Cloud mask: A cloud mask product providing information
from clear ones in satellite data. Numerous satellite
on pixel
clear/cloudy
statusinformation.
and aerosol
products
require
cloud mask
Bydetection.
way of
Cloud
mask
data
are
created
from
Himawari-8’s
example, Sea Surface Temperature (SST) (Yasudaoutput,
and
Shirakawa
Sky of
Radiance
(CSR) (Uesawa
and are 1999),
also anClear
element
the fundamental
cloud
2009)
and Aerosol
Detection
al. 2003)
product.
Cloud mask
data(Okawara
contain et
only
high- are
and
derived
in
clear
pixels.
Conversely,
Cloud
Grid
low-quality flags, while the cloud type product
Information (Tokuno 2002) is processed in cloudy pixels.
provides a much wider range of data on quality.
There were no MSC cloud mask products before the
Daytime: The period during which the sun zenith angle is
start of Himawari-8’s operation; each satellite product
than
80 degrees.
hadsmaller
its own
cloud
mask process. As cloud mask is also
Effective
emissivity
at 11.2products,
micronsMSC
(ε(11.2)):
required for most Himawari-8
decidedThe
to
develop
CMP
as a resource
available
effective
emissivity
of 11.2
micronsforatHimawari-8
the center
products
in common.
CMP is of
a part
the Fundamental
wavelength
is the product
cloudofamount
and cloud
Cloud
Product
(FCP),
which
contains
cloud
phase,
type
emissivity as estimated in consideration of cloud at
the
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
Basis (Pavolonis 2010).
began Look-up
operation.table containing radiance values to support
LUT:
This
paper describes
the algorithm
basiswater
for
discrimination
between
ice phasetheoretical
cloud and
CMP.
phase cloud. The radiances were calculated using
（一行）
RSTAR
(see
“RSTAR”
in this section) under various
2. Algorithm
conditions for ice phase and water phase.
MODIS:
Moderate resolution imaging spectroradiometer
2.1
Overview
Cloud
CMP
algorithm is based on
on
boarddetection
the AquainandtheTerra
satellites.
comparison
of observation
and clear sky data
MSC: Meteorological
Satellitedata
Center.
calculated
from
numerical
weather
prediction
(NWP)is
Nighttime: Period during which the sun
zenith angle
data. If observation data differ from clear sky data, the
larger than 93 degrees.
pixel is cloudy. To discriminate between cloudy and clear
NWCSAF: Nowcasting satellite application facility.
pixels, it is desirable for observation variables in clear
λ: Center
wavelength
of the
and
cloudy
conditions
to AHI
differbands
and in
formicrons.
the physical
:
Effective
radius
of
cloud
particles.
𝑟𝑟𝑟𝑟reason
behind the difference to be explicit (e.g.,
𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
temperature,
reflectance
and emissivity of the top and
reflectance.
𝑅𝑅𝑅𝑅𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 : Bi-directional
transmittance
relating
to clouds
the wavelength
atmosphere).of
R (λ): Observed
reflectivity
at theand
central
Rather
than
encompassing
retrieval
of
these five
λ microns, where λ is 0.64 or 1.6.
variables, most of the tests in the CMP algorithm involve
RSTAR: A radiative transfer code developed by the
the use of brightness temperature or reflectance data with
Center for Climate System Research at the University
similar tendencies. Cloud mask tests are based on the
of Tokyo
1986).
cloud
mask(Nakajima
algorithmandofTanaka
the NoWCasting
(NWC)
RTTOV:
A
radiative
transfer
code
developed
by
Satellite Application Facility (SAF) (Meteo-France,
2012)
and the(Eyre
National
EUMETSAT
1991). Oceanic and Atmospheric
Administration
Environmental
Surface category: (NOAA)/National
MSC uses the Global
Land Cover
Satellite,
Data,
and
Information
Service
Characteristics Data Base Ver 2.01 of USGS(NESDIS)
to identify
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--20
1 --
－ 20 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
earth surface attributes. The cloud type product
Cloudy pixels in Cloud Mask (Imai and Yoshida 2016)
（二行）
features four main categories: 1. water area; 2. lowland,
are processed in the cloud type product.
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
urban or crop area; 3. mountain, desert or rock area;
（二行）
and 4. forest area.
τ: Optical thickness.
2. Outline of cloud type and phase product
IMAI Takahito* and YOSHIDA Ryo**
（三行）
𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 : Sun zenith angle (the angle from the sun to the
zenith at an arbitrary point on the earth’s surface).
The cloud type/phase product provides information on
cloud type (opaque, semi-transparent or fractional) and
𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 : Satellite zenith angle (the angle from the satellite
the phase of cloud drop (water, ice or unknown/mixed).
Abstract
Thisat paper
describespoint
the algorithm
theoretical The
basismethod
for theofHimawari-8
Mask
Product technique
to the zenith
an arbitrary
on the earth’s
analysis is Cloud
based on
a threshold
(CMP)
developed
by
the
Meteorological
Satellite
Center.
CMP
is
part
of
the
Fundamental
surface).
largely modeled on the NWCSAF cloud Cloud
type algorithm.
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
𝑇𝑇𝑇𝑇(𝜆𝜆𝜆𝜆) : Observed
brightness
Cloudy pixels in Cloud Mask (Imai and Yoshida 2016)
operational
since 7temperature
July 2015. at the center
wavelength of
where isisbased
3.9, on
6.2,the
7.3,
8.6,maskare
processed
in the cloud for
typeMSG/SEVIRI
product.
The microns,
CMP algorithm
cloud
technique
of NWC-SAF
and
for GOES-R/ABI. Most cloud detection tests involve threshold methods based on
11.2, 12.4NOAA/NESDIS
or 13.3.
radiative brightness
transfer calculation
using
data as atmospheric
The thresholds are modified
temperature
at NWP
the center
𝑇𝑇𝑇𝑇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 (𝜆𝜆𝜆𝜆): Estimated
2.1. Input profiles.
data
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
wavelength of λ microns as calculated under the
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
assumption
of cloud
presence
with data,
a height
i. AHI
observation
algorithm
was applied
to SEVIRI
with results
showing
hit ratiosdata
with the MODIS product
corresponding
to the 85%
surface.
corresponds to the
were around
for allλseasons.
AHI observation wavelength. By way of example, （三行）
it
Reflectances at 0.64 and 1.6 microns and brightness
may be 6.2, 7.3, 8.6 and so on. Plev is a constant value
1. Introduction
for pressure between the surface and the tropopause.
By way of example,（一行）
it may be surface, 850 hPa, 500
Himawari-8 is the new geostationary satellite of the
hPa and so on. A surface value indicates that there is no
Japan Meteorological Agency (JMA). It was launched on
cloud. RTTOV
is used
estimation
outlined
7 October
2014 and
beganforoperation
on as
7 July
2015.later.
The
Estimated
brightnessHimawari
temperature
at the(AHI),
center
𝑇𝑇𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (𝜆𝜆𝜆𝜆):
satellite
carries
the Advanced
Imager
wavelength
λ microns
which
is greatlyofimproved
over as
pastcalculated
imagers inunder
terms the
of
its assumption
number of bands
and its
temporal/special
of opaque
cloud.
In regard toresolution
radiative
(Bessho
2016).
AHI 𝑇𝑇𝑇𝑇𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝data,
JMA’s
transfer,ettheal.cloud
top isUsing
at height
𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 (𝜆𝜆𝜆𝜆), which
Meteorological
Satellite
Center
(MSC)
developed
the
accords with 𝑇𝑇𝑇𝑇(𝜆𝜆𝜆𝜆). λ corresponds to the observation
Himawari-8 Cloud Mask Product (CMP).
wavelength. By way of example, it may be 6.2, 7.3, 8.6
Cloud mask is used to discriminate cloudy pixels
and so on.
from clear ones in satellite data. Numerous satellite
Terra: NASA’s
low-earth-orbit
science satellite.
products
require
cloud mask earth
information.
By way of
*
http://edc2.usgs.gov/glcc/globdoc2_0.php
example, Sea Surface Temperature (SST) (Yasuda and
Shirakawa
1999),
Clear
Skywhich
Radiance
(CSR)
Twilight time:
Period
during
the sun
zenith(Uesawa
angle is
2009)
and than
Aerosol
Detection
2003) are
smaller
93 degrees
and (Okawara
larger thanet80al.degrees.
derived in clear pixels. Conversely, Cloud Grid
Information (Tokuno 2002) is processed in cloudy pixels.
2. Outline of cloud type and phase product
There were no MSC cloud mask products before the
start of Himawari-8’s operation; each satellite product
Theitscloud
on
had
own type/phase
cloud maskproduct
process.provides
As cloudinformation
mask is also
cloud
type
semi-transparent
fractional)
required
for (opaque,
most Himawari-8
products,orMSC
decidedand
to
develop
as a drop
resource
available
for Himawari-8
the
phaseCMP
of cloud
(water,
ice or unknown/mixed).
products
in common.
CMP
is a part
the Fundamental
The
method
of analysis
is based
on aofthreshold
technique
Cloud
Product
(FCP),
which
contains
cloud
phase,
type
largely modeled on the NWCSAF cloud type algorithm.
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
temperatures at 3.9, 7.3, 8.6, 11.2 and 12.4 microns are
beganinoperation.
used
the product.
This paper describes the algorithm theoretical basis for
CMP.
ii. Radiative transfer calculation data based on JMA’s
（一行）
global
spectral
model
2. Algorithm
Wavelengths
2.1
Overview correspond to those of the AHI observation
detection
dataCloud
described
above.in the CMP algorithm is based on
comparison of observation data and clear sky data
calculated
from data
numerical weather prediction (NWP)
iii.
Cloud mask
data. If observation data differ from clear sky data, the
iv. Satellite navigation data
pixel is cloudy. To discriminate between cloudy and clear
Valuesit represent
longitude,variables
satellitein zenith
pixels,
is desirablelatitude,
for observation
clear
angle,
satellite
azimuthtoangle,
sun
and cloudy
conditions
differ sun
and zenith
for theangle,
physical
azimuth
angle andthe
scanning
time for
reason behind
difference
to each
be pixel.
explicit (e.g.,
temperature, reflectance and emissivity of the top and
transmittance
relating to clouds and the atmosphere).
v.
Ancillary data
Rather than encompassing retrieval of these five
variables, most of the tests in the CMP algorithm involve
Values represent elevation on land, surface category and
the use of brightness temperature or reflectance data with
white sky albedo. The surface category is described in
similar tendencies. Cloud mask tests are based on the
Subsection
The white
of the (NWC)
MODIS
cloud mask1.2.
algorithm
of sky
the albedo
NoWCasting
albedo
(Moody
at al. (SAF)
2005) is(Meteo-France,
used for the
Satelliteproduct
Application
Facility
2012) andproduct.
the National Oceanic and Atmospheric
type/phase
Administration
(NOAA)/National
Environmental
Satellite,
Data,
and
Information
Service
(NESDIS)
2.2. Process flow of the cloud type/phase product
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--21
1 --
－ 21 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
– Level 4 (Relatively low cloud in the middle cloud)
（二行）
As the cloud phase product is associated with the cloud
0.5 × 𝑇𝑇𝑇𝑇400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) + 0.5 × 𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) ≤ 𝑇𝑇𝑇𝑇(11.2) < 𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2)
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
type product, the cloud type procedure is run first (Fig.
（二行）
1).
– Level 5 (Low cloud)
IMAI Takahito* and YOSHIDA
**
𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2)Ryo
≤ 𝑇𝑇𝑇𝑇(11.2)
< 0.5 × 𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) + 0.5 × 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2)
（三行）
– Level 6 (Very low cloud)
0.5 × 𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) + 0.5 × 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2) ≤ 𝑇𝑇𝑇𝑇(11.2) < 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2)
Abstract
This paper describes the algorithm theoretical basis for the Himawari-8 Cloud Mask Product
(CMP) developed by the Meteorological Satellite Center.
CMP is part of the Fundamental Cloud
– Level 7 (Extremely low cloud)
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2) ≤ 𝑇𝑇𝑇𝑇(11.2)
operational since 7 July 2015.
The CMP algorithm is based on the cloud mask technique of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS for GOES-R/ABI. Most cloud detection
tests
involve
threshold methods based on
ii. Cloud
phase
information
radiative transfer calculation using NWP data as atmospheric profiles. The thresholds are modified
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
Cloud phase is identified as ice, water
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
unknown/mixed.
algorithm was applied to SEVIRI data, with results
showing hit ratios with the MODIS product
were around 85% for all seasons.
Fig. 1. Process flow of the cloud type/phase （三行）
3. Cloud type/phase product algorithm
product
1. Introduction
（一行）
Himawari-8 is the new geostationary satellite of the
2.3. Output data
Japan Meteorological Agency (JMA). It was launched on
7 October 2014 and began operation on 7 July 2015. The
i.
Cloud carries
type information
satellite
the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
itsClouds
numberareof identified
bands andasitsopaque,
temporal/special
resolutionor
semi-transparent
(Bessho etTheal.opaque
2016).andUsing
AHI data,categories
JMA’s
fractional.
semi-transparent
Meteorological
Satellite
Center
(MSC)
developed
have seven levels based on brightness temperaturetheas
Himawari-8 Cloud Mask Product (CMP).
observed at 11.2 microns, and the fractional category has
Cloud mask is used to discriminate cloudy pixels
only one level. A total of 15 types of information are
from clear ones in satellite data. Numerous satellite
provided.
The seven
are classifiedBybased
on
products require
cloud levels
mask information.
way of
comparison
between
T(11.2)
and Tplev(11.2).
example, Sea
Surface
Temperature
(SST) (Yasuda and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
Aerosolhigh
Detection
–2009)
Level and
1 (Extremely
cloud) (Okawara et al. 2003) are
derived
in
clear
pixels.
Cloud Grid
𝑇𝑇𝑇𝑇(11.2) < 0.5 × 𝑇𝑇𝑇𝑇400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) + Conversely,
0.5 × 𝑇𝑇𝑇𝑇200ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2)
Information (Tokuno 2002) is processed in cloudy pixels.
There were no MSC cloud mask products before the
– Level 2 (High cloud)
start of Himawari-8’s operation; each satellite product
(11.2) + 0.5 × 𝑇𝑇𝑇𝑇200ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) ≤ 𝑇𝑇𝑇𝑇(11.2) < 𝑇𝑇𝑇𝑇400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2)
0.5
𝑇𝑇𝑇𝑇400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃
had× its
own
cloud mask process. As cloud mask is also
required for most Himawari-8 products, MSC decided to
CMP as high
a resource
available
for Himawari-8
–develop
Level 3 (Relatively
cloud in the
middle cloud)
products
in common. CMP is a part of the Fundamental
𝑇𝑇𝑇𝑇
400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 ≤ 𝑇𝑇𝑇𝑇(11.2) < 0.5 × 𝑇𝑇𝑇𝑇400ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2) + 0.5 × 𝑇𝑇𝑇𝑇600ℎ𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (11.2)
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
or
began
operation.
3.1.
Cloud
type product
This paper describes the algorithm theoretical basis for
CMP.
Cloud type determination for the product is based on the
（一行）
threshold
technique.
Each
pixel is identified as opaque
2. Algorithm
cloud, semi-transparent cloud or fractional cloud as
described
in the previous section depending on observed
2.1 Overview
Cloud detection
in the reflectance,
CMP algorithm
based on
brightness
temperature,
andis brightness
comparison of
observation
data two
andwavelengths.
clear sky data
temperature
difference
between
The
calculated
from
numerical
weather
prediction
(NWP)
tests performed for each level are described in this section,
data. If observation data differ from clear sky data, the
and the algorithm is described in the subsequent
pixel is cloudy. To discriminate between cloudy and clear
subsections.
pixels, it is desirable for observation variables in clear
Thecloudy
cloud conditions
types identified
in and
this for
product
do not
and
to differ
the physical
correspond
to thethe
ten basic
types commonly
referenced
reason behind
difference
to be explicit
(e.g.,in
temperature,
reflectance
and emissivity
of theonly
top and
surface
weather
observation;
essentially,
the
transmittance
relating cloud
to clouds
and the atmosphere).
categories
of opaque
and semi-transparent
cloud
Rather
than
encompassing
retrieval
of
these
five
are used. MSC of JMA also provides High-resolution
variables, most of the tests in the CMP algorithm involve
Cloud Analysis Information (HCAI) (Suzue et al. 2016)
the use of brightness temperature or reflectance data with
containing cloud type information similar to that
similar tendencies. Cloud mask tests are based on the
indicating
ten basic
cloud of
types.
cloud mask
algorithm
the NoWCasting (NWC)
Satellite Application Facility (SAF) (Meteo-France,
2012) Opaque
and theor semi-transparent
National Oceanic cloud
and atAtmospheric
3.1.1.
Levels 1, 2,
(NOAA)/National
Environmental
3Administration
and 4
Satellite, Data, and Information Service (NESDIS)
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--22
1 --
－ 22 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
A 𝑇𝑇𝑇𝑇(11.2) − 𝑇𝑇𝑇𝑇(12.4) test is utilized to discriminate
daytime only, and the 𝑇𝑇𝑇𝑇(3.9) − 𝑇𝑇𝑇𝑇(11.2) test is for lower
（二行）
opaque cloud and semi-transparent cloud in these levels.
opaque cloud at nighttime. The 𝑇𝑇𝑇𝑇(8.6) − 𝑇𝑇𝑇𝑇(11.2) test is
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
The semi-transparent cloud test is carried out first. If the
for separation of lower fractional
（二行）
cloud
and
result is positive (i.e., the pixel is semi-transparent), the
semi-transparent cloud.
IMAI
opaque cloud test is not carried out. If
the Takahito
result *isand YOSHIDA Ryo**
negative, the pixel is given an opaque cloud status.
– Semi-transparent cloud
（三行）
i. Daytime
The opaque cloud test is carried out first. This is
Abstract
This >
paper
describes the algorithm theoretical followed
basis for by
thethe
Himawari-8
Cloudtest,
Mask
𝑇𝑇𝑇𝑇(11.2) − 𝑇𝑇𝑇𝑇(12.4)
𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
fractional cloud
andProduct
remaining cloud
(CMP) developed by the Meteorological Satellite Center.
CMP
is
part
of
the
Fundamental
Cloud
is classified as semi-transparent.
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
– Opaque cloud
operational since 7 July 2015.
Opaque cloud
𝑇𝑇𝑇𝑇(11.2) − 𝑇𝑇𝑇𝑇(12.4)
≤ 𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
The CMP
algorithm is based on the cloud mask-technique
of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS for GOES-R/ABI. Most cloud detection
tests
involve threshold
methods based on
𝑇𝑇𝑇𝑇(11.2)
− 𝑇𝑇𝑇𝑇(12.4)
< 𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
radiative
transfer
calculation
using
NWP
data
as
atmospheric
profiles.
The
thresholds
are modified
And
The threshold for 𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 is derived from
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
𝑅𝑅𝑅𝑅(0.64) > 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚064
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
algorithm was applied to SEVIRI data, with results showing hit ratios with the MODIS product
(11.2)
𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (12.4)
- Fractional cloud
𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠
were =
around
85%
for all−seasons.
+ 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_112_𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
（三行）
𝑇𝑇𝑇𝑇(8.6) − 𝑇𝑇𝑇𝑇(11.2) ≤ 𝑇𝑇𝑇𝑇86𝑇𝑇𝑇𝑇112𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
1.
Introduction
Here,
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_112_𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 is a constant value
（一行）
considered to exhibit the
highest skill score in opaque
Himawari-8 is the new geostationary satellite of the
cloud separation. The offset is taken from an LUT (see
Japan Meteorological Agency (JMA). It was launched on
the
Appendix
forand
information
on the offset).
7 October
2014
began operation
on 7 July 2015. The
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
its number of bands and its temporal/special resolution
(Bessho et al. 2016). Using AHI data, JMA’s
Meteorological Satellite Center (MSC) developed the
Himawari-8 Cloud Mask Product (CMP).
Cloud mask is used to discriminate cloudy pixels
from clear ones in satellite data. Numerous satellite
Fig. 2. require
Flow of cloud
cloud type
products
maskdetermination
information. for
ByLevels
way of
example,
Sea 4Surface Temperature (SST) (Yasuda and
1, 2, 3 and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
2009) and Aerosol Detection (Okawara et al. 2003) are
derivedOpaque
in clear
pixels. Conversely,
Cloud
3.1.2.
or semi-transparent
cloud at
LevelsGrid
5, 6,
Information (Tokuno 2002) is processed in cloudy pixels.
and 7 and fractional cloud
There were no MSC cloud mask products before the
start of Himawari-8’s operation; each satellite product
In its
these
, 𝑅𝑅𝑅𝑅(0.64)
had
own categories,
cloud mask 𝑇𝑇𝑇𝑇(11.2)
process. −
As𝑇𝑇𝑇𝑇(12.4)
cloud mask
is also ,
𝑇𝑇𝑇𝑇(3.9) −for
𝑇𝑇𝑇𝑇(11.2)
and 𝑇𝑇𝑇𝑇(8.6)
− 𝑇𝑇𝑇𝑇(11.2)
tests are
required
most Himawari-8
products,
MSC decided
to
develop
as a resource
available for Himawari-8
utilized CMP
to discriminate
semi-transparent,
opaque and
products
common.
is a test
part for
of the
Fundamental
fractionalincloud.
The CMP
𝑅𝑅𝑅𝑅(0.64)
semi-transparent
Cloud
Product
(FCP),
which
contains
cloud
cloud covering lower-layer cloud is carriedphase,
out intype
the
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
operation. cloud
-began
Semi-transparent
This
paper
describes
the algorithm
theoretical
basis for
All pixels
other
than those
in the above
categories
CMP.
（一行）
2. Algorithm
2.1 Overview
Cloud detection in the CMP algorithm is based on
comparison of observation data and clear sky data
calculated from numerical weather prediction (NWP)
data. If observation data differ from clear sky data, the
pixel is cloudy. To discriminate between cloudy and clear
pixels, it is desirable for observation variables in clear
and
conditions
to differ
and for for
the Levels
physical
Fig.cloudy
3. Flow
of cloud type
determination
reason behind the difference to be explicit (e.g.,
5, 6 and 7 in the daytime
temperature, reflectance and emissivity of the top and
transmittance relating to clouds and the atmosphere).
Rather
retrieval
of these
five
The than
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚encompassing
𝑚𝑚𝑚𝑚064 threshold
is used
to allow
variables, most of the tests in the CMP algorithm involve
semi-transparent cloud to be distinguished from opaque
the use of brightness temperature or reflectance data with
cloud in the NWCSAF algorithm (Meteo France 2012).
similar tendencies. Cloud mask tests are based on the
This
based on
the concept
that NoWCasting
semi-transparent
cloud
cloudis mask
algorithm
of the
(NWC)
and
opaque
cloud areFacility
separated
by the
maxCiR064
Satellite
Application
(SAF)
(Meteo-France,
2012) and reflectance
the National
Oceanic
Atmospheric
threshold
derived
fromand𝑇𝑇𝑇𝑇(11.2)
. The
Administration
(NOAA)/National
Environmental
maxCiR064 threshold
and 𝑇𝑇𝑇𝑇(11.2) have
a linear
Satellite,
Data,
and
Information
Service
(NESDIS)
relationship.
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--23
1 --
－ 23 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
A line can be drawn through two points:
（二行）
・(223.15, 0.35): Coldest point at 11.2 microns
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
・(𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2),𝑅𝑅𝑅𝑅𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 ): Warmest (surface) point at 11.2
microns
（二行）
IMAI Takahito
* and YOSHIDA Ryo**
The bi-directional effect of the overcast model
proposed
（三行）
by Manalo-Smith et al. (1998) is used for calculation of
𝑅𝑅𝑅𝑅𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 .
Abstract
This
describes
the algorithmbetween
theoretical basis for the Himawari-8 Cloud Mask Product
Figure 4 shows
an paper
example
of the relationship
(CMP)
developed
by
the
Meteorological
Center. CMP is part of the Fundamental Cloud
𝑇𝑇𝑇𝑇(11.2) and 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚064. The straight line is usedSatellite
to
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
Fig. 5. Flow of cloud type determination for Levels 5, 6
distinguish semi-transparent
opaque cloud.
operational sincecloud
7 Julyfrom
2015.
and 7 at twilight
The CMP algorithm is based on the cloud mask technique
of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS for GOES-R/ABI. Most cloud detection tests involve threshold methods based on
radiative transfer calculation using NWP data as atmospheric profiles. The thresholds are modified
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
iii Nighttime
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
algorithm was applied to SEVIRI data, with results showing hit ratios with the MODIS product
The procedure for nighttime testing involves application
were around 85% for all seasons.
1. Introduction
（三行）
of the 𝑇𝑇𝑇𝑇(3.9) − 𝑇𝑇𝑇𝑇(11.2) test to detect lower-layer
（一行）
Himawari-8 is the new geostationary satellite of the
Japan
Agency𝑇𝑇𝑇𝑇(11.2)
(JMA). It
was𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
launched
Fig. 4.Meteorological
Relationship between
and
𝑚𝑚𝑚𝑚064on
7 October 2014 and began operation on 7 July 2015. The
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
its
number of bands and its temporal/special resolution
ii Twilight
(Bessho et al. 2016). Using AHI data, JMA’s
Meteorological
Center
developed
The procedure Satellite
for twilight
tests (MSC)
is the same
as thatthe
for
Himawari-8 Cloud Mask Product (CMP).
the daytime except for the 𝑅𝑅𝑅𝑅(0.64) test.
Cloud mask is used to discriminate cloudy pixels
from clear ones in satellite data. Numerous satellite
- Opaque require
cloud cloud mask information. By way of
products
𝑇𝑇𝑇𝑇(11.2) −Sea
𝑇𝑇𝑇𝑇(12.4)
< 𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
example,
Surface
Temperature (SST) (Yasuda and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
2009)
and Aerosol
- Fractional
cloud Detection (Okawara et al. 2003) are
derived
in
clear≤ pixels.
Conversely,
𝑇𝑇𝑇𝑇(8.6) − 𝑇𝑇𝑇𝑇(11.2)
𝑇𝑇𝑇𝑇86𝑇𝑇𝑇𝑇112𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 Cloud Grid
Information (Tokuno 2002) is processed in cloudy pixels.
There were no MSC cloud mask products before the
- Semi-transparent cloud
start of Himawari-8’s operation; each satellite product
Allits
pixels
those
in the above
categories
had
ownother
cloudthan
mask
process.
As cloud
mask is also
required for most Himawari-8 products, MSC decided to
develop CMP as a resource available for Himawari-8
products in common. CMP is a part of the Fundamental
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
opaque cloud.
began operation.
This paper
describes the algorithm theoretical basis for
- Opaque
cloud
CMP.
𝑇𝑇𝑇𝑇(11.2) − 𝑇𝑇𝑇𝑇(12.4) < 𝑇𝑇𝑇𝑇112𝑇𝑇𝑇𝑇124𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
（一行）
And
2. Algorithm
𝑇𝑇𝑇𝑇(3.9) − 𝑇𝑇𝑇𝑇(11.2) < 𝑇𝑇𝑇𝑇39𝑇𝑇𝑇𝑇112𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
2.1 Overview
Cloud detection
- Fractional
cloud in the CMP algorithm is based on
comparison
of observation
data and
clear sky data
𝑇𝑇𝑇𝑇(8.6)
− 𝑇𝑇𝑇𝑇(11.2)
≤ 𝑇𝑇𝑇𝑇86𝑇𝑇𝑇𝑇112𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
calculated from numerical weather prediction (NWP)
data. If observation data differ from clear sky data, the
- Semi-transparent cloud
pixel is cloudy. To discriminate between cloudy and clear
All pixels other than those in the above categories
pixels, it is desirable for observation variables in clear
and cloudy conditions to differ and for the physical
reason behind the difference to be explicit (e.g.,
temperature, reflectance and emissivity of the top and
transmittance relating to clouds and the atmosphere).
Rather than encompassing retrieval of these five
variables, most of the tests in the CMP algorithm involve
the use of brightness temperature or reflectance data with
similar tendencies. Cloud mask tests are based on the
cloud mask algorithm of the NoWCasting (NWC)
Satellite Application Facility (SAF) (Meteo-France,
2012) and the National Oceanic and Atmospheric
Administration
(NOAA)/National
Fig. 6. Flow of cloud
type determinationEnvironmental
for Levels
Satellite, Data, and Information Service (NESDIS)
5, 6 and 7 at nighttime
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--24
1 --
－ 24 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
The
𝑇𝑇𝑇𝑇39𝑇𝑇𝑇𝑇112𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
and
𝑇𝑇𝑇𝑇86𝑇𝑇𝑇𝑇112𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
（二行）Cloud particle phase discrimination involves the use of
thresholds are used to detect opaque cloud and fractional
cloud type and satellite data on variables such as
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
cloud, respectively, at nighttime.
brightness temperature. For opaque
（二行）
𝑇𝑇𝑇𝑇39𝑇𝑇𝑇𝑇112𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
cloud,
the
relationships linking 𝑇𝑇𝑇𝑇(11.2) , 𝑇𝑇𝑇𝑇(8.6) − 𝑇𝑇𝑇𝑇(11.2) ,
IMAI Takahito* and YOSHIDA
Ryo**
𝑅𝑅𝑅𝑅(0.64) and
𝑅𝑅𝑅𝑅(1.6) are used to determine the particle
= 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (3.9) − 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2)
+ 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_39_𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
（三行）
phase. For other types, the phase is the same as that of the
local radiative center (LRC) if the LRC is opaque;
otherwise, the ice phase is attributed to semi-transparent
Abstract
This𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂paper
basis for
Cloud
MasktoProduct
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 describes the algorithm theoretical cloud
𝑇𝑇𝑇𝑇86𝑇𝑇𝑇𝑇112𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
and the
the Himawari-8
water phase is
attributed
fractional cloud.
(CMP) developed
by
the
Meteorological
Satellite
Center.
CMP
is
part
of
the
Fundamental
Cloud
= 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (8.6) − 𝑇𝑇𝑇𝑇𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (11.2)
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
3.2.1. Cloud type discrimination results
+ 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
operational
since 7 July 2015.𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
The CMP algorithm is based on the cloud mask technique of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS
for GOES-R/ABI.𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂Most
cloud detection
tests involve
methods
based on involves
For opaque
cloud threshold
pixels, phase
discrimination
and 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 are
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_39_𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
radiative
transfer
calculation
using
NWP
data
as
atmospheric
profiles.
The
thresholds
are
modified
the use of the satellite data described in this section. The
constant values considered to have the highest skill score
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
phase is ice for semi-transparent cloud pixels and water
in separation of opaque cloud. These biases are contained
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
forshowing
fractionalhitcloud
pixels.
in an LUT (see
the Appendix
for further
information
algorithm
was applied
to SEVIRI
data, on
with results
ratios
with the MODIS product
biases).
were around 85% for all seasons.
（三行）
3.2.2. 𝑻𝑻𝑻𝑻(𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏. 𝟐𝟐𝟐𝟐) test
3.2. Cloud phase product
1. Introduction
This section describes （一行）
cloud particle phase classification,
Himawari-8 is the new geostationary satellite of the
which is also based on the threshold technique. The flow
Japan Meteorological Agency (JMA). It was launched on
shown in2014
Fig. and
7. began operation on 7 July 2015. The
7is October
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
its number of bands and its temporal/special resolution
(Bessho et al. 2016). Using AHI data, JMA’s
Meteorological Satellite Center (MSC) developed the
Himawari-8 Cloud Mask Product (CMP).
Cloud mask is used to discriminate cloudy pixels
from clear ones in satellite data. Numerous satellite
products require cloud mask information. By way of
example, Sea Surface Temperature (SST) (Yasuda and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
2009) and Aerosol Detection (Okawara et al. 2003) are
derived in clear pixels. Conversely, Cloud Grid
Information (Tokuno 2002) is processed in cloudy pixels.
There were no MSC cloud mask products before the
start of Himawari-8’s operation; each satellite product
had
its7.own
cloud
maskphase
process.
As cloud mask is also
Fig.
Flow
of cloud
processing
required for most Himawari-8 products, MSC decided to
develop CMP as a resource available for Himawari-8
products in common. CMP is a part of the Fundamental
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
began
operation.
For opaque
cloud, the phase is ice when 𝑇𝑇𝑇𝑇(11.2) is
This
paper
describes
the algorithm
basis for
below freezing temperature
without theoretical
an ice nucleus
and
CMP.
water when 𝑇𝑇𝑇𝑇(11.2) is above freezing temperature.
（一行）
2. Algorithm
・Ice particle cloud
𝑇𝑇𝑇𝑇(11.2)
< 243.15
2.1 Overview
Cloud detection in the CMP algorithm is based on
comparison
of observation
data and clear sky data
・
Water droplet
cloud
calculated
from
numerical
weather
prediction (NWP)
𝑇𝑇𝑇𝑇(11.2) > 273.15
data. If observation data differ from clear sky data, the
pixel is cloudy. To discriminate between cloudy and clear
3.2.3 𝑻𝑻𝑻𝑻(𝟖𝟖𝟖𝟖. 𝟔𝟔𝟔𝟔) − 𝑻𝑻𝑻𝑻(𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏. 𝟐𝟐𝟐𝟐) test
pixels, it is desirable for observation variables in clear
and cloudy conditions to differ and for the physical
As described
2.3,difference
the ice or to
water
reason
behind inthe
be droplet
explicitphase
(e.g.,is
temperature,
and of
emissivity
of the−top
and
attributed onreflectance
the basis
the 𝑇𝑇𝑇𝑇(8.6)
𝑇𝑇𝑇𝑇(11.2)
transmittancedifference,
relating towhich
clouds
and the atmosphere).
temperature
is generally
larger for ice
Rather
than
encompassing
retrieval
of
these five
cloud than for water cloud.
variables, most of the tests in the CMP algorithm involve
the use of brightness temperature or reflectance data with
・Ice particle cloud
similar tendencies. Cloud mask tests are based on the
𝑇𝑇𝑇𝑇(11.2) of the NoWCasting (NWC)
cloud𝑇𝑇𝑇𝑇(8.6)
mask −algorithm
𝑇𝑇𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (8.6)
− 𝑇𝑇𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐(Meteo-France,
Satellite Application >Facility
(SAF)
𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (11.2)
2012) and the National
Oceanic and Atmospheric
+ 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
Administration
(NOAA)/National
Environmental
Satellite,
Data,
and
Information
Service
(NESDIS)
The value of 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 is set in reference
to an
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--25
1 --
－ 25 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
・Satellite zenith angle (𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ): every 10 degrees from 0 to 90
LUT, and is a constant considered to have the highest
（二行）
skill score in ice particle cloud detection. The score is
・ Azimuth difference between sun and satellite ( 𝜑𝜑𝜑𝜑{𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠} )
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
(degrees) : 0, 2.5, 5, 10, 15, 25, 35, 45, 55, 65, 75, 85, 95, 105, 115,
calculated from the Calipso cloud phase product (Hu el al.
（二行）
125, 135, 145, 155, 165, 170, 175, 177.5, 180
2009) and NWP data (see the Appendix for information
IMAI Takahito* and YOSHIDA Ryo**
on the constant).
（三行）
ii. Cloud
・Water droplet cloud
・Optical thickness (τ): 0.25, 0.50, 1, 2, 4, 8, 16, 32, 64, 96, 128
・Cloud height: 3 km
𝑇𝑇𝑇𝑇(8.6) − 𝑇𝑇𝑇𝑇(11.2)
Abstract
This paper
the algorithm theoretical basis for the Himawari-8 Cloud Mask Product
≤ 𝑇𝑇𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐describes
𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (8.6) − 𝑇𝑇𝑇𝑇𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (11.2)
(CMP) developed
by
the
Meteorological Satellite Center.
CMP radius
is part (𝒓𝒓𝒓𝒓
of the) and
Fundamental
+ 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
shape forCloud
ice cloud
iii. Effective
𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
・Effective radius (μm): 10, 20, 30, 40
operational since 7 July 2015.
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂_86_𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤
set inon
reference
The value of The
is based
spherical
CMP algorithm is
the cloudtomask・Shape:
technique
of NWC-SAF for MSG/SEVIRI and
for GOES-R/ABI.
Most
cloud detection tests involve threshold methods based on
an LUT, andNOAA/NESDIS
is a constant considered
to have the
highest
radiative
transfer
calculation
using
NWP
datais as atmospheric
profiles.
modified
iv. Effective
radiusThe
andthresholds
shape for are
water
droplet cloud
skill score in
ice particle
cloud
detection.
The
score
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
・Effective radius (μm): 4, 8, 16, 32
calculated from the Calipso cloud phase product (Hu et al.
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
・Shape:
spherical
2009) and NWP
data was
(see applied
the Appendix
for information
algorithm
to SEVIRI
data, with results
showing
hit ratios with the MODIS product
on the constant).
were around 85% for all seasons.
（三行）
v. Other conditions
3.2.4. 𝑹𝑹𝑹𝑹(𝟎𝟎𝟎𝟎. 𝟔𝟔𝟔𝟔𝟔𝟔𝟔𝟔) and 𝑹𝑹𝑹𝑹(𝟏𝟏𝟏𝟏. 𝟔𝟔𝟔𝟔) tests
1. Introduction
（一行）
If the values of 𝑅𝑅𝑅𝑅(0.64)
for water droplet cloud and ice
Himawari-8 is the new geostationary satellite of the
particle cloud are the same, the value of 𝑅𝑅𝑅𝑅(1.6) for ice
Japan Meteorological Agency (JMA). It was launched on
will 2014
be smaller
thanoperation
that for on
water
droplet
7cloud
October
and began
7 July
2015.cloud.
The
This characteristic
discrimination
the two
cloud
satellite
carries theallows
Advanced
Himawari ofImager
(AHI),
types. is Based
on this,overconsistent
cloud
phase
which
greatly improved
past imagers
in terms
of
its
number of bands
and its temporal/special
resolutionof
determination
is achieved
through comparison
(Bessho
al. 2016).
Using values
AHI and
data, radiative
JMA’s
𝑅𝑅𝑅𝑅(0.64) / et𝑅𝑅𝑅𝑅(1.6)
observation
Meteorological
Satellite
Center
(MSC)
developed
the
transfer calculation.
Himawari-8 Cloud Mask Product (CMP).
In this method, an LUT is created using the RSTAR
Cloud mask is used to discriminate cloudy pixels
radiative transfer model with various conditions of
from clear ones in satellite data. Numerous satellite
𝑅𝑅𝑅𝑅(0.64) and
𝑅𝑅𝑅𝑅(1.6).
the radiative
transferBycalculation,
products
require
cloudIn mask
information.
way of
the
reflectances
of
ice
and
water
phase
cloud
at
0.64
and
example, Sea Surface Temperature (SST) (Yasuda and
Shirakawa
Clear Sky
(CSR)
(Uesawa
1.6 microns1999),
are acquired.
TheRadiance
reflectance
is varied
with
2009)
and Aerosol
Detection
(Okawara
al. 2003)cloud
are
the conditions
of cloud
altitude,
cloud et
thickness,
derived
in
clear
pixels.
Conversely,
Cloud
Grid
particle shape/radius, positional relationship between the
Information (Tokuno 2002) is processed in cloudy pixels.
sun and the satellite, atmospheric profile and surface
There were no MSC cloud mask products before the
reflectance. The calculation conditions are summarized
start of Himawari-8’s operation; each satellite product
below.
had
its own cloud mask process. As cloud mask is also
required for most Himawari-8 products, MSC decided to
develop
CMPconditions
as a resource available for Himawari-8
i.
Geometric
products in common. CMP is a part of the Fundamental
Cloudzenith
Product
contains
phase,
・Sun
angle(FCP),
(𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ):which
every 10
degrees cloud
from 0 to
90 type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
・Surface albedo (𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ) (𝑅𝑅𝑅𝑅(0.64)): 0.02, 0.05, 0.1, 0.2, 0.3, 0.5
began operation.
・Surface
albedo (𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 ) (𝑅𝑅𝑅𝑅(1.6)): 0.02, 0.05, 0.1, 0.2, 0.3, 0.5, 0.7,
1.0This paper describes the algorithm theoretical basis for
CMP.
・Atmospheric profile: standard for mid-latitude summer
（一行）
2. Algorithm
The value shown below is then set to the LUT.
2.1𝑅𝑅𝑅𝑅Overview
{𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖/𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤} �𝜆𝜆𝜆𝜆, 𝜏𝜏𝜏𝜏, 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 , 𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 , 𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 , 𝜑𝜑𝜑𝜑{𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠} , 𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 �
Cloud detection in the CMP algorithm is based on
comparison
observation
sky data
The cloud of
particle
phase isdata
set and
fromclear
the LUT.
The
calculated
from
numerical
weather
prediction
(NWP)
procedure is as outlined below.
data. If observation data differ from clear sky data, the
pixel is cloudy. To discriminate between cloudy and clear
i. The values of 𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 , 𝜃𝜃𝜃𝜃𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 , 𝜑𝜑𝜑𝜑{𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠−𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠} and 𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for
pixels, it is desirable for observation variables in clear
every
pixel conditions
are determined
based
and
and cloudy
to differ
andonfornavigation
the physical
white(e.g.,
sky
ancillary
data. Here,
𝐴𝐴𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 is from
reason behind
the difference
to MODIS
be explicit
temperature,
reflectance
emissivity
of the�𝜆𝜆𝜆𝜆,top
𝜏𝜏𝜏𝜏, 𝑟𝑟𝑟𝑟and
albedo
data (Moody
et al.and
2005).
𝑅𝑅𝑅𝑅{𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖/𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤}
𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 �
transmittance
clouds and the atmosphere).
is
also obtainedrelating
from thetoLUT.
Rather than encompassing retrieval of these five
variables, most of the tests in the CMP algorithm involve
ii. Values of 𝜏𝜏𝜏𝜏 corresponding to each value of 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 are
the use of brightness temperature or reflectance data with
determined based on comparison of 𝑅𝑅𝑅𝑅(0.64) and
similar tendencies. Cloud mask tests are based on the
𝜏𝜏𝜏𝜏, 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 �.of the NoWCasting (NWC)
𝑅𝑅𝑅𝑅{𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖/𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤}
cloud
mask�0.64,
algorithm
�1.6, 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒(SAF)
� corresponding
to 𝜏𝜏𝜏𝜏 is
A
value of
𝑅𝑅𝑅𝑅{𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖/𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤}
Satellite
Application
Facility
(Meteo-France,
2012)determined.
and the National Oceanic and Atmospheric
then
Administration
(NOAA)/National
Environmental
Satellite,
Data,
and
Information
Service
(NESDIS)
for
iii. 𝑅𝑅𝑅𝑅(1.6) and 𝑅𝑅𝑅𝑅{𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖/𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤𝑤} �1.6, 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 � are compared
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--26
1 --
－ 26 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
all values of 𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 . If the results all indicate water, the
Cloud phase is expressed as 1-byte integers as indicated
（二行）
phase is marked as water; if they all indicate ice, the
below.
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
phase is marked as ice; otherwise, the phase is marked as
unknown or mixed.
4. Data format
（二行）
-128: Error pixel associated with satellite data errors and
IMAI Takahito* and YOSHIDA
similar Ryo**
（三行）
0: Clear pixels in cloud mask
1: Ice phase
Data are provided in flat binary format, with those Abstract
of
2: Water phase
theasalgorithm
theoretical 3:
basis
for the orHimawari-8
Cloud Mask Product
cloud type andThis
cloudpaper
phasedescribes
expressed
1-byte integers.
Unknown
mixed phase
(CMP)
developed
by
the
Meteorological
Satellite
Center.
CMP
is
part
of
the
Fundamental Cloud
Data are written in files following the order from
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
5. Validation
northwest tooperational
southeast with
since5,500
7 Julypixels
2015. and 5,500 lines.
The pixel sequence
is thealgorithm
same as is
that
of the
Himawari
The CMP
based
on the
cloud mask technique of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS
for GOES-R/ABI. Most cloud detection
involve threshold
methodsCenter
based onof Japan
The tests
Meteorological
Satellite
Standard Data
format.
radiative transfer calculation using NWP data as atmospheric
profiles.
The
thresholds
are
modified
Meteorological Agency carried out cloud phase validation
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
4.1. Cloud type values
using the MODIS level 2 product over a period of two
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
weeks
fromhit
20 ratios
July to
2 August
2015.product
The comparison
algorithm was applied to SEVIRI data, with results
showing
with
the MODIS
Cloud typewere
is expressed
as 1-byte
integers as indicated
indices used were the capture ratio (Eq. 5-1), the false
around 85%
for all seasons.
below.
（三行）
alarm ratio (Eq. 5-2) and the undetected error ratio (Eq.
1. Introduction
-128:
Error pixel associated with satellite data errors and
（一行）
similar
Himawari-8 is the new geostationary satellite of the
0: Clear pixel in cloud mask
Japan Meteorological Agency (JMA). It was launched on
1:
Extremely
high
7 October
2014
andopaque
begancloud
operation on 7 July 2015. The
2:
High opaque
satellite
carries cloud
the Advanced Himawari Imager (AHI),
3:
Relatively
highimproved
opaque cloud
theimagers
middle in
layer
which
is greatly
over in
past
terms of
its Relatively
number oflow
bands
and cloud
its temporal/special
resolution
4:
opaque
in the middle layer
(Bessho
et al.
2016). Using AHI data, JMA’s
5:
Low opaque
cloud
Meteorological
Satellite
6: Very low opaque cloud Center (MSC) developed the
Himawari-8 Cloud Mask Product (CMP).
7: Extremely low opaque cloud
Cloud mask is used to discriminate cloudy pixels
8: Extremely high semi-transparent cloud
from clear ones in satellite data. Numerous satellite
9:
High semi-transparent
cloud information. By way of
products
require cloud mask
10:
Relatively
high semi-transparent
cloud in
the middle
example,
Sea Surface
Temperature (SST)
(Yasuda
and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
layer
2009)
and Aerosol
Detection (Okawara
et al.
2003)
are
11:
Relatively
low semi-transparent
cloud
in the
middle
derived
in
clear
pixels.
Conversely,
Cloud
Grid
layer
Information (Tokuno 2002) is processed in cloudy pixels.
12: Low semi-transparent cloud
There were no MSC cloud mask products before the
13: Very low semi-transparent cloud
start of Himawari-8’s operation; each satellite product
14:
low semi-transparent
had Extremely
its own cloud
mask process. Ascloud
cloud mask is also
15:
Fractional
cloud
required for most Himawari-8 products, MSC decided to
develop CMP as a resource available for Himawari-8
products
in phase
common.
CMP is a part of the Fundamental
4.2.
Cloud
values
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
5-3) as calculated from a contingency table. Table 1
began operation.
shows
an example of an ice phase contingency table.
This paper describes the algorithm theoretical basis for
CMP.
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
（一行）
(𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 5 − 1)
Capture ratio =
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋
2. Algorithm
False
alarm ratio =
2.1 Overview
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
(𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜5 − 2)
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
Cloud detection in the CMP algorithm is based on
𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 data and clear sky data
comparisonerror
of ratio
observation
(𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜5 − 3)
Undetected
=
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
calculated from numerical+ 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋
weather prediction (NWP)
data. If observation data differ from clear sky data, the
pixel is cloudy. To discriminate between cloudy and clear
pixels, it is desirable for observation variables in clear
Table 1. Contingency table in case of ice phase
and cloudy conditions to differ and for the physical
OO represents
the number
of cases
which both
reason
behind the
difference
to bein explicit
(e.g.,
temperature,
reflectance
and
emissivity
of
the
top
and
Calipso or MODIS and cloud phase results indicate
transmittance
relating
to clouds
and the ofatmosphere).
ice phase. OX
represents
the number
cases in
Rather than encompassing retrieval of these five
which cloud
ice CMP
phasealgorithm
but Calipso
or
variables,
most type
of theindicates
tests in the
involve
MODIS
does not. XO
and XX or
represent
the data
inverse
the
use of brightness
temperature
reflectance
with
similar
tendencies. Cloud mask tests are based on the
situations.
cloud mask algorithm of the NoWCasting (NWC)
Satellite Application Facility (SAF) (Meteo-France,
2012) and the National Oceanic and Atmospheric
Administration
(NOAA)/National
Environmental
Satellite, Data, and Information Service (NESDIS)
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--27
1 --
－ 27 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
Figures 8 (a) and (b) show the results of cloud phase
References:
（二行）
validation. Although the capture ratio for the ice phase is
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
Baum, B. A., W. P. Menzel, R. A. Frey, D. Tobin, R. E.
（二行）
comparatively favorable, the undetected error ratio and
false alarm ratio are large for the unknown/mixed phase.
Holz, S.A. Ackerman, A. K. Heidinger, and P. Yang,
Ryo** cloud top property refinements for
The capture ratio for the latter phase is alsoIMAI
low.Takahito* and YOSHIDA
2012: MODIS
（三行）
MSC has not carried out validation for the cloud type
product because data in the Calipso and MODIS level 2
Collection 6.
J. Appl. Meteor. Clim., 51, 1145-1163.
Eyre J. R., 1991: A fast radiative transfer model for
products are not appropriate for comparison.
Abstractsatellite sounding systems., ECMWF Research Dept.
This paper describes the algorithm theoretical basis
forMemo.
the Himawari-8
Cloud Mask Product
Tech.
176.
(CMP) developed by the Meteorological Satellite Center.
CMP
is
part
of
the
Fundamental
Cloud
Hu Y., D. Winker, M. Vaughan,
B. Lin,
A. Omar, C.
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
Trepte, D. Flittner, P. Yang, S. L. Nasiri, B. Baum, R.
operational since 7 July 2015.
Holz, W.ofSun,
Z. Liu, Z.
S. Young,and
K. Stamnes,
The CMP algorithm is based on the cloud mask technique
NWC-SAF
forWang,
MSG/SEVIRI
NOAA/NESDIS for GOES-R/ABI. Most cloud detection
tests involve
based on
J. Huang,
and R.threshold
Kuehn, methods
2009: CALIPSO/CALIOP
radiative transfer calculation using NWP data as atmospheric
profiles.
The
thresholds
are
modified
Cloud Phase Discrimination Algorithm. J. Atmos.
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
Oceanic Technol., 26, 2293-2309.
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
(a) results
Imai
T., and hit
R. Yoshida,
2016:
theoretical basis
algorithm was applied to SEVIRI data, with
showing
ratios with
the Algorithm
MODIS product
for Himawari-8 Cloud Mask Product. Meteorological
were around 85% for all seasons.
（三行） Satellite Center Technical Note, 61, 1-17.
1. Introduction
（一行）
Himawari-8 is the new geostationary satellite of the
Japan Meteorological Agency (JMA). It was launched on
7 October 2014 and began operation on 7 July 2015. The
(b)
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
its number of bands and its temporal/special resolution
(Bessho et al. 2016). Using AHI data, JMA’s
Fig. 8 (a) Satellite
Comparison
of JMA
cloud
phase the
Meteorological
Center
(MSC)
developed
Himawari-8
CloudL2Mask
Product (CMP).
to MODIS
product
Cloud mask is used to discriminate cloudy pixels
Fig. 8 (b) Comparison of JMA cloud phase
from clear ones in satellite data. Numerous satellite
to Calipso
product
products
require L2
cloud
mask information. By way of
example, Sea Surface Temperature (SST) (Yasuda and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
2009) and Aerosol Detection (Okawara et al. 2003) are
derived in clear pixels. Conversely, Cloud Grid
Acknowledgements:
Information (Tokuno 2002) is processed in cloudy pixels.
There were no MSC cloud mask products before the
The authors gratefully acknowledge the two anonymous
start of Himawari-8’s operation; each satellite product
reviewers
helpful
this
had its ownwho
cloudprovided
mask process.
As comments
cloud mask on
is also
manuscript,
thank theproducts,
NASA Langley
Research
required for and
mostalso
Himawari-8
MSC decided
to
developAtmospheric
CMP as a resource
available
for for
Himawari-8
Center
Science Data
Center
providing
products
in common.
CMP
a part
of the
Fundamental
the
MODIS
and Calipso
dataisused
in the
study.
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
Meteo France, 2012: Algorithm Theoretical Basis
began
operation.
Document
for “Cloud Products” (CMa-PGE01 V3.2,
This
paper
describes
the algorithm v2.2).
theoretical basis for
CT-PGE02 v2.2, & CTTH-PGE03
CMP.
Moody, E. G., M. D. King, S. Platnick, C. B. Schaaf, and
（一行）
F.
Gao,
2005:
Spatially
complete global spectral
2. Algorithm
surface albedos: Value-added datasets derived from
MODIS land products. IEEE Trans. Geosci.
2.1Terra
Overview
Cloud detection
the CMP algorithm is based on
Remote
Sens., 43, in
144-158.
comparison
clearformulations
sky data
Nakajima, T.,ofandobservation
M. Tanaka,data
1986:and
Matrix
calculated
from
numerical
weather
prediction
(NWP)
for the transferof solar radiation in a plane-parallel
data. If observation data differ from clear sky data, the
scattering atmosphere. J. Quant. Spectrosc. Radiat.
pixel is cloudy. To discriminate between cloudy and clear
Transfer,
13-21. for observation variables in clear
pixels,
it is 35,
desirable
Pavolonis,
2010: GOES-R
Advanced
Imager
and
cloudyM.,
conditions
to differ
and forBaseline
the physical
(ABI)behind
Algorithm
For
reason
the Theoretical
difference toBasis
be Document
explicit (e.g.,
temperature,
emissivity of the top and
Cloud Typereflectance
and Cloud and
Phase.
transmittance
relating
to clouds
and the
Smith, M. N., G.
L. Smith,
S. N. Tiwari,
andatmosphere).
W. F. Staylor,
Rather
than
encompassing
retrieval
of
five
1998: Analytical forms of bidirectional these
reflectance
variables, most of the tests in the CMP algorithm involve
functions for application to Earth radiation budget
the use of brightness temperature or reflectance data with
studies, Journal of Geophysical Research, 103 (D16)
similar tendencies. Cloud mask tests are based on the
19733-19751.
cloud
mask algorithm of the NoWCasting (NWC)
Strabala,
K, S. A. Ackerman,
W. P.(Meteo-France,
Menzel, 1994:
Satellite I.Application
Facility and
(SAF)
2012)
the National
and Data.
Atmospheric
Cloudand
Properties
inferred Oceanic
from 8-12-μm
J. Appl.
Administration
(NOAA)/National
Environmental
Meteor., 33, 212-229.
Satellite,
Data,
and
Information
Service
(NESDIS)
Suzue, H., T. Imai, and K. Mouri, 2016: High-resolution
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--28
1 --
－ 28 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
Cloud Analysis Information derived from Himawari-8
（二行）
data. Meteorological Satellite Center Technical Note,
61, 43-51.
Skill score varies with the offset (Fig A-1). The offset
Algorithm Theoretical Basis for Himawari-8
Cloud Mask Product
value with which the highest skill score is achieved is
（二行）
adopted. In this example, the offset value is 2K.
IMAI Takahito* and YOSHIDA Ryo**
（三行）
Appendix:
Abstract
This paper describes the algorithm theoretical basis for the Himawari-8 Cloud Mask Product
Values for 𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶_𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏𝟏_𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐,
(CMP)
developed
by the Meteorological
𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶_𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖_𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆
𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆, Satellite Center. CMP is part of the Fundamental Cloud
𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶_𝟑𝟑𝟑𝟑𝟑𝟑𝟑𝟑_𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐,
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶_𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖_𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇
and since
𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶𝑶_𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖_𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘𝒘
operational
7 July 2015.
The CMP algorithm is based on the cloud mask technique of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS
for GOES-R/ABI.
Most cloud detection tests involve threshold methods based on
of offset values
is based on comparison
Determination
radiative
transfer
calculation
using
NWP
data as atmospheric profiles. The thresholds are modified
between the cloud classification of the Calipso cloud
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
phase product (Hu et al. 2009) and cloud type identified
Fig. A-1. was
Biasmore
values
for85%.
T39T112opaque
results showed that the CMP hit ratio derived from such comparison
than
The CMP and
using the algorithm
method described
in
Subsections
3.1.1
and
was applied to SEVIRI data, with results showing
ratios This
with the
skill hit
scores.
test MODIS
is for product
opaque cloud
3.1.2. The were
skill around
score is85%
adopted
asseasons.
a comparison index.
for all
detection at nighttime using T(3.9) – T(11.2)
（三行）
By way of example, the 2 x 2 contingency table shown
as described in 3.1.2.
below is used for the opaque cloud detection
began operation.
1.(𝑇𝑇𝑇𝑇39𝑇𝑇𝑇𝑇112𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜)
Introduction
described in Chapter 3.1.2. Skill
（一行）
This paper describes the algorithm theoretical basis for
score is defined using the following equation:
Himawari-8 is the new geostationary satellite of the
CMP.
In practice, the offset value is acquired for each value of
Japan Meteorological Agency (JMA). It was launched on
（一行）
𝑇𝑇𝑇𝑇(11.2)
and
surface
land
type. This offset tables are
(𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂
+
𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋)
−
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆
7𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆October
2014 and began operation on 7 July 2015. The
2. Algorithm
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 =
𝑁𝑁𝑁𝑁 𝑁Advanced
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆
shown in Figs A-2 (a), A-2 (b), A-2 (c), A-2 (d) and A-2
satellite carries the
Himawari Imager (AHI),
(e).
In this study, data from Meteosat-9’s SEVIRI
which is greatly improved over past imagers in terms of
2.1 Overview
(𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋)
(𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋)
× (𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 +
𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂)its+ temporal/special
× (𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 +
𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋)
Sc =number of bands
Cloudwere
detection
in the CMP
algorithm
is based The
on
its
and
resolution
imager
used because
𝑇𝑇𝑇𝑇(8.6)
was required.
𝑁𝑁𝑁𝑁
𝑁𝑁𝑁𝑁
comparison
of an
observation
data of
andeight
clearweeks
sky data
(Bessho et al. 2016). Using AHI data, JMA’s
study covered
overall period
(two
𝑁𝑁𝑁𝑁 = 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 + 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋
+ 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 Center (MSC) developed the
calculated
from
numerical
weather
prediction
(NWP)
Meteorological
Satellite
weeks during each of winter 2011 – 2012 (December and
data. If observation data differ from clear sky data, the
Himawari-8 Cloud Mask Product (CMP).
January), spring 2012 (March and April), summer 2012
pixel is cloudy. To discriminate between cloudy and clear
Cloud mask is used to discriminate cloudy pixels
(June and July) and autumn 2012 (September and
pixels, it is desirable for observation variables in clear
from
ones in satellite
Tableclear
A-1. Contingency
table data. Numerous satellite
October)).
the future, to
thediffer
offset and
valueforwill
updated
and
cloudyInconditions
thebephysical
products
require cloud
mask of
information.
By way
OO represents
the number
cases in which
bothof
based onbehind
additional
of accumulated
data. (e.g.,
reason
theyears
difference
to be AHI
explicit
example, Sea Surface Temperature (SST) (Yasuda and
Calipso
and
cloud
type
results
indicate
opaque
cloud.
temperature, reflectance and emissivity of the top and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
OX represents
theDetection
number of
cases inetwhich
cloud
transmittance relating to clouds and the atmosphere).
2009)
and Aerosol
(Okawara
al. 2003)
are
Rather than encompassing retrieval of these five
derived
in
clear
pixels.
Conversely,
Cloud
Grid
type indicates opaque cloud but Calipso does not. XO
variables, most of the tests in the CMP algorithm involve
Information (Tokuno 2002) is processed in cloudy pixels.
and XX represent the inverse situations.
the use of brightness temperature or reflectance data with
There were no MSC cloud mask products before the
similar tendencies. Cloud mask tests are based on the
start of Himawari-8’s operation; each satellite product
cloud mask algorithm of the NoWCasting (NWC)
had its own cloud mask process. As cloud mask is also
Satellite Application Facility (SAF) (Meteo-France,
required for most Himawari-8 products, MSC decided to
2012) and the National Oceanic and Atmospheric
develop CMP as a resource available for Himawari-8
Administration
(NOAA)/National
Environmental
products in common. CMP is a part of the Fundamental
Satellite, Data, and Information Service (NESDIS)
Cloud Product (FCP), which contains cloud phase, type
(Heidinger and Straka III, 2012).
and top height for Himawari-8/AHI pixels (Mouri et al.
Radiative transfer calculation using NWP data for
2016a, 2016b). FCP has been produced since Himawari-8
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--29
1 --
－ 29 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
（二行）
Algorithm Theoretical Basis for Himawari-8 Cloud Mask Product
（二行）
IMAI Takahito* and YOSHIDA Ryo**
（三行）
Abstract
This paper describes the algorithm theoretical basis for the Himawari-8 Cloud Mask Product
(CMP) developed by the Meteorological Satellite Center. CMP is part of the Fundamental Cloud
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
operational since 7 July 2015.
The CMP algorithm is based on the cloud mask technique of NWC-SAF for MSG/SEVIRI and
NOAA/NESDIS for GOES-R/ABI. Most cloud detection tests involve threshold methods based on
radiative transfer calculation using NWP data as atmospheric profiles. The thresholds are modified
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
algorithm was applied to SEVIRI data, with results showing hit ratios with the MODIS product
were around 85% for all seasons.
（三行）
1. Introduction
began operation.
（一行）
This paper describes the algorithm theoretical basis for
Himawari-8 is the new geostationary satellite of the
CMP.
Japan Meteorological Agency (JMA). It was launched on
（一行）
7 October 2014 and began operation on 7 July 2015. The
2. Algorithm
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
2.1 Overview
Cloud detection in the CMP algorithm is based on
its number of bands and its temporal/special resolution
comparison of observation data and clear sky data
(Bessho et al. 2016). Using AHI data, JMA’s
calculated from numerical weather prediction (NWP)
Meteorological Satellite Center (MSC) developed the
data. If observation data differ from clear sky data, the
Himawari-8 Cloud Mask Product (CMP).
pixel is cloudy. To discriminate between cloudy and clear
Cloud mask is used to discriminate cloudy pixels
pixels, it is desirable for observation variables in clear
from clear ones in satellite data. Numerous satellite
and cloudy conditions to differ and for the physical
products require cloud mask information. By way of
reason behind the difference to be explicit (e.g.,
example, Sea Surface Temperature (SST) (Yasuda and
temperature, reflectance and emissivity of the top and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
transmittance relating to clouds and the atmosphere).
2009) and Aerosol Detection (Okawara et al. 2003) are
Rather than encompassing retrieval of these five
derived in clear pixels. Conversely, Cloud Grid
variables, most of the tests in the CMP algorithm involve
Information (Tokuno 2002) is processed in cloudy pixels.
the use of brightness temperature or reflectance data with
There were no MSC cloud mask products before the
similar tendencies. Cloud mask tests are based on the
start of Himawari-8’s operation; each satellite product
cloud mask algorithm of the NoWCasting (NWC)
had its own cloud mask process. As cloud mask is also
Satellite Application Facility (SAF) (Meteo-France,
required for most Himawari-8 products, MSC decided to
Fig. A-2
2012) and the National Oceanic and Atmospheric
develop CMP as a resource available for Himawari-8
(a)
Bias
values
for
T3.9T11.2opaque
(b)
Bias
values
for
T8.6T11.2fractional
Administration
(NOAA)/National
Environmental
products in common. CMP is a part of the Fundamental
Satellite,
Data,
and
Information
Service
(NESDIS)
Cloud
Product
(FCP),
which
contains
cloud
phase,
type
(c) Bias values for T11.2T12.3opaque (d) Bias values for ice phase (e) Bias values for water phase
(Heidinger and Straka III, 2012).
and top height for Himawari-8/AHI pixels (Mouri et al.
Determination of these values is based on land type and T(11.2)
Radiative transfer calculation using NWP data for
2016a, 2016b). FCP has been produced since Himawari-8
* Office of Observation Systems Operation, Observation Department, Japan Meteorological Agency
** System Engineering Division, Data Processing Department, Meteorological Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
--30
1 --
－ 30 －
METEOROLOGICAL SATELLITE CENTER TECHNICAL NOTE, No.61, MARCH, 2016
雲タイプ・相アルゴリズム記述書
（二行）
Algorithm
Theoretical Basis for Himawari-8 Cloud Mask Product
毛利 浩樹*、泉 敏治**、鈴江 寛史*、吉田 良*
（二行）
IMAI Takahito* and YOSHIDA Ryo**
（三行）
要旨
雲タイプ、相プロダクトは基本雲プロダクトを構成する１つである。基本雲プロダクトは雲
マスク、雲タイプ、雲の相、および雲頂高度からなる。気象庁ではひまわり８号を機に様々な２
Abstract
次プロダクトが含んでいる雲マスク、雲タイプ、相、雲頂高度プロダクトを独立させ、基本雲プ
This paper describes the algorithm theoretical basis for the Himawari-8 Cloud Mask Product
ロダクトとして一つにまとめ上げた。雲タイプ、相プロダクトは不 f 透明、半透明、断片雲とい
(CMP) developed by the Meteorological Satellite Center. CMP is part of the Fundamental Cloud
った雲タイプと、水、氷、未定・混合といった雲の相を出力する。解析手法は過去に実績があり、
Product, which incorporates cloud phase, type and top altitude from Himawari-8/AHI and has been
比較的理解しやすい NWCSAF の雲タイプアルゴリズムに主に基づいている。雲の相に関して、
operational since 7 July 2015.
2015 年 7 月 20 日から 8 月 2 日において気象庁気象衛星センタープロダクトと MODIS
（MYD06L2）
The CMP algorithm is based on the cloud mask technique of NWC-SAF for MSG/SEVIRI and
プロダクト及び Calipso（CAL_LID_L2_01kmCLay-ValStage1-V3-30）との比較を実施したところ、
NOAA/NESDIS for GOES-R/ABI. Most cloud detection tests involve threshold methods based on
氷相では 80%以上の捕捉率となった。
radiative transfer calculation using NWP data as atmospheric profiles. The thresholds are modified
using offsets determined on the basis of comparison with the MODIS cloud mask product. Initial
results showed that the CMP hit ratio derived from such comparison was more than 85%. The CMP
algorithm was applied to SEVIRI data, with results showing hit ratios with the MODIS product
were around 85% for all seasons.
（三行）
1. Introduction
（一行）
Himawari-8 is the new geostationary satellite of the
Japan Meteorological Agency (JMA). It was launched on
7 October 2014 and began operation on 7 July 2015. The
satellite carries the Advanced Himawari Imager (AHI),
which is greatly improved over past imagers in terms of
its number of bands and its temporal/special resolution
(Bessho et al. 2016). Using AHI data, JMA’s
Meteorological Satellite Center (MSC) developed the
Himawari-8 Cloud Mask Product (CMP).
Cloud mask is used to discriminate cloudy pixels
from clear ones in satellite data. Numerous satellite
products require cloud mask information. By way of
example, Sea Surface Temperature (SST) (Yasuda and
Shirakawa 1999), Clear Sky Radiance (CSR) (Uesawa
2009) and Aerosol Detection (Okawara et al. 2003) are
derived in clear pixels. Conversely, Cloud Grid
Information (Tokuno 2002) is processed in cloudy pixels.
There were no MSC cloud mask products before the
start of Himawari-8’s operation; each satellite product
had its own cloud mask process. As cloud mask is also
required for most Himawari-8 products, MSC decided to
develop CMP as a resource available for Himawari-8
products in common. CMP is a part of the Fundamental
Cloud Product (FCP), which contains cloud phase, type
and top height for Himawari-8/AHI pixels (Mouri et al.
2016a, 2016b). FCP has been produced since Himawari-8
* 気象衛星センターデータ処理部システム管理課
began operation.
This paper describes the algorithm theoretical basis for
CMP.
（一行）
2. Algorithm
2.1 Overview
Cloud detection in the CMP algorithm is based on
comparison of observation data and clear sky data
calculated from numerical weather prediction (NWP)
data. If observation data differ from clear sky data, the
pixel is cloudy. To discriminate between cloudy and clear
pixels, it is desirable for observation variables in clear
and cloudy conditions to differ and for the physical
reason behind the difference to be explicit (e.g.,
temperature, reflectance and emissivity of the top and
transmittance relating to clouds and the atmosphere).
Rather than encompassing retrieval of these five
variables, most of the tests in the CMP algorithm involve
the use of brightness temperature or reflectance data with
similar tendencies. Cloud mask tests are based on the
cloud mask algorithm of the NoWCasting (NWC)
Satellite Application Facility (SAF) (Meteo-France,
2012) and the National Oceanic and Atmospheric
Administration
(NOAA)/National
Environmental
Satellite, Data, and Information Service (NESDIS)
(Heidinger and Straka III, 2012).
Radiative transfer calculation using NWP data for
気象研究所気象衛星・観測システム研究部
***
Office
of Observation Systems Operation, Observation Department, Japan Meteorological Agency
- 31 ** System Engineering Division, Data Processing Department, Meteorological
Satellite Center
(Received September 1, 2015, Accepted 20 November 2015)
-1-
－ 31 －