Background on passive microwave, visible and infrared images from

Background on passive microwave, visible and
infrared images from a meteorological point of view.
Study and application of a rain rate retrieval algorithm
and a numerical weather prediction model.
Author: Lic. Lighezzolo Rafael Andrés
Supervisor: Dr. Vincenzo Levizzani
June 2012
Abstract
This report was performed in the Istituto di Scienze dell'Atmosfera e del Clima (ISAC) of the
Consiglio Nazionale delle Ricerche (CNR) di Bologna ITALY in the framework of Progetto
Sistema Italo-Argentino per la Gestione delle Emergenze (SIASGE).
The report shows the work performed based on the work program proposed which is attached in the
end of the report and begins with a general introduction about satellites and their meteorological
application from the point of view of precipitation retrievals, after that the report can be divided in
three parts. The first part is about the use of a precipitation retrieval algorithm based in passive
microwave images analysis from the AMSU-B sensor. In this part describes basics concepts of
microwave sensed from satellite, the sensitivities of the different AMSU-B channels with the
precipitation, the study of the 183-WSL fast rain rate retrieval algorithm, an analysis of passive
microwave images and the application of the 183-WSL algorithm. The supervisor in these topics
was Dr. Sante Laviola. The second part is an analysis of visible and infrared images from the
SEVIRI sensor channels. The properties of each channel and their uses in meteorological
applications are described. The most used RGB combinations are described too. In some cases
EUMETSAT data (processed with a visualization tool exe_tools_msgreader-11) are used as further
examples. The supervisor in these topics was Dr. Elsa Cattani. In the third part, there is a brief
introduction to numerical weather predictions and a description of different models the hydrostatic
model called BOLAM and the non-hydrostatic model called MOLOCH. Then some study cases are
performed over Ligurian Region in Italy. These study cases were edited and will be present in the
Congreso Argentino de Teledectección 2012 (CAT). Both models were run over Argentina and
together with other information as, meteorological data, infrared and passive microwave images
were considered as a study case over Argentina. The supervisor in these topics was Dr. Silvio
Davolio.
Acknowledgments
I would like to thanks all the ISAC team who from the first day made me feel very comfortable.
Thanks:
Dr. Vincenzo Levizzani
Dr. Sante Laviola
Dr. Elsa Cattani
Dr. Silvio Davolio
Claudia Acquistapace
Alice Malvaldi
Contents
1. Introduction
1.2. Measurement principles of remote sensors
1.3. Satellite orbits
1.4. Satellite precipitation retrievals
1.5. Precipitation retrieval methods
1.5.1. VIS/IR methods
1.5.2. Passive microwave methods
1.5.3. Active microwave methods
1.5.4. Multi-sensor techniques
2. Passive microwave remote sensing of rain satellite sensors
2.1. Fundamental principles of microwave remote sensing from satellite sensors
2.2. The Radiative Transfer Equation
2.3. Impact of precipitation on microwave measurement
2.4. The sensitivity at 183.31 GHz to surface emissivity and rainy cloud altitudes
2.5. High frequency method to retrieve rainrates (183-WSL)
3. The 183-WSL algorithm
3.1. Window frequencies: rain/no-rain identification and convective-stratiform
rain classification
3.2. Rain rate estimation
3.3. Application of a Probability Of Rain Development Function (PORDF)
3.4. Liquid water path
4. Training in passive microwave images and the application of the 183-WSL algorithm
4.1. Observational analysis from infrared and microwaves of a cloudy scenario
4.2. AMSU-B channels
4.3. Land and cloud features with AMSU-B channels
4.3.1. Surface features
4.3.2. Cloud features
4.3.3. Combination of AMSU-B channels
4.4. Precipitation producs
5. Visible and Infrared images from SEVIRI instrument, Description of characteristics
and content of the channels
5.1. Description and Interpretation of the channels.
5.1.1. Solar channels
5.1.2. The Ch4: 3.9 µm (IR)
5.1.3. Window Channels
5.1.4. Absorption Channels
5.1.5. Ozone channel
5.1.6. CO2 channel
5.1.7. The Seviri high-resolution visible (HRV) channel
6.1. RGB images composite
6.2. Cloud Physical Properties represented by the MSG Channels
6.3. Recommended schemes for RGB image composites
1
1
1
2
3
3
3
4
4
5
5
7
12
12
13
13
14
17
18
20
20
20
23
25
25
26
29
33
36
38
38
41
44
45
45
46
46
46
47
47
6.3.1. RGB 03, 02, 01 ("Day Natural Colours")
6.3.2. RGB 02,04r,09 ("Day Microphysical")
6.3.3. RGB 02, 03, 04r ("Day Solar")
6.3.4. RGB 05-06, 04-09, 03-01 ("Convective Storms")
6.3.5. RGB 10-09, 09-04, 09 ("Night Microphysical")
6.3.6. RGB 10-09, 09-07, 09 ("Day and Night and Desert Dust")
6.3.7 RGB 05-06, 08-09, 05i ("Airmass")
7. Numerical weather prediction
7.1. BOLAM and MOLOCH Models
7.2. BOLAM and MOLOCH schemes
7.3. Running BOLAM and MOLOCH
7.4. Study cases applying BOLAM and MOLOCH simulations
7.4.1. Meteorological analysis
7.4.2. Precipitation
7.4.3. Numerical simulations
7.4.4. Common features in MCS events
7.5. BOLAM and MOLOCH running over Argentina, setup the models
7.6. Study case over Argentina applying BOLAM and MOLOCH simulations
7.6.1 Meteorological analysis
7.6.2. Numerical simulations
7.6.2. 5 of April, accumulated precipitation study case
8.1. General conclusion
Appendix
A.1 Weighting Function
A.2 Information about Advanced Microwave Sounding Unit-B (AMSU-B)
48
48
49
50
50
51
52
53
54
55
56
60
60
61
63
67
69
73
73
76
81
85
87
87
88
1. Introduction
Information from remote sensors and remote sensing system is today an important and often
indispensable data source for meteorologists working in operational weather forecasting. Here, the
primary use of the data is for visualizing and monitoring the present weather situation with high
spatial and temporal resolution. The impact of the data on forecast quality is large for the very short
forecast range (0-12 hours) but it is not easy to assess it in detail due to the subjective way the data
are used.
Substantial progress has also recently been made in the quantitative use of remotely sensed
information. This concerns the use in numerical weather prediction (NWP) models and the use for
diagnosing environmental and climate conditions. A third important aspect is the use in research
applications, investigating the detailed structure of the atmosphere, especially the structure of
hazardous and violent weather phenomena.
1.2. Measurement principles of remote sensors
Remote sensors can be grouped into two categories depending on the used method to measure
radiation. These are: passive sensors and active sensors.
Passive sensors measure the incoming radiation to a detector without any manipulation or
modulation of radiation amounts. The only objective here is to measure incoming radiation
emanating from the earth surface and the atmosphere itself. Passive sensors, normally termed
radiometers, dominate on space-based platforms where they are used for radiation budget
investigation, for imaging purposes and for extraction of atmospheric vertical profiles. They are
also frequently used from aeroplanes and upward-looking instruments on the ground. Measurement
are usually confined to very limited spectral intervals (often named spectral channels or spectral
bands) but this can be extended either by introducing a multi-channel capacity or by introducing a
broad-band sensor response.
Active sensors are designed to measure radiation from targets that are difficult to study when
applying passive measurements. The naturally emitted or reflected radiation amounts from such
targets may be too small to be measurable even if there are no intervening radiation sources
between the target and the sensor. Typical examples of such targets are large hydrometeors
(precipitation) and earth surface. The idea of an active sensor is to transmit a short pulse of
electromagnetic radiation and measure the returned reflected radiation (the backscatter signal). The
frequency of the transmitted radiation must naturally be chosen in a region where the specific
targets reflect and where contributions from the intervening atmosphere can be expected to be
small. Energy levels must additionally be high enough to dominate over “background” radiation
levels.
1.3. Satellite orbits
Geostationary (GEO) satellites are in an orbit such that they rotate around the Earth at the same
speed as the Earth rotates, thus appearing stationary relative to a location on the Earth. This enables
each geostationary satellite to view about one third of the Earth’s surface on a frequent and regular
basis. A total five geostationary satellites provide operational imagery: these currently include the
Meteosat Second Generation satellites (MSG) from the European Organization for the Exploitation
of Meteorological Satellites (EUMETSAT), two Geostationary Operational Environmental
Satellite (GOES) and the Japanese Multifunctional Transport Satellites (MTSAT) series. They share
a number of common attributes: they typically carry visible (VIS) and infrared (IR) sensors with
nadir resolutions of about 1×1 km and 4×4 km, respectively, acquiring images nominally every 30
min. These satellites provide an unrivalled platform for continual observation but are limited in the
Polar Regions by the unfavorable viewing angle at high latitudes.
Low-Earth orbiting (LEO) satellites generally cross the Equator at the same local time on each
orbit, providing about two overpasses per day. These satellites carry a range of instruments capable
of precipitation retrievals, including multichannel VIS/IR sensors, and passive microwave (PMW)
sounders and imagers. Current operational polar-orbiting satellites include the National Oceanic and
Atmospheric Administration (NOAA) series of satellites with NOAA-19 and EUMETSAT’s
MetOp series. Observations made by these satellites are typified by wide swaths (2800 km) with
resolutions of about 1 km.
1.4. Satellite precipitation retrievals
Precipitation is a major component of the global water and energy cycle, helping to regulate the
climate system. Moreover, the availability of fresh water is vital to life on Earth. Measurement of
global precipitation through conventional instrumentation uses networks of rain (or snow) gauges
and, where available, weather radar systems. However, the global distribution of these is uneven:
over the land masses the distribution and density of gauges is highly variable with some regions
having “adequate” coverage while others have few or no gauges. Over the oceans few gauges exist,
and those located on islands might be subject to local influences and therefore not representative of
the surrounding ocean. The availability of historical precipitation data sets can also be problematic,
varying in availability, completeness and consistency as well availability for near real-time analysis.
Satellites offer an unrivalled vantage point to observe and measure Earth system processes and
parameters. Precipitation (rain and snow) in particular, benefit from such observations since
precipitation is spatially and temporally highly variable and with satellites overcoming some of the
deficiencies of conventional gauge and radar measurement.
A number of other operational and “research and development” satellite missions are also used for
precipitation estimation. The Defense Meteorological Satellite Program (DMSP) satellites series of
Special Sensor Microwave/Imager (SSM/I) and Special Sensor Microwave Imager-Sounder
(SSMIS) provide measurements of naturally emitted microwave (MW) radiation from the Earth and
its atmosphere. The AQUA satellite includes the PMW Advanced Microwave Scanning
Radiometer-Earth Observing System (EOS) (AMSR-E). PMW sensors typically have swath widths
in the order of 1500 km and resolutions, variable with observational frequency, of between 4 km
and about 50 km. In addition, the AQUA mission also carries the VIS/IR MODerate-resolution
Imaging Spectroradiometer (MODIS), the latter also carried onboard the TERRA satellite. The
MODIS instrument, with 36 VIS/IR spectral bands enables information on cloud properties (such as
cloud drop radii and phase) to be retrieved.
The Tropical Rainfall Measuring Mission (TRMM), launched in 1997, was the first dedicated
precipitation satellite and has been key in the development and improvement of satellite rainfall
estimation techniques. It carries a range of instruments which allow direct comparisons to be made
between VIS, IR, PMW and active MW observations. The Precipitation Radar (PR) was the first
spaceborne precipitation radar, capable of sampling precipitation both vertically and horizontally,
although with a limited swath of 210 km. Other instruments include the TRMM Microwave Imager
(TMI), the Visible and InfraRed Scanner (VIRS), and the Lightning Imaging Sensor (LIS). The non
sun-synchronous nature of its orbit allows samples across the full diurnal cycle to be made. Other
current active MW sensors include the nadir-pointing Cloud Profiling Radar (CPR) on the CloudSat
mission.
1.5. Precipitation retrieval methods
Satellite estimates of precipitation can be derived from a range of observations from many different
sensors. The retrieval methodologies fall primarily into three main categories based upon type of
observation, primarily VIS/IR techniques, passive and active microwave techniques, and multisensor techniques.
1.5.1. VIS/IR methods
In the VIS part of the spectrum clouds appear relatively bright against the surface of the Earth due
to their high albedo. Rainfall can be inferred from VIS images since bright clouds tend to be thick,
and thick clouds are more likely to be associated with rainfall. However, the relationship between
brightness and rainfall is poor and consequently VIS imagery is usually only used in conjunction
with other observations. Cloud-top texture derived from VIS imagery can provide useful
information: stratus is typically smooth, while convective clouds tend to be more “lumpy”.
Properties of cloud-top particles, (size, phase) can be obtained from multi-channel near-IR (nIR)
data. The use of reflected/emitted radiances around 1.6, 2.1 and 3.9 μm have proved very useful in
studying the microphysical properties of clouds, despite some being restricted to full daylight
operations. IR imagery that measures the thermal emissions from objects is potentially more useful,
and is available night and day. Heavier rainfall tends to be associated with larger, taller clouds with
colder cloud tops. By observing cloud top temperatures (CTT) a simple rainfall estimate can be
derived. However, the CTT to rainfall relationship is indirect, with significant variations in the
relationship during the lifetime of a rainfall event, between rain systems, and between
climatological regimes. Nevertheless, IR-based techniques benefit from a degree of simplicity
coupled with the 24 h availability of the data.
1.5.2. Passive microwave methods
The Earth naturally emits low levels of MW radiation which may be passively sensed by
radiometers. This radiation is primarily attenuated by the presence of precipitation-sized particles.
Two processes can be used to identify precipitation: emission from rain droplets which lead to an
increase in MW radiation and scattering caused by precipitating ice particles which leads to a
decrease in MW radiation. The background radiometric signal over water is low and constant
(emissivity,  ≈ 0.4–0.5), therefore additional emissions from precipitation can be used to identify
and quantify the rainfall using low-frequency channels (<20 GHz). Over land, the surface has a
much higher background emissivity ( ≈ 0.9) and consequently emissions from hydrometeors
cannot be reliably measured. Here, scattering caused by ice particles, resulting in a decrease in
received radiation at high frequencies (>35 GHz), must be utilized. Over land stringent schemes
must be employed to ensure the surface background does not contaminate any precipitation
retrieval.
1.5.3. Active microwave methods
Active MW techniques are perhaps the most direct method of precipitation estimation. However,
the use of spaceborne precipitation estimation has been very limited with only the TRMM PR being
specifically designed for retrieving precipitation characteristics. As with all radar systems, the PR
relies upon the interpretation of the backscatter of radiation from precipitation, which is broadly
proportional to the number of precipitation-sized particles and therefore the intensity. However, the
precipitation intensity to backscatter relationship is not constant. Nevertheless, the PR has been
extensively used as a primary source of high-quality rainfall estimates for evaluating the differences
of rainfall regimes over land and over the ocean. Combined radar-radiometer algorithms have been
developed to exploit their intrinsic ambiguities. Grecu et al. (2004) investigated vertical profiles of
precipitation from multifrequency, multi-resolution active and passive MW observations. Viltard et
al. (2006) proposed a rainfall retrieval technique for the TMI using a database of about 35000
brightness temperature (Tb) vectors collocated with the corresponding PR rain rate profiles.
Recently the CloudSat Cloud Profiling Radar (CPR; Stephens et al., 2008) has shown a
precipitation retrieval capability. The CPR operates at 94 GHz and is much more sensitive to cloud
hydrometeors, tending to saturate in regions of dense cloud or rainfall. However, through the use of
attenuation-correction algorithms and surface reflectivity modelling, it has proved to be very useful
in the identification of light rainfall and snowfall. Haynes et al. (2009) demonstrated a notable
rainfall retrieval potential at mid-latitudes over the ocean. For rain rates less than 0.8mmh−1 the
CPR produces nearly three times the rainfall occurrence than that sensed by the TRMM PR,
resulting in the production of a more representative precipitation intensity distribution function.
1.5.4. Multi-sensor techniques
A growing number of techniques are being developed to exploit the synergy between the polarorbiting PMW retrievals (infrequent, more direct) with the geostationary observations (frequent,
less direct). Techniques have been developed to adjust the IR, or generate calibration curves to map
IR radiances using the other data sets, such as radar, gauge or other satellite data set.
These techniques are ultimately limited by the indirectness of the IR to sense rainfall itself.
However, the IR data can provide a reasonable measure of cloud movement, which can then be used
to advect or morph the more direct PMW data between the successive satellite overpasses. Such
techniques came under the category of motion-based techniques.
Lightning data, from ground based detection networks or from space sensors, has proved useful in
improving precipitation retrievals.
2. Passive microwave remote sensing of rain satellite sensors
Remote sensing from satellite-based sensors has established itself as the key method for the
observation and monitoring of the planet Earth because of several reasons. From a practical point of
view, observations from orbital platforms guarantee continuous measurements of very wide regions
of the globe especially over inaccessible areas such as impervious mountains or open oceans where
measurement campaigns are often difficult, expensive and even dangerous. Scientifically speaking,
since a satellite platform generally ”carries” a large instrumental suite covering several frequencychannels, very diverse observations and studies can be planned at same time exploiting data from a
single space mission. Finally, especially for last generation of geostationary and low-orbiting
satellites, orbital sounding can be done at high temporal and spatial resolutions both day and night
allowing for a continuous monitoring of a phenomenon during all its evolutional stages. Satellite
rainfall retrieval acquires even more value when associated to more conventional observing
systems, which were inadequate to correctly quantify precipitating phenomena both at the local and
global scales. A variety of schemes using VIS and IR satellite data have been applied to the problem
of precipitation estimation (Levizzani, 2003; Levizzani et al., 2007; Hsu et al., 2007) and to retrieve
cloud microphysical properties (Rosenfeld, 2007; Cattani et al., 2007). Nevertheless, due to the very
low penetrating capabilities of dense medium approaches based on these are restricted to
observations of cloud top. For this reason, their implementation into retrieval techniques of
precipitation often result in erroneous estimations of rain amount especially during light rain
episodes when cloud top temperatures are comparable to those of non-rainy clouds.
Microwave radiation is, in general, affected in a minor way by most clouds, particularly for
wavelengths above the centimeters. From the precipitation retrieval perspective, microwaves offer
the enormous advantage to guarantee a direct relationship between the radiation field and the bulk
of hydrometeors “illuminated” by the upwelling radiation. On this basis, many retrieval techniques
originally thought for VIS and IR wavelengths were implemented with microwave frequencies
when they became available at the end of the 1970s. Skillful combinations of observations from
high spatial and temporal resolution VIS-IR geostationary sensors with rainfall-measuring
capabilities of microwave radiometers are employed in the so called ”blended techniques”. Such
methods, based on the calibration of IR brightness temperatures with microwave rain signatures
(e.g., Turk & Mehta, 2007) or on the advection of microwave precipitation by using thermal IR
signal (Joyce et al., 2004), are normally used to reconstruct precipitation tracks over regions not
regularly covered by polar orbiting satellites. Therefore, it is quite intuitive to adopt improvement
strategies microwave precipitation retrievals for their input to statistical methods as well as to
numerical models.
2.1. Fundamental principles of microwave remote sensing from satellite sensors
Passive microwave remote sensing, which has largely developed over the last few years thanks to
the improvement in the spatial resolution and in the variety of channel-frequencies on board of last
generation sensors, can be summarized using a few keywords: quasi-transparency, surface
emissivity, absorption-scattering. This oversimplification of the problem is not due to the fact that
microwave sounding is simpler than that based on other approaches such as that based on IR or VIS
wavelengths, but it refers to some aspects of the observed scene that are more unambiguously
identified with microwave frequencies than with other shortwave methods. This is due to the
existing direct link between the measured observable and the perturbation induced to the radiation
field along the wave propagation path. This signal displacement at the nominal frequency value is
directly correlated to the bulk of the observed parameter.
Microwave measurements from space are extremely sensitive to a wide range of surface and
atmospheric properties. Surface type, different terrain coverage and local variation of atmospheric
parameters, can deeply affect space observations often introducing errors in the retrieval methods
because of the incorrect evaluation of local conditions.
Unlike the thermal IR spectral region where the black body approximation often well describes the
real behavior of emitters, in the microwaves an emitting surface must be considered as a grey body
so that its emissivity value is usually lower than unit. The observed variability in microwave
radiances for homogeneous land surfaces is normally caused by variations in skin temperature and
surface emissivity, while the variability for open seawater is attributed to the atmospheric
constituents such as columnar water vapor, temperature profiles and presence of cloud liquid water.
The land surface emissivity being higher (ε ≈ 0.80-1.00) than ocean’s (ε ≈ 0.40-0.60) appears as a
“warm” object. Nevertheless, unlike for the ocean, land emission variability is strictly linked to the
strong temporal and spatial variations of soil features as roughness, vegetation cover and moisture
content. It is thus very complex to model surface properties in the microwave from arid surfaces to
dense vegetation or snow and consequently it is difficult to discern between the surface and
atmospheric contributors to the upwelling radiation. Over open ocean the substantially stable and
uniform “cold” background emphasizes more the extinction of upwelling radiation by atmospheric
constituents and the contribution of various elements to the total radiation depression are reasonably
well separated. Sea surface emissivity is largely determined by dielectric properties of seawater
through the Fresnel equation and, especially for a drier atmosphere, the surface has a larger effect
on the measured radiance. Many authors have developed models to predict the dielectric constant of
seawater in order to improve the retrieval method of atmospheric parameters.
An example of the radiometric response of the NOAA/AMSU-B frequency range 89-190 GHz to
the emission for several surface and atmospheric contributors is reported in Fig. 1. An analysis of
the images in the window frequencies at 89 GHz (left) and 150 GHz (middle) evidences the striking
contrast between land and open water numerically denounced by a brightness temperature
discrepancy over 50 K at 89 GHz. Similarly, the coldest background widely enhances the presence
of cloud liquid water at 89 GHz close to Spanish, Italian and Northern Europe coastlines. This
characteristic is attenuated at 150 GHz, whose weighting function “peaks” (see appendix A.1)
around 1 km above surface, and thus warmer atmospheric layers partially mask cloud liquid
signatures. An interesting aspect of Fig. 1 is related to the land emissivity changes. Observing the
image at 150 GHz a brightening structure is extensively distributed in the middle of the image. In
the same location but at 89 GHz this region is related to the Alps and Apennines whereas at 190
GHz (right) it almost disappears except over higher mountain tops. The similarity between satellite
images and daily snow cover map unmistakably suggests that snowy terrain is the main responsible
of significant reduction of the Earth’s emissivity. This assertion is also corroborated by mixing ratio
measurements retrieved by three sample radiosonde stations. As a consequence of these drier
conditions the weighting function lowers close to the surface largely enhancing the effects of
scattering by ice particles of fallen snow. The final result is that the brightness temperature of the
upwelling radiation reaching the satellite drastically decreases from 40 K to 70 K over the Alps and
Apennines. In addition, it must be said that the signal extinction of snow cover at 150 GHz is quite
similar to that of scattering by ice on cloud top with an enormous errors during rain pixel
classification. A different behavior is shown in the 89 GHz channel, where the upward radiance
varies from 20 K to 70-80 K over mountain with increasing surface roughness. Finally, the 190
GHz channel sounding the absorption of water vapor around 2 km in general is less affected by
surface emissivity variations. Nevertheless, when local dry condition establish this frequency senses
closer to the surface and it can sense more surface effects.
Figure 1
2.2. The Radiative Transfer Equation
The radiative transfer equation is a mathematical description of the spatial-angular distribution of
monochromatic radiation intensity Iν which, at a certain instant t and at the frequency band ν,
propagates into a medium across cross section A, in the observation direction Ω along the path s.
The intensity of radiation varies while this passes through the medium. In particular, the energy of
the incoming beam will decrease due to the absorption by the medium substance and to the
deviation of a fraction of the radiation from the original trajectory due to the scattering in all
directions. At the same time, the thermal radiation emission by the volume of material will enhance
the energy balancing the net energy flux losses by the extinction processes.
If we consider an elementary volume dAdS in the form of a cylinder with the main axis coincident
with the radiation path s (Fig. 2), the variation of flux intensity when the incoming radiation passes
through the elementary path ds is represented by the quantity:
where dA, dΩ, dν and dt correspond to elementary crossed surface, solid angle of energy
propagation direction, frequency band in the vicinity of ν and unit of time, respectively.
Let us indicate with W the increase of the radiation Iν passing through the above considered
volume. The quantity:
represents the enhanced energy of an incident beam into the elementary cylindrical volume dAds
with respect to the direction Ω and relative to the time interval dt and frequency band dv. From the
combination of the (2.2.1) and (2.2.2), the quantity Wν is derived in terms of the incoming intensity
energy variation to unit path:
By considering an absorbing, emitting and scattering medium, the quantity Wν can be written in the
explicit formulation of interaction mechanisms as follows:
This relationship represents the balance equation between the increment (positive terms) and
decrement (negative terms) of the energy during the interaction whit material substance.
The first term (WE) to right-hand side represents the increasing of radiation energy per unit time,
volume, solid angle and frequency due to the emission of radiation. The second term (WA) to righthand corresponds to the energy losses caused by the radiation absorption by a medium that, for a
volume element in the unit time, solid angle and frequency. The third and fourth terms describe the
balance of radiation energy diffused in all direction by the scattering mechanisms. Specifically, the
quantity (WIS) takes into account the radiation scattered by the medium in the direction of the
observer (positive) that, for an isotropic medium and purely coherent scattering. Finally the quantity
(WAS) is related to radiation losses for the reason that the energy beams are deflected along the main
direction Ω.
The explicit relationship in the compact formulation is:
Where the first term is (Wr), the second (WA + WAS), the third (WE) and the last is (WIS).
That, in more compact form, could be written as:
Where
In these relations Sν(s) is called the source function, βν(s) is the spectral extinction coefficient and
ων(s) is the spectral albedo.
Purely scattering medium
If we observe a hypothetical real purely scattering medium, namely where the thermal radiation
does neither absorb nor emit such is the case of the frozen top of cold rainy clouds, the last equation
will be banally simplified as ων(s)= 1. With this simplification, the term related to Planck’s
emission completely disappears and the total extinction coefficient becomes βν(s)= σν(s). Equation
can be rewritten as:
This is an integro-differential equation and its analytical solution does not exist. Several methods
often based on approximated formulation of this equation could be found in literature. Generally
scattering effects increase for ν > 60 GHz.
Purely absorption/emission medium
Absorbing and emitting media differ from purely scattering ones because they absorb external
radiation and re-emit it in the same direction without scattering extinction by the substance
constituents. Small cloud droplets, water vapor and precipitating clouds with few ice crystals on top
or totally deprived of them (warm rain) can be virtually considered as an absorbing/emitting
medium. For such media, where ων(s)= 0, the equation (2.2.11) assumes the form:
Where the first term represents the amount of absorption of external radiation by the medium
described by the boundary intensity radiation I0 and an exponential decreasing law of incoming
radiation into the medium; the integral term expresses the radiation variation emitted from the
surface at the temperature T along the path length s. For ν < 60 GHz the radiation extinction is
conventionally attributed to the absorption processes.
Considering that AMSU-B channels are ranged in the scattering domain many sensitivity studies
(Bennartz & Bauer, 2003) have demonstrated that in absence of strong scatterers liquid cloud
droplets largely absorb radiation at 89 GHz while ice hydrometeors on the top of clouds act as
scatterers more at 150 GHz and 190 GHz than at the other frequencies. Furthermore, experiments
demonstrate that when a light precipitation episode is sensed, usually associated to stratiform rain
with raindrop sizes comparable to non-rainy cloud droplets, the sensitivity to the absorption at 89
GHz is more marked than the scattering signal at 150 GHz. Therefore, making use of these basic
considerations we report an example of intense scattering by large ice crystals during the
evolutional stages of a Mesoscale Convective System (MCS) over the Mediterranean Sea and an
example of absorption by light stratiform rain and cloud liquid water over the North-Eastern
England Sea. In the case of deep convection (Figure 3) it is worth noting how the ice particle bulk
depresses upwelling radiances, expressed as brightness temperatures in unit of Kelvin, at all
frequencies from surface (89 GHz and 150 GHz) to the top of the troposphere (at 184 GHz the
weighting function “peaks” at about 8 km lowering down to 2 km at 190 GHz) denouncing a system
vertically well developed. Besides, it is interesting observe that the signal depression is enhanced at
150 GHz (comparable with measurement at 190 GHz) where the signal extinction is quantifiable
over 100 K with respect to the channel’s nominal value. In the practical use of satellite remote
sensing, the properties of this frequency combined to those of other channels such as the 89 GHz
and 190 GHz are often exploited to discern ice signature in the clouds and possibly correlate
probability information related to the conversion of melting ice into rainfall at the ground (Bennartz
et al., 2002; Laviola & Levizzani, 2008).
Figure 3
Figure 3: Mesoscale Convective System over Southern Italy on 22 October 2005 as observed by the NOAA-15/AMSU-B at 89 , 150 ,
184, 190 and 186 GHz clockwise from top left panel.
Referring to Fig. 4, the observed satellite radiance attenuation is mainly due to the absorption and
emission of small cloud particles and hydrometeors, warm cloud spots at 89 GHz due to the
absorption of water clouds over open water (black arrows) and stratiform rainy clouds (blue arrow)
over the coastline also sounded at 150 and 190 GHz can be clearly distinguished. It is interesting to
compare extinction intensities at 89 and 150 GHz both from the absorption and scattering point of
view by using in that order open sea liquid cloud and snow cover over the Alps (white arrow) as
terms of comparison. At 89 GHz over the sea the strong contrast between cold sea surface (≈ 200
K) and warmer liquid clouds (≈ 250) bring to a net difference of about 50 K whereas over land the
difference due to the scattering of snowy terrain is quantified in about 60-70 K. At 150 GHz the
discrepancy between the cold sea surface and liquid water clouds can be evaluated in about 10 K
while the change in surface emissivity over land induces a satellite brightness temperature
depression up to 70 K increasingly describing the strong sensitivity of that frequency to the
scattering.
Figure 4
Quasi-pure stratiform system over Belgium and cloud liquid water over North-Eastern England Sea on 17 January 2007 as observed by
the NOAA-15/AMSU-B at 89 , 150, 184, 190 and 186 GHz clockwise from top left panel. The strong contrast at 89 GHz allows to
observe water clouds over open sea (black arrows) whereas the change in emissivity highlights snow cover over Alps both at 89 and 150
GHz (also at 190 GHz). At higher opaque frequencies (186 and 184 GHz) the absorption of middle and high layer water vapor can only
detected.
2.3. Impact of precipitation on microwave measurement
From the radiative point of view, when incident radiation interacts with precipitating hydrometeors
all particles present in any elementary volume are totally irradiated and consequently the incoming
radiation is extinguished both by absorption and scattering processes at the same time. Passive
microwave rainfall estimations are carried out by exploiting either absorbed or scattered signals
from raindrops or a combination of the two. In the hypothesis of warm rain rainfall is estimated
through the emission associated with absorption by liquid hydrometeors through Kirchoff’ s law. In
this case, raindrops absorption and emission provide a direct physical relationship between rainfall
and the measured microwave radiances. With increasing precipitation intensities, scattering by large
drops becomes dominant with respect to absorption and the observed radiation appears drastically
depressed for a downward-viewing observer. A more realistic situation is represented by rainclouds
formed by a mixture of liquid, frozen and eventually supercooled hydrometeors. Since scattering is
primarily caused by ice hydrometeors aloft the emitted signal by liquid drops is substantially
blocked by intense scattering and its contribution to the total extinction significantly decreases with
the increase of the frozen bulk. Measured radiances are therefore indirectly related to the rain mass
and consequently the estimations become less correlated with falling rain below cloud base.
This situation is often observed during the development of intense convections typically associated
with heavy rain events. The case of liquid rain drops discussed before can be roughly associated
with the stratiform systems whose light precipitation is linked more to the absorption of water
droplets than to the scattering of small crystals which form on cloud top.
2.4. The sensitivity at 183.31 GHz to surface emissivity and rainy cloud altitudes
The 183.31 GHz bands are mainly dedicated to the sounding of the atmospheric water vapor
amount (Kakar, 1983; Wang et al., 1989). However, several studies have demonstrated the effects
of clouds on these frequencies and their possible application into rainfall retrieval schemes. Note
that the use of PMW information is necessary to detect rainy systems or correct and integrate IR
measurements, for instance in the blended techniques. However, their use is limited because of the
variability of surface emissivity (ε). Grody et al. (2000) proposed a few algorithms based on
different land type studies to evaluate surface emissivity using AMSU data. Figure 5 shows the
simulated brightness temperatures for all AMSU-B frequencies as a function of surface emissivity
in clear sky conditions. As expected, the signal around 89 and 150 GHz has strong surface
contributions showing a deep depression near low emissivity values and converging about to the
same brightness temperature when ε=1 (dry-land). Therefore, the decreasing surface emissivity
from dry-land values to water bodies enhances the influence of atmospheric moisture on these
channels. Another significant aspects of Fig. 5 is that, since their weighting functions are peaked
beyond 2 km altitude, the three moisture channels are little or not at all affected by different surface
emissivities thus suggesting their application both over land and over the sea. Precipitating cloud
altitude is another important variable affecting the AMSU-B brightness temperatures. In agreement
with the weighting function distribution, which peak between 2 and 8 km, our results show that only
rainy clouds positioned above 2 km altitude interact with three AMSU-B opaque frequencies at
183.31 GHz and the interaction will be always more intense as the cloud becomes thicker.
Figure 5
2.5. High frequency method to retrieve rainrates (183-WSL)
A new rainfall estimation method, named 183-WSL (Laviola & Levizzani, 2008, 2009a), is now
described based on the high frequency water vapor absorption bands at 183.31 GHz of AMSU-B
sensors on board NOAA and EUMETSAT Polar System (EPS) satellites, which is conceived to
retrieve rainrates over land and sea. AMSU-B is the second module of the AMSU passive MW
across-track scanner operating into the frequency range from 90 up to 190 GHz with a spatial
resolution of 16 km at nadir view (Saunders et al., 1995; Hewison & Saunders 1996).
An emission approach at 183.31 GHz is adopted to infer surface precipitation because one of our
major targets is the estimation of warm rain. The 183-WSL retrieval scheme (Laviola & Levizzani,
2009b), based on a suite of brightness temperature (BT) thresholds, distinguishes and classifies
convective and stratiform precipitation while filtering out condensed water vapor and snow cover
on mountain top, which particularly affects more opaque superficial channels (i.e., 190 GHz).
3. The 183-WSL algorithm
The Water vapour Strong Lines at 183 GHz (183-WSL) fast retrievalmethod retrieves rain rates and
classifies precipitation types for applications in nowcasting and weather monitoring. The retrieval
scheme consists of two fast algorithms, over land and over ocean, that use the water vapour
absorption lines at 183.31 GHz corresponding to the channels 3 (183.31±1 GHz), 4 (183.31±3 GHz)
and 5 (183.31±7 GHz) of the Advanced Microwave Sounding Unit module B (AMSU-B) and of the
Microwave Humidity Sounder (MHS) flying on NOAA-15-18 and Metop-A satellite series,
respectively. The method retrieves rain rates by exploiting the extinction of radiation due to rain
drops following four subsequent steps (Figure 6).
Figure 6
Flow chart of the 183-WSL algorithm: 1) ingestion, processing, land/sea discrimination; 2) module 183-WSLWV for water vapour and
snow cover filtering; 3) modules 183-WSLC and 183-WSLS for convective and stratiform rain classification, respectively; and 4) module
183-WSL for total rain rate estimation.
The first step is dedicated to ingesting and processing the satellite data stream. All relevant
information, namely TBs, surface type (land/sea/mixed), satellite local zenith angles, and
topography are separated from the data stream and arranged for input into the 183-WSL processing
chain. Pixel data are limb-corrected according to the sensor scanning geometry. The second step
consists of land/sea/other pixel detection that classifies all sounded pixels; at this stage a water
vapour and snow cover filter is also applied. The third step is dedicated to the estimation of the
convective and stratiform components of rainfall and the last step computes the total rain rates as a
sum of the convective and stratiform intensities.
3.1. Window frequencies: rain/no-rain identification and convective-stratiform rain
classification
A sensitivity study on the behaviour of the 89 and 150 GHz AMSU-B channels in clear-sky and
rainy conditions was conducted by Laviola (2006b) using a synthetic dataset. The dataset was
generated using RTTOVSCATT (Burlaud et al., 2007; O'Keeffe et al., 2004), a version of the
radiative transfer model RTTOV (Eyre, 1991; Matricardi et al., 2004; Saunders et al., 2009) that
handles the scattering of radiation by hydrometeors through the delta-Eddington approximation
(Kummerow, 1993; Wu and Weinman, 1984). The Eddington approach to scattering approximates
the radiance vector and the phase function to the first order so that only one angle is required for
the scattering calculations. The Marshall–Palmer size distribution (Marshall and Palmer, 1948)
function for rain droplets and a modified gamma for ice crystals (Evans and Stephens, 1995a,b) are
assumed. RTTOVSCAT was applied to 20,000 atmospheric profiles in clear and rainy conditions
from the model of the European Centre for Medium-Range Weather Forecasts (ECMWF). As an
example, Fig. 7 shows the scatter plots of the 89 and 150 GHz synthetic data of rainy vs clear sky
conditions over the ocean (note that a subset of 9500 ECMWF atmospheric profiles was used); the
“rainy” profiles refer to atmospheric columns containing cloud droplets, cloud ice, and liquid and
solid precipitation hydrometeors. The atmospheric columns associated with rainy conditions
significantly absorb the 89 GHz radiation increasing the TBs of about 40 K with respect to the clear
sky conditions. At 150 GHz the scattering from precipitation hydrometeors depresses the TBs of
about 30 K.
Figure 7
Scatter plot of rainy vs clear sky TBs over the ocean using RTTOVSCAT for 9500 ECMWF different atmospheric profiles: a) 89, and b)
150 GHz.
The radiation at 89 GHz is markedly extinguished by increasing water content. Cloud liquid water
or low-layered clouds absorb the upwelling radiation generally increasing the outgoing signal when
the surface emissivity ε is low (Bennartz et al., 2002). On warmer surfaces (ε > 0.90) the opposite
situation will usually be observed. In a very humid atmosphere the 150 GHz frequency behaves
more like a “lowlevel” water vapour channel than as a window channel with its weighting function
peaking higher up around 850 hPa (Bennartz and Bauer, 2003). In the case of drier profiles the
channel goes back to its quasi-transparency.
These facts suggest a way to exploit the effects of cloud liquid water amount variations and cloud
positioning in the frequency range 89 - 150 GHz (Bennartz and Bauer, 2003; Muller et al., 1994).
Liquid water clouds, corresponding to low rain rate intensities, generally tend to suppress the effect
of surface emissivity, particularly at 89 GHz, whereas higher clouds influence more the 150 GHz
signal (Laviola, 2006a). For example, a 1 km-thick cloud located at 2–3 km over land absorbs
radiation near 89 GHz resulting in a slight decrease of TBs while it attenuates the signal down by ≈
60 K more in the case of colder clouds at 10 km height. At 150 GHz low clouds impact a few K less
than the nominal (clear sky) channel value and the extinction by colder clouds depresses the signal
drastically more than 30 K in dependence of cloud altitude. It is thus conceivable that, by
appropriately combining the radiometric features of the 89 and 150 GHz channels, precipitating
areas can be delineated both over land and over water surfaces on the basis of absorption and
scattering induced by different hydrometeor states.
The SI approach (Bennartz et al., 2002) and the successive modification by Laviola (2006a,b) is the
basis of the 183-WSLW (the final W standing for water vapour) screening module, which
discriminates between rainy and non-rainy clouds on the basis of a rainfall dataset at 5 km
resolution of the UK MetOffice Nimrod radar network, http://badc.nerc.ac.uk/data/nimrod/. It is
found that over land, when Δwin < 3 K the observed pixels are classified as non rainy and therefore
removed from the computation. Over open water, where the impact of the atmospheric parameters
is greater than over land, the previous threshold is reduced to 0 K. The thresholds are reported in
Table 1.
The SI approach based on Δwin is also applied to differentiate between convective and stratiform
rain types. The strong scattering by growing ice hydrometeors, which typically characterize
convective cell formations, induces a clear signature at 150 GHz depressing the TB150 of several K
with respect to what happens to TB89. These different sensitivities to the presence of ice has been
exploited to calculate a fixed threshold value based on the Nimrod radar database. Generally, Δwin <
10 K values are associated with stratiform precipitation whereas values higher than these can be
correlated to the more scattering convective cells. Table 1 details the various empirical thresholds
over land and sea for the two categories. Fig. 8 report the performance of the modules 183-WSLC
(convective - left side) and 183-WSLS (stratiform - right side), respectively.
Figure 8
Note that, due to the presence of ice into the cloud cores, just the borders of precipitating clouds are
“warmer” and thus recognized as stratiform rain. In fact, in these areas the coexistence of large
amount of droplets, typically of a few microns size, with drops associated with light-rain stratiform
clouds of comparable dimensions is crucial to distinguishing rain from no-rain pixels.
In addition, tests carried out during the winter reveal that the scattering signal at 190 GHz relative to
the snow cover on mountain tops is analogous to the ice signature on convective cloud tops.
Therefore, a snow threshold as a combination of Δwin values over land and topographic
information from a digital elevation model (DEM) is applied to further remove false rain signals
from the 183-WSL estimations. The derived thresholds may not be valid for climate regions
characterized by low temperature profiles and possible frozen soil. At very high latitudes, for
example, the sounding at 89 and 150 GHz is strongly influenced by the high variability of surface
emissivity except for the open waters scenario (Mathew et al., 2006; Todini et al., 2009).
3.2. Rain rate estimation
The fast algorithm 183-WSL is based on a linear combination of the AMSU-B opaque channels at
183.31 GHz and is the result of a multiple linear regression between the TBs of the 183.31 GHz
channels of the AMSU-B radiometer and the radar-derived rainfall rates of the Nimrod network
(Laviola, 2006 a,b).
Rain rates are retrieved in mmh−1 both over land and sea by sounding cloud features from 1 to 2 km
up to the top of the troposphere according to the channel weighting function. Rain rate values
between 0.1 and 20mmh−1, representing the 183-WSL sensitivity to different rain types, are
employed to infer the precipitation amount for the latitude range ± 60°. However, further
investigations have demonstrated a posible variation of the threshold values when the 183-WSL is
applied over regions where atmospheric conditions such as temperature lapse rate and humidity
profiles are extremely variable, e.g. over tropical areas where rain rates up to 30mmh−1 are
observed. On the other hand, over regions located at latitudes above 60°, characterized by low
temperatures especially closer to the surface, when the latitude becomes > 70°, the estimation of
rainfall intensities < mmh−1 becomes crucial.
The land algorithm is as follows
RRl = A + B × (TB190 - TB184) + C TB186
(1)
where A=19.12475, B= -0.206044 and C= - 0.0565935 are coefficients of the multiple regression
analysis. The following adjustment is used to improve estimations with Eq. (1):
RRl = RRl − 0.6972
(2)
The sea algorithm has the following form
RRs = D + E × (TB190 - TB184) + F TB186
(3)
where D=9.6653, E=−0.3826 and F=−0.01316 are coefficients of the multiple regression analysis.
Again, two adjustments are applied to improve the algorithm in (3):
RRs = (RRs + 4) × 0.5 - 1.3510
(4)
Eqs. (2) and (4) represent an adjustment of the first calibration of the retrieval algorithm derived
using an independent radar dataset over the same area of the initial calibration.
3.3. Application of a Probability Of Rain Development Function (PORDF)
A further improvement for a better delineation of the rain areas is introduced. The basic concept is
that more pixels associated with high values of condensed water vapour, particularly during light
rain events, can be associated with clouds in a growing stage depending on their water vapour
amount. With increasing drop size the freshly nucleated droplets can develop into light stratiform
rain. To describe this process a Probability Of Rain Development Function (PORDF) has been
conceived.
The PORDF stems from the concept of an evolutional logistic model (Hilbe, 2009) where the
starting element of the population are pixels associated with rain rate values > 2mmh−1, as shown in
Fig. 9 (left). The coefficients of the PORDF model were previously calculated using an ad hoc
model, which links rain clouds observed in the IR (200 < TB < 253 K) and the 183-WSLWretrieved values (Fig. 9 right).
Figure 9
(Left side) Evolutional logistic function PORDF. When the rainfall intensity from the 183-WSLW module is b2 mm h−1, PORDF is not
activated. (Right side) 10 June, 2007. Distribution of MSG SEVIRI TBs at 10.8 μm vs the 183- WSLW rain rate values. The dots are the
observed values while the four lines represent the best fits. The time is that of the MSG-SEVIRI scan over the area.
The PORDF is activated when discarded pixels, classified as cloud droplets (i.e., non rainy),
correspond to rain rates > 2 mm h−1. This limit value is considered as a crucial threshold between
cloud droplets (non precipitating) and the beginning of stratiform rain development (light
precipitation).
The effects of the application of the PORDF on the 183-WSL rainfall retrievals are documented for
two events: 18 January 2007 and 23–24 March 2008 (Fig. 10).
Note the discontinuity of the rainfall field in the areas of low rain rates due to obvious
underestimation where the 183-WSLW module discards cloud areas deemed non-precipitating, but
that are in reality characterized by low rain intensities. The overall result is visually appreciated as a
better continuity of the systems from the centre to the borders.
Figure 10
183-WSL rain rate retrievals (in mm h−1). Left column: without the application of the PORDF function to the low rain rate end of the
scale. Right column: with the application of the PORDF.
3.4. Liquid water path
As a side product of the 183-WSL algorithm the liquid water path (LWP) is computed for the
clouds over the open sea that are identified as non-precipitating by applying the −20 < Δwin < 0 K
interval. The synthetic dataset already mentioned was used by Laviola (2006a) to derive the Δwin
Vs LWP relationship described by Fig. 11. The Δwin values all below 0 K denote the cloud type,
i.e. nonprecipitating clouds over the open sea. Moreover, the shape of the scattergram is that of the
water vapour continuum as it is to be expected.
Figure 11
Δwin as a function of liquid water path for the non-precipitating clouds over open sea as derived from the simulations of Laviola (2006b).
The LWP values increase from the outermost boundaries of the cloud towards the interior where the
cloud gradually becomes precipitating. On the border between the nonprecipitating section of the
cloud and the low-intensity rainfall sector the LWP retrieved value is around 0.5 × 103g m−2
(Karstens et al., 1994).
4. Training in passive microwave images and the application of the 183WSL algorithm
4.1. Observational analysis from infrared and microwaves of a cloudy scenario
The following analysis is based in infrared images (10.8 µm) from SEVIRI-MSG sensor.
4.1.1. Discrimination between “cold” cloud and “warm” cloud in infrared
In the infrared (IR) region we can consider surfaces like blackbodies (absorption = emission), so we
can associate the radiation received by the satellite with the temperature of the radiating body (call
brightness temperature Bt). In clouds we can decide, based on radiation received, if a cloud is
“cold” or “warm”, since we can calculate the Bt from the cloud and then the warm cloud will have
higher brightness temperature than the cold one.
In a grayscale, the warm temperatures are showing in dark shades (land and sea) and the cold
temperatures, like clouds, are showing in grey shades and in white in the case of a very cold cloud.
In general the signal received in IR is from the top of the clouds, so the Bt is related to the top of
these. Therefore, if an observed cloud has deep vertical development we cannot extract information
from the base.
The following images (Fig 12) are examples of different brightness temperatures. On the left image
enclosed by a red circle we can see an example of cold cloud, by the other hand, on the right image
enclosed by a yellow circle is a cloud more warm than the first case.
Figure 12
In general, the cloud temperature is associated with its height, thus a cold cloud is higher than a
warm cloud. Therefore in an IR analysis is possible to estimate qualitatively the cloud height
relative to the brightness temperatures observed. However the very lower clouds are difficult to see
in IR.
4.1.2. Are cold clouds always precipitating? And warm cloud?
The cold clouds do not always precipitate. Cirrus clouds are an example of this, they are high and
cold clouds but do not precipitate. Nor is true that warm clouds always precipitate, cloud liquid
water is a warm cloud does not precipitate.
Analysis of AMSU-B channels images (microwave channels MW) and Comparison between IR
images
4.1.3. Discrimination between “cold” cloud and “warm” cloud in MW
In MW region is also possible to relate the radiation sensed with the brightness temperature from a
radiating body (related to a gray body), so high values of radiation sensed means a high Bt of the
emitting body, it is thus possible distinguish between cold clouds and warm clouds, since high Bt
indicates warm cloud and lower Bt indicates cold cloud. The operational frequency of the MW
sensor is very important because different clouds or different cloud features can be observed for
different sample frequencies. In the case of AMSU-B the relationship between different frequencies
(named in channels) and the altitude ranges of view is governed by the weighting functions. These
are represented in the next figure (Fig 13).
Figure 13
The first two channels (Cha1 = 89 GHz; Cha2 = 150 GHz) named windows channels can describe
surface features. The remaining tree has weighting function peaks in:
Cha3 = (183.31 ± 1) ≈ 184 GHz - altitude range 6-8 Km (from the surface)
Cha4 = (183.31 ± 3) ≈ 186 GHz - altitude range 4-6 Km (from the surface)
Cha5 = (183.31 ± 7) ≈ 190 GHz - altitude range 2-4 Km (from the surface)
Consider the images (Fig 12) of cold and warm clouds and let us compare with the MW images
(Fig. 14).
Figure 14
In MW is not possible clearly identify cold or warm cloud with only one channel, in fact is not easy
identify a cloud with a only channel. However, using several channels we can determine the altitude
of the cloud and then to estimate its temperature. Based on this, a warm cloud has to be observed in
the first channels, while a deep vertical development cloud should be seen in all the channels.
In order to do a comparation between the applications of IR and MW frecuencies, in figure 12 it is
shown a cold cloud from IR frecuencie (left side, red circle). In the other hand, in the figure 14 we
have the same system in MW (all channels, left side and blue circle). As you can see, it is not easy
to distinguish the cold cloud in MW since the observation of the enclosured area by the blue circle
in every channel. However this could be explained because thin clouds, like cirrus clouds, are
transparent to MW, so the selected cloud probably is a cirrus. In case of the warm cloud (right side)
we can see it through MW in channel 5 clearly.
4.1.4 Are cold clouds always precipitating? And warm cloud?
In the MW regions, as was mentioned before, cirrus and thin clouds cannot be see. Due to this, a
cold cloud detection (with lower Bt) probably be associated with a precipitating cloud. For warm
clouds the behavior observed in MW is the same than IR.
4.2. AMSU-B channels
Considering the features of the AMSU-B channels (Table 2) and the Figure 15, estimate:
- Cloud height
- Kind of cloud (liquid water/precipitating)
- Cloud phase (ice/water/mix)
Make a general conclusion.
Figure 15
Ch. 1
surface
Ch. 2
surface
Ch. 3
6-8 Km
Ch. 4
4-6 Km
Ch. 5
2-4 Km
Cloud height
The decreace of the brightness temperature due of the presence of the cloud is highlighted to the
level of channel 5 (2-4 Km), so this cloud is at medium height level.
Kind of cloud (liquid water/precipitating)
The first three channel show a moderate scattering that could be cuased by precipitation. So this
clous is classified as a precipitating cloud.
Cloud phase (ice/water/mix)
If well we can detect a scattering from the cloud, the decrease of the brightness temperature is
not enough to be considered as ice scattering. Thus we may conclude that the cloud is composed of
water or at most mixed but with small ice crystals.
General conclusion
Based on the above reasons, we can conclude that this is a cloud system with a incomplet vertical
development, which could be causing moderate rainfall.
4.3. Land and cloud features with AMSU-B channels
Based on a given feature, we propose the best channel to describe it.
4.3.1. Surface features
Surface types (Land/Sea) and coast effects
To watch this features we consider the following images (Fig 16).
Figure 16
From the above images we can conclude that the different surfaces such as land and sea are more
highlighted in channel 1 (89 GHz), however in channel 2 (150 GHz) we can watch the differences
but are less pronounced. The coast effects can be observed in channel 1 (89 GHz).
Land and mountains snow cover
The next images (Figure 17) are plotted to watch this features.
Figure 17
From the images we can see that the snow cover is more clearly enhanced at 150 GHz with respect
to other frequencies. Whereas that the mountain snow cover can be observed from the surface
channels to the channels that correspond to the height of the montain.
4.3.2. Cloud features
Liquid water clouds and precipitation clouds
In next images (figure 18) we choose examples of liquid water clouds and precipitation clouds.
Figure 18
In the images, the liquid water clouds are enclosed by white circles, which can be detected in the
windows channels. These clouds do not produce a significant decrease (as the scattering) in the
brightness temperature and for this reason we consider that they are not a precipitating clouds. By
the other hand, we see a significant decrease in the brightness temperature in the border between
France and Spain and in the region enclosed with a black circle. The attenuation of the signal sensed
is due to the scattering from de clouds and this effect can be detected in all the channels in the first
example and the channels: (Cha1:89GHz; Cha2:150GHz; Cha5:190GHz; Cha4:186GHz) in the
second example. These clouds are examples of precipitation clouds.
Stratiform precipitation
Consider again the last images, but now we select another region as an example of stratiform
precipitation (Figure 19).
Figure 19
Ch1
Ch2
Ch5
Ch4
Ch3
With the channels (Cha1:89GHz; Cha2:150GHz; Cha5:190GHz) we can detect stratiform
precipitation, since channels 2 and 5 sensed a decrease in the the brightness temperature due the
scattering.
Convective precipitation
As an example of convective precipitation we considered, again, the last images (Figure 20).
Figure 20
The first case (black circle) the scattering due to convective precipitation is sensed for all the
channels as a significant decrease in the brightness temperature like we expected, this cloud has a
vertical development above 8 km. In the second case (white circle), the scattering is sensed until
channel 4 (186 GHz), so the altitude of the cloud is around 4-6 Km. We conclude that in first case
the precipitation will be more strong that in the second case.
Water vapor effect
The water vapor effect can be observed in the opaque channels (184, 186 and 190 GHz). In these
channels the water vapor variations can be detected as variations in the brightness temperatures. In
the presence of water vapor the brightness temperature is decreased (orange colors), whereas that
the minimal presence or absence of vapor water are represented in red colors. Sometimes the strong
presence of water vapor can mask the scattering from clouds with incomplete vertical development.
Example of the mask effect is observed in the next figure (Figure 21)
Figure 21
4.3.3. Combination of AMSU-B channels
Based on a given feature, we propose the best combination to describe it. The combinations
considered are:
- deltab= cha1 - cha2 = 89 GHz - 150 GHz
- delta89= cha1 - cha5 = 89 GHz - 190 GHz
- delta150= cha2 - cha5 = 150 GHz - 190 GHz
-
deltaWV= cha5 - cha4 = 190 GHz - 186 GHz
Surface types (Land/Sea) and coast effects
Next images (Figure 22) are the different combinations cited
Figure 22
Surface types and coast effects are well highlighted with the combination deltab = 89 GHz - 150
GHz.
Land and mountains snow cover
Next images (Figure 23) are the different combinations cited
Figure 23
Watching the images we can see that deltab and delta150 combinations described the land snow
cover well. The mountain snow cover can be highlighted with the combination deltaWV.
Stratiform precipitation
We take for this comparison (Figure 24) the example of stratiform precipitation above used.
Figure 24
From the observation, the deltab combination seems describe well the stratiform precipitation.
Convective precipitation
Now consider the examples of convective clouds previously used (Figure 25)
Figure 25
In the most of images we can see the presence of convective precipitation, however the best
combination seems to be the deltaWV.
Water vapor effect
Consider Figure 25, the combination deltaWV can describe the vapor water effects very well.
4.4. Precipitation producs
Visualize and interpret the following precipitation products:
- Convective precipitation (183-WSLC)
- Stratiform precipitation (183-WSLS)
- Liquid water cloud (183-LWP)
- Water vapor absortion and cloud dropplets (183-WSLW)
- Total precipitation (183-WSL)
Convective precipitation
We compare (channel 2, arbitrarily chosen) the example of convective precipitation with the
product (183-WSLC) (Figure 26).
Figure 26
The algorithm classifies as convective precipitation both selected regions. It´s also possible to see
that the intensity of the precipitation is different between the two regions and it is different within
each region too. Watching the scale of rain rates we can deduce that, when the precipitation is more
intensive the brightness temperature is lower.
Stratiform precipitation
Applying the same proceeding (Figure 27)
Figure 27
The selected region as stratiform precipitation is in agreedment with the classification made by the
algorithm. The intensity of the precipitation retrieved in this case is less than for convective
precipitation. We can also observe stratiform precipitation zone surrounding the convective
precipitation.
Liquid water cloud
We use here the same images (89 GHz) used to discriminate precipitating and not precipitating
clouds. The non precipitating clouds chosen are enclosed in white circles (Figure 28).
Figure 28
The algorithm does not classify the selected regions as liquid water clouds, but an example of these,
is enclosed in red circle on the right image and near of the convective precipitation.
Water vapor absortion and cloud droplets
Again we used the image above to compare with the classication of the algorithm (Figure 29).
Figure 29
The algorithm classified only one of the regions as water vapor absortion or cloud droplets, so the
unclassified region was choose wrong as cloud or water vapor.
Total precipitation
The contribution of the all precipitation products are summarized in the following image (Figure
30):
Figure 30
5. Visible and Infrared images from SEVIRI instrument, Description of
characteristics and content of the channels
For more than 40 years, meteorological satellites have been the best way to observe the changing
weather on a large scale. Advances in satellite technology have led to improved observational
capabilities. Particularly, the first generation of European geostationary meteorological satellites
dates to 1977, with the launch of Meteosat-1. From this, it have been many improvements in
satellite technology until the present days, an examples of this is the new generation of Meteosat
satellites, known as Meteosat Second Generation (MSG), which was launched in August 2002 from
the Kourou launch site in French Guiana. The current Meteosat series, MSG is spin-stabilized, and
capable of greatly enhanced Earth observations. The satellite’s 12-channel imager, known formally
as the spinning enhanced visible and infrared imager (SEVIRI), observes the full disk of the Earth
with an unprecedented repeat cycle of 15 minutes in 12 spectral wavelength regions or channels.
The accompanying table (Table 3) provides more details of the characteristics of these channels,
and indicates how each channel is used: for observations of clouds and surface temperatures, water
vapour or ozone.
Now referring to SEVIRI, it is important to mention that the Most of their spectral channels have
been build upon the heritage from other satellites, which has the great advantage that the operational
user community can readily use existing know-how to utilize SEVIRI radiance observations. Figure
3 shows the weighting functions of the thermal IR channels at 3.9, 6.2, 7.3, 8.7, 9.7, 10.8, 12.0, and
13.4 μm. The weighting functions demonstrate that the channels have been selected such that they
provide good information on clouds and the earth’s surface, water vapor, and ozone. A combination
of channels provides useful information on atmospheric instability. The heritage of the MSG
channels can be summarized as follows:
- VIS0.6 and VIS0.8 (channel 1 and 2 respectively): Known from the Advanced Very High
Resolution Radiometer (AVHRR) of the polar-orbiting NOAA satellites. They are essential for
cloud detection, cloud tracking, scene identification, aerosol, and land surface and vegetation
monitoring.
- NIR1.6 (channel 3): Discriminates between snow and cloud, ice and water clouds, and provides
aerosol information. Observations are, among others, available from the Along Track Scanning
Radiometer (ATSR) on the Earth Remote Sensing Satellite (ERS).
- IR3.9 (channel 4): Known from AVHRR. Primarily for low cloud and fog detection (Eyre et al.
1984; Lee et al. 1997). Also supports measurement of land and sea surface temperature at night and
increases the low level wind coverage from cloud tracking (Velden et al. 2001). For MSG, the
spectral band has been broadened to longer wavelengths to improve signal-to-noise ratio.
- WV6.2 and WV7.3 (channel 5 and 6 respectively): Continues mission of Meteosat broadband water
vapor channel for observing water vapor and winds. Enhanced to two channels peaking at different
levels in the troposphere (see Fig. 4). Also supports height allocation of semitransparent clouds
(Nieman et al. 1993; Schmetz et al. 1993).
- IR8.7 (channel 7): Known from the High Resolution Infrared Sounder (HIRS) instrument on the
polar-orbiting NOAA satellites. The channel provides quantitative information on thin cirrus clouds
and supports the discrimination between ice and water clouds.
Table 3
Channel
No.
Spectral
Band
(µm)
Characteristics of Spectral Band (µm)
cen
min
max
Main observational application
1
VIS0.6
0.635
0.56
0.71
Surface, clouds, wind fields
2
VIS0.8
0.81
0.74
0.88
Surface, clouds, wind fields
3
NIR1.6
1.64
1.50
1.78
Surface, cloud phase
4
IR3.9
3.90
3.48
4.36
Surface, clouds, wind fields
5
WV6.2
6.25
5.35
7.15
Water vapor, high level
clouds, atmospheric instability
6
WV7.3
7.35
6.85
7.85
Water vapor,
instability
7
IR8.7
8.70
8.30
9.1
Surface, clouds, atmospheric
instability
8
IR9.7
9.66
9.38
9.94
Ozone
9
IR10.8
10.80
9.80
11.80
Surface, clouds, wind fields,
atmospheric instability
10
IR12.0
12.00
11.00
13.00
Surface, clouds, atmospheric
instability
11
IR13.4
13.40
12.40
14.40
Cirrus cloud height,
atmospheric instability
12
HRV
Broadband (about 0.4 – 1.1 µm)
atmospheric
Surface, clouds
- IR9.7 (channel 8): Known from HIRS and current GOES satellites. Ozone radiances could be used
as an input to numerical weather prediction (NWP). As an experimental channel, it will be used for
tracking ozone patterns that should be representative for wind motion in the lower stratosphere. The
evolution of the total ozone field with time can also be monitored.
- IR10.8 and IR12.0 (channel 9 and 10 respectively): Well-known split window channels (e.g.,
AVHRR). Essential for measuring sea and land surface and cloud-top temperatures; also for
stability.
Figure 31
The radiation sources are sun and earth. Solar radiation is dominant for wave length < 5 µm and the
earth radiation is dominant for wave length > 5 µm. the channels are classified in three groups.
• Ch01, 02, 03, 12: only sun radiation
• Ch04: both: radiation from sun and earth
• Ch 05, 06, 07, 08, 09, 10, 11: only thermal earth radiation
Figure 32 shows the radiation sources and the location of the channels in the spectrum.
Figure 32
5.1. Description and Interpretation of the channels.
In this section is shown a wide description of the featuring of the SEVIRI Channels, also it is used a
set of image from the MSG obtained through the EUMESAT web site of a torment scene in the
Mediterranean sea, for show how to do an interpretation of this phenomenon with the SEVIRI
channels more used for meteorological operations.
5.1.1. Solar channels
The channels, Ch1: 0.6 µm (VIS); Ch2: 0.8 µm (VIS) y Ch3: 1.6 µm (NIR) are the solar channels
and their areas applications are:
• Recognition of cloud because of reflected sun radiation
• Recognition of earth surface characteristica (soil, vegetation)
• Discrimination of water and ice cloud
• Recognition of snow/ice because of reflected sun radiation
The following are examples of these applications.
Recognition of cloud with Ch1: 0.6 µm (VIS); Ch2: 0.8 µm (VIS).
Figure 33 shows different clouds, snow, land and sea. The different reflectivities from different
clouds are visualized in different grey shades for channel 1 and 2.
Figure 33
Referens Figure 33: white: optically thick cloud
; grey: transparent cloud
; white: snow ; grey to darkgrey: land
; black: sea
Recognition of earth surface characteristica Ch1: 0.6 µm (VIS); Ch2: 0.8 µm (VIS).
The channel Ch2: 0.8 µm (VIS), allow to a better recognition of surface structures because of higher
reflectance of soil and leafs. However, transparent clouds better visible in Ch1: 0.6 µm (VIS) due to
the low reflectivity of surface as is shown in Figure 34.
Figure 34
Discrimination of water and ice cloud with Ch3: 1.6 µm (NIR)
In 1.6 µm the absorption from water is different respect to absorption from ice. The absorption in
the ice phase is higher as can be observed in Figure 35, and this let to distinguish water clouds from
ice clouds.
Figure 35
The water clouds show in white and the ice cloud show in dark gray.
Recognition of snow/ice because of reflected sun radiation with Ch3: 1.6 µm (NIR)
The different reflectivities (Figure 36) between the snow/ice (low) and clouds (high) allows
discriminate optically thick clouds from snow over land.
Figure 36
References figure 7:
white: water cloud
grey to dark grey: ice cloud
black: snow/ice
grey to darkgrey: land
black: sea
Now consider next images from EUMETSAT data. These show a storm moving across the
Mediterranean Sea. Such as in the previous images, analyze the new images highlighting in each
case the respective characteristics.
Figure 37
References figure 37:
white: water cloud
grey to dark grey: ice cloud
black: snow/ice
grey to darkgrey: land
black: sea
Ch1 (right top), Ch2 (right bottom), Ch3 (left)
5.1.2. The Ch4: 3.9 µm (IR)
Cooling effect
The Ch4: 3.9 µm (IR) is a window channel but close to the CO2 absortion band at 4-5 µm. The CO2
porduce an effect on brightness temperature of this channel call (cooling effect). The cooling effect
of the CO2 absortion on channel Ch4 depends on: Surface temperature and lapse rate in the lower
troposphere (large for hot desert surfaces during daytime), height of the cloud (small for high
clouds) and satellite viewing angle (so called "limb cooling" effect, large for large satellite viewing
angles).
Solar and thermal contribution
Signal in Ch4 comes from reflected solar and emitted thermal radiation.
Thus during day-time the temperature is not representative of the Planck relation between radiance
and temperature (Figure 38).
Figure 38
During daytime this channel has a thermal and a solar contribution. Therefore, applications and
algorithms are different for night- and day-time (Figure 39).
Figure 39
Meteorological use of the Ch4: 3.9 µm (IR)
This channel can be used to:
- Detection of low clouds and fog [day and night]
- Detection of thin Cirrus [day and night] and multi-layer clouds [day]
- Cloud phase & particle size [day and night]
- Sea and land surface temperature [night]
- Detection of forest fires [day and night]
- Urban heat island [night]
- Super-cooled clouds [day and night]
- Cloud top structures (overshooting tops) [day]
- Sunglint [day]
Below there are some examples of cloud phase and particle size.
Due to the high reflection from water droplets at IR3.9, low-level water clouds are much darker
than high-level ice clouds (during day-time) (Figure 40).
Figure 40
IR3.9 shows cloud top structures (very sensitive to particle size) (Figure 41).
Figure 41
Recognition exercises with images from EUMETSAT data: sensitivity to ice particle sizes. (Figure
42)
Figure 42
Reference:
1= ice clouds with very small particles
2= ice clouds with small particles
3= ice clouds with large ice particles
5.1.3. Window Channels
The channels, Ch7: 8.7 µm (IR); Ch9: 10.8 µm (IR) y Ch10: 12 µm (IR) are the window channels
and their areas applications are the recognition of cloud systems because of the thermal radiation of
cloud and earth surface and extraction of high thin Cirrus from the whole cloud systems.
Some characteristics are:
Ch7: 8.7 µm (IR)
• Indicative for high, thin Cirrus cloud
– improved recognition of Cirrus
– best use through preparation of RGBs and differences
• Support for discrimination between ice and water cloud
• From optical point of view (at the time being) no typical new features recognisable
Ch9: 10.8 µm (IR) y Ch10: 12 µm (IR)
• Usual evaluation of cloud in both channels
• Not much difference between the apperance in the three window channels
• Ch10 is more sensitive for high thin Cirrus
– not easy to recognise through checking by eyes
– difference images give information
Examples for visualisation of high Cirrus (Figure 43): Differences of Ch7, Ch9, Ch10
Figure 43
5.1.4. Absorption Channels
The channels, Ch5: 6.2 µm (IR); Ch6: 7.3 µm (IR) are the absorption channels or Water Vapour
(WV) channels. These channels have an absorption band around 6 m.
Ch05 is more in the centre of the absorption band with strong absorption, consequently radiation
only from higher levels comes to the satellite. Ch06 is more to the wings of the absorption band
with less strong absorption consequently radiation also from lower layers comes to the satellite.
Application areas for the WV channels:
• Identification of areas with high upper level WV
– Identification of jet axes
– Cloud intensification areas within jet streaks (left exit region)
– Wave developments within jet streaks (right entrance region)
• Identification of dry areas in WV
– special interest are areas of stratospheric air protruding downward
• tropopause folding
• PV anomalies
• release of secondary cyclogenesis
Figure 44 shows tipycal images of Ch5 (6.2) and Cha6 (7.3)
Figure 44
5.1.5. Ozone channel
The channel, Ch8: 9.7 µm (IR) is the ozone channel and its areas applications are:
• Qualitative application of images:
– No well-known additional features compared to the IR window channels
• But new feature detected: White stripes
– Stripes are white in case of the inverted presentation of the image
– They are a measure for high stratospheric Ozone concentrations which have
protruded relatively far down into the troposphere
• Quantitative applications:
– The computation of quantitative Ozone values will be a future task
5.1.6. CO2 channel
The channel, Ch11: 13.4 µm (IR) is the CO2 channel and its areas applications are:
• Qualitative application of images:
•
– No additional features compared to the IR window channels
– No special application for qualitative image interpretation (at the time being)
Quantitative applications:
– temperature profiles
– instability
– winds
5.1.7. The Seviri high-resolution visible (HRV) channel
Characteristics:
 Monitoring of small-scale features
 High-resolution wind vectors
 Broadband visible channel, as the current Meteosat VIS channel, but with an improved
sampling interval of 1 km
Example (Figure 45)
Figure 45
6.1. RGB images composite
Definition of RGB Colours:
Red, Green, Blue: The three colours of light which can be mixed to produce any other colour.
Coloured images are often stored as a sequence of RGB triplets or as separate red, green and blue
overlays though this is not the only possible colour representation. These colours correspond to the
three "guns" in a colour cathode ray tube and to the colour receptors in the human eye.
RGB colors are defined using a decimal or hexadecimal notation for the combination of Red,
Green, and Blue color values (RGB). The lowest value that can be given to one light source is 0
(hex #00). The highest value is 255 (hex #FF) (Figure 46).
Figure 46
6.2. Cloud Physical Properties represented by the MSG Channels
VIS0.6: optical thickness and amount of cloud water and ice.
VIS0.8: optical thickness and amount of cloud water and ice "greeness" of vegetation.
NIR1.6, IR3.9r: particle size and phase.
WV6.2, WV7.3: mid- and upper level moisture.
IR8.7, IR10.8, IR12.0: top temperature.
IR8.7 - IR10.8: phase and optical thickness.
IR12.0 - IR10.8: optical thickness.
IR3.9 - IR10.8: optical thickness, phase, particle size.
IR13.4 - IR10.8: top height.
WV6.2 - IR10.8: top height, overshooting tops.
6.3. Recommended schemes for RGB image composites
RGB Composite
6.3.1. RGB 03,02,01
6.3.2. RGB 02,04r,09
6.3.3. RGB 02,03,04r
6.3.4. RGB 05-06,04-09,03-01
6.3.5. RGB 10-09,09-04,09
6.3.6. RGB 10-09,09-07,09
6.3.7. RGB 05-06,08-09,05
Applications
Vegetation, Snow, Smoke, Dust, Fog
Clouds, Convection, Snow, Fog, Fires
Snow, Fog
Severe Convection
Clouds, Fog, Contrails
Dust, Thin Clouds, Contrails
Severe Cyclones, Jets, PV Analysis
Time
Day
Day
Day
Day
Night
Day&Night
Day&Night
In the following, describes the above combinations using the next labels:
(A) Vegetation; (B) Thin clouds of small ice particles; (C) Water clouds with small droplets; (D) Ice
clouds; (E) Optically thick clouds with small ice particles near their tops; (F) Snow; (G) Dust; (H)
Deserts; (I) Bare ground; (J) Ocean; (K) thin clouds composed of large ice crystals; (L) Dust over
the sea; (M) water clouds that do not precipitate; (N) large drops that are typical to precipitating
clouds; (O) Supercooled water clouds.
6.3.1. RGB 03, 02, 01 ("Day Natural Colours")
R = Channel 03 (NIR1.6)
G = Channel 02 (VIS0.8)
B = Channel 01 (VIS0.6)
Figure 47 shows an RGB composition of the “Day Natural Colors” color scheme: 1.6, 0.8, and 0.6
μm reflectance in the red green and blue beams respectively and shows the spectral response
functions of the three channels of the “Day Natural Colors” color scheme and spectral signatures of
several surface types. In this color scheme vegetation appears greenish (A) because of its large
reflectance in 0.8 μm (the green beam) compared to 1.6 μm (red beam) and to 0.6 μm (blue beam).
Water clouds with small droplets (C) have large reflectance at all three bands and hence appear
whitish, while snow (F) and ice clouds (D) appears cyan because ice strongly absorbs in 1.6 μm (no
red). Bare ground (I) appears brown because of the larger reflectance in the 1.6 μm than at 0.8 μm,
and the ocean (J) appears black because of the low reflectance in all three bands.
Figure 47
Combination “Day Natural Color” applied to image from EUMETSAT data.
6.3.2. RGB 02,04r,09 ("Day Microphysical")
R = Channel 02 (VIS0.8)
G = Channel 04r (IR3.9, solar component)
B = Channel 09 (IR10.8)
Figure 48
The RGB composition for the “Day Microphysical” color
scheme (Fig. 48) was inherited from Rosenfeld and Lensky
(1998): the 0.8 μm reflectance in red approximates the
cloud optical depth and amount of cloud water and ice; the
3.9 μm solar reflectance in green is a qualitative measure
for cloud particle size and phase, and the 10.8 μm
brightness temperature modulates the blue. This color
scheme is useful for cloud analysis, convection, fog, snow,
and fires. In this color scheme water clouds that do not
precipitate appear white (C,M) because cloud drops are
small, whereas large drops that are typical to precipitating clouds appear pink (N), because of the
low reflectance at 3.9 μm manifested as low green. Supercooled water clouds (O) appear more
yellow, because the lower temperature that modulate the blue component. Cold and thick clouds
with tops composed of large ice particles, e.g., Cb tops, appear red (D). Optically thick clouds with
small ice particles near their tops appear orange (E). Thin clouds composed of small ice particles
such as contrails appear as faint green (B), and thin clouds composed of large ice crystals appear
dark (K).
6.3.3. RGB 02, 03, 04r ("Day Solar")
R = Channel 02 (VIS0.8)
G = Channel 03 (NIR1.6)
B = Channel 04r (IR3.9r, solar component)
Figure 49
Figure 49 shows an RGB composition for the “Day Solar” color scheme: 0.8, 1.6 and 3.9 μm solar
reflectance in the red green and blue beams respectively and the spectral response functions of the
three channels of the ”Day Solar” color scheme and spectral signatures of several objects. In this
color scheme vegetation (A) appears greenish because of the large reflectance in the 1.6 and 0.8 μm
channels (the red and green beams). Snow (F) appears red because of the strong absorption in 1.6
and 3.9 μm (no green and blue), small particle ice cloud (E) appears orange, while large particle ice
cloud (D) appears with greater red component. Snow on the ground (F) appears as full red, because
its grains are usually much larger than cloud ice particles. Water clouds with small droplets (C, M)
have large reflectance in the 3.9 μm channel appears bright yellow. The dynamic range of reflected
sun light from different ice particles is larger at 1.6 μm than in 3.9 μm, enabling us to discriminate
ice particle size. Typically ice particles that form by mixed phase process in a super cooled water
cloud grow quickly to much larger sizes than crystals forming by vapor deposition in ice-only
clouds. This helps separating convective precipitating clouds from non-precipitating or layer ice
clouds. However, small ice particles can occur also at the tops of severe convective storms or high
base convective clouds, because cloud drops freeze directly into ice crystals by mechanism of
homogeneous nucleation (Rosenfeld et al., 2006). Deserts (H) appear bright cyan because of the
larger reflectance in the 1.6 and especially at 3.9 μm. Sea surface (J) appears black because of the
low reflectance in all three bands. This color scheme is very sensitive to cloud microphysics, but
there is no information from the thermal channels about the temperature (vertical extent) of the
clouds. The main applications of this color scheme are: cloud analysis, convection, fog, snow, and
fires.
6.3.4. RGB 05-06, 04-09, 03-01 ("Convective Storms")
R = Difference WV6.2 - WV7.3
G = Difference IR3.9 - IR10.8
B = Difference NIR1.6 - VIS0.6
Figure 50
In the “Convective Storms” color scheme
(Fig. 50) the brightness temperature
difference (BTD) between 6.2 and 7.3 μm
channels (BTD 6.2-7.3) is regulating the red,
which is modulated by the mid level
moisture. Overshooting Cb clouds have near
zero or even slightly positive BTD 6.2-7.3
(red). Cb tops (D) that do not extend to the
tropopause show small negative values
while the surface (A) shows large negative
BTD 6.2-7.3 values (no red). The BTD between 3.9 and 10.8 μm channels (BTD 3.9-10.8) modulates the
green and indicates the microphysics. At daytime, the large solar reflectance from small cloud
particles at 3.9 μm is added to the thermal contribution at this wave length, contributing to a larger
BTD3.9-10.8. This effect is stronger for cold clouds because the fraction of the added solar radiation is
larger when the thermal radiation is relatively small. The difference between the 1.6 and 0.6 μm
solar reflectances (RD1.6-0.6) modulates the blue, where large negative RD1.6-0.6 indicates ice clouds
and much larger RD1.6-0.6 is typical for the surface. Severe convective storms (E) appear bright
yellow in this color scheme because of the near zero BTD6.2−7.3 of overshooting Cb clouds (high
red). The strong updrafts in these clouds produce small ice particles at cloud tops due to
homogeneous freezing of cloud drops (Rosenfeld et al., 2006), resulting with large BTD3.9-10.8 (high
green). Finally, large negative RD1.6-0.6 because of the large absorption at 1.6 μm by ice particles
keeps the blue very low. Small ice crystals of Cirrus clouds (E in Fig. 50b) should not be confused
for vigorous convection (as in Fig. 50a).
6.3.5. RGB 10-09, 09-04, 09 ("Night Microphysical")
R = Difference IR12.0 - IR10.8
G = Difference IR10.8 - IR3.9
B = Channel IR10.8
The BTD between 10.8 and 3.9 μm channels (BTD10.8 - 3.9) modulates the green beam in the “Night
Microphysical” color scheme (Fig. 51). Lensky and Rosenfeld (2003a) showed how sensitive is the
BTD10.8 - 3.9 to particle size, and used this information to delineate precipitation (Lensky and
Rosenfeld, 2003b). Figure 52 shows that the 3.9 μm emissivity is much smaller for clouds
composed of small drops (0.85) than for large drops (0.99) while at 10.8 μm both show large values
(0.98 - 0.99), therefore large BTD10.8- 3.9 will indicate clouds with small drops. Nighttime shallow
clouds or fog with small drops appears in this color scheme in white (A). If the water clouds or fog
are colder (B, C) they appear more in yellow colors. Water clouds with larger drops have greater
red component, as shown in Fig. 51. The radiation emitted from the tops of cold Cb (<220 K) in 3.9
μm is very low, resulting with low signal to noise ratio and sprinkled color at the Cb tops. Therefore
deep Cb clouds (D) at night appear in sprinkled orange red color.
Figure 51
Figure 52
6.3.6. RGB 10-09, 09-07, 09 ("Day and Night and Desert Dust")
The “Day and Night” (Fig. 53a) and the “Desert Dust” (Fig. 53b) are the same color schemes, just
with RGB dynamic range adjusted for maximum sensitive for clouds in the “Day and Night” color
scheme, and for dust in the “Desert Dust” color scheme. The BTD between 10.8 and 8.7 μm
channels (BTD10.8 - 8.7) modulates the green beam in these color schemes. The emissivity of large
particles of quartz mineral (125 - 500 μm) is very low in 8.7 μm (see Fig. 54), thus the BTD10.8 - 8.7 is
very large over sands with quartz mineral in the desert (H), together with relative high BTD 12.0 - 10.8
and warm temperatures, the color of the desert sand is whitish. The emissivity in 8.7 μm of small
quartz mineral particles (0 - 45 μm) in the desert dust (G) is much larger, resulting in smaller
BTD10.8−8.7 and very strong pink colors. This color scheme is also useful for day and night detection
of contrails (B). As in the “Day Microphysical”, deep Cb clouds (E) appear red: positive
BTD12.0−10.8 revealing opaque clouds and high red, large ice particles resulting with small BTD10.8−8.7
and low green and cold tops – low blue. Thick water clouds (C) appear yellow: larger BTD10.8−8.7 –
higher green, warmer tops – higher blue. Thin Ci clouds appear black, while clouds with small
particles appear green. The colors of thin clouds may be affected by the underlying surface
properties (temperature and emissivity).
Figure
Figure 54
53
6.3.7 RGB 05-06, 08-09, 05i ("Airmass")
The RGB composition for the “Air Mass” color scheme (Fig. 55) is: BTD 6.2-7.3 modulates the red as
in the “Convective Storms” color scheme, where dry atmosphere has values close to zero (reddish),
and moist atmosphere has negative values. The BTD between 9.7 (ozone) and 10.8 μm channels
(BTD9.7−10.8) modulates the green, where rich ozone polar air mass has large negative BTD 9.7−10.8
(bluish), while low ozone tropical air mass has small negative BTD9.7−10.8 (greenish). The 6.2 μm
water vapor channel modulates inversely the blue beam, where moist atmosphere shows bluish
colors. Dry descending stratospheric air related to an advection jet appears in reddish colours. This
color scheme is useful for studying Rapid Cyclogenesis, Jet Stream Analysis, and PV Analysis.
Figure 55