Content uploaded by Martin Setvák
Author content
All content in this area was uploaded by Martin Setvák on Feb 08, 2023
Content may be subject to copyright.
PROCEEDlNGS
EUM P 06
lssN 1015-9576
4TH
AVHRR DATA
USERS,MEETING
Rothenbuí9, F. R. Germany,
5-8 September 1989
,>^
Eu/LlETsAT \ai
CALCULATTON 0F THE CHANNEL 3REFLECTlVTTY ITS APPLICATION *
T0 CONVECTIVt STORMS RESEARCH AND T0 SNOl^/ COVER RECOGNITION
MARTTN SETVÁK and KAREL HLAVATÝ
Czech Hydrometeorologic91 Institute
pracoviště Libuš, Na Sabatce Ll
l4306 Prague 4, Czechoslovakía
ABSTRACT
A method for separation of the reflected solar component and for
con5ecutive computation of the NOAA / AVHRR channe1 J ref}ectivity
t,vas developed ín Czech Hydrometeorologica1 Institute at Prague.
The method combining caIibrated data from AVHRR channels ] and 4
a5sumes severa1 simplifications.
The infIuence of those simplificatíons is the 1east when applying
this method to cloud tops of convective storms. These exhibit from
time to time an increased channe1 J cloud top ref}ectivity (CH ] CTR)
which can be explained by difference5 in cloud top microphysica1
composition only. There ale probably two basic mechanisms which
produce particles responsible for an increased CH ] CTR, a11 the
observed patterns of an increased CH ] CTR appear to be produced by
one of these mechanisms oL by their combination. There seems to be
a 1ink between increased CH ] CTR and an occurence of hailstorms.
The channe1 3 reflectívity can also be successfully used for snow
cover recognition. The advantage of use of the AVHRR channef ] data
for these purposes is in much srnaller dependence on a type of
vegetation cover when compared with visible or near infrared data
( AVHRR channels 1 and 2 ).
CALCULATION 0F THE AVHRR CHANNEL ] REFLECTTVITY
Denoting eJ as the emissivity related to AVHRR channe1 ] andu1 as
the channe1 ] ref}ectivity, the tota1 radiance N3 measured by the
channe1 ] during daytime can be expressed as
N3 - N3(ref ) * €a,N3(T) ( 1 )
where N3(ref) stands for the reflected component of the channe1 3,
and el.N3(T) for the component emitted by a body at temperature T
and havíng the emissivity of L3. The reflected component can be
specified as
*paper submitted but not presented
171
Nl(ref ) - a3.N3(5B00 K)
o3 =,§;!r$|
(x)," cos I(2)
where N3(5B00 K) ls radiance from the solar photosphere, R is the
radius of the Sun, _r is !h" radius of the Earth's orbit and f is
the zenith angle of the Sun. Then, denoting
S;(r,|) = N](5B00-' (l) 2,cosP (])
the equation ( 1 ) can be rewtitten as
N3 = &J.T(r,|) + 9l.N3(T) (4)
Note that the method assumes the independence of €a and al3 on V
wíthin the range of channe1 ]. Tf now, fol sufficiently dense cIouds
or for the Earth's surface, a zeTo transmissivity ís a5sumed,
Kirchhoff 1aw wi1} hold
% **3 = ] (5)
From (4) and (5), relations fol computation of the emissivity and
reflectivity in the channe1 J are obtained :
Nl - Ni(T)
-) 53(r,P) - N1(I)
(6)
(1)
counts and radiances for each pixel , while S] ( r,l ) is calculated
fron (3) if the actua1 values of r and P are known'.
The vafue of N3 is obtained from the ca]ibration relation between
A certain simplificatíon has to be made in order to determine the
vaIue of Nl(T). To be able to determine it, rnie need to know the
temperature T. This can be found from channe1 4 data plovided that
0+ = 1. Since rea1 values of O4 ale always smaller than l, this
step is a soulce of certain error in the calcu}ations.
What are the simplifications and inaccuracies of this method ?
a) absorption and dispersion of the incident, reflected and emitted
radiation by the atmosphere are neglected
b ) emissivíty is neglected when determining the temperature
from channe1 4
c) effect of water vapour on data measured by channe1 4 is neglected
d) zero transmissivity is assumed
e) diffuse reflection ( Lambertian surface ) is assumed.
2. USE 0F THE CHANNEL ] REFLECTTVITY FOR CONVECTIVE STORMS RE5EARCH
Ice Particles which make up the cIoud tops of convective storms
should behave in IR channels almost as a black body, i.e. their
channel ] reflectivity should approach zero ( Scorer , L'9B7 ). Using
a LUT where ].ower count values ( higher radiance5 ) are depicted aŠ
darker in comparison with higher count vaIues ( Iower radiances ),
convective storms shouId appear in channeI } images as nearly white,
homogeneous area5. However, when a suitable LUŤ is used, channe1 3
images sometimes show convective storms containing darker parts,
or darker as a whole when compared with the surrounding convective
storms (Lil3as, 1987). By way of comparison with enhanóed channel 4
172
images which depict only temperatures ranging from -40" to -70"C,
higher temperatures, can ímmediately be eliminated as a reason for
observing higher channe1 3 radiances in these dark parts. From (4)
it then fo11ows that the explanation shoulri be sought in the increased
channe1 ] c]oud top refIectivity ( CH l CTR ).
The increased CH ] CTR has been registered for early sta9es of storm
deve]opement as we1] as for storms in mature or dissipating stages.
From the observations it is obvious that there are at 1east two
independent mechanisms which produce particles responsible for the
increased CH ] CTR ( Setvák, 1989 ). In the first case the area of
increased CH ] CTR is associated with overshooting tops, penetrating
towers, smaller single - ce11 storms or it extends over the whole
area of the cloud top of a storm. In the second case the area of
íncreased CH ] CTR resembles a plume or a V-shape pattern which is
produced by a source síze of whích is comparable with one pixel or
smaller. The first type occurs mOre frequently than the second one,
only one case was registered when both types occured simu}taneously.
The highest values of CH ] CTR as determined by our method amount
to 12%, whereas common storms show CH ] CTR values of L -3%. Let us
note that the impact of the simplifications mentioned above is the
1east when the method is appIied to cloud tops of convective storms
( Setvák, L989 ). The reason for such 9reat differences in CH ] CIR
shou]d be looked for in the microphysica] composition of cloud tops,
most 1ikely in the size and / or shape of ice particles present in
cloud tops. At the same tirne, both the size and shape of ice partícles
Figure 1. NOAA11 BJULl9B9 1]15UTC. AVHRR channel2 (left) andchannel} (right)
images, showing convective storms over south Germany. One ce}l of the storm in
the ríght exhibits increased cH ] cTR of the first type (see text) and at the same
time generates fronr íts centre a plume with CH ] CTR only a little lower than its
own, but stíll much higher than is CHJCTR of rest of the storms in the area.
173
are affected by the environment and the conditions ( e.g. vertica1
velocities) insíde the storm under which ice particles are formed.
From this point of víew we can regard CH ] CTR as a source of ínfor-
mation about processes inside a storm.
It seems that there can be a connection between the increased
CH ] CTR and the occurence of hailstorms. Tf a si9nificant hailstorm
is registered either by a public source or by a ground meteorologica1
station and if satellite data are available for that period, the
relevant convective storm exhibits an increased CH ] CTR. However,
due to a 1ack of ground based data and insufficient radar measure-
ment5, this could not be up to date proved unambiguously.
j. USE 0F THE CHANNEL ] REFLECTIVIl Y FOR SNOliJ COVER RECOGNlTION
A number of automatic c]oud detection and classification methods
t,las developed in the 1ast years where great importance was given to
channe1 ] data ( ".g. Saunders, 19BB; LiIjas, L9B1 1 Turner, L981 ),
eSpecially when detecting c]ouds over snotnl-covered 1and. But channe1 ]
Can a}so be successfully used for snow cover lecognition as the
channel ] reflectivity of snow is 1ower than that of the bare 1and.
]f the above described method is applied to these purposes, the
impact of the simplificatíons of the method is much greater than ín
the case of convective storms. Nevertheless, the method ,lvas tested
on NOAA 9 data from ] wínter / spring 19B1 and 19BB cloud - free
situations. 0btained u3 values were cOmpared wíth snow reports from
about 65 stations from tvestern part of Czechos]ovakia. Even though
Some negative vaIues of U, |Ýele obtained (which is an erlor resu1-
ting from the simpIifications ) , this test clearly showed that the
values of oa1 6 l% corre5pond a]rnost allvays to snow - covered 1and,
values between 1 and j% to partly snow - covered 1and and those
higher than 3% to snotnl - f ree ]and. l,ťhat appears to be a great advan-
tage of this method for the snow cover Lecognition - compared to
methods using visible and near infrared data - is its almost entire
independence on the slope of the examíned aLea and on forest masking.
No significant colrelation between uó value and snow cover height
t^las f ound.
REFERENCES
LILJAS, E., (1981): Multispectral methods for cloud classifícation. Proc.
and Radar Imagery Interpretation, Reading, July I9B1, p. 475 - 493
SAUNDERS, R.W., KRTEBEL, K.T., (l9BB): An improved method for detecting
and cloudy radiances from AVHRR data. Tnt.J.Remote Sensing, Vo1.9, No.1,
SateIlite
clear sky
p. L23-L50
SETVÁK, M., (1989): Convective storms-the AVHRR channel] c}oud top reflectlvity
as a consequence of internal processes. Proc. Weather Modification and Appl íed
Cloud Physics, Beijing, May I9B9 , WMO / TD - No . 269 , p. I09 - II2
SCORER, R.S., (L9B7): Convectíve rain as seen by channel }. Proc. Sate}lite and
Radar Imagery Tnterpretation, Reading, July 1987, p.509 -5L9
Ti]Ri',lER, J., (L9B7): Detection of cloud over ice. Proc. Satellite and Radar Inragery
Tnterpretation, July I9B] , p. 52I - 5JJ
174