TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 10, Octobe
r 20
14, pp. 7287
~ 729
8
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.597
9
7287
Re
cei
v
ed Ma
rch 1
4
, 2014;
Re
vised July
14, 2014; Accepted Augu
st
3, 2014
Land Surface Temperature Retrieval from the Mediu
m
Resolution Spectral Imager (MERSI) Thermal Data
Hailei Liu
1,2
*, Shenglan Z
h
ang
1,2
1
Colle
ge of Atmosph
e
ric Sou
ndi
ng, Ch
eng
d
u
Univ
ersit
y
of Information T
e
chno
log
y
,
Che
ngd
u 61
02
25, Chi
na;
2
Institute of Atmosph
e
ric Ph
ysics,
Chin
ese
Academ
y of Sc
ienc
es,
Beiji
ng 1
0
0
029
, China
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: liuh
a
il
ei@c
uit
.
edu.cn
A
b
st
r
a
ct
A singl
e cha
n
n
e
l la
nd surfac
e
temp
eratur
e (LST
) retrieval
alg
o
rith
m n
a
m
ed Sin
g
l
e
Ch
a
nne
l W
a
ter
Vapor
De
pe
nd
ent (SCW
VD)
meth
od
w
a
s p
r
esente
d
fo
r
Medi
um R
e
sol
u
tion
Sp
ectral
Ima
ger
(MER
S
I)
thermal i
n
frare
d
ban
d ab
oar
d
FengYun-
3A (
F
Y-3A) sate
llit
e. W
a
ter Vapor
Content (W
VC
) is the only in
put
para
m
eter in the al
gorith
m
a
ssumi
n
g
the surface e
m
iss
i
vi
ty is k
now
n. NCEP rean
alysi
s mont
hly
me
a
n
datasets
are
u
s
ed to
d
e
vel
o
p
the S
C
W
V
D
alg
o
rith
m. So
m
e
te
sts, i
n
cl
ud
i
n
g
gl
o
bal
n
u
mer
i
cal
si
mu
lat
i
on
s
and
val
i
dati
o
n
s
w
i
th both
in
-situ
meas
ure
m
e
n
ts
a
nd M
O
DIS LST
pro
duct at
Lake
T
ahoe,
USA,
w
e
re
carried
out to
eval
uate th
e a
l
gorit
hm perfor
m
a
n
ce. C
o
mp
ared w
i
th
NCE
P data
an
d U.
S. standar
d
mi
d
-
latitud
e
su
mme
r atmosp
here
mo
de
l, the r
e
tri
e
ved
LST
fro
m
si
mul
a
ted
ME
RSI brig
htness
temper
ature w
i
th
MODT
RAN ha
d a RMSE ab
o
u
t 0.8 K. In the valid
atio
n,
MERSI Level 2 w
a
ter vapor pro
d
u
c
t w
a
s employ
ed,
and the MER
S
I band e
m
iss
i
vity w
a
s evaluated us
i
ng th
e MODIS band 31 an
d 32 emissivity w
i
th
a
n
empiric
a
l expr
essio
n
. T
he results show
that the
di
ffere
nce
betw
e
e
n
th
e r
e
trieve
d MERS
I LST
a
n
d
the
in-
situ me
asur
e
m
ents is less tha
n
1 K in most situatio
ns. T
he comparis
on w
i
th
the MODIS LST
products (V5
)
show
s that the RMSE
is abo
ut 2.3 K.
Ke
y
w
ords
:
MERSI, land surf
ace te
mper
atur
e, w
a
ter vapor content, sin
g
le
chan
nel; b
and
emissivity
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Medium
Re
solution Spe
c
tral Image
r (M
ERSI)
is on
e
of the 11 instrument
s abo
ard FY-
3
A
sp
ac
ec
r
a
ft which i
s
th
e first satellite of the se
cond gene
ration of Chine
s
e polar-orbiting
meteorological satellites l
aunched on
27 May 2008.
MERSI is
a multispectral and medium-
resolution sp
ectral imager.
It has 20 cha
nnels, of
which there are fo
ur VIS and one TIR chann
e
l
s
with a high sp
atial resolution of 250 m, which enabl
es i
m
aging of the Earth with high resolution
in
natural color
during the da
y and high re
solution TIR imaging durin
g the night [1
, 2]. T
h
e
s
e
da
ta
improve
ou
r unde
rsta
ndin
g
s of
glob
al dynamics an
d processe
s
occurrin
g
on
the
land, oce
ans,
and
in
the
lower
atmos
p
here. Det
a
ils
about
some
spect
r
al p
r
op
ert
i
es of
ME
RS
I
are list
ed in
Table 1.
LST is a key para
m
eter of the surfa
c
e p
h
ysic
al pro
c
e
s
ses on regio
nal and glob
a
l
scale
s.
It plays an importa
nt role
in many app
lication
s
such
as ag
riculture, geoscie
n
ces, clim
ate a
n
d
other environ
mental
studi
e
s
[3
-5
]. Depe
nding
on
the
regio
n
whe
r
e
land
surfa
c
e
processe
s
a
r
e
monitored, hi
gher
spatial a
nd tempo
r
al resol
u
tion
s are need
ed, wh
ich can be off
e
red by FY-3
A
MERSI. However, like th
e
Land
sat mi
ssion
s
, on
e
of
the main li
mitations of
MERSI thermal
informatio
n is the presen
ce of only o
n
e
cha
nnel
i
n
t
he TIR
sp
ect
r
al regio
n
. It can
not u
s
e t
h
e
split-wind
o
w
techni
que, th
e multi-cha
n
nel meth
od
o
r
the m
u
lti-a
ngle m
e
thod,
whi
c
h m
a
ke
s it
more difficult to perform LST retrieval.
Several atte
mpts have
b
een d
one to
perfo
rm LST
retrieval fo
r the La
nd
sat 5
TM and
Land
sat 7 ETM+ TIR b
a
n
d
[6-8], but few have b
e
e
n
repo
rted fo
r MERSI data. What is m
o
re,
most of
tho
s
e previo
us method
s req
u
ire
i
n
fo
rm
ation fro
m
atm
o
sp
heri
c
ra
di
oso
undi
ngs t
o
perfo
rm atm
o
sp
heri
c
correction for LS
T retrie
val. Qin et al. developed a m
ono-win
d
o
w
LST
retrieval
algo
rithm for
Lan
dsat TM
6 da
ta usin
g gro
und emi
s
sivity, atmosphe
ric tran
smittan
c
e
and effe
ctive mean
atmo
sph
e
ri
c temp
eratu
r
e a
s
i
n
put pa
ramete
rs [9]. In th
e
mono
-wi
ndo
w
algorith
m
, the
water va
po
r
conte
n
t (WVC) i
s
de
sig
n
e
d
as
0 to 3 g/
cm
2
, whic
h limits
LST retrieval
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 728
7
– 7298
7288
whe
n
the a
c
t
ual WV
C b
e
yond 3
g
/cm
2
.
More
over, ai
r temperature
s
a
r
e n
o
t ava
ilable
when
o
n
e
wishe
s
to retrieve LST over larg
e are
a
s.
Jim
ё
ne
z-Mu
ñoz a
nd Sob
r
ino develope
d a gene
rali
zed
singl
e-cha
n
n
e
l method u
s
i
ng WV
C as t
he only input
para
m
eter,
which mi
nimize
s the inp
u
t data
requi
re
d and
then provid
es a
n
ope
rat
i
onal meth
od
ology to retri
e
ve LST fro
m
the Lan
dsat 5
thermal ba
nd
[10, 11]. Th
ey used thre
e param
eters (
Ψ
1,
Ψ
2
and
Ψ
3) d
epen
ding on WV
C to
retrieve
LST.
Each
of the
p
a
ram
e
ter ha
s a
relatio
n
ship with
WV
C,
whi
c
h
ha
s b
e
en exp
r
e
s
sed
by
statistical fits. As three mi
ddle pa
ram
e
ters
(
Ψ
1,
Ψ
2 and
Ψ
3)
hav
e been
used
in the algo
rithm,
more u
n
certai
nties would
b
e
introd
uced
durin
g fi
tting the middl
e pa
rameter to
WV
C re
sp
ectively.
An erro
r in th
e water vap
o
r source
coul
d
lead to
anoth
e
r erro
r in the
three pa
ram
e
ters, which will
dram
atically prop
agate
to the
LST retri
e
vals.
T
h
is problem i
s
com
m
on to
any t
e
ch
niqu
e b
a
sed
on a dire
ct si
ngle-ch
ann
el inversi
on of the radi
ation tran
sferring e
quati
on (RTE
), in which the
final retrieval
s
are ve
ry se
nsitive to unc
ertaintie
s
on the input pa
ra
meters [12].
In this pap
er,
an advan
ce
d ope
rative si
ngle chan
nel
LST retri
e
val algorith
m
for
MERSI
TIR d
a
ta
wa
s p
r
o
p
o
s
ed.
Assu
ming
th
at land
su
rfa
c
e emi
ssivity (LSE) is kn
own, LST ca
n be
retrieve
d by t
h
is n
e
w
adva
n
ce
d alg
o
rith
m usi
ng
WV
C a
s
the
onl
y input pa
ra
meter. Comp
ared
with the
p
r
evi
ous meth
od
s, we
mai
n
ly focu
se
d
o
n
i
m
provin
g the
accu
ra
cy of
retrieve
d LST
by
decrea
s
in
g u
n
ce
rtaintie
s i
n
trodu
ce
d in
the three p
a
rameters fitting to WVC, a
nd the validit
y o
f
this algo
rithm
when the
WVC in atmosp
here b
e
yond
3g/cm
2
.
Table 1. MERSI Channel
Chara
c
te
risti
c
s (partial
)
Chan
nel
W
a
v
e
le
ng
th
(
μ
m
)
Band
w
i
dth(
μ
m)
Sub-
poi
nt
R
e
s
o
lu
t
i
on
(m
)
NE
∆
T /
Ρ
(%
)
K (300K
)
Primar
y
use
3
0.650
0.05
250
0.4
Land surface emi
ssivity
4 0.865
0.05
250
0.45
5
11.25
2.5
250
0.5K
Land surface te
mperatu
r
e
17 0.905
0.02
1000
0.10
W
a
t
e
r v
a
p
o
r
content
18
0.940
0.02
1000
0.10
19
0.980
0.02
1000
0.10
2. Theor
y
an
d Method
olo
g
y
2.1. Atmosp
heric Ra
diati
v
e
Transfer
Gene
rally
sp
eaki
ng, the
g
r
oun
d i
s
n
o
t
a bla
c
kbody,
thus groun
d
emissivity h
a
s to
be
con
s
id
ere
d
fo
r computin
g t
herm
a
l radia
n
ce
emi
tted
by gro
und. Al
so atm
o
sphe
re ha
s imp
o
rt
ant
effects
on th
e re
ceive
d
radian
ce
at remote
sen
s
o
r
level. F
o
r
a pla
ne-pa
ra
llel cl
oud
fre
e
atmosp
he
re
unde
r lo
cal
therm
odynami
c
e
quilib
rium
, ignori
ng
scattering i
n
flu
ence, the
RTE
descri
b
ing
th
e radiation
int
ensity o
b
serv
ed in
chan
nel
i
at
ze
nith a
ngle
θ
,
can
b
e
form
ulated
by
inclu
d
ing the
radia
n
ce emit
ted by the ground, t
he u
p
w
elling
radi
ation emitted by
the atmosph
e
re
towards the
sensor, and th
e downwellin
g radiatio
n
e
m
itted by the atmosp
here that rea
c
he
s
the
Earth’s
surfa
c
e an
d is th
en refle
c
ted
toward
s the
sen
s
o
r
. Therefore, the T
O
A radia
n
ce
Ii(
θ
)
measured by
the satellite
sen
s
o
r
in
chann
el
i
at t
he zenith
an
gle
θ
can b
e
app
roximat
e
ly
expre
s
sed a
s
[13]:
I
i
(
θ
)=
B
i
(T
i
)=
τ
i
(
θ
)
ε
i
B
i
(T
s
)+I
i
↑
+
τ
i
(
θ
)(
1-
ε
i
)I
i
↓
(1)
Whe
r
e
T
s
i
s
the LST.
T
i
(
θ
),
τ
i
(
θ
)
and
ε
i
(
θ
)
are
th
e at-sen
so
r
brightn
e
ss te
mperature, t
h
e
atmosp
he
ric transmittance
and gro
und
emissivity in cha
nnel
i
at
zenith
θ
.
I
i
↑
and I
i
↓
i
s
t
h
e
atmosp
he
ric
path a
nd d
o
w
n
w
ard r
adi
ance, re
sp
ectively. To ob
ta
in LST, three atmo
sp
he
ric
para
m
eters (
τ
, I
i
↑
and I
i
↓
) an
d one ba
nd a
v
erage e
m
issivity should b
e
determi
ned.
In TIR band, the LST retrie
val problem
can
be viewed
as two interd
epen
dent pro
c
e
s
ses:
corre
c
tion fo
r the effects of
the atmosp
h
e
re, an
d the uncouplin
g of
the surfa
c
e t
e
mpe
r
ature a
n
d
emissivity. As for MERSI TI
R data, the
g
eneral obj
ecti
ve of atmosp
heri
c
corre
c
tion alg
o
rithm
s
is
to rem
o
ve th
e atmo
sph
e
ri
c effe
cts, e
s
p
e
cially of
wat
e
r vap
o
r
ab
sorption. A
nd t
hen
an a
c
cu
rate
estimation of
the surf
ace temperatur
e and emissivity
will be obtained.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Land Su
rface
Tem
perature
Retrie
val fro
m
the Medi
um
Resolution Spectral
Im
a
ger… (Hailei Liu)
7289
2.2. The simplification of Planck's Fu
nction
In orde
r to derive
T
s
from
Equation (1
), it is cruci
a
l to simplify the Planck'
s functio
n
esp
e
ci
ally for the
singl
e
chann
el meth
od a
nd
split
-wind
o
w alg
o
rithm [14, 1
5
]. According
to
Qin
[9], there i
s
an a
p
p
r
oxi
m
ate line
a
rit
y
bet
we
en
LST an
d Pl
anck’
s radia
n
ce
in 1
1
.2
5
μ
m.
Therefore
，
the simplifi
c
ati
on of Planck’
s functi
o
n
ca
n be expre
ssed as follo
ws:
B
(
T
)
=
a
+
b
T
(2)
Whe
r
e
a, b
is the
reg
r
e
ssi
on
coeffici
ent
s, an
d
ca
n
be
assig
ned
-2
3
.
87, 0.109
9
resp
ectively
with
a RMSE of 0.06, Whe
n
T
is in the ran
g
e
of 260~3
00K
.
2.3. The Deri
v
a
tion of Sin
g
le Chann
e
l Algorithm fo
r MERSI Data
The derivatio
n
of singl
e ch
annel
alg
o
rith
m
is
ba
se
d o
n
radian
ce
transfe
r
(1
). Accordin
g
to the simplification of pla
n
k
functio
n
me
ntioned a
bov
e, the (1) can
be rewritten
as:
a + b*T
i
=
ε
i
τ
i
(a+
b
*T
s
)+
I
i
↑
+
τ
i
(
θ
)(1
-
ε
i
)I
i
↓
(3)
Solving for Ts, we obtain th
e algorith
m
for LS
T retri
e
va
l from MERSI TIR data as f
o
llows:
Error!
(
4
)
Equation (5) i
s
re
written in
(6) fo
r simplifi
c
ation
：
T
s
=
A
T
i
+
B
(5)
Whe
r
e,
Error!
,
Error!
.
Wate
r vapor i
s
the major a
b
so
rbe
r
in the TI
R, and WVC in the atmosp
he
re varies b
o
th
spatially and
temporally, its effect on
transmi
ss
io
n in the TIR can al
so vary [16]. So it is
importa
nt for
the algo
rithm
to use water vapor
as
an
input vari
able
to improve t
he a
c
curacy
of
the LST
retrieval [17]. As
sho
w
n i
n
Figure 1,
co
efficients (
A
and
B
) h
a
v
e a qu
ad
ra
tic
depe
nden
ce
on WV
C (
w
)
respe
c
tively. Thus, th
e rel
a
tionship b
e
twee
n coeffici
ents
(
A
,
B
) a
n
d
WVC can
be expre
s
sed
a
s
:
A=
a1*w2+
a
2*w+a3
(6a
)
B=
b1*w2+
b
2*w+b3
(6b
)
Figure 1.
Rel
a
tionship bet
wee
n
the Two Coeffici
ents (A, B) and WVC
0
1
2
3
4
5
6
7
8
0
2
4
6
8
C
oef
f
i
c
i
ent
A
W
a
t
e
r (g
/
c
m
2
)
C
oef
f
i
c
i
ent
A
and
B
v
s
.
w
a
t
e
r
v
a
por
0
1
2
3
4
5
6
7
8
-8
00
-6
00
-4
00
-2
00
0
C
oef
f
i
c
i
ent
B
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 728
7
– 7298
7290
Combi
n
ing
(6
a) and
(6b
)
wi
th (5) results i
n
a new relati
on to be deriv
ed between
Ts
,
Ti
and
w
:
T
s
= (a1*
w
2
+a
2*w+a3
)Ti +
b1*w
2
+b2
*
w+b3
(7)
Usi
ng a n
onl
inear
reg
r
e
ssion techniqu
e, six co
efficients
a
i
an
d
b
i
(i=1,2,3)
can be
determi
ned from (7
). In the next se
ctio
n, we
will a
d
d
re
ss th
e pro
g
re
ss fo
r det
ermin
a
tion of
th
e
regres
s
i
on coeffic
i
ents
in details
.
2.4. Dete
rm
ination of
the
SCWVD
Co
efficie
n
ts
Global
-ba
s
e
d
simulation
d
a
taset
s
, inclu
d
ing atmo
sph
e
ric
pro
fi
le
s, surfa
c
e tem
p
eratu
r
e,
and surfa
c
e e
m
issivity, were use
d
to develop SCWVD algorithm.
The atmo
sp
h
e
ric pro
fi
l
e
s
(geo
potential
height, air tempe
r
ature, and h
u
midity) were
derived
from
monthly m
ean p
r
o
d
u
c
ts (2.5 g
r
id-point spa
c
in
g) fro
m
NCEP climate
data
assimilation
system (CDAS
)
rea
nalysi
s
p
r
oje
c
t [18
]. We sele
cted
46
7 pixels u
n
ifo
r
mly over lan
d
on glob
al sca
l
es in Janu
ary and July fro
m
2000 to 20
07 (see Fig
u
re 2(a
)). Th
us,
there a
r
e 74
72
sampl
e
s in e
i
ght years. T
hen we
ca
rri
ed out
clou
d
detection
s o
v
er the 7472
sample
s u
s
i
n
g
MODIS mont
hly fraction p
r
odu
cts [19] b
y
setting a cr
i
t
erion a
s
0.3.
If the cloud fraction in
a pi
xe
l
wa
s la
rge
r
th
an 0.3, it
was co
nsi
dered
a
s
clou
d
conta
m
inated,
and
the pixel
wa
s eli
m
inated.
At
last, 6757
sa
mples
und
er
clou
d cle
a
r
condition
s were sele
cted. A
s
sh
own in F
i
gure
2(b
)
, 4
2
7
pixels were
retained after
clou
d dete
c
tion in Ja
nua
ry 2001.
In orde
r to en
large th
e validity of calcul
a
t
ed co
efficien
ts, we did th
e
following thi
n
gs: (1
)
LST wa
s pro
v
ided by addi
ng -6, -3, 0, 3 and 6K to
the su
rfa
c
e ai
r tempe
r
ature
of each p
r
ofi
l
e.
(2)
LSE wa
s
set from
0.90
to 1.00 with
0.01 interval
s increa
se. (3
) The view
ze
nith angle
wa
s
set to
be
the
value
s
: 0
°
, 1
5
° a
n
d
30
°. (4) T
h
e
surfa
c
e elevatio
n a
t
each
pixel
wa
s ta
ken
from
USGS (U.S.
Geolo
g
ical Survey) GT
OPO30, and
the
satellite altitude is a
s
sum
ed to be 705
km.
At last, 1,114,905 (6
757
×5
×11
×
3
)
pai
rs
of LST and
the at-sen
so
r brightn
e
ss te
mperature
s
f
o
r
the MERSI T
I
R ba
nd
s a
r
e
gen
erate
d
from the
ra
di
at
ive tran
sfer calcul
ation
s
(MODT
R
A
N
4
.
0).
The d
a
tasets we
re
split i
n
to a trai
nin
g
data
s
et u
s
ed for
cal
c
ul
ating coeffici
ents
(78
0
,43
4
pattern
s)
and
an ind
epen
d
ent test set u
s
ed to
ev
alua
te its pe
rform
ance (3
34,47
1 pattern
s). T
he
SCWV
D coef
ficients
we
re
calculated
b
y
a leas
t sq
uare m
e
thod.
Table 2
sho
w
s th
e deriv
ed
coeffici
ents.
The
RMSE for ea
ch S
C
WVD equ
ation
r
ang
es f
r
om
0.81K to 0.91
K at the different
emissivity (0.90 to 1.00).
(a)
(b)
Figure 2. (a)
Global di
strib
u
tion of the 467 pixels
; (b
)
Global di
strib
u
tion of retain
ed 427 pixel
s
after clou
d de
tection in Jan
uary 200
1
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Land Su
rface
Tem
perature
Retrie
val fro
m
the Medi
um
Resolution Spectral
Im
a
ger… (Hailei Liu)
7291
Table 2. Coe
fficients an
d RMSE (in Kel
v
in
) of Equation (7
) for ME
RSI TIR Cha
nnel
Emissi
v
i
t
y
a1
a2
a2
b1
b2
b3
RMSE (in K
)
1.0000
0.014139
0.023359
1.0284
-4.1175
-5.4869
-5.4909
0.81
0.9900
0.015181
0.02238
1.0331
-4.4023
-5.3201
-6.1495
0.83
0.9800
0.016371
0.02088
1.0371
-4.7394
-4.9526
-6.6638
0.86
0.9700
0.016847
0.02063
1.0418
-4.8643
-4.9873
-7.3307
0.88
0.9600
0.016545
0.02212
1.0454
-4.788
-5.4445
-7.7097
0.89
0.9500
0.013779
0.02618
1.0497
-4.006
-6.6615
-8.2341
0.88
0.9400
0.012322
0.02883
1.0553
-3.5843
-7.5506
-9.0678
0.88
0.9300
0.008616
0.03033
1.0612
-2.5221
-8.1346
-9.9687
0.91
0.9200
0.002974
0.03149
1.0676
-0.88283
-8.7275
-10.964
0.87
0.9100
0.001608
0.02303
1.0742
-0.057323
-6.5891
-12.084
0.84
2.5. Sensitiv
e Analy
s
is
Provided t
hat
gro
und
emi
s
sivity is
kno
w
n,
the
SCWVD
algo
rith
m for ME
RSI req
u
ire
s
WVC a
s
the only one pa
ra
meter. Sensiti
v
ity and erro
r
analysi
s
in term of the unce
r
tainty of WVC
in the atmosp
here a
r
e p
r
e
s
ented in this
se
ction.
The sen
s
itivity analysis
of retrieve
d LST
wa
s ca
rri
ed
out with the
chang
e of WV
C in the
stand
ard atm
o
sp
here (Mid
-latitude sum
m
er atmo
sph
e
re, WV
C =
2.92 g/cm2
)
simulate
d wit
h
MODT
RA
N 4
.
0. As sh
own
in Figu
re 3, th
e LST er
ro
r li
nearly in
crea
se
s with th
e
WVC e
r
ror
rising
in all ca
se
s,
esp
e
ci
ally
wh
en t
he
WVC
is sm
alle
r tha
n
the truth va
lue. Wh
en e
m
issivity is lo
w,
the increa
sin
g
of LST e
r
ror i
s
mu
ch
slo
w
er. T
he
maximum e
r
ror d
oes not
excee
d
0.8
K
whateve
r
the
emissivity when
the erro
r on WVC i
s
less than 0.
5 g/cm2. In this ca
se, the
maximum
error
obtaine
d
depe
nd
s mo
re on
the
em
issivity. If on
e con
s
ide
r
s
only the
ca
ses
emissivity larger th
an 0.9
5
,
the maximu
m error
on th
e retri
e
ved te
mperature
do
es n
o
t exce
e
d
0.6 K when th
e error on
W is less than 0.
5 g/cm2.
As a
con
c
lu
sion, the se
nsitivity of the SCWV
D
method to errors o
n
SCWV
D increa
se
s for greater
WVC and lo
we
r emissivity.
Figure 3.
Rel
a
tionship
s
be
tween the retrieved LST
error and
WVC
error at given
different
emissivity
2.6. Land Surfac
e Emissiv
i
t
y
and Water Vapor Co
nten
t
2.6.1. Land Surfac
e Emissiv
i
t
y
It is very
ch
allengin
g
to
accurately e
s
timate lan
d
su
rface
emi
s
sivity (LSE) at th
e glo
b
a
l
scale.
F
o
r water su
rface whi
c
h
i
s
com
parativ
ely h
o
m
ogen
eou
s,
a con
s
tant e
m
issivity
can
be
assume
d; for land
surfa
c
e, the LSE
dynamics
ha
ve wid
e
r
ra
n
ge a
nd
ca
n
vary over sh
ort
distan
ce. S
e
veral
metho
d
s have
bee
n
reporte
d
on th
e ba
si
s of
eit
her the
norm
a
lize
d
differe
nce
vegetation in
dex (NDVI) o
r
the lan
d
cover in
fo
rmatio
n [20-2
4
]. Re
d and
NIR
ch
annel
s of ME
RSI
in Table 1 ca
n be used to derive NDVI for cal
c
ul
ating
LSE.
In this resea
r
ch,
an e
m
p
i
rical
expressi
on
wa
s b
u
ilt to evaluate the ME
RSI band
emissivity usi
ng the
MO
DI
S band
31
a
nd 3
2
e
m
issi
vity. Figure 4
(
a) give
s ME
RSI an
d MO
DIS
-1
.
5
-1
-0
.
5
0
0.
5
1
1.
5
-2
-1
.
5
-1
-0
.
5
0
0.
5
1
1.
5
W
a
t
e
r
v
a
por
c
o
n
t
en
t
e
r
r
o
r
(
g/
cm
2
)
LST
e
r
ror
(
K
)
Em=
0
.9
2
Em=
0
.9
4
Em=
0
.9
6
Em=
0
.9
8
Em=
1
.0
0
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 728
7
– 7298
7292
thermal i
n
fra
r
ed ba
nd spe
c
tral re
sp
on
se
s fun
c
tion. M
E
RSI thermal
band i
s
lo
cat
ed in the
regi
on
8~1
4
μ
m [25,
26]. It is possible to
estim
a
te the MERSI band emi
s
sivity using
MODIS ba
nd
31
and 32 e
m
issivity due to th
e lowe
r emi
s
sivity values
variation in 8
-
14
μ
m. In ord
e
r to analy
z
e
the
relation
shi
p
betwe
en th
e
MERSI a
n
d
MO
DIS th
ermal
infra
r
e
d
ba
nd
emi
ssivity. Surfa
c
e
emissivity
was provid
ed by
usi
ng 55
materi
als
(water,
snow/i
ce, veg
e
tatio
n
, and
soil,
etc.)
selected f
r
om a spectral
library (http://speclib
.jpl.nasa.gov/) [27]. Then, the band
average
emissivity
wa
s cal
c
ulated
usin
g
the
sel
e
cted
JPL
e
m
issivity spe
c
tra
co
nvolve
d with th
e M
E
RSI
and
MO
DIS respon
se
fun
c
tion. Fi
gure
4(b)
sh
ow
s
the relation
ship b
e
twe
en
MERSI thermal
band emi
s
si
vity and the average
e
m
issivity
of MODIS ban
d
31 and b
a
nd 32. The
fina
l
expre
ssi
on for MERSI LSE is given by:
ε
mersi
=
0.791(
ε
m
odi
s31+
ε
mo
dis32
)
+0.2
04. In our
validation, we will
use the MO
DIS Land emi
s
sivity product
(M
OD11_L2) to derive MERSI
emissivity maps for the LS
T retrieval.
Figure 4. (a)
Therm
a
l ban
d respon
se fu
ncti
on
s for M
E
RSI and M
O
DIS; (b)
Rel
a
tionship
betwe
en ME
RSI Therm
a
l band emi
s
sivity and the
averag
e emi
ssi
vity of MODIS band 31 an
d
band 3
2
2.6.2. Water Vapor Con
t
e
n
t
As sho
w
n i
n
Table
1, ME
RSI ha
s thre
e chann
els which
ca
n b
e
u
s
ed
to e
s
tima
te WVC,
inclu
d
ing st
ro
ng wate
r vap
o
r ab
sorption
line at 0.940
μ
m, weak
wa
ter vapor ab
sorption lin
e at
0.905
μ
m a
n
d
atmosp
he
ric wind
ow at 0
.
980
μ
m. The
WVC
can
b
e
derive
d
u
s
i
ng the reflect
ed
sola
r ra
dian
ce mea
s
ureme
n
t [28, 29]. In
this re
sea
r
ch
, the MERSI L2 PWV prod
uct is u
s
e
d
. The
details of the
MERSI L2 PWV algorith
m
can be fo
und in [30]. The metho
d
adopte
d
here
for
PWV retri
e
val is based on the ra
tio of
reflected solar radi
ance
detected by
satellite between
water vap
o
r
absorptio
n chann
els an
d
atmosph
e
ri
c windo
w ch
a
nnel
s. By e
m
ploying cha
nnel
ratios, the
a
e
ro
sol extin
c
tion distri
buti
on an
d t
he
variation effe
ct of su
rfa
c
e
reflecta
nce
are
partially
remo
ved, and
the
atmosp
he
ric tran
smi
ttan
c
e
of wate
r vap
o
r
chan
nel
s i
s
approximatel
y
obtaine
d. Th
e PWV is de
rived from th
e atmo
sphe
ri
c tran
smittan
c
e ba
se
d on
a Loo
k up T
able
whi
c
h is
pre
-
cal
c
ul
ated u
s
ing a
radi
ation tran
sfer model.
The sen
s
itivities of
atmosp
he
ri
c
transmissio
n
in ea
ch
NI
R
water vap
o
r
cha
nnel
s
of
MERSI to th
e total p
e
rce
p
tible
water v
apor
are
also
simulated. The result indi
cates t
hat
FY-3A/M
E
RSI has and good
ability in detecting
NIR
water va
por,
and can d
e
m
onst
r
ate fin
e
cha
r
a
c
teri
st
ic of PWV spatial dist
ribu
tions
with 20
%
relative e
r
ror
to the
sou
ndi
ng. Th
ey hav
e a
s
sesse
d
t
he P
W
V L2
WVC a
c
cura
cy, the
retri
e
ved
WVC from
M
E
RSI NIR a
r
e comp
are
d
with the
g
r
ou
nd b
a
sed
so
undin
g
d
a
ta.
Over
clo
ud f
r
ee
area, the
r
e is
a good a
g
re
e
m
ent betwe
e
n
them in
vari
ation trend a
n
d
spatial di
stri
bution.
3. Results a
nd Discu
ssi
on
3.1. Numeric
a
l Tests o
f
the Algorith
m
3.1.1. Standa
rd atmosp
he
re Simulation Resul
t
s
In this secti
on, we
appl
y the SCWVD algo
rith
m to retri
e
ve LST to e
v
aluate its
perfo
rman
ce.
The
be
st way to validat
e an
algo
rith
m is to
com
pare
the i
n
-situ gro
und
truth
8
9
10
11
12
13
14
15
0
0.
1
0.
2
0.
3
0.
4
0.
5
0.
6
0.
7
0.
8
0.
9
1
W
a
v
e
l
e
ngt
h(
m)
R
e
l
a
ti
v
e
s
p
e
c
tr
a
l
r
e
s
p
o
n
s
e
(a
)
M
E
RS
I
ban
d
5
M
O
DI
S
ban
d 3
1
M
O
DI
S
ban
d32
0.
9
6
0.
9
6
5
0.
97
0.
97
5
0.
9
8
0.
9
8
5
0.
9
9
0.
995
1
0.
96
0.
9
6
5
0.
97
0.
9
7
5
0.
98
0.
9
8
5
0.
99
0.
9
9
5
1
A
v
e
r
ag
e em
i
s
s
i
v
i
t
y
of
M
O
D
I
S
ba
nd
31 a
nd
32
E
m
i
s
s
i
v
i
t
y
of
M
E
RS
I
ba
nd 5
y
=
0
.
79
1x
+
0
.
2
0
4
R=
0
.
9
1
6
(b
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Land Su
rface
Tem
perature
Retrie
val fro
m
the Medi
um
Resolution Spectral
Im
a
ger… (Hailei Liu)
7293
measurement
s of
LST
with
the
retri
e
ved
one.
Howev
e
r, thi
s
i
s
not
feasi
b
le, fo
r
it is
extremel
y
difficult to obtain the in-situ
groun
d truth
meas
urement
s whi
c
h m
u
st
be com
para
b
le to the pixel
size of MERSI data at the
satellite pass. A practi
cal way is to use the simul
a
ted
data generat
ed
by atmosph
e
ric simul
a
tion
prog
ram
s
such
as LO
WT
RAN, MODT
RAN or RTTO
V [31].
The sim
u
latio
n
with the mid-latitude
su
mme
r atmo
sphere wa
s carri
ed out to test our
algorith
m
. MODT
RAN 4.0
was
used in
the cal
c
ulat
i
on. Detaile
d results a
r
e li
sted in Ta
ble
3,
whe
n
the
WVC in mid
-
la
titude sum
m
e
r atmo
sp
here
is 2.9
2
g/cm2. The results indi
cate
the
algorith
m
is a
b
le to provide
a quite accu
rate es
tim
a
tio
n
of LST, with the differen
c
e bet
wee
n
the
assume
d LS
T and th
e retrieved le
ss tha
n
0.5 K in
mo
st case
s. It is
encouragin
g
that the result
is
good at several different e
m
issivity.
Table 3. Valid
ation of SCWVD Algorithm
for the Mid-la
titude Summer Atmosphe
re
Emi
s
s
i
v
i
ty R
Tb
Ts
_truth
Ts
_truth-
T
b
L
STSCW
V
D
T
s
_
truth-
LSTSC
W
VD
(Erro
r)
1.00 7.865503
288.4949
295.00
6.5051
294.5252
0.4748
0.98 7.771243
287.7112
295.00
7.2888
294.5644
0.4356
0.96 7.676987
286.9221
295.00
8.0779
294.5562
0.4438
0.94 7.582730
286.1276
295.00
8.8724
294.5204
0.4796
0.92 7.488467
285.3274
295.00
9.6726
294.3825
0.6175
3.1.2. NCEP/CDAS Re
ana
l
y
s
is Simulation Results
Global
-ba
s
e
d
simul
a
tion datasets as mentione
d in
se
ction
2.4
we
re
used
to test
SCWV
D algo
rithm. Figure 5(a
)
depi
cts t
he diffe
ren
c
e
betwee
n
LST retrieved u
s
ing the SCWV
D
method an
d LST get from global a
s
simil
a
tion data.
Figure 5
(
b
)
re
prese
n
ts the rel
a
tions b
e
twe
e
n
the erro
rs of
retrieve
d LST
usi
ng th
e S
C
WV
D m
e
th
od a
nd
WVC.
As
sho
w
n
in
Figu
re
5(b
)
,
the
absolute erro
rs of ret
r
ieve
d LST in glo
bal are
a
are mainly con
c
e
n
trated in the
rang
e of±1.5
K,
with a
RMSE
of 0.87K. We
also fo
und th
at the retri
e
val errors
are
mainly locate
d in the
ran
g
e
of
±1K, wh
en
WVC i
s
less than 1.5g/cm2. Ho
weve
r, Whe
n
WV
C is la
rge
r
than 3g/
cm2,
the
retrieval errors
c
a
n reac
h up to 2K.
Figure 5. (a)
Relatio
n
ship betwe
en the truth
LST (NCEP) and the retrieved LST
by SCWVD
method. (b
) Relation
ship b
e
twee
n LST error an
d WV
C
3.2. Tests
Us
ing In-situ M
easur
e
ments in Lake Ta
hoe and MO
DIS Product
The o
b
je
ctive of the prese
n
t wo
rk i
s
to
esti
mate th
e
LST from th
e
se
con
d
ge
ne
ration
of
China's polar-orbiting
met
eorol
ogi
cal satellite
FY
-3A MERSI observation
ov
er the
cloud-free
area.
The T
O
A brightness tem
peratures
are di
rectl
y
extracted f
r
om
the ME
RSI L1B
satellite
data. The L
SE map wa
s de
rived fro
m
the MO
DIS LSE prod
uct a
c
cordi
n
g to the met
hod
motioned a
b
o
v
e. The WVC are obtain
e
d
from MERSI L2 total perceptible water
prod
uct, whi
c
h
can be
accessed from
FE
NGYUN
S
a
tellite Data
Center (http://fy3.
satellite.cm
a.gov.cn/). FY
-3
A
200
22
0
240
260
280
300
320
20
0
22
0
24
0
26
0
28
0
30
0
32
0
Tr
ut
h L
S
T
LS
T
r
e
t
r
i
v
ed by S
B
W
V
D
m
e
t
h
o
d
(a
)
0
1
2
3
4
5
6
-2.
5
-2
-1.
5
-1
-0.
5
0
0.
5
1
1.
5
2
2.
5
W
a
te
r v
a
p
o
r c
o
n
t
e
n
t e
rro
r(g
/
c
m
2
)
L
S
T
e
rro
r
(K
)
(b
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 728
7
– 7298
7294
MERSI Level
2 wate
r vap
o
r pro
duct i
s
u
s
ed a
s
the in
p
u
t paramete
r
for the SCWV
D meth
od, an
d
a 2-D d
a
ta in
terpolatio
n proce
dure is a
pplied in
o
r
d
e
r to match t
he MERSI L1
B data in spa
t
ia
l
resolution.
In orde
r to validate our p
r
e
s
ente
d
algo
rithm, the Lake
Tahoe,
CA
/
N
V
,
US
A
,
is
sele
cte
d
as the
study
area.
NASA sci
entist
s
selected Lake
Tahoe
as
a validation sit
e
just before the
Terra
satellit
e was launched in 1
999 on a 15-year
missi
on to
st
udy Earth's environm
ent [32].
Equippe
d
with a
suite
of in
strum
ents tha
t
con
s
t
antly
monitor the l
a
ke
enviro
n
me
nt, the rafts a
n
d
buoys
provid
e inform
ation
that help
s
m
a
ke
su
re
th
at Earth-ob
serv
ing satellites
are
getting th
eir
temperature
measurement
s right.
Me
asurem
ents
at
t
he
site a
r
e
made f
r
om f
our
pe
rman
e
n
tly
moored bu
oys on the la
ke
, referred to
as TB1, TB
2,
TB3, and TB
4. Each bu
o
y
has a
cu
sto
m
-
built radiom
e
t
er that measures the
ski
n tem
peratu
r
e and seve
ral temperatu
r
e se
nsors that
measure the bulk
wat
e
r tempe
r
at
ure. The
a
u
tomated va
lidation site,
whe
r
e g
r
o
und
measurement
s of
la
ke
skin
tempe
r
atu
r
e
have
b
een
m
ade on a nea
r contin
uou
s basi
s
(eve
ry 2
min) si
nce 1
999 an
d use
d
to calibrate
and vali
date
TIR data an
d prod
uct
s
from airb
orn
e
and
satellite in
stru
ments, incl
udi
ng the ASTER, MO
DIS, Land
sat 5 TM, the Land
sat 7 TM and AT
SR
[32-37]. Figu
re 6 gives the
curre
n
t locati
on of the mea
s
ureme
n
t site
s on a map.
As SCWV
D a
l
gorithm
is de
veloped
for cl
oud
cle
a
r co
n
d
itions, clo
u
d
dete
c
tion sh
ould be
done
first. In
this
se
ction,
the SCWV
D meth
od i
s
validated
using ME
RSI scen
es a
c
qui
red
arou
nd
La
ke
Tahoe
in 2
0
0
9
. Twelve
M
E
RSI and M
O
DIS sce
n
e
s
we
re
sele
ct
ed from
June
to
Octob
e
r. Ta
b
l
e 4 give
s th
e Data
acqui
sition d
e
tails
of the vario
u
s
daytime M
E
RSI and M
O
DIS
scene
s.
T
he differen
c
e
of MERSI
an
d MODIS over
pass tim
e
was
within
30
minutes in
m
o
st
ca
se
s. Ta
kin
g
into a
c
cou
n
t the la
ke
surface i
s
rela
tively uniform
, LST ch
ang
es
cau
s
e
d
by
the
time differen
c
e is n
egligibl
e
. The MO
DIS LST pro
d
u
c
ts have be
en v
a
lidated
withi
n
1K in multip
le
validation
sites i
n
relative
ly wide
ra
ng
es
of
surfa
c
e an
d atmo
spheri
c
co
nditi
ons.
We
test
ed
SCWV
D algo
rithm g in more than 10 cle
a
r-sky ca
se
s according to
comp
are the MERSI retrie
ved
LST with in-si
t
u measu
r
em
ent data and
MODIS LST prod
uct.
Table 4. Dat
a
Acqui
sition
Detail
s of Various
Daytime Imagery of M
E
RSI and M
O
DIS
Date
Julian Da
y
MERSI overpass time
MODIS over
pass time
2009.6.9
160
18:30
19:00
2009.6.16
167
19:10
18:55
2009.6.18
169
18:30
19:00
2009.7.2
183
19:10
19:00
2009.7.4
185
18:30
18:55
2009.8.3
215
19:10
18:55
2009.8.5
217
18:30
18:55
2009.8.28
240
18:00
18:55
2009.9.13
256
18:30
19:00
2009.9.29
272
19:35
18:55
2009.10.9
282
18:10
19:30
2009.10.22
295
19:05
19:00
Figure 6.
Current Lo
cation
of the Measu
r
ement Sites o
n
the Map (from
http://lak
etahoe.jpl.nas
a
.gov)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Land Su
rface
Tem
perature
Retrie
val fro
m
the Medi
um
Resolution Spectral
Im
a
ger… (Hailei Liu)
7295
The differe
nce of MERSI and MO
DIS over
pa
ss time wa
s within
30 minute
s
in most
ca
se
s.
Ta
kin
g
into a
c
cou
n
t the la
ke
surface i
s
rela
tively uniform
, LST ch
ang
es
cau
s
e
d
by
the
time differen
c
e is n
egligibl
e
. The MO
DIS LST pro
d
u
c
ts have be
en v
a
lidated
withi
n
1K in multip
le
validation
sites i
n
relative
ly wide
ra
ng
es
of
surfa
c
e an
d atmo
spheri
c
co
nditi
ons.
We
test
ed
SCWV
D algo
rithm g in more than 10 cle
a
r-sky ca
se
s according to
comp
are the MERSI retrie
ved
LST with in-si
t
u measu
r
em
ent data and
MODIS LST prod
uct.
(a)
(b)
Figure 7. (a)
Retrieve
d LS
T using S
C
WVD algo
ri
thm from MERSI L1B data ove
r
Lake Tah
o
e
at
03h UT
C, 3 Aug. 2009;
(b
)
MODIS LST prod
uct ove
r
Lake Taho
e a
t
same time
Figure 7
(
a
)
d
epict
s the
surface te
mpe
r
at
ure
dist
ributio
n retrieved
by
SCWVD
met
hod fo
r
MERSI scen
e
acq
u
ire
d
around 0
3h UT
C on 3 Aug
u
s
t 2009. Fig
u
r
e 7(b) give t
he MO
DIS LST
prod
uct
(p
ro
vided by
NA
SA) aroun
d
Lake T
ahoe.
Obviou
sly, t
he
su
rface t
e
mpe
r
atures in
Qingh
ai La
ke ra
nge i
n
1
6
~1
8
℃
a
nd
the tempe
r
at
ure
distri
buti
on is quite
uniform
with
the
averag
e valu
e of
16.7K
whi
c
h i
s
nea
r
the
value
measured by
the buoy (1
7.5
◦
C).
The
LST
arou
nd
the
L
a
ke
Ta
hoe
is obviou
s
ly hi
gher tha
n
wa
ter fa
ce
with t
he ave
r
ag
e v
a
lue
abo
ut 3
0
℃
.
Takin
g
into
a
c
count th
e spatial-re
soluti
on
differen
c
e
betwe
en
M
E
RSI and
M
O
DIS, match
-
up
wa
s g
ene
rat
ed e
m
ployin
g the
2-D i
n
terpol
ation.
Fi
gure
8
(
a
)
giv
e
s th
e
error
distrib
u
tion
map
betwe
en ME
RSI LST wit
h
MODIS L
S
T prod
uct.
For the
wa
ter su
rfa
c
e t
e
mpe
r
ature, the
maximum diff
eren
ce
is
abo
ut 1.2 K, most of t
he differences
are around
0.
5 K, a
nd the
RMSE
is
0.35 K. The
l
a
rge
s
t temp
e
r
ature differe
nce
wa
s
obt
a
i
ned o
u
tsid
e l
a
ke
with
an e
rro
r of
4.7 K. We
think the
mai
n
rea
s
o
n
may
be the effe
ct of spatial
-
resolution differe
nce
and th
e i
m
pleme
n
t of the
interpol
ation. Figure
8(b) gi
ve
s the
scattering
plot of the re
t
r
ieved MERSI
LST and
M
O
DIS LST
prod
uct with an RMSE
of 2.3K.
The re
sults
in
dicate
t
hat mo
st of th
e ret
r
ieved
L
S
Ts from ME
RSI
data a
r
e
a litt
l
e hig
her tha
n
that from M
O
DIS LS
T
p
r
odu
cts
aro
u
n
d
the l
a
ke. It
is
wort
h n
o
ticing
that MERSI
LST ha
s g
r
e
a
t impr
ovem
ent than
MO
DIS in the
spatial resoluti
on, for
exam
ple,
MERSI can
e
a
sily g
e
t the t
e
mpe
r
ature
chang
es in
formation of
the
sm
all
water
body n
earby
the
Lake Taho
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 10, Octobe
r 2014: 728
7
– 7298
7296
Figure 8. (a)
Difference between ME
RSI LST with
MODIS LST product (p
rovided by NASA); (b)
Comp
ari
s
o
n
betwe
en the
derived ME
RSI LST and MODIS LST
The LSTs re
trieved by the SCWV
D algorit
hm have
also bee
n compa
r
ed wit
h
in-situ
measurement
s in
La
ke
Ta
hoe. Fig
u
re
9
give
the
differen
c
e
bet
we
en ME
RSI, MODIS an
d b
u
o
y
measurement
wate
r surfa
c
e skin
tem
peratu
r
e
ove
r
the
12 d
a
y
s (4
buoy
s mea
s
ureme
n
ts
averag
e pe
r
day). The
re
sults indi
cate t
hat the
accu
racy of the ret
r
ieved ME
RSI LSTs is le
ss
than 1.5K. The MO
DIS LST accuracy i
s
better tha
n
MERSI’s.
Figure 9.
Co
mpari
s
o
n
of the LST Error
betwe
en t
he
MERSI, MODIS LSTs and In-situ M
e
a
s
ured
LSTs in La
ke
Tahoe
4. Conclusio
n
As a new generation of polar
orbiting meteorologi
cal
satellite,
FY-3 series consi
s
ts of
two experim
e
n
tal and at least four op
erational sate
llit
es, whi
c
h is e
x
pected to ha
ve a service life
until 2020. L
aun
che
d
re
spectively on
27 May 200
8
and 5
Nove
mber 2
011, F
Y
-3A and FY
-3B
are
de
sign
ed
with the
sam
e
a
ssi
gnme
n
ts a
nd
equi
pp
ed
with 1
1
p
a
y
loads.
The
o
n
ly differen
c
e
is
that FY-3A is a mornin
g-o
b
se
rvation sa
tellite
and FY-3B is an aft
e
rno
o
n
-
ob
se
rvation satellit
e
.
These two sa
tellites can p
r
ovide glo
bal
observati
on
o
f
the Earth L
and
su
rface tempe
r
ature
with
high spatial resol
u
tion (2
5
0
m) fou
r
times pe
r day, which i
s
a gre
a
t improveme
n
t comp
ari
s
o
n
to
the current LST product
s of other sate
llite in the temporal and spatial.
Based
on the
uppe
r ba
ckg
r
oun
d, we
de
veloped a S
C
WV
D alg
o
ri
thm for LST
retrieval
from MERSI TIR data. The derivation o
f
this algor
ith
m
is based o
n
the thermal
radian
ce tra
n
sfer
equatio
n and
the linearization of Planck'
s radia
n
ce
function. T
o
tally there are two criti
c
al
para
m
eters i
n
the
algo
rith
m: emissivity and
WVC.
On givin
g
th
ose
two
pa
ra
meters, it
will
be
very ea
sy to use thi
s
alg
o
rithm fo
r LS
T estimatio
n
not only from
MERSI data
but also fro
m
285
290
295
300
305
31
0
315
32
0
325
285
290
295
300
305
310
315
320
325
M
O
DI
S
LS
T
P
r
o
duc
t
s
(
N
A
S
A
)
MER
S
I
L
S
T
RM
S
E
=
2
.
3
K
1
2
3
4
5
6
7
8
9
10
11
12
-1
.
5
-1
-0
.
5
0
0.
5
1
1.
5
Da
y
LS
T
e
r
r
o
r
(
K
)
ME
R
S
I
MO
D
I
S
Evaluation Warning : The document was created with Spire.PDF for Python.