relationship between water vapor and atmospheric transmit-
tance. In contrast, the polynomial fitting can better describe
this relationship.
Comparison
The retrieved
IST
obtained by the proposed method was
compared with the
MOD
29 product and the
AWS
data. The first
comparison was conducted based on the data from Zhong-
shan Station (Figure 6). This figure presents the results of the
retrieved
IST
, the
MOD
29 product, and the
AWS
data
.
Referring to the
AWS
data, the bias and root-mean-square
error (
RMSE
) for the retrieved
IST
and the
MOD
29 product are
provided in Table 2
.
The retrieved
IST
exhibits a much higher accuracy than the
MOD
29 product. In addition, the differences between the air
temperature and the retrieved
IST
, and those between the air
temperature and the
MOD
29 product, are presented in Figure 7
.
In Figure 7, the difference (absolute value) values of eight
sets of samples are much higher than those of the others (Ta-
ble 3). After visual interpretation by employing
MOD
021KM
data, it was found that two of these scenes were clear and six
scenes were affected by suspected fog or cloud. Taking these
six scenes out of the calculation, the revised bias and
RMSE
are
reported in Table 4. The accurate results imply that the preci-
sion of the results is related to the precision of the
MOD
35
product. However, it is difficult to determine the accuracy of
the
MOD
35 product due to the similar spectral characteristics
between cloud and snow/ice
.
Hall
et al.
(2004) found that the differences were slightly
greater when wind speeds were lower, and were less when
Figure 6. A comparison of satellite-derived ISTs (retrieved IST and MOD29 product) and
in situ
AWS ISTs from Zhongshan Station. From Decem-
ber 2004 to December 2013, 33 cloud-free scenes with a corresponding wind speed of greater than 4 ms
-1
were chosen for the comparison.
Figure 7. The differences between the air temperature and the retrieved IST, and those between the air temperature and the MOD29
product, for Zhongshan Station.
T
able
2. T
he
A
ccuracy
of
the
R
etrieved
IST
and
the
MOD29 P
roduct
for
Z
hongshan
S
tation
Data
Accuracy
Bias (K)
RMSE (K)
Retrieved IST
−0.62
1.32
MOD29
−1.34
1.81
T
able
3. T
he
E
ight
S
cenes with
the
G
reatest
D
ifference
(A
bsolute
V
alue
)
V
alues
. I
t was
E
valuated
through
V
isual
I
nterpretation
W
hether
E
ach
S
cene
was
I
mpacted
by
F
og
/C
loud
Number
The differences
between the
air temperature and
the retrieved IST (K)
The differences
between the
air temperature and
the MOD29 (K)
Comments
2004-359
−1.76
−2.87
Fog/cloud
2004-363
−2.48
−2.82
Fog/cloud
2005-353
−3.02
−3.58
Fog/cloud
2008-360
−1.98
−2.87
Fog/cloud
2009-365
−1.74
−2.02
Clear
2010-352
−2.59
−2.56
Fog/cloud
2011-352
−2.18
−2.67
Fog/cloud
2012-365
−1.61
−2.2
Clear
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
November 2015
867
