the dual-window algorithm, as compared to 1.83 K from the split-window
algorithm.
However, in general, all of accuracy comparisons show that using the dualwindow to derive LST is fairly feasible, though its total accuracy is slightly worse
than those from the split-window algorithm. These results show that the lack of
split-window channels on the GOES M-Q series may degrade the performance of
GOES-LST retrieval.
45
40
35
30
25
20
15
10
5
0
-5
-5 0
5 10 15 20 25
LST InSltu (Unit:°C)
LST InSltu (Unit:°C)
30 35
1:1 Line
1:1 Line
Linear Fit Line
Linear Fit Line
40 45
-10
-5 0 5 10 15 20 25 30 35 40 45
-10
-10
45
40
35
30
25
20
15
10
5
0
-5
-10
GOES-W 3.7 um + 11 um LST (Unit:°C)
GOES-W 11 um + 12 um LST (Unit:°C)
Corr.=0.985
Corr.=0.972
Bias=-0.630
Bias=-0.862
STD=1.839
STD=2.536
RMS=2.282
RMS=3.049
b
a
Fig. 19.18 Scatter plots of LST derived from the GOES vs. in situ observations for (a) splitwindow algorithm and (b) dual-window algorithm
FPK
TBL
ARM
SGP
DRA
-125
25
30
35
40
Latitude
45
50
-120 -115 -110 -105 -100
-95
-90
-85
-80
-75
-70
BON
PSU
GWN
Longitude
Fig. 19.19 Distribution of the SURFRAD and ARM sites
326
D. Sun and Y. Yu