Model II Regression Mosaic
The model II regression procedure creates cloud-free, seam-
less, mosaics. The quality of the images selected for the model
II regression procedure is vital to the quality of the mosaics
produced. The images selected to create the western and east-
ern mosaics came from late summer and early fall to avoid
issues with clouds and shadows (Table 5).
Figure 4 shows the frequency of
NDVI
values for zones 16,
23, 48, 54, and 59 generated from the model II regression
mosaic (solid line), median composite (dashed line), and the
maximum
NDVI
composite (dotted line). Except for zone 23,
the
NDVI
values for the model II regression mosaic were lower
than the
NDVI
values for the other image composites.
Plates 1 and 2 shows examples of maximum
NDVI
com-
posites (a) and (b), median composites (c) and (d), and model
II regression mosaics (e) and (f) for path/row 38/34 (Plate 1)
and path/row 20/36 (Plate 2). The maximum
NDVI
composite
is very pixelated while the median composite and model II
regression mosaic look like a normal Landsat scene. The scat-
tering of pixels observed in the maximum
NDVI
composites is
caused by the interspersion of various dates of imagery. Also
(a)
(b)
(c)
(d)
Figure 3. Linear regressions for median NDMI (a), NDVI (b), and
Tassel Cap (c-e) values calculated from 15 Landsat 5 scenes and
NDMI, NDVI, and Tassel Cap values derived from a Landsat 5 me-
dian composite image for WRS-2 path/row 36/33.
(e)
T
able
4. P
ercent
of
P
ixels
U
sed
for
E
ach
D
ate
in
the
M
aximum
NDVI
C
omposite
for
WRS-2 P
aths
/R
ows
36/33
and
20/36
Path 36 Row 33
Path 20 Row 36
Acquisition
Date
Percent of
Total Pixels
Acquisition
Date
Percent of
Total Pixels
18-Jul-09
36% 31-May-09
39%
06-Oct-09
22% 25-Aug-11
12%
12-Oct-11
21% 04-Oct-08
11%
06-Aug-10
7% 06-Jun-11
10%
03-Aug-09
5% 16-Jun-09
8%
10-Sep-11
3% 04-Sep-09
4%
22-Jun-11
2% 03-Jun-10
4%
25-Aug-11
1%
18-Jul-09
4%
07-Sep-10
1% 23-Sep-10
3%
05-Jul-10
0% 03-Aug-09
3%
24-Jul-11
0% 09-Oct-10
2%
09-Aug-11
0%
05-Jul-10
1%
19-Jun-10
0%
21-Jul-10
1%
06-Jun-11
0% 10-Sep-11
1%
19-Aug-09
0%
02-Jul-09
0%
204
March 2016
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
