and pervious surface from the summer image result had high
PAs
and
UAs
(greater than 87 percent). For the winter image
result, the
PA
and
UA
of impervious surface were very high
(95.24 percent and 98.13 percent, respectively), whereas the
UA
of pervious surface was relatively low (64 percent). This
is possibly caused by the impervious surface accounting for a
major proportion of the study area. Therefore a small misclas-
sification of a section (omission) of impervious surface would
lead to a large disagreement proportion (commission) of
pervious surface.
The quantity and allocation disagreement of the summer
image classification were 1.59 percent and 3.73 percent, and
2.67 percent and 3.29 percent for the winter image, respec-
tively. For both seasons, the quantity and allocation disagree-
ments were less than 4 percent and the allocation disagree-
ment was slightly higher than the quantity disagreement.
The quantity and allocation disagreements indicate that the
classifications correctly assigned the total number of pixels
to each cover class (low quantity disagreement), and also al-
located them in the right location on the map (low allocation
disagreement). Thus, the estimated proportion and allocation
of impervious surface on the classification map from each im-
age was close to those derived from the reference map.
The final impervious surface extraction results from the two
season images for the Tianjin area are shown in Figure 4. In
general, the impervious surface area is greater than that of the
pervious surface from the summer image result, and, in par-
ticular, in the result from the winter image. By comparing the
results from summer and winter images, the area of impervious
surface extracted from the winter image is more than that from
the summer image (Figure 4). It is also very clear that most
streets and sidewalks obscured by deciduous tree canopies
in the summer image were correctly identified as impervious
surface in the winter image, e.g., the street shown in the grey
rectangle (Figure 4B). In some residential areas the proportion
of pervious surface in the winter image was also less than that
in the summer image, as rooftops and pathways (impervious
surface) obscured by tree canopies were exposed and identified
as impervious surface (e.g., grey ellipse in Figure 4).
Area proportions of the impervious surface class from the
summer and winter image results were also estimated using
the method proposed by Stehman (2014), taking into account
the classification error. The results showed that the area of the
impervious surface from the summer image result accounted
of 72.64 percent of the entire image (with a standard error of
0.71 percent, and 71.26 ~ 74.02 percent at the 95 percent con-
fidence interval), whereas the estimated area proportion of the
impervious surface from the winter image result was 90.68
percent (with a standard error of 0.63 percent, and 89.45 ~
91.91 percent at 95 percent confidence interval). Thus, the
area of impervious surface extracted from the winter image
was greater than that from the summer image by 18.04 per-
cent. Furthermore, an analysis of the proportions of impervi-
ous surface in the shaded area indicated that the impervious
surface extracted from the shaded area (16.19 percent in total)
in the summer image accounted for 12.67 percent of the study
area, while the impervious surface extracted from the shaded
area (31.14 percent in total) in the winter image accounted for
24.82 percent of the study area. The area of impervious sur-
face in shaded area from the winter image was considerably
greater than that from the summer image.
As with the example from the Beijing area, by compar-
ing impervious surface extraction results from summer and
winter images for Tianjin (as difference map shown in Plate
6), we found that differences in area in impervious surface ex-
traction between summer and winter seasons made up 20.15
percent of the entire image. The major difference (16.97 per-
cent of the entire image area, shown as red in Plate 6) is in the
areas classified as pervious surface from the summer image
and as impervious surface from the winter image. This type of
difference is mainly distributed along roads, sidewalks, and
around buildings (Plate 6). The difference is mainly caused by
seasonal variations in deciduous trees, where the areas were
obscured by tree canopies in the summer image and exposed
as impervious surface in the winter image. The minor differ-
ence is in the areas classified as impervious surface from the
summer image and as pervious surface in the winter image
(accounting for 3.18 percent of the image) (shown as blue in
Plate 6). This type of difference is mainly caused by varia-
tions in the view angle and the illumination between the
Figure 4. The impervious surface extraction results from (A) the summer image, and (B) from the winter image. 1, impervious surface;
2, pervious surface. The grey rectangle shows some road areas and the grey ellipse shows some rooftop and pathway areas which were
obscured by deciduous trees in summer and classified as pervious surface using summer image while extracted as impervious surface
using winter image.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2016
345
