Geoscience Reference
In-Depth Information
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
14
35
10
6
16
6
9
14
30
12
5
5
8
12
25
10
7
4
4
10
6
20
8
5
3
3
8
6
15
4
6
2
2
3
10
4
4
2
1
1
5
2
2
1
0
0
0
0
0
0
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
Patch Density
Largest Patch Index
Mean Patch Size
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
60
12
10
2
120
30
9
1.8
105
50
10
25
8
1.6
90
7
1.4
40
8
20
75
6
1.2
30
6
5
1
60
15
4
0.8
45
10
20
4
3
0.6
30
2
0.4
5
10
2
15
0.2
1
0
0
0
0
0
0
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
Edge Density
AWM Shape Index
Nearest Neighbor Distance
Mean
Std. Dev.
Mean
Std. Dev.
Mean
Std. Dev.
1
0.16
0.6
0.12
70
7
0.9
0.14
60
6
0.1
0.5
0.8
0.12
50
5
0.7
0.08
0.4
0.1
0.6
40
4
0.5
0.08
0.3
0.06
30
3
0.4
0.06
0.2
0.04
0.3
20
2
0.04
0.2
0.1
0.02
10
1
0.02
0.1
0
0
0
0
0
0
1
2
3
12 3
1
2
3
123
1
2
3
1
2
3
Shannon's Diversity Index
Simpson's Diversity Index
Contagion Index
Figure 15.1
The mean and standard deviations for nine selected landscape indices for three accuracy groups;
1 = lowest accuracy, 2 = intermediate accuracy, 3 = highest accuracy.
= 0.98), but the errors
of all other indices did not show a simple relationship with REA (Figure 15.5). The REA seemed
to have a better control over landscape indices errors than did overall accuracy; the variations of
landscape index errors corresponding to REA were smaller than those corresponding to overall
accuracy (Figure 15.4 and Figure 15.5). Also, the lowest errors of landscape indices normally
occurred when REA reached zero (Figure 15.5). Both overall accuracy and REA were not reliable
indicators for explaining variations of spatially sophisticated landscape indices, such as MNN
and MPI.
The relative errors of %LAND for the forest from the 20 maps ranged from 12 to 25% before
calibration (Figure 15.6a). Based on Equation 15.11, the values of %LAND for the forest were
calibrated and resulting errors of %LAND for the forest were between 2 and 5% (Figure 15.6b),
much lower than the errors before calibration.
The errors of %LAND have a perfect linear relationship with REA (R
2
15.4 DISCUSSION
Methods used for image classification determine thematic maps' classification content and
quality. Although different statistics are used for assessing the accuracy of image data classifications,
most are derived directly or indirectly from error matrices. Indices of thematic map accuracy indicate
how well image data are classified but do not tell how thematic maps correspond to a landscape's
structure and function. This is partly because there is no effective approach to quantify classification
