)
CLASS
No. of pixels
in class
Sample size
No. of sample
correctly
classified
% accuracy
for each
class
95%> confidence
limits by class
(per cent)
Weighting
Water
192
60
60
100.0
94 - 100
0.0059
Shallow Water
31
31
31
100.0
90 - 100
0.0009
Woodland
2067
120
106
88.3
82.4 - 94.3
0.0631
Heathland
5430
60
50
83.3
73.1 - 93.6
O.1656
Upland Scrub
10621
60
56
93.3
86.2 - 100.0
0.3239
Grassland
11295
60
52
86.7
77.2 - 96.1
0.3445
Cropland
2847
60
51
85.0
75.1 - 94.9
0.0868
Urban
118
60
29
48.3
34.8 - 6l.8
0.0036
Fire Damaged/
Clear Fell
Areas
90
60
58
96.7
91.3 - 100.0
0.0027
Direct Shadow
26
26
11
42.3
21.4 - 63.2
0.0008
TOTALS
32727
+ 59_
unclassi
fied
597
0.9979
+ 0.0021
unclassi
fied
= 32786 = 1.000
Overall map accuracy is the sum of the individual class accuracies, each being weighted
according to the area of the map covered.
Accuracy value = 87.5% and 95% confidence limits for this figure = 83.5 - 91.5%
(See Snedecor and Cochran, 1977, p.527).
TABLE 2
RESULTS FROM SAMPLING GRANTOWN-ON-SPEY BLOCK FOR ACCURACY TESTING (BLOCK SIZE 194 x l69 = 32786 pixels)