2004 
  
  
  
  
  
  
  
55] 
lines 
es are 
bands 
  
ferent 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
3.2 Thematic Accuracy 
For the year 2000, 12816 polygons have been created. That 
causes the thematic accuracy investigation not applicable for a 
full investigation. Because of that, sampling groups have been 
created randomly from database for all the landuse types used 
in database. The rule for sampling is illustrated in Table 4. 
  
  
  
  
  
  
  
Number of Sampling size 
Polygon 
30 or more 30 
Between 20 — 29 20 
Between 10 - 19 10 
Between 1 — 9 All 
  
  
Table 4. Sampling rule for thematic accuracy investigation 
Applying the rule to the database for the year 2000, the 
sampling sizes has been calculated as in Table 5. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Landuse |Sampling| | Landuse |Sampling| | Landuse Sampling 
Type Size Type Size Type Size 
2.1.1.1 30 1.2.2.6 30 2.4.4 7 
1.1.1.2 30 12.27 20 311 30 
1.4.1.9 30 1.2.3 30 3.1.2 30 
1.1.21 30 1.2.4 4 3.13 30 
11.22 30 1.31 30 3.1.3.1 
1.1.2.3 30 132 1 3.2.1 2 
1.1.2.4 3 1.3.3 30 32.11 i 30 
1.21.4 30 1.3.4 30 3.2 1.2 30 
1.21.2 30 1.4.1 30 32.13 10 
1.21.3 30 1.4.1.1 30 3.2.4 10 
1.2.1.4 20 1.4.2 30 3.2.4.4 2 
1215 10 24.1 30 3.3.1.1 
1.2.1.6 20 2.1.1.1 30 3.3.1.2 10 
12.1.7 10 21.12 5 3.322 30 
1.2.1.8 10 2.1.1.3 3 3.3.3.1 2 
1.2.1.9 30 2.1.2 30 4.1.2 
1.2.1.10 10 2:3.1.2 4 5814.1 
122.1 30 2.4.2.1 4 5112 8 
1222 30 2.4.2.2 30 5.121 10 
1223 10 2.4.3 4 5.1.2.2 20 
1224 1 2.4.3.1 30 52.3 5 
1225 1 2.4.3.2 2 Total 1168 
  
  
  
  
  
  
  
  
  
Table 5 Sampling sizes for landuse types 
After calculating the sampling sizes, samples have generated 
using the random property of computer. The sample polygons 
are identified with their unique number in database. 
For all the polygons in the sampling group, a thematic accuracy 
process has been done manually. This was a very long and 
Monotonous work but in order to make an assessment about the 
thematic accuracy, it was a must. All the polygons have been 
controlled manually if they defined correctly by digitalisation 
process. As a result, some defining errors about the landuse 
have been detected but the amount of the errors was acceptable. 
Errors are given in Table 6 with their percentage to the 
sampling groups. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Landuse | Error Landuse | Error Landuse | Error 
Type Yo Type Yo Type Yo 
1.1.1.1 6,7 1.2.2.6 33 2.4.4 0,0 
1.1.1.2 6.7 1.227 5,0 3 1.1 0,0 
1.1.4.3 6,7 1.2.9 0,0 3.1.2 0,0 
1.1.2.1 1.10.0 1.2.4 0,0 3.1.3 0,0 
1.1.22 | 10.0 13.1 10,0 3.1.3.1 0,0 
1.1.2.3 6.7 1.3.2 0,0 32.1 0,0 
1.1.2.4 0,0 1.3.3 0,0 3.2.1.1 6,7 
1.2.4.1 0,0 1.3.4 10,0 32124 67 
12.12] 133 1.4.1 6,7 3.2.1.3 0,0 
1.2.4.9 6,7 1.4.1.1 0,0 3.2.4 0,0 
1.2.1.4 0,0 1.4.2 0,0 3.2.4.4 0,0 
1.245 0,0 21.1 6.7 3.3.1.1 0,0 
1216 0,0 2.1.1.1 0,0 3.3.1.2 0,0 
1.2.4.7 0,0 24.1.2 0.0 3.2.2.2 0,0 
1.2.1.8 0,0 2.1.1.3 0,0 3.3.3.1 0,0 
1 2 1.9 0,0 2.1.2 0,0 4.1.2 0,0 
12.31.1041. 00 2.3.1.2 0,0 5.1.1.1 0,0 
1.2.2.1 0,0 2.4.2.1 0,0 5.1.1.2 0,0 
2222 0,0 2422 97 54:21 0,0 
1.2.2.3 0,0 2.4.3 0,0 5.1.2.2 0,0 
1.2.2.4 0,0 2.4.3.1 0,0 523 0,0 
1.2.2.5 0,0 2.4.3.2 0,0 Total 3.3 
  
  
  
  
  
  
  
  
  
Table 6 Error percentages of landuse classes 
Table 6 gives information about the accuracy of the results. For 
example, it can be said that the landuse type 1.1.1.1 
(Residential continuous dense urban fabric) has been defined 
with an error of 6.7% where the landuse type 3.1.3 (Mixed 
forest) has been defined without error. 
In Table 6, 48 landuse types have been shown as “totally true”. 
30 from these 48 landuse types have 10 or less sampling size 
i.e. they are few in total amount. This causes to define them 
much more correct than other landuse types. Other 18 landuse 
types, which are defined as “totally true”, have whether specific 
information about their landuse on 1:25000 maps (e.g. 
vegetated cemeteries) or they are clearly visible and detectable 
from aerial or satellite imagery (e.g. industrial areas or Scas). 
Using the additional data such as 1:25000 maps or 1:16000 city 
plans by digitalisation and defining the areas causes accurate 
results. Another parameter, which causes the results to be 
accurate, is the knowledge of operator on Istanbul. Defining the 
fast transit road and associated land can be a suitable example 
for this parameter. 
In Table 6, it can be seen that a total error percentage is given 
as 703.3. This value is only a rough result for total accuracy. In 
order to give more precise result about accuracy, another 
investigation must be done with the values in Table 6. In this 
investigation, error percentages for individual landuse types 
must be taken with their weight. The weight is selected as the 
percentage of number of landuse type to the total number of 
defined polygons. 
In order to derive a total accuracy result, Bayes theorem must 
be used for the results. 
1029