INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT 3. Results 3.1. The differences of results from different input maps The study began with processing the different input maps. There were 6 input maps from 3 different years. After overlaying (intersection) for each year, the first map in 1985 (85a), 1990 and the first map in 2000 (00a), result 1 (R1) was obtained meaning forest cover 85a-90-00a. The same process was run for other sets until getting R6. Maps and tables that follow show the results coded as R, forest areas coded as F, and non-forest areas coded as NF. Table 3: Results of different combination of forest cover in 3 years Name of result Set Number of polygons R1 Forest 85a – 90 – 00a 339 R2 Forest 85b – 90 – 00a 362 R3 Forest 85a – 90 – 00b 464 R4 Forest 85b – 90 – 00b 504 R5 Forest 85a – 90 – 00c 355 R6 Forest 85b – 90 – 00c 381 The differences of preview for all results are depicted in figure 10 below. ( Result 1 – R1) (Result 2 – R2 ) 31 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT (Result 3 – R3) (Result 4 – R4) (Result 5 - R5) (Result 6 - R6) Figure 10:Six results of overlaying forest cover maps in 3 different years A simple visual impression would show that R1, R3 and R5 almost have the same preview, even though they have different number of polygons. The same is true for R2, R4, and R6 which have nearly the same preview. When it is seen clearly, R2 and R4 relatively have the same preview but actually they have different number of polygons, R4 having 504 while R2 has only 362 polygons. The more polygons mean the more information in the attributes even though they have the same boundary. R1 and R3 also relatively have the same preview, even though they have different number of polygons in that, R3 with 464 which is more than 339 polygons for R1. The extent the differences among the results maps cannot be differentiated only from the map display. It is needed to assess the differences among the maps using Kappa Index of 32 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT Agreement to find out the extent of agreement and disagreement between 2 maps of results on forest cover. 3.2. Kappa Index of Agreement With 6 results of forest cover combinations in 3 different years, assessing the differences between 2 maps produced 15 combination of Kappa Index of Agreement. Some combination of calculations can be seen in the tables and maps below. Other combinations are presented in Appendix 12. Kappa index of agreement for this research consider forest and non forest area to examine from 2 maps having different interpretation in which areas are called forest but non forest in the other and vice versa. Table 4:Calculation of Kappa index of agreement for R1 and R2 R1 F (ha) % NF (ha) % Total (ha) % F (ha) 1,986,876.8 62.2 0.0 0.0 1,986,876.8 62.2 R2 NF (ha) 201,340.1 6.3 1,003,735.4 31.4 1,205,075.5 37.8 Total (ha) 2,188,217.0 68.6 1,003,735.431.4 3,191,952.4 100.0 R1 – R2 Observed Chance Kappa 0.937 0.545 0.861 Table 4 depicts the agreement between R1 and R2 in that the agreement of forest area from both maps is 62.2% (dark green on the map-figure 11) and the agreement of non forest area is 31.4 % (pink area). The rest is disagreement (error1) that is classifying forest for R1 but non forest for R2 (6.3%, light green), and 0% for error2 (non forest for R1 but forest for R2). The kappa index/coefficient is 0.861 which is interpreted as very good agreement (Cohen’s Kappa interpretation). 33 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT Figure 11:Agreement of R1 – R2 Figure 12:Agreement of R1 – R3 Table 5 shows the agreement between R1 and R3, forest area has 68.5% of agreement (dark green on the map-figure 12) and the agreement of non forest area is 31.4 % (pink area). The rest is disagreement (error1 and error 1, which are 0.00%) or in other words there is no error/disagreement between the two maps. The kappa index/coefficient is exactly equal to 1.000 meaning complete agreement for both maps. Table 5:Calculation of Kappa index of agreement for R1 and R3 R1 F (ha) % NF (ha) % Total (ha) % F (ha) 2,187,702.3 68.5 0.0 0.0 2,187,702.3 68.5 R3 NF (ha) 514.7 0.0 1,003,735.4 31.4 1,004,250.1 31.5 Total (ha) 2,188,217.0 68.6 1,003,735.4 31.4 3,191,952.4 100.0 R1 – R3 Observed Chance Kappa 1.000 0.569 1.000 The complete kappa index of agreement between 2 maps of results for all combination is showed in table 6. In general, all kappa indices for 15 combinations of agreement between 2 maps revealed that even though they have different results from different input maps, the differences of the results is relatively low. Two combinations of results show complete agreement between each other (kappa index = 1), and the rest combinations of results showed by kappa index are classified into very good agreement (kappa index more than 0.800) meaning that they have relatively no big differences among input maps. 34 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT Table 6: Recapitulation of Kappa Index of Agreement for all combination of results Combination Kappa Index Interpretation Combination Kappa Index Interpretation R1-R2 0.861 Very good agreement R2-R5 0.812 Very good agreement R1-R3 1.000 Complete agreement R2-R6 0.949 Very good agreement R1-R4 0.861 Very good agreement R3-R5 0.941 Very good agreement R1-R5 0.941 Very good agreement R3-R6 0.809 Very good agreement R1-R6 0.808 Very good agreement R4-R5 0.812 Very good agreement R2-R3 0.862 Very good agreement R5-R6 0.864 Very good agreement R2-R4 1.000 Complete agreement 3.3. Change detection of forest cover 3.3.1. Definite Forest Definite forest generated from intersection of all Results (R1 until R6) means that in every set (6 set of changes from 1985, 1990, and 2000) there are always forest areas no change. During 15 years (1985 until 2000), even though all sets have different results of forest cover (R1-R6), after overlaying of all sets it was shown that there is a forest area that definitely has never changed. It is clearly seen on the map-figure 13 with dark green area and F-Ri in appendix 2. 35 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT Figure 13: Definite Forest/Non Forest, probability, and changes 3.3.2. Forest probability The probability of forest not changing or being changed 1985, 1990, and 2000 could be categorized into 6 probability classes: (a) 6/6: definite forest, (b) 5/6: probably still forest, (c) 4/6: probably/possibly still forest, (d) 3/6: uncertain whether still forest, (e) 2/6: probably still forest, (f) 1/6: probably/possibly still forest. The probability of forest no change was calculated from the remaining area (Rm) of each result (F R1 – F R6) compared to the common definite forest, as follows: x4/6 probably no change or still forest (purple color in figure 13) - (Rm F R1_4 meaning the remaining forest from Result 1, 2, 3, and 4). x3/6 uncertain whether still forest (pink color) - (Rm F R1, 3, 5). x2/6 probably still forest (light green) - (Rm F R5, 6). x2/6 probably still forest (red) - (Rm F R1, 3). x1/6 probably still forest (dark blue) - (Rm F R1). There are 5 different probabilities of forest, showing the chance/probability of those areas whether they were still forest during 3 years (1985, 1990, and 2000) or not. 36 INFLUENCE OF FOREST COVER LOSS ON ORANG UTANS HABITAT 3.3.3. Definite Non forest Non forest areas were derived from the remaining area of forest cover and admin boundary for each input map (F 1985a, F 1985b, F 1990, F 2000a, F 2000b, and F 2000c). Six sets of overlaying 3 different years from 6 different maps of non forest area were run to obtain non forest for 6 results (NF R1 until NF R6). The common area of non forest derived from intersection of 6 results is the area of non forest that did not changed from 1985 to 2000. This area is presented on figure 13 with yellow area and symbol NF_Ri in appendix 4. 3.3.4. Non forest probability The probability of non forest area was obtained from the remaining area of each result (NF R1 until NF R6) compared to the common definite non forest (NF Ri), namely: x4/6 probably no change or still non forest during 1985, 1990, 2000 (light blue color) - (Rm NF R1_4 meaning remaining non forest area from Result 1,2,3,4) x2/6 probably still non forest (magenta) - (Rm NF R4,6). x1/6 probably still non forest (dark purple) - (Rm NF R6).