4766 =
Thailand/Miller, Nualchawee and Tom
(a) topographic aspect (b) M.D. to roads and trails
FIGURE 7.7. REPRESENTATIVE GRAYMAPS OF THE PHYSIOGRAPHIC (a) AND CULTURAL FEA-
TURES (b) DATA PLANES OVERLAYING THE LANDSAT MSS BANDS. M.D. = minimum distance.
The topographic aspect data plane was computed from the topographic elevation data plane at the 1 hec-
tare resolution. The minimum distance data plane was computed from an airphoto interpretation of
cultural features.
TABLE 7.1. COMPARISON OF THE VERIFIED ACCURACIES FOR THE LANDSAT FOREST COVER
CLASSIFICATIONS WITH ANCILLARY DATA OVERLAYS. The numbers represent the percentage
of the total number of the 29,290, 1 ha. cells which were correctly classified into their respective cover
type when checked against the 1972 land cover map. The second set of percentages in italics represents
only those cells correctly classified into shifting cultivation. These italicized values for shifting cultiva-
tion fluctuate widely as the various classification approaches were optimized to achieve the best overall
accuracy for all 8 land covers and not optimized for this specific cover type.
MSS-5 and 7,
those above plus MSS-4 and 6,
those above plus topographic slope, aspect, elevation, and distance to drainage,
those above plus distance to roads and trails, permanent housing, and temporary
2 variables used
4 variables used
8 variables used
11 variables used
housing.
TYPE OF TRAINING SETS
VARIABLES RECTANGULAR ASPECT GRID
USED Equal Proportional Equal Proportional Equal Proportional
2 26.0 43.6 19.0 41.4 22.1 41.7
15.8 67.6 14.1 61.9 13.5 66.7
4 26.2 44.5 33.6 39.0 28.0 48.8
16.6 64.8 87.9 99.0 27.4 83.3
8 33.4 43.9 22.6 29.6 32.9 52.2 ]
34.9 47.4 34.7 54.4 35.6 79.7
11 36.0 46.2 38.0 7 47.8 36.3 55.3
37.4 48.2 36.2 46.8 34.9 74.0