Tom 
'orner inter- 
iaccessible 
n agriculture, 
| landscape. 
ing primarily 
] forest sites. 
the field 
grow into 
xcept for the 
types. The 
nce of more 
by the pro- 
ufficiently 
1creasing 
ime of shift- 
yiven time. 
eing cut and 
y used more 
ts are dis- 
> large non- 
Its of this 
ited in the 
letermine 
n of spatial 
rch discussed 
forested 
information 
f shifting 
ind March 
nths. The 
heral aspects 
riat of the 
uses and 
uthors. 
it 45 kilom- 
and lies just 
- 1753 - 
Thailand/Miller, Nualchawee and Tom 
south of the Burma border. Approximately 12 good Landsat images exist for the site which 
lies within the frame designated as path 36 and row 47. The five* scenes which represent 
annual, anniversary coverage of the site have been employed in the analysis to date. The spe- 
cific forest area to be mapped and modeled is enclosed in a rectangle of 432 square kilometers 
(24 km E-W and 18 km N-S) which was cellularized with a resolution of one hectare (approx- 
imately 2.5 acres) yielding 43,200 cells. The closed, irregular bounded study site within this 
rectangle was limited to an area common to several of the special maps made available from 
earlier analysis of the site by others. It consisted of an area of 292.9 km? or 29,290 square 
cells of 1 ha. resolution and is an anti-watershed bounded by major water courses or drain- 
ages on all sides with a mountain ridge in the center. 
Approach and Analysis: The landscape model of this site consists to date of 103 image and 
map variables overlaid such that each constitutes a cellular data plane of 1 ha. resolution cov- 
ering a common area of 29,290 ha. (Fig. 7.1). Utmost care was exerted to insure that each of 
these data planes registers upon all others to the nearest 1 ha. resolution cell. Input into the 
landscape model of those data planes derived from existing maps and airphotos was com- 
pleted entirely by a manual dot sample method. Area planes, such as topographic elevation 
and geology, were directly sampled cell-by-cell from 1:50,000 scale maps. Additional area 
planes, such as topographic slope and aspect, were computed from elevation. Point feature 
planes, such as the location of temporary huts and permanent dwellings, were interpreted us- 
ing four different dates of low altitude, black and white aerial photography. These planes 
were then computed into minimum distance area planes. Linear features, such as drainage 
and roads and trails, were similarly interpreted from each set of airphotos. They were also 
converted by computation into minimum distance area planes. 
Airphoto interpretation maps depicting nine forest and agricultural land cover types were pre- 
pared for the four different dates and overlaid onto the model via the area dot sampling 
procedure. All four of these forest cover maps were interpreted by a single individual in a 
consistent fashion. The nine types of land cover interpreted from airphotos for each of the 
four dates (Fig. 7.2) include 
À hill evergreen forest 
A dry evergreen forest 
A teak plantations 
À dry dipterocarp 
A dry dipterocarp forest with pine 
A mixed deciduous forest with teak 
together with three agricultural covers of 
A shifting cultivation A irrigated rice paddies A tea plantations. 
The direct visual inspection of the resultant forest cover maps illustrates the marked increase 
in the area occupied by active shifting cultivation over the past 20 years. This comparison 
illustrates that the nominal 20 year cycle from forest to shifting cultivation and back to forest 
has been drastically shortened. 
  
*Landsat scenes of 27 January 1973, 22 January 1974, 13 February 1975, 26 February 1976, and 
2 February 1977.