58
independence states that the distribution of the classes in
one band is random to the distribution of the classes in the
the other band. If there is no independence, as is shown
by a large calculated chi-square value, an interaction is
said to be present. Critical chi-square values are avail
able in most statistical text books, although a more re
liable rule-of-thumb can be used: if the calculated value
exceeds three times the number of degrees-of-freedom, the
distribution is dependent. This rule is preferred because
of the random noise that is introduced between bands due to
different mapping scales, projections, and orientations.
By examining the results summarized in Table 2, it is appar
ent that there was a high degree of dependence between the
vegetation band and the soil, contour, aspect, and slope
bands, respectively. None of these results are startling,
for plant ecologists have known for years that there is a
close interaction between the distribution of soils, veg
etation, aspect, elevation, and slope.
Selection of the stratification scheme. Although the
chi-square analyses indicated a high degree of dependence
between the vegetation, slope, aspect and contour bands, we
selected the vegetation band for the stratification because
this band best represented the range of ground conditions
present in the Park to the highest ground resolution. Use
of the hypsographic bands was rejected due to relatively
poor horizontal accuracy; and use of the soil band was re
jected due to relatively poor horizontal accuracy; and use
of the soil band was rejected due to its large minimum map-
ping areas. A
total of
402 ground
plots were proportion-
ately allocated
to the
19 spectral
classes
(Table 3).
SPECTRAL
NUMBER OF
PERCENT OF
NUMBER OF
STRATIFICATION
PIXELS*
TOTAL AREA
GROUND PLOTS
VEGETATION TYPE
Conifer-1
300
2
8
Conifer-2
1106
7
28
Conifer-3
1824
11
46
Redwood Forest
Conifer-Hardwood-1
1994
12
50
8379 acres
Conifer-Hardwood-2
2151
14
54
Hardwood-Conifer
2040
13
51
Hardwood-1
1538
10
39
Hardwood-2
885
6
22
Hardwood-3
308
2
8
Douglas Fir Fori
Hardwood-Woodland-1
84
1
2
Hardwood-Woodland-2
546
4
14
4379 acres
Hardwood-Woodland-3
484
3
12
Hardwood-Woodland-4
192
1
5
Hardwood-Woodland-5
562
4
14
Hardwood-Woodland-6
321
2
8
99
1
2
Chaparral
Brush-2
543
3
14
Shrub
436
1
11
1444 acres
Grassland
545
3
14
15958
100
402
1 pixel = .89 acres
Table 3. Allocation of 402 ground samples to the 19 spec
tral class stratification, and the grouping of
these classes into three vegetation types based
on these samples.