"shrub patches
ly fire units.
-3
26
12
31
40
during the study
) are blamed for
nangrove in the
contained the
etation in 1982.
> area declined
and from 45 ha
y account for a
| those units.
shrub patches
vith some shrub
eption to this
nce classified as
1982
Bl Shrub Categories
[ .] Non-Shrub Categories
250m 0 125m 250m
500 m
1992
Figure 1. Shrub areas in fire unit T-17 at Merritt Island National Wildlife Refuge
3.2 Error analysis
Classification of vegetation types was
81% accurate, with the largest errors
occurring in differentiating types of between
types of grasses (STG and SPJ). Combined,
the shrub types were classified correctly 37 of
40 times (see Table 5). This included one case
of shrub being classified willow (also in shrub
group). Overall, shrubs were classified
correctly 92% of the time.
Because of the number of patches
present within each fire unit the overall
variance for the estimate of total shrub area
was low, ranging from 0.4% to 4.3% of the
total shrub area estimate. The largest
confidence intervals, as a proportion of
estimated area, were in fire units T-10-J and
T-10-L. These had the fewest number of
polygons and, along with T-10-K, also tended
113
to have smaller (high PAR’s) polygons than
other units.
4. CONCLUSIONS
Due to the small number of fire units
sampled and the uncontrolled nature of this
study, it is not possible ascertain a relationship
between fire frequency and shrub growth.
Fire units that had only one burn during the
ten year period had large increases in shrub
area. However, large increases were also
recorded in two of the five units with four or
five burns. Overall, significant increases were
found in units with one burn, and in three of
five units with four or five burns. In the other
two units, T-10-K showed an increase of 0.4
ha, and T-10-L with a minor decrease of 0.9
ha. It is apparent that, if a goal of managing
MINWR is to limit shrub growth, efforts
seemed to have failed in the most study units
during the period of 1982 to 1992