verall |
Kappa
0.46
060 |
y Site
lata set (Ahl,
)y examining
ere evaluated
iyers was the
. They were
Similarly to
> four groups
1Omogeneous
ISer accuracy
Kappa were
st among all
0 study site
. Orono by
r agreement.
grated mod!
phasizes the
> aggregation
pected result
ce ratio. The
and Kapp
able 1). This
nfluenced by
est of Kappa
es when tle
he integrated
ents on bof
| significant
5.0 CONCLUSIONS
The integrated model provided a flexible and promising
approach for the mapping of forested wetlands from satellite and
ancillary data (hydric soil. slope, NWI maps. hydrology). The
formulated algorithm can be extended through a GIS expert
svstem for an automated classification approach. However, the
ancillary data set needs to be investigated further to broaden the
consideration of GIS layers. For example, the slope layer may
need to be referenced to terrain positions or shape because
forested wetlands are less likely to occur on convex shape slopes
and ridge topographic positions, and are more likely to occur on
concave, flat terrain. Skidmore's (1989) topographic position
model of ridge, side slope, toe slope, bottom was reviewed but a
suitable program was not available to this study. The poor
contribution of the stream buffer layer suggested that this
variable can be ignored or possibly combined with a topographic
variable. This investigation and Ahl's (1994) work suggest that a
hybrid classification and the integrated GIS model are the most
promising for identification and mapping of forested wetlands.
On the basis of the two Maine study sites and the variables tested
the level of effort to develop and implement a forested wetland
mapping model does not appear to be justified over the
conventional hybrid classification approach. Furthermore, the
satellite and GIS methods may not be an improvement over NWI
(aerial photo methods) for mapping forested wetlands in Maine.
However, the satellite mapping approach can provide more
complete inventory of land cover types than is available from
>
Table 2 Analysis Matrix of the Integrated Model with Weighted GIS Layers:
Acadıa Study Site
Integrated Model Analysis Matrix
of Weighted GIS Layers
H|S|N|B F Combi Aerial Photo Sample Plots
Y|L|W|U S nation Forested Other Forested Other Assigned
SIPIIIÓIT T Level Wetland Wetland Upland Upland Total Categories
0 [0:40:40 0 0 6 35 112 385 538 Other
01:01 0 d 0 1 0 0 0 0 0 Upland
0 10} 2} 0 0 2 0 8 15 15 38
0102/1 0 3 0 0 0 0 0
074 | 0:| 09 0 4 6 23 18 75 122
014 |9]7 0 5 3 6 3 9 23
0141210 0 6 4 7 0 0 11
Q442 | 0 7 0 0 0 0 0
21010190 0 8 11 10 10 0 31
8707011 0 9 1 7 2 0 10 Other
$191 219 0 10 0 1 0 1 Wetland
$101.21 1 0 11 0 0 0 0 0
814109010 0 12 12 18 12 18 60
$1401 0 13 ] 19 0 0 20
8141270 0 14 5 15 0 0 20
1412] 1 0 15 0 0 0 0 0
0101060 16 16 85 26 1153 120 1384
ÿ 10 P604-1 16 17 6 9 112 3 130
9101210 16 18 4 0 1 0 5
910/211 16 19 0 0 0 0 0 Forested
0147010 16 20 49 9 209 18 285 Upland
014101 16 21 9 1 20 3 33
01412 10! 16 22 10 0 4 2 16
8144211 16 23 0 0 0 0 0
810} 010 16 24 117 9 155 ] 282
8101011 16 25 ...18 4 36 0 58 Forested
819]2. 01 16 26 5 0 3 0 8 Wetland
850[2]1-1. ig 27 4 0 2 0 6
[$ 41010 16 28 167 11 56 17 251
214 /01 1) 16 29 20 8 7 0 35
131342 [0] 46 30 54 9 4 0 67
En $12 11 16 31 4 0 0 0 4
Total 603 234 1935 666 3438
Threshold Control by 408/603 113/234 1475/1935 385/666
Balance Ratio 0.68 0.48 0.76 0.58
423
International Archives of Photogrammetry and Remote Sensing. Vol
. XXXI, Part B7. Vienna 1996