Sanders, Marlies
classes and the vegetation types to be investigated. The vegetation types were grouped into wetness classes to make a one
to one comparison possible (Sanders & Clevers, 1999). Random points were used to compare the vegetation types with
the spectral classes because it would have cost too much effort to digitize the vegetation map. For each point, the
accompanying vegetation type was looked up on the map, digitized and combined with the spectral classes in a GIS.
2.3 Plant species habitat
The plant species used in this study were selected for their expected ecological significance regarding to hydrology and
vegetation management. Scorpidium scorpioides (Hedw.) Limpr and Erica tetralix L. were used to evaluate the model.
Scorpidium is a rare moss indicative of wet base-rich conditions and an early stage of terrestrialization. The species is
characteristic for the plant community Scorpidio-Caricetum diandrae. This typical rich-fen vegetation is a very species-rich
and valuable vegetation type. To ensure its survival, management focuses on the supply of base-rich water, to prevent the
accumulation of acidic rainwater (De Vegetatie van Nederland, 1995-1999). Erica is a dwarf shrub indicative of drier base-
poor conditions and an advanced stage of terrestrialization. The species is characteristic for the plant community Sphagno
palustris-Ericetum. This is a typical fen-bog vegetation. The best management of this vegetation is to mow it in summer to
prevent Betula seedlings developing into woodland (De Vegetatie van Nederland, 1995-1999). Point-distribution maps
(Figure 5) of Scorpidium and Erica were made for the management plan (Staatsbosbeheer, 1988) by vegetation mapping in
1985. Their scale is 1:25000.
Statistical tests were used to assess the relationship between observed plant species sites and the hydrological and
management variables. The statistical tests require presence and absent points of species occurrence. The distribution of
absence points must be representative of the coverage of all site factor classes, to prove that species distribution is not
evenly distributed over all site factor classes. The absences were obtained by generating many random points (Sanders,
1999). The presences were on the point distribution maps. The point maps were combined with maps of the site factors.
The statistical tests were executed in GENSTAT (Genstat 5 Committee, 1987). The marginal test is based on a simple
regression, which takes account of only one site factor at a time. The conditional test analyses the significance of a variable
taking in account all other variables at the same time. The statistical tests were applied to management type (reed cutting,
summer mowing), wetness classes (wet, intermediate wet, dry), the hydrological isolation patterns (numeric) and to the
components of the model such as distance to the surface water (numeric), the amount of hinterland (numeric) and the raft
thickness (very thin, thin, thick). The significant site factor classes that determine the probable habitat characteristics were
obtained with the predict command in GENSTAT.
3 RESULTS AND DISCUSSION
The result of the remote sensing interpretation and classification is a map of wetness classes (wet, intermediate wet, dry),
forest, watercourses and peat baulks (Figure 6). The geometric accuracy is ca. 5 m. The wetness classes were evaluated by
comparing them to a vegetation map (Staatsbosbeheer, 1988). The overall classification accuracy of the remotely sensed
wetness classification is 76%. This percentage is only indicative because there is a time gap of ten years between collecting
the field data and recording the photographs. Ideally, the recordings of the photographs and field data would be within the
same year, which would probably result in a higher overall accuracy. Yet in spite of the influence of such interference, the
spectral classes correspond reasonably well to the vegetation types.
The result of the hydrological model is a map of the hydrological isolation (Figure 7). It was to complicated to evaluate the
map with field data. The accuracy depends on the input remote sensing interpretation and is therefore assumed to be
around 5 m. The thematic content was evaluated by comparing it's patterns with the distribution of indicative plant species.
The results of the evaluation executed with statistical tests are given in Table 1.
Marginal test Conditional test
Site factor Erica Scorpidium Erica Scorpidium
Distance to open water + Ns Ns Ns
Distance to hinterland Ns Ns Ns Ns
Wetness classes he pax ev Kee
Raft thickness LEE kk LII "kk
Management indo Ns e Ns
ydrological isolation ee Tuy Ns Ns
Table 1. Statistical significance of site factors predicting the species occurrence obtained with marginal en conditional
tests (* = 0.01<p<=0.05; ** = 0.001<p<=0.01; *** = p<=0.001; Ns = p > 0.05).
1312 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.
rer
spe
site
me
ma
reg
wil
spe
or.
AC
I ai
val
RE
Bo:
pro
