s, 
1e 
el 
1€ 
Sanders, Marlies 
Hi, == dh (3) 
xeRy 
Hi, = the hydrological isolation of site y 
Ry = the set of sites (x) that connect site y with the nearest water source 
(the set Ry includes site y). 
The parameters mentioned above were used to model hydrological isolation (equation 5) in the grid module of Arc/Info 
(Esri, 1994). The resolution of the grid was chosen 5 * 5 m?, which is quite detailed and still manageable in a GIS. A site 
was considered synonym to the grid cell. Average values for precipitation (P), evapotranspiration (E) and infiltration (D) 
for the entire reserve were used because spatial data was not available. An average D of 1 mm/day was obtained from 
Hoogendoorn & Vernes (1994). The dimensions of the grid cell determine the cross section (A) and the length (dL) of the 
flow channel. A is the grid cell width (5 m) multiplied by the depth to the sandy subsoil (ca. 2.5 m). 
The following factors determine spatial variability in hydrological isolation: 
e The thickness of the floating raft determines hydraulic resistance because the water below it has a higher permeability 
than the raft. A map with floating raft thickness classes (Haans & Hamming, 1962) was reclassified taking the 
permeability classes of Van der Perk and Smit (1975): very thin floating raft (1500 m/day), thin floating raft (178 
m/day), thick floating raft (44 m/day). 
e The hinterland area: the area between any site and the most hydrologically isolated site. De hinterland area was 
obtained by distance measures in GIS. 
* The position of peat baulks: peat baulks are much more resistant to water through flow than floating rafts (Van der Perk 
& Smit, 1975), because they consist of clay or original peat that is denser than the root mat of the floating raft. Their 
permeability was considered negligible and they, therefore, form hydrological barriers. The peat baulks were mapped 
with remote sensing. 
* The topology of the surface water network: this network of channels, petgaten and ditches supplies base-rich water. 
The distance from the network was considered to be one of the factors that determines the magnitude of the surface 
water influx. The surface water network was mapped with remote sensing. 
e Irrigation was considered to be a hydrological short cut, as it enables the surface water to reach a site without taking 
the driving force or hydraulic resistance into account. F looding has a similar effect. Irrigation can be mapped with 
remote sensing. 
2.2 Remote sensing 
Vegetation classification based on aerial photography is a very difficult task because fen ecosystems are characterized 
by many plant species and gradients. When variables obtained by remote sensing are indicative for hydrological 
conditions, they offer information on species composition in an indirect way. Interpretation based on expert knowledge 
of the spectral values coincide with differences in biomass (Box et al. 1989; Tomer et al, 1997) of the vegetation and 
wetness of the floating raft. Wetness is defined as the amount of water on the floating raft. The best and most objective 
combination of methods was used to derive significant information from remote sensing data without relying too much 
on the skill of an interpreter. Aerial photographs can be used for digital image processing (Lobo et al., 1998; Tomer et 
al., 1997; Sanders et al., 1997). 
False colour photographs 1:22000 from May 1995 (Figure 4) were used for digital image processing and analogue 
interpretation. In early spring the vegetation is not developed and wetness is therefore the mean feature. The 
photographs were scanned with 600 dots per inch and the image was georeferenced to the Dutch coórdinate system and 
resampled to 1 m resolution in Imagine (Erdas Imagine Fieldguide, 1991). Radiometric distortion due to light fall off 
and sun angle was minimized by using only the middle parts of the images for analysis and by applying a histogram 
match. Comparison of absolute reflectance between images would require calibration by placing ground reference 
targets with known reflectance values. This means that the digital values are relative and best interpreted in terms of 
variability observed within the image. Visual photo interpretation was used to map watercourses and petgaten, peat 
baulks, reed fen and woodland because they were easily recognized thanks to their contrasting reflectance, texture and 
sharp boundaries. Within the reed fen segment digital image processing supplied information on wetness gradients. 
Unsupervised classification was used to classify the scanned photographs in “wet”, “intermediate wet" and “dry”. 
Contemporaneous ground truth on wetness was not available to evaluate the results. However, vegetation types mapped by 
fieldwork (Staatsbosbeheer, 1988) implicitly contain information on wetness. The vegetation types were defined by 
species responding to certain environmental conditions including wetness. This enabled the similarity between the spectral 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1311