312 
4 - THE VEGETATION STUDY 
4.1. Introduction 
Much research focussed on the possibilities of remote sensing for mapping different land cover units such as 
forest, crops and grasshtnds. To distinguish several floristic types within one unit is often more difficult. This 
is especially true for the very similar types within the heterogenous (semi) natural grass and herb vegetation 
where boundaries between types are vague in comparison to for instance the distinct and straight line 
boundaries of agricultural fields. 
Studies on remote sensing of salt marshes have been done e.g. by Drake (1976), Jensen (1980), Bartlett and 
Klemas (1981), Hardisky et al. (1983), Hardisky et al. (1984), Gross et al. (1987) but they concentrated 
mainly on biomass assessment. Bartlett and Klemas (1981) concluded that seasonal divergence and 
convergence of reflectance signatures is a significant process affecting spectral discriminability of important 
wetland vegetation types. They found that potential spectral discrimination of "low marsh" (S. alterniflora) 
from "high marsh" (S. patens and D. spicata) plant communities in Delaware is greatest in early winter and 
poorest during late spring and early summer if Landsat-MSS spectral bands were used. Budd and Milton 
(1982) studied the potential of Landsat TM bands for estimating above-ground biomass and discriminating 
species groups of an english salt marsh, using ground spectral data collected with a portable multiband 
radiometer. Certain species could be separated using transformations of these data. 
The objective of the vegetation study presented here was to determine whether different species composition 
(vegetation types) within the natural grass and herb vegetation of the relatively species rich salt marshes in 
the Netherlands can be detected by remote sensing. 
To gain insight in the spectral characteristics of some of the main vegetation types, detailed spectra were 
recorded with a field spectrometer (PSII) of each type at different times in the year. These spectra were 
compared to detect type by type differences and temporal changes in reflection. Furthermore, the possibilities 
of spacebome (Landsat TM) remote sensing for mapping this vegetation were studied by analyzing and 
classifying images of May, July and October. Analyzis of the casi images will be presented in a later paper. 
4.2 Methods 
4.2.1. PSII measurements. Vegetation spectra were gathered in 5 main vegetation types with the following 
dominant species: Agrostis stolonifera, Elymus pycnanthus, Phragmites australis, Trigochlin maritima and 
Scirpus maritimus. About 10 reflectance measurements (for procedure see 3.1.2) were taken within each 
vegetation type in May 1993. The location of each measurement was accurately determined using a 
Tachymeter. The measurement series was repeated in September at the same locations. The sensor was held 
3 meters above the ground so that an area with 1 meter diameter was measured. 
Averages and standard deviations of the field spectra per vegetation type were calculated. The detailed 
spectra from the different vegetation types and the different times were compared graphically. 
4.2.2. Landsat-TM. The Landsat image of July was geometrically corrected with known 'ground control 
points’ and resampled with the nearest neighbourhood method. The May and October images were corrected 
towards the edited image of July and also nearest neighbourhood resampled. 
A detailed digital vegetation map (1 : 10.000) of the salt marsh based on false colour photographs (1 : 
10.000) and fieldwork was available. On the screen this map was overlaid on the Landsat-TM images to 
select the training sites. In all three images (May, July and October) training sites were selected at the same 
locations. As training samples ca. 6 pixels were selected where a vegetation type covered a sufficiently large 
.aid homogeneous area. 
Vegetation types with one of the following dominant species were sampled: Phragmites australis, 
Elymus pycnanthus, Scirpus maritimus-gtva&d, Scirpus maritimus-nol grazed, Spartina anglica, Aster 
tripollum, Puccinellia maritima grazed and not grazed. The vegetation type with Scirpus was separated in 
one type with Atriplex prostata .and one without this species. Samples were also taken in gulleys, on mud 
and on sand. 
The variation of spectral characteristics of the different samples within one vegetation type was checked 
in scatter diagrams with ellipses, based on the means and standard deviations of the samples. Three outlying 
samples of Aster, Phragmites and Elymus were excluded. As the ellipses of Scirpus-gtna&d and not grazed 
were overlapping they were combined to one Scirpus sample. The remaining samples of each vegetation type