Figure 6. Scatter diagrams with ellipses of the training samples of the different vegetation types of the 
Landsat TM image of July. 1= Spartina , 2= Puccinellia- not grazed, 3= Puccinellia-gnxzed, 4= Aster, 5= 
Phragmites, 6=Scirpus without Atriplex, 7= Scirpus with Atriplex, 8= Ely mu s, 10= sand, mud and gullies. 
4.4 Conclusions and future research 
In this study it has been demonstrated that salt marsh vegetation types of the largest part of the studied area 
are spectrally separable but reflectance changed considerably during the growing season. Besides canopy 
structure, the soil coverage and amount of dead material of each vegetation type seemed to be a very 
important factor for spectral separation of vegetation types. In many studies it is tried to eliminate the 
influence of soil or dead material but for classification of natural vegetation it could be an discriminating 
characteristic of a vegetation type. 
The patterns of the maximum likelihood classification of the July Landsat image corresponded very well to 
the patterns of the vegetation map and the false colour photographs for the vegetation types with the 
following dominant species: Elymus pycnanthus, Scirpus maritimus, Phragmites australis and grazed parts. 
Discrimination between the types of Aster tripolium , Spartina anglica and Puccinellia maritima on the lower 
part was poor which was attributed to the lower vegetation cover and the smaller patches of these vegetation 
types in this area 
The May, October and multitemporal classifications were worse and they did not add any extra information 
to the July classification. This could explain the lower vegetation cover in May and October. In October 
pixel values were also lower due to a lower sun angle. 
The suitability of the intermediate months is not yet investigated. Next year we will try to complete the field 
spectral measurements with the months of June, July and August to assess the temporal changes of reflection 
during this period in the growing season. Special attention will be payed to the Aster, Spartina and 
Puccinellia vegetation types which were difficult to separate on the Landsat-TM images: maybe with higher 
resolution remote sensing they will be distinguishable in a certain period of the year. 
At present the airborne multi spectral casi data of September 1993 are being processed to further analyze the 
spectral characteristics of salt marsh vegetation and the possibilities and limitations of airborne remote 
sensing for monitoring this kind of areas. 
Acknowledgements 
Both the sediment and the vegetation study were sponsored by the Dutch Remote Sensing Board (BCRS). 
We are very greatful for the skilfulness with which Cathy Wrightson of ITS (Canada) operated the casi. The 
aircraft for the airborne experiment and the GPS were provided by the NERC Equipment pool. Geosea 
performed the sediment grain size analysis. Furthermore we would like to thank the people of the Tidal 
Waters Division-Middelburg for their guidance during field work, Mrs. Jordans in Stampersgat for giving the 
opportunity to install the differential GPS in her backyard. We wish to thank Rob Jordans and Renaat van 
Rompaey for their assistance during field work; Rob Jordans and Harm Bakker for their help with the image 
analysis.