5. CONCLUSIONS 
In crop monitoring GIS systems can be use at two different 
levels- In a national or regional decision making process 
GIS systems act as integrator for the different type of 
information of spatially distributed, geocoded relevant 
parameters. At this level the analysis and modelling 
capability of the system determines its potential. 
At the same time the GIS is also very useful when involved 
in supporting the analysis of remotely sensed data. Mainly 
those systems are the best for this purpose that have 
similar basic concepts as that of digital image processors. 
To emphasize the necessary capability to build and analyze 
complex models that require to take many features into 
account at one geographic location (cell) the concept of 
geographic information and modelling system was proposed. 
With the help of these a balance can be reached in the 
cooperation of IAS and GIS in favor of image analysis. Two 
examples were shown how the customary way of agricultural 
land use mapping can be supported by GIS and that new 
concept classifiers can be devised which in some cases (e.g. 
in Hungary) are superior to others making use of strong 
a'priori information: the digital field boundary maps. 
Some further research can be proposed in the topic using 
more intensively GIS in the thematic map creation. These 
should be the computer aided stratification, the selection 
of representative training and test samples from a large 
area to be surveyed. The above outlined two classification 
schemes should also be improved. 
Summarized, a remote sensing based National Crop Information 
System requires stable and high PCC and confidence values in 
crop identification and inventories. The comparison of per 
point and the new GIS supported classification based 
inventory studies in Hungary showed the definite advantage 
of the latter ones. The benefits of increased PCCs seem to 
exceed the cost of DFBM and the digital field boundary maps 
can be used in many other problems. It has been observed 
that with the greater spectral and radiometrical resolution 
of Landsat TM, the within class variability of pixels 
definitely increases, that causes a drop in PCCs compared to 
Landsat MSS. With the increased spectral capability of TM 
combined with these per field methods the overall perfor— 
mance may be remarkably better.These or similar methods can 
be used in different environment and countries. The methods 
need further improvement but seem promising in a national 
crop monitoring system. The general idea of integrating more 
à priori information from a GIS into the classification 
procedure is worth further utilizing.