2001). The workflow of the MDA to generate enhanced land 
use and land cover data consists basically of a GIS- and a RS- 
part (Fig.1). The task of the remote sensing part is to classify 
multitemporal (and multisensoral) satellite imagery and to 
provide a classification assessment in terms of data quality. 
For the analysis of crop rotations, multitemporal satellite 
imagery are of key importance to consider phenological 
characteristics (Rohierse and Bareth, 2004). The results are 
imported into a GIS environment. Here, the classified data are 
combined with additional relevant and available topographical 
and/or land cover data. These are usually official data provided 
by national surveying and mapping bureaus. The idea is to use 
high quality topographic spatial information e.g. about 
residential area to improve the land use classification. In the 
latter case, all affected land use classes of the remote sensing 
analysis will not be considered any more. 
Besides official land use data which are available in topographic 
information systems like the German official topographic- 
cartographical information system called ATKIS 
(www.atkis.de) or e.g. the official land use database of China 
(http://ngcc.sbsm.gov.cn/english/), numerous land use infor- 
mation are stored in various spatial databases. These spatial 
biotope/biodiversity databases (e.g. in Germany), spatial 
databases of national parks, research projects, water protection 
areas etc., can be used for such an approach. Additionally, data 
from official statistics like agricultural or land use data have to 
be considered (Bareth 2009). The MDA is described in detail by 
Bareth (2008). 
  
pu o un 
images 
   
       
   
   
oi 
   
  
  
Gis: ] 
residental, forest, a.m. 
       
  
e E .URST08 — 
  
  
  
Figure 1. Schematic procedure of the MDA 
3. CASE STUDIES 
3.1. Dairy Farm Region “Wiirttembergisches Allgiu” 
The first case study is the intensive dairy farm region of the 
"Württembergisches Allgáu" in Southern Germany (Bareth 
2000; Bareth et al. 2001). For the MDA, an IRS-1C image (1% 
September 1997) was analyzed for land use and was overlaid 
with numerous available official GIS data. The most important 
GIS data source is the ATKIS, the German topographic vector 
database in a scale of 1:25,000. Besides ATKIS, available data 
on biotopes, water and nature protection areas as mentioned 
above were used for overlay analyses. Hence, the final MDA 
land use data contain numerous spatial information besides the 
RS derived classification. In Fig. 2, the IRS-1C image is 
visualized with the selected topographical GIS vector data. Only 
by displaying urban and forest polygons, it is very obvious that 
a significant proportion of the image is covered by spatial 
knowledge from an additional source. 
  
Figure 2. IRS-1C image with topographical GIS vector data 
The generated final MDA land use data serve in a next step as 
the base for spatially disaggregating agricultural management. 
Here, the N-fertilizer input is presented as an example. The N- 
fertilizer input for grassland and arable land is calculated for 
each administrative unit (township) by using agricultural 
statistics (fertilizer amount, livestock number etc.). The 
distribution of the fertilizer amount within a township is based 
on the land use data. But most important, some land use classes 
in the MDA land use indirectly contain limitations by law for 
N-fertilizer input e.g. special water protection zones or special 
biotopes. Therefore, following four rules were applied for the 
distribution (Bareth 2000): 
- extensive (0 - 100 kg N ha'!yr ^): 
all land use polygons within the biotope zones and the 
water protection zones I and II 
-  moderate-extensive (101 - 150 kg N ha ‘yr!): 
all grassland polygons of township Tettnang 
- moderate (151 - 200 kg N ha!yr!): 
all grassland polygons of the other townships 
- intensive (201 - 250 kg N ha'!yr!): 
all arable land polygons 
  
N-Fertilizer in kg ha/yr 
[7] extensiv (0 - 100) —— Township Border 
[7] 101 - 150 Water 
151 - 200 
  
    
    
    
    
Residential 
Forest 
0 10 Kilometer 
»———— 
  
  
  
Figure 3. Spatially disaggregated regionalization of N-fertilizer 
input for the “Wiirttembergisches Allgäu” 
   
The resul 
regionaliz 
Fig. 3. I 
between 
which ha 
livestock 
manure. 
footprint 
maps. Tl 
amount 0 
the area « 
lower inp 
like the ' 
study. 
3.2. Aral 
The “Kra 
West of 
of Germ: 
Forest an 
the Rhin 
the study 
was inve 
(2004). 
For the ^ 
2000 anc 
carried 0 
the analy 
land use 
were ide 
and sumi 
generaliz 
Land Use 
  
  
Figure 4 
The ger 
regional 
method 
(2004) 
MDA p 
to link 
spatial-t 
carried |