REGIONALIZATION OF AGRICULTURAL MANAGEMENT 
BY USING THE MULTI-DATA APPROACH (MDA) 
    
  
  
G. Bareth*"* and G. Waldhoff* 
? Institute of Geography (GIS & Remote Sensing Group), University of Cologne, 50923 Cologne, Germany - 
(g.bareth, guido.waldhoff)@uni-koeln.de 
^ ICASD - International Center for Agro-Informatics and Sustainable Development (www.icasd.org) 
KEY WORDS: crop, agriculture, management, GIS, remote sensing, multi-data approach, land cover, land use, regional modelling 
ABSTRACT: 
Regional process-based (agro-)ecosystem modelling depends mainly on data availability of land use, weather, soil, and agricultural 
management. While land use, weather, and soil data are available from official sources or can be captured with monitoring systems, 
management data are usually derived from official statistics for administrative units. For numerous spatial modeling approaches, 
these data are not satisfying. Especially for process-based agro-ecosystem modeling on regional scales, spatially disaggregated and 
land use dependent information on agricultural management is a must. Information about date of sowing, dates of fertilization, dates 
of weeding etc. are required as input parameters by such models. These models consider nitrogen (N)- and carbon (C)-matter fluxes 
but essential amounts of N-/C-input and N-/C-output are determined by crop management. Therefore, in this contribution a RS- and 
GIS-based approach for regional generation of management data is introduced. The approach is based on the Multi-data Approach 
(MDA) for enhanced land use/land cover mapping. The MDA is a combined RS and GIS approach. The retrieved information from 
multitemporal and multisensoral remote sensing analysis is integrated into official land use data to enhance both the information 
level of existing land use data and the quality of the land use classification. The workflow of the MDA to generate enhanced land use 
and land cover data consists basically of two components: (a) the methods and data of the remote sensing analysis and (b) the 
methods and data of the GIS analysis. The MDA results in disaggregated land use data which can be used to link crop management 
information about the major crops and especially crop rotations like date of sowing, fertilization, irrigation, harvest etc. to the 
derived land use classes. Consequently, depending on the land use, a distinct management is given in a spatial context on regional 
scale. In this contribution, three case studies of different regions in Germany will be presented: (i) the dairy farm region 
“Württembergisches Allgäu”, (ii) the arable land region “Kraichgau”, and (iii) the diverse Rur-Watershed in Western Germany. For 
each of the study regions, a different MDA-based approach for regionalizing agricultural management is applied and will be 
discussed. 
1. INTRODUCTION 
Since decades, regional modelling of ecosystems is in the focus 
of numerous research activities and publications (Li, 1996; Li, 
2007; Wolf et al, 2012). For agro-ecosystems models which 
focus on plant growth, yield forecast, and matter fluxes are 
developed and are available for regional applications (Klar et 
al., 2008; Leip et al., 2008; Lenz-Wiedemann et al., 2010). The 
most important limitations of regional application of agro- 
ecosystem models are described by Kersebaum et al. (2007). 
The authors state that the "applications of agro-ecosystem 
models on field or regional scale are mostly characterized by a 
high uncertainty of input data, especially regarding soil and 
management information". This fairly strong statement is a 
result of comparing 18 different models on the same input 
dataset. Bareth and Angenendt (2003) and Bareth and Yu 
(2005) present similar findings for regional economic- 
ecological agro-ecosystem modelling. In this context, Bareth 
(2009) proposes the establishment of spatial environmental 
information systems (SEIS) which provides such data. 
In this contribution, the focus is on generating spatial 
management data for regional agro-ecosystem modelling. 
Process-based models like the DNDC model (www.dndc.sr. 
unh.edu) need detailed input data such as date of sowing, of 
fertilizing, of weeding, of tillering etc. Additionally, the 
amounts of the inputs are required (e.g. irrigation, fertilizer 
etc.). Due to the above mentioned problem and limitation of the 
  
* Corresponding author. 
availability of management data, different approaches for 
providing spatial management are described. Bareth et al. 
(2001) use administrative units to distribute and calculate 
average fertilizer inputs for townships while Neufeldt et al. 
(2006) use crop specific inputs for spatial modelling units. 
Similar disaggregated approaches are described by Leip et al. 
(2008) and Smith et al. (2010), especially the latter focus on a 
modelling tool for management data. 
In contrast to the described and available approaches to 
regionalize management data on administrative or modelling 
units, we propose in this contribution a new method to produce 
spatially disaggregated management data. The proposed method 
is based on the Multi-data Approach (MDA) (Bareth, 2008; 
Bareth and Waldhoff, 2009) and builds on crop type maps or 
even crop rotation (CR) maps (Waldhoff et. al, 2012). 
Therefore, we (i) introduce the general approach of the MDA, 
we (ii) present results of the land use classification for three 
agricultural regions in Germany, and we (iii) evaluate the 
potential of using the MDA to produce spatially disaggregated 
agricultural management data. 
2. MULTI-DATA APPROACH (MDA) 
The basic idea of the MDA is integrating remote sensing 
classification data into official land use data and overlaying 
additional available and appropriate data sources (Bareth,