Spatial-Temporal Mapping of Agro-Ecosystems 
and the need to build Thematic Legends 
Dr. C.A.J.M. de Bie 
ITC, Enschede, The Netherlands; debie(@ itc.nl 
TS: WG VII/2 - Sustainable Agriculture & EcoSystem Approach 
KEYWORDS: Remote Sensing, Agriculture, Mapping, Land Use, Classification, Land Cover, Generalization, Monitoring 
ABSTRACT 
During the past 10 years several inter-related topics were discussed 
at different forums. They were: (i) land cover and land use classi- 
fication systems, (ii) geomatics, being the scientific management of spatial information, and (iii) monitoring of climate, land cover 
and land use changes. Simultaneously the spatial and temporal quality of remote sensing platforms and of GPS-technology improved 
substantially and provided new possibilities to study our environment in detail. 
Progress made by the forums and the quality increases of RS-products are only worthwhile if gathered spatial GIS-information can 
be integrated and processed, i.e. if the used survey, generalization and classification methods are scientifically based. In other words: 
the quality of gathered spatial information, at various scales, must be sufficiently high to support modeling, monitoring, and prepara- 
tion of decision-support systems for sustainable land management. 
In spite of this, available standards for the collection, storage, generalization, classification and presentation of geo-information on 
climate, soil and terrain conditions, water resources, land use, land cover and bio-diversity, and on social and economic conditions, 
are often poorly implemented, or re-defined on a project-by-project basis. The root causes are: (i) non-familiarity with underlying 
concepts, (ii) use of terminology without considering their scientific definitions, (iii) complications faced when standards and rules 
must be followed, (iv) ignoring future uses of prepared products, and (v) institutional conservatism. 
To integrate available standards, a “root” concept of a comprehensive information system is presented. Each theme presented at the 
root can be linked to a theme-based relational database. Joint use of each thematic layer creates a full description of an agro- 
ecosystem or of an agro-ecosystem class (full descriptions help 
to study the system functioning). Map legends must like-wise contain 
entries on a theme-by-theme basis, possibly nested, or must remain theme specific (theme specific maps are required for monitoring). 
Following this, legends covering several themes should not be based on a simple list of suggestive names that cross theme boundaries 
and represent an incomplete mix of class entries of the themes covered. 
To further substantiate the above, the theme *land use' is elaborated upon in some detail. Concepts to describe agricultural land use 
and the subsequent method to classify collected ‘land use’ data are presented. Shown are the key concepts: ‘crop calendar’ and 
‘cropping system’ (classification system appended). They form important links between RS-images, land cover, land use, and the 
subsequent land use legend (= classification application). The links are demonstrated by discussing spectral classification results of 
multi-temporal RS-images. Derived map units link to monothemat 
phenology while units having agro-ecosystems differ in crop calend 
ic legend classes: units having natural vegetation differ in plant 
ar. The used method, following only two key concepts, provides 
support to further improve mapping of agro-ecosystems (e.g. by describing the operation sequences followed), to improve crop- 
monitoring methods, and to build spatial information systems that adhere to sound survey, generalization and classification methods. 
1. INTRODUCTION 
Many scientists have conceptual problems differentiating (agri- 
cultural) land use information from information on other the- 
matic ecosystem components, e.g. many maps, and other survey 
products, tend to confuse information on land use with, for 
instance, land cover. The impact of land use on land cover can 
hardly be studied if such products are used. It is a precondition 
for any exercise involving detection of change that causes (e.g. 
change in land use) and effects (e.g. change in cover) are kept 
apart. This principle is not open to compromise. 
This paper discusses aspects to describe (agricultural) land use 
at plot level to optimize options to cluster, generalize and clas- 
sify collected primary date, and to extrapolate the results spa- 
tially through modern RS/GIS techniques into map units that 
have an attached legend, in which the generalization or 
classification results are applied by theme. 
Archiving properly collected primary land use data provides 
options to re-use them when new (e.g. monitoring) studies are 
called for. Different, study-specific classification rules can then 
be applied on the available primary data for alternative cluster- 
ing, generalization, classification and extrapolation. 
The approach followed here is intended to be both practical and 
conceptually correct. The bottom-up approach that was adopted 
leads to the holding-level where actual decision-making by 
individual land users takes place. Studies of biophysical land 
use system performance generates inputs for socio-economic 
evaluation, culminating in the definition of planning scenarios 
that conserve land resources and are rewarding for both primary 
and secondary stakeholders. 
Agro-ecosystem (land use system) studies must include studies 
of the land (Figure 1). Management activities (operations) at 
plot level aim at modifying one or more aspects of land, e.g. the 
soil, flora/fauna, or infrastructure. Operations are carried out to 
support one or more land use purpose(s), e.g. to harvest a good 
crop but they can also have negative side effects that affects the 
sustainability of the system. Often, operations are pre-planned, 
but they can be of a remedial nature depending on dynamic 
land processes, for example, incidence of pests and diseases, 
weeds infestation, water and nutrient deficiencies, etc. 
1148