International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
popular application of GIS in agriculture, which may be
classified as micro application nowadays is digital agriculture,
tailoring soil and crop management to fit the specific conditions
found within a field with the aim to improve production
efficiency and/or environmental stewardship.
2.3 The Complexity of Agricultural Field Management
Agriculture production is a spatial ecological system that shows
uncertain, fuzzy characters in management. To get maximal
benefit and minimal side effect, it is necessary to vary
management method on different agriculture farm field due to
the variation on field properties. There are two kinds of
agriculture field model: regular grid field in digital agriculture
and irregular field grid typical in rugged regions. Digital
agriculture, also known as precision agriculture, usually
regularly partitions a large area into groups of small cells (Fig.
2 (a)). In practice, however, it is almost impossible to get
regular farm field due to scatted farm location and rugged land
(as in mountainous region). Figure 2 (b) shows decision-making
grid for large-scale farm-level fields. Those fields are irregular,
scatted and ill shaped. Because of this fact, it is much more
complicated for farm-level agricultural field management than
that for digital agriculture because the diversity of living
condition for crop should be considered when decision on field
management is made. To get a valid and acceptable result it is
essential to have the support of expert knowledge, which
impose great challenge to users, as they must acquire full
knowledge through various ways.
2130456] 7] 89! wn]
(b Irregular farm-level field distributio
(a) Regular grid in digital agriculture
Fig. 2 Comparison between two kinds of field model
3. INTEGRATION OF EXPERT KNOWLEDGE AND
GIS MODELS
3.1 Knowledge acquisition and representation
Knowledge acquisition (KA) is the first step to make an expert
system. KA is the process of transferring conceptual knowledge
from the knowledge source to knowledge engineer (or expert
system builder) To acquire the required knowledge, we
followed the KA procedure discussed in Morpurgo and Mussi
(2001) and Wada et al. (2001). The expert knowledge can be
obtained from specific literature and descriptions and the rules
from domain experts. In this process knowledge engineer is the
centre of the task. The following steps will be included to
complete the knowledge base construction and application
(Figure 3).
188
y ( 9 (Conception |
es : te la 2 k
a ó puentes n
{ ? j | a
a hon lll: ee
Problems ô Formalization
-1— template
m ait. TERA
ó
20M. Knowledge
y ects engineer
Books 4 A
<> Knowledge bas
nn
Knowledge maintenance
input | check |
Domain expert
Fig. 3 knowledge acquisition and application
3.2 Model organization
Model organization must be efficient to simplify the real world.
In the past few years, varied models have been developed and
integrated with GIS to solve difficult problems. Valuable
models have been put in practice as environment quality
evaluation, resource allocation, etc.. Models can also be used to
agriculture resource planning, soil erosion, and plant growing.
Three ways are put in use: database expression (DE), logical
expression (LE) and program-based expression (PE), with each
has its advantages and disadvantages. The basic principle to
decide which way should be used depends on the to be solved
problem feature. Generally speaking, if the problem needs
continuous numerical calculation, DE will be adopted. In the
other hand if the problem is discrete numerical calculation or
logic judgment, LE is better and more suitable to realize
complex reasoning. In this sense LE will integrates numerical
calculation and logic judgment. In this study both DE and LE
methods are used to cooperate and database is used to store
them, as models can be easily maintained in this way.
In the process of models construction through database,
grammar and implication should be defined first. Grammar and
implication bridge the communication channels for model
constructor and user. This model structure can express complex
mathematical functions like this:
F(x (,X2,X3,-- Xn)=ax p""+asxap"""a;x3p""+. …+anXap"*
Where X;, X2, X3>
a, coefficient for each factor.
The most distinct feature of LE lies on its power to relate with
expert knowledge. Productive rules can be represented in LE,
which paves the possibility to use more complex rules stored in
expert knowledge base. In this way LE model representation
opens the gate into expert knowledge utilization and makes the
ill-structured problem diagnosis possible in GIS applications.
x, are continuous factor name , a,, 22,93,
International .
ee
3.3 Integrati
Model is a $
problem solvi
character of
important in |
and most spa
(also referrec
mathematical
knowledge b:
they must co
knowledge is
sophisticated
knowledge to
In this way
together. In a
masterstroke :
base. Contrar)
A model uni
interface and
The function
knowledge e
decision supp
knowledge is
mode knowle
redesigned to
the second |
introduced an
4(b)). This m
and expert kn
in SDSS. In t
because expe
knowledge 1
operation can
control unit,
realizes direc
(Figure 4(c)).
directly and
interaction.
according to i
models. Com
integration be
