2004 
—— 
catio 
  
'orld. 
] and 
uable 
1ality 
ed to 
wing. 
gical 
each 
le to 
olved 
needs 
n the 
on or 
saltze 
erical 
d LE 
store 
\base, 
r and 
nodel 
nplex 
3n 
> with 
n LE, 
red in 
tation 
es the 
ns. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
3,3 Integration of GIS models and domain knowledge 
Model is a simplified reality and is designed for computer 
problem solving. Knowledge utilization is the most prominent 
character of spatial decision system. Expert knowledge is 
important in problem solving but model is also indispensable 
and most spatial problems solving depends on spatial models 
(also referred as GIS models), esp. in tasks that can be 
mathematically expressed. In the intelligent system, although 
knowledge base and model base are dependently organized, 
they must cooperate. The fusion of spatial analysis and expert 
knowledge is an effective way to realize their cooperation in a 
sophisticated problem solving. Model can be used by expert 
knowledge to solve some structured and well-formed problems. 
In this way expert knowledge and models are connected 
together. In a concrete problem solving, knowledge founds the 
masterstroke and is able to use any models organized in model 
base. Contrary to this, models can also be used directly by users. 
A model unique identical number is input through model use 
interface and the corresponding model then can be driven to run. 
The function relation between spatial analysis model and expert 
knowledge experiences three periods. The original spatial 
decision support system depends totally on models and expert 
knowledge is embedded in model during the first period. In this 
mode knowledge is a fixed component of model and cannot be 
redesigned to adjust a changed environment (Figure 4(a)). In 
the second period, more supplicated expert knowledge is 
introduced and put in use under control unit in SDSS (Figure 
4(b)). This mode emphasizes the importance of analysis model 
and expert knowledge and they are parallel in a problem solving 
in SDSS. In this way, SDSS is developed into intelligent SDSS 
because expert knowledge plays a crucial role. But domain 
knowledge is separately stored with model base; model 
operation can be driven by the knowledge. But the middle ware, 
control unit, is indispensable at this period. The third period 
realizes direct communication between model and knowledge 
(Figure 4(c)). This mode relates model with domain knowledge 
directly and offers a mechanism for model and knowledge 
interaction. Models can retrieve and use domain knowledge 
according to its need and expert knowledge can retrieve and use 
models. Compared with the second period, this period is a real 
integration between knowledge and model. Control unit is no 
189 
  
  
Proposed |. 
problems 
  
  
  
  
c User interface | EIL Rn 
—— // Stored spatial ^ 
| bo Control ads analysis model | 
TP >> : A} 
se (integrated domain | 
1 T x knowledge) 
  
| Domain knowledge | 
\ EA 
(a) spatial decision based on spatial analysis model 
User interface r 
»<<Control unit >—#— OECD | 
3 | analysis model | 
Suit au) ; 
) | 
/ | 
(b) spatial decision based on integration of model 
and knowledge coordinated by control unit 
Proposed 
problems 
| 
| 
| 
  
   
  
  
I : X 
Domain 
| knowledge | 
| AN. 
User interfac | nh i x 
Proposed 
problems 
  
f Stored spatial | 
| analysis model ) 
Ad 2 
  
   
M 
  
  
(c) spatial decision based on close integration of model and knowledge 
Fig. 4 Modes of integration of GIS models and domain 
knowledge 
longer a need in the third period thus model and knowledge has 
a closer and flexible integration. 
4. CASE STUDY 
4.1 Research Area 
The central theme of this paper is helping conduct farm-level 
decisions including water and fertilizer application, with varied 
planted crops in different farms, usually with small area. 
Located in the Guangdong Province, China, with 113.5+4.0 east 
longitude and 22.8+2.5 north latitude, Guangzhou City has a 
total land area of 7400 km2. As one of the riches city in China, 
it has a total output value of 1.072 billion US$ in industry and 
12b in agriculture in 1993 (TCSB, 1993). Most of the area for 
agriculture production covers the suburb region of the city and 
because the semitropical climate of the city meets required 
living conditions of many crops, the cropping pattern is very 
complex, with a large diversity of plants such as fruits, flowers, 
vegetables and rice. In addition, a great part of area is 
mountainous and this makes the farm field fragment, which 
adds even more difficulty in farm-level decision-making as 
water and fertilizer application. 
4.0 Framework of Software Structure 
Three layers, viz. data layer, application layer and browser 
layer, logically composes GZ-AgriGIS framework. This multi- 
layer structure makes the system maintenance easier and the 
system service range wider. Anyone who can connect to 
Internet can be authorized to use the system and aided to 
manage his farm fields in any place at anytime through graphic 
user interface (GUI) of the system. GUI is built in dynamic Java