ISPRS Workshop on Service and Application of Spatial Data Infrastructure, XXXVI(4/W6), Oct. 14-16, Hangzhou, China 
271 
A MULTI-SCALE GIS DATABASE MODEL BASED ON PETRI NET 
Zhangcai Yin 
School of Resources and Environmental Engineering, Wuhan University of Technology, 122 Luosi Road, Wuhan, China 
yinzhangcai@ 163 .com 
KEY WORDS: Petri Net, Scale, Cartographic generalization 
ABSTRACT: 
In cartographic generalization, cartographic entities would change with scale. State and event are two basic elements to describe 
these changes. However, current hierarchical multi-scale representation models put more emphasis on the representation of state, 
rather than on the event that causes the state to change. In this paper, elementary concepts such as scale reference system, scale event 
are firstly discussed, and then a Petri-net-based multi-scale representation model is proposed, in which place, transition, and arc of a 
Petri net are used for describing state, event of and relations between cartographic entities. Finally, by making some experiment 
benefits of this model are analyzed in great detail. 
1. OVERVIEW OF PREVIOUS STUDIES 
As applications for geographic information systems have been 
continuously extended and the demanding level has been 
increasingly on the rise, it is more and more demanded that 
geographic phenomena should be observed, interpreted, and 
described at different resolutions, and at different spatial scales 
(Sun, 2002). Current study shows that multi-scale 
representation of map features is implemented in existing basic 
map databases by representing, organizing, and managing their 
information at a single scale (Wang, 2003). This hierarchical 
multi-scale model can represent snapshots at a specified scale, 
which ignores relations between different representation 
instances of the same cartographic entity at different scales, thus 
makes it unsuitable for updating cartographic data (Sun, 2002) 
and spatial analyzing (Kilpelainen, 1997; Parent, 1998) and also 
difficult to guarantee consistency (Wu, 2002) among different 
representations of the same cartographic entity. Therefore, it is 
highly demanded that different representations of the same 
cartographic entity should be organically associated with each 
other (Kilpelainen, 1997; Parent, 1998). Current GIS and 
database management systems seldom provide functionality of 
processing multi-scale spatial data, and as a result data updating 
takes a great deal of time (Jones, 1996). Moreover, the process 
of generalizing spatial data should be saved (Hae-Kyong, 2004) 
for better updating multi-scale spatial data, managing its 
consistency, and further decreasing manual interruption (Jones, 
1996) involved in data update. Thus current study should lay 
much importance on how to effectively create a multi-scale 
representation model (Stefanakis, 2003) for cartographic 
entities and also how to effectively deal with them. In this paper, 
Petri net is used to effectively describe our spatio-temporal 
model (Yin, 2004), and also represent scale-based spatial data 
model, thus clearly describing states of representation instances, 
scale events, and relations between them, and finally providing 
theoretical basis for the solving of multi-scale representation 
problems. 
Cartographic generalization is the perfect method used in multi 
scale representation. But it is difficult to be fully realized in a 
short period (Wang, 2003). At present, the most primary mean 
is using hierarchical tree/graph structure to realize maps in a 
series of specific scales (Wu, 2002; Stefanakis, 2003; Peter, 
1995). Take Series of Maps for Nation Basal Scales for example, 
representation instances of the same cartographic entity under 
different scales are linked together with the help of some 
models such as Stefanakis’s semi-structured data model (SSD) 
(Stefanakis, 2003), Wang Yanhui’s extended E-R models 
(Wang, 2003) and so on. Their disadvantages are as follows: (1) 
There exists “state” of a representation instance at a “specific 
scale”, overlooking the scale events between different 
representation instances; (2) Redundancy exists in Multi-scale 
map data in that the same representation instance is stored 
several times in GMS (Generalized Map Series); (3) 
Representation instances have no scale zones, which makes it 
difficult to realize continuous scalar changes in representation 
instances. 
The scale granularity of a hierarchical model is the same to its 
hierarchy level, while it disagrees with generalization events of 
cartographical objects. In this way, it ignores details of 
cartographic generalization, and makes it hard to represent 
cartographic objects’ generalization process, thus leading to 
discontinuous changes in the same cartographic objects at 
different scales. Meanwhile, as result of subjectivity in 
cartographic generalization, different experts may generalize the 
same map differently (Guo, 1998). Therefore, it is rather 
necessary to record, compare, and evaluate the process of 
generalization. As for the simulating of cartographic 
generalization, there are some advantages, here are: (1) helpful 
for recording methods and techniques used by cartographic 
experts; (2) helpful for comparing the process of cartographic 
generalization made by different cartographic experts, and also 
for summarizing knowledge and rules of cartographic 
generalization; (3) helpful for simulating changes in the same 
cartographic entities at different scales; (4) helpful for 
evaluating specific methods of cartographic generalization; (5) 
helpful for map updating, so that, when some geographical 
features at a certain scale change, map updating at other scales 
can be achieved by quickly finding other changed 
representation instances with the help of correlations between 
representation instances. As for our Petri-net based multi-scale 
representation model, it not only shows its strength in the 
research, evaluation, and also simulation of cartographic 
generalization, but also in recording cartographic generalization 
and its formal representation.