Geoscience Reference
In-Depth Information
number of studies have continued to emerge since the irst edition of this topic was published in
2000 (see Table 11.1).
Although the research side of spatial ES is clearly strong, these systems have rarely or never been
used in practice (Keyes 1989; Uran and Janssen 2003). Leung (1997) argues that the limited qual-
ity of decision support provided by existing systems is due to the fact that a greater emphasis has
been placed on the technical side of development, that is, software engineering and the integration
of databases with models, rather than on the knowledge engineering side, which involves capturing
human knowledge and turning that into intelligent decision-making. As a result, the outputs may
not be reliable. These inherent limitations of ES suggest the need for improving existing knowledge
acquisition methods towards the direction of self-learning systems (Jackson 1999) and improving
representation methods to better model the complexity of real-world spatial problems. In addition,
most systems focus on the academic environment because they are research prototype products
rather than being developed in a professional environment aimed at satisfying the needs of the users.
Furthermore, the integration approach (of GIS with ES) is not efficient and user-friendly because of
a lack of appropriate tools. Many spatial planning professionals, for example, are not aware of the
prospects of integrating these technologies, that is, GIS and ES, and hence, they prefer conventional
ways of decision-making including customised computer assistance. Similarly, when spatial plan-
ning professionals are aware of these technologies, many believe that the latter could never fully
capture the expertise and deeper, tacit, knowledge used in difficult decision-making situations.
From the earlier literature, four main models of integration have emerged:
1.
Loose coupling
: Where the GIS and the ES remain as two separate applications and com-
munication is facilitated through data exchange from one system to another. The advan-
tages of a loose coupling approach include improved synergy, ease of system development
and simplicity of design. However, the main disadvantages are a reduction in the speed of
operation, interoperability issues in data exchange and poor overall system maintainability.
2.
Tight coupling (or close coupling)
: In this model, the ES works as a shell of the GIS or
vice versa
, although both systems are still separate independent modules. The main system
calls the other system and then control is returned back to the main application so there is
only virtual seamless integration. Communication is accomplished by direct parameter or
data passing. This integration model is only practical when a small number of the func-
tions from the second system or module are needed. Advantages of this approach include
improved runtime operation, retention of modularity, flexibility in design and improved
robustness. Disadvantages are increased development and maintenance complexity, redun-
dant parameters and data processing and issues of system validation and verification.
3.
Full integration
: Here the ES tools are integrated as a standard GIS operation. Both sys-
tems share data and knowledge representation, offer communication via their dual struc-
ture and allow cooperative reasoning. With this integration model, interactive exchange
between system elements is done in real time and in a seamless way without user interac-
tion. Advantages include robustness and improved problem-solving potential. The system
development and the operation are contained in one common environment, which results
in better design and implementation of the overall application. Moreover, a more uniform
user interface and reduced system maintenance can be achieved. The main disadvantages
are the increased development time and complexity, validation and verification issues and
maintenance compatibility.
4.
System enhancement
: This approach represents a type of full integration where any of
the two systems is enhanced with functions from the other system. In other words, the
GIS can be enhanced with rule-based system capabilities or the ES can be enhanced with
GIS operations, where the latter is more difficult than the former. The advantage of this
approach is that the user works in only one environment, but there is added development
complexity with this approach.
