on which to base decisions. All parties must understand the structure and
limitations of a proposed model because models with the appearance of
a “black box” will create suspicion and reduce cooperation. Modelers
must clearly communicate to nonmodelers the structure and relationships
within the model, and provide a method by which participants can suggest
improvements to model design.
8.2.1 Approaches and Technologies
8.2.1.1 Logic-Based Model Specifications
Since the 1960s, ecological modeling has emphasized simulation of process.
Early implementations were procedural and based on flow charts. Later
implementations, trying to better cope with ecological complexity, have
tended toward object-oriented models based on the universal modeling
language (UML) (Boggs and Boggs 1999) or similar semantic models for
object-oriented analysis and design. In either case, these implementations
are fundamentally process oriented. However, ecosystem evaluation based
on knowledge-based systems theory and logical abstraction shows promise
for improving the tractability of ecosystem evaluation (Reynolds et al.
2000). Logic-based networks, flowcharts, and UML are all semantic models
(Booch 1994), but logic-based networks are distinct from conceptual
models by having a formal grammar and syntax. Two examples of logicbased approaches are fuzzy network models and Bayesian belief networks.
8.2.1.1.1 Fuzzy Network Models
Fuzzy logic networks are a powerful form of knowledge representation,
ideally suited to the abstract problems posed by ecosystem evaluation.
Similar in concept to a metadatabase, a knowledge base is a formal
specification for interpreting information (Walters and Nielsen 1988).
NetWeaver is such a knowledge base, having a formal grammar and syntax
that makes the knowledge base an executable specification (Reynolds
1999). A NetWeaver knowledge base graphically represents the ecosystem
state as linked networks of propositions.Two key properties of a NetWeaver
proposition are its measure of truth (i.e., the degree of support for the
proposition) and its logical specification, which is graphically constructed
from operators (fuzzy, Boolean, and arithmetic), data, and other propositions. The implementation of fuzzy math in NetWeaver facilitates compact
and efficient representation of large, abstract problems. For example, a
prototype knowledge base evaluates forest ecosystem sustainability as
prescribed by the Montreal Process (Reynolds 2001). Also, fuzzy math
provides a set-theoretic implementation of uncertainty (see Section 8.4.1.3)
as an alternative to the more familiar notion based on probability theory
(Zadeh and Kacprzyk 1992).
136
Eric Gustafson et al.
limitations of a proposed model because models with the appearance of
a “black box” will create suspicion and reduce cooperation. Modelers
must clearly communicate to nonmodelers the structure and relationships
within the model, and provide a method by which participants can suggest
improvements to model design.
8.2.1 Approaches and Technologies
8.2.1.1 Logic-Based Model Specifications
Since the 1960s, ecological modeling has emphasized simulation of process.
Early implementations were procedural and based on flow charts. Later
implementations, trying to better cope with ecological complexity, have
tended toward object-oriented models based on the universal modeling
language (UML) (Boggs and Boggs 1999) or similar semantic models for
object-oriented analysis and design. In either case, these implementations
are fundamentally process oriented. However, ecosystem evaluation based
on knowledge-based systems theory and logical abstraction shows promise
for improving the tractability of ecosystem evaluation (Reynolds et al.
2000). Logic-based networks, flowcharts, and UML are all semantic models
(Booch 1994), but logic-based networks are distinct from conceptual
models by having a formal grammar and syntax. Two examples of logicbased approaches are fuzzy network models and Bayesian belief networks.
8.2.1.1.1 Fuzzy Network Models
Fuzzy logic networks are a powerful form of knowledge representation,
ideally suited to the abstract problems posed by ecosystem evaluation.
Similar in concept to a metadatabase, a knowledge base is a formal
specification for interpreting information (Walters and Nielsen 1988).
NetWeaver is such a knowledge base, having a formal grammar and syntax
that makes the knowledge base an executable specification (Reynolds
1999). A NetWeaver knowledge base graphically represents the ecosystem
state as linked networks of propositions.Two key properties of a NetWeaver
proposition are its measure of truth (i.e., the degree of support for the
proposition) and its logical specification, which is graphically constructed
from operators (fuzzy, Boolean, and arithmetic), data, and other propositions. The implementation of fuzzy math in NetWeaver facilitates compact
and efficient representation of large, abstract problems. For example, a
prototype knowledge base evaluates forest ecosystem sustainability as
prescribed by the Montreal Process (Reynolds 2001). Also, fuzzy math
provides a set-theoretic implementation of uncertainty (see Section 8.4.1.3)
as an alternative to the more familiar notion based on probability theory
(Zadeh and Kacprzyk 1992).
136
Eric Gustafson et al.