general laws in ecosystem ecology (Lawton 1999). Currently, ecosystembased approaches to practical ecological problems suffer from vagueness
and circularity (Goldstein 1999). The single-species models are obviously
not useful for all questions; models at community and ecosystem levels are
often needed to address different types of questions. Nevertheless, it seems
that single-species dynamics will remain one of the major practical methods
for environmental conservation and management in the next decade and
beyond, until the increased understanding of ecosystem dynamics allows
more generic and practical models to be built.
At this point, the question of model selection, or the selection of the
modeling approach appropriate for a particular case, arises. Three important factors are the question to be addressed, the quantity and quality of
the data available, and the ecology of the system involved. In some cases,
these criteria may point to different models. For example, the question may
require a complex model, but the data may allow only a simple model. In
such a case, the common approach of using a complex model and making
assumptions for the parameters for which data are not available is not the
most productive approach. Instead, two approaches can be taken, in many
cases, simultaneously. On the one hand, a simple model can be used to
explore other (more fundamental or more general or simpler) questions.
On the other hand, more data can be collected, guided by sensitivity and
uncertainty analyses with the more complex model.
Other factors important in model selection include generality and transportability. Other factors being equal, more generic models are easier to
apply to new cases with minimal or no new programming, whereas more
case-specific models often require additional programming to be applicable to a new location or species. In some cases, this additional programming
can be as substantial as creating a new model.
13.2.3 Integration of Existing Models
In the recent past, significant model development has involved the integration of existing models rather than models created from scratch. Of
course, in some sense, all models are created by combining basic building blocks, such as components that implement basic functions for dose–
response relationships, density-dependence functions, random-variate generators, etc. However, what we mean by “integration” here is the linking of
two or more fully developed models or generic modeling platforms.
We will first discuss examples of such integration and then list some potential future developments that may involve the integration of existing
models.
Two of the most commonly used approaches in population modeling
are matrix models (Leslie 1945; Caswell 1989) and metapopulation
models (Levins 1970; Gilpin and Hanski 1991). Models that integrate these
two approaches have included multiregional models [e.g., Fahrig and
254
Lev Ginzburg and H. Resit Akçakaya
and circularity (Goldstein 1999). The single-species models are obviously
not useful for all questions; models at community and ecosystem levels are
often needed to address different types of questions. Nevertheless, it seems
that single-species dynamics will remain one of the major practical methods
for environmental conservation and management in the next decade and
beyond, until the increased understanding of ecosystem dynamics allows
more generic and practical models to be built.
At this point, the question of model selection, or the selection of the
modeling approach appropriate for a particular case, arises. Three important factors are the question to be addressed, the quantity and quality of
the data available, and the ecology of the system involved. In some cases,
these criteria may point to different models. For example, the question may
require a complex model, but the data may allow only a simple model. In
such a case, the common approach of using a complex model and making
assumptions for the parameters for which data are not available is not the
most productive approach. Instead, two approaches can be taken, in many
cases, simultaneously. On the one hand, a simple model can be used to
explore other (more fundamental or more general or simpler) questions.
On the other hand, more data can be collected, guided by sensitivity and
uncertainty analyses with the more complex model.
Other factors important in model selection include generality and transportability. Other factors being equal, more generic models are easier to
apply to new cases with minimal or no new programming, whereas more
case-specific models often require additional programming to be applicable to a new location or species. In some cases, this additional programming
can be as substantial as creating a new model.
13.2.3 Integration of Existing Models
In the recent past, significant model development has involved the integration of existing models rather than models created from scratch. Of
course, in some sense, all models are created by combining basic building blocks, such as components that implement basic functions for dose–
response relationships, density-dependence functions, random-variate generators, etc. However, what we mean by “integration” here is the linking of
two or more fully developed models or generic modeling platforms.
We will first discuss examples of such integration and then list some potential future developments that may involve the integration of existing
models.
Two of the most commonly used approaches in population modeling
are matrix models (Leslie 1945; Caswell 1989) and metapopulation
models (Levins 1970; Gilpin and Hanski 1991). Models that integrate these
two approaches have included multiregional models [e.g., Fahrig and
254
Lev Ginzburg and H. Resit Akçakaya