world might and could do, but not necessarily what it will do. Some sort of
validation is useful to determine if the model produces realistic projections.
Model results always contain uncertainties because they are based on (1)
current understanding of interactions and (2) field and laboratory studies.
That is why we call model results projections (i.e., estimates of future
possibilities) rather than predictions, something that is declared in advance
(Dale and Van Winkle 1998). Great caution is required in basing decisions
solely on model results. Models produce approximations to real situations
and are only as good as the assumptions upon which they are based.
Because these assumptions are typically specific to each situation, caution
must be used in applying a model developed for one circumstance to
another case. The appropriate application of a model has time implications
as well. Thus, a corollary to a dictum often adopted by modelers that
“Reality Is a Special Case” is that “Reality (t) π Reality (t + 1)” (Dale and
Van Winkle 1998). Until information is available to validate a model for the
situation at hand, model results should be considered with caution; they are
the logical extensions of existing data produced via a process that assimilates and applies current understanding.
Current understandings of complex environmental systems, as reflected
in models, will rarely be adequate alone to provide simple answers to
environmental questions. The caution required in interpreting model
calculations is illustrated by an example documented by Christensen et al.
(1981) and Barnthouse et al. (1984). Under the scrutiny of legal proceedings, two computer simulation models were developed to determine the
potential impacts of several power plants on fish populations. One model,
emphasizing a particular theory of population dynamics, concluded that
there would be little impact and that changes in the fish population could
be explained by natural factors. The second model, relying on a different
understanding of how fish populations interact with their environment,
concluded that significant impacts would occur. Both models were subjected to intense scrutiny, but the difference in conclusions remained. Such
cases notwithstanding, model projections often remain our best source of
information for extrapolating limited theory and field and laboratory data
to the real-world decision arena.
Although some would argue that models should be used as a crystal ball
to gaze into the future, we think such use of models is an inappropriate goal.
Models should not be believed more than any other scientific hypothesis
(Dale and Van Winkle 1998). “Belief” suggests a faith or trust based on
incomplete information. Instead, models should be used to improve understanding or insight about the ecological relationships and management
implications. When the process of modeling inappropriately emphasizes
belief rather than understanding, the failure of a model to predict a specific
reality reflects, in part, unrealistic expectations.
The discussion and examples in this book build upon experiences in
applied ecology, where industries or agencies are looking for models to help
1. Opportunities for Using Ecological Models for Resource Management
13
validation is useful to determine if the model produces realistic projections.
Model results always contain uncertainties because they are based on (1)
current understanding of interactions and (2) field and laboratory studies.
That is why we call model results projections (i.e., estimates of future
possibilities) rather than predictions, something that is declared in advance
(Dale and Van Winkle 1998). Great caution is required in basing decisions
solely on model results. Models produce approximations to real situations
and are only as good as the assumptions upon which they are based.
Because these assumptions are typically specific to each situation, caution
must be used in applying a model developed for one circumstance to
another case. The appropriate application of a model has time implications
as well. Thus, a corollary to a dictum often adopted by modelers that
“Reality Is a Special Case” is that “Reality (t) π Reality (t + 1)” (Dale and
Van Winkle 1998). Until information is available to validate a model for the
situation at hand, model results should be considered with caution; they are
the logical extensions of existing data produced via a process that assimilates and applies current understanding.
Current understandings of complex environmental systems, as reflected
in models, will rarely be adequate alone to provide simple answers to
environmental questions. The caution required in interpreting model
calculations is illustrated by an example documented by Christensen et al.
(1981) and Barnthouse et al. (1984). Under the scrutiny of legal proceedings, two computer simulation models were developed to determine the
potential impacts of several power plants on fish populations. One model,
emphasizing a particular theory of population dynamics, concluded that
there would be little impact and that changes in the fish population could
be explained by natural factors. The second model, relying on a different
understanding of how fish populations interact with their environment,
concluded that significant impacts would occur. Both models were subjected to intense scrutiny, but the difference in conclusions remained. Such
cases notwithstanding, model projections often remain our best source of
information for extrapolating limited theory and field and laboratory data
to the real-world decision arena.
Although some would argue that models should be used as a crystal ball
to gaze into the future, we think such use of models is an inappropriate goal.
Models should not be believed more than any other scientific hypothesis
(Dale and Van Winkle 1998). “Belief” suggests a faith or trust based on
incomplete information. Instead, models should be used to improve understanding or insight about the ecological relationships and management
implications. When the process of modeling inappropriately emphasizes
belief rather than understanding, the failure of a model to predict a specific
reality reflects, in part, unrealistic expectations.
The discussion and examples in this book build upon experiences in
applied ecology, where industries or agencies are looking for models to help
1. Opportunities for Using Ecological Models for Resource Management
13