Finally, terminology can be an important barrier to model communication. It is important that the terminology associated with the model is
described in sufficient detail so that the user understands what is being said.
A generic problem is that modeling terminology (and, for that matter, risk
assessment terminology) includes many terms that have specific meaning
in their respective contexts but that are also used by others to mean
different or less explicit things. Unless these terms are made very clear,
miscommunications and misunderstandings will inevitably ensue, clearly
creating another barrier to the effective use of the models.
5.2.7 Uncertainty
What are the uncertainties, and how are they characterized and communicated to nonscientists? Scientists and especially modelers are used to
dealing with uncertainties, but managers and the public are not. Uncertainties can arise from a host of sources, including inadequate understanding of the specific system or stress–response relationship, inadequate
databases to parameterize the model, important relationships that are
not included in the model or have lost their reliability through over
aggregation, and natural variability in physical environmental conditions.
Some uncertainties can be addressed through improved data collection or
improved model development; other uncertainties cannot be reduced, such
as variability in weather events. In any case, the modeler must characterize
uncertainties in terms that are understandable to the various audiences and
must provide an evaluation of the significance of the uncertainties to the
decision to be made. For the ideal case, there may be many sources of uncertainty, but the model outputs are so rigorous that the uncertainties would
not alter the conclusions. In other cases, the uncertainties are very significant and may result in an incorrect decision; this result is common and is a
risk all decision makers face. Finally, how uncertainty is handled within the
model is an issue. For example, will the model use a Bayesian statistical
approach, use Monte Carlo simulations, or bound the range of potential
parameter values? This topic is much-debated and one that deserves equal
attention to how we explain the uncertainties. Therefore, the uncertainty in
risk assessments needs to be made explicit to the decision maker, and
models must have a truth-in-packaging aspect, in terms of both the uncertainties and the analytical approaches used to quantify them. If both of
these elements are made explicit, then the decision makers can have far
greater confidence in the results.
5.2.8 Data and Extrapolation
What are the sources, reliability, density over time and space, and applicability to the specific problem at hand of the data used in the model? A
universal issue is the question of the appropriateness of the data used to
5. Overcoming Barriers to the Use of Models
99
described in sufficient detail so that the user understands what is being said.
A generic problem is that modeling terminology (and, for that matter, risk
assessment terminology) includes many terms that have specific meaning
in their respective contexts but that are also used by others to mean
different or less explicit things. Unless these terms are made very clear,
miscommunications and misunderstandings will inevitably ensue, clearly
creating another barrier to the effective use of the models.
5.2.7 Uncertainty
What are the uncertainties, and how are they characterized and communicated to nonscientists? Scientists and especially modelers are used to
dealing with uncertainties, but managers and the public are not. Uncertainties can arise from a host of sources, including inadequate understanding of the specific system or stress–response relationship, inadequate
databases to parameterize the model, important relationships that are
not included in the model or have lost their reliability through over
aggregation, and natural variability in physical environmental conditions.
Some uncertainties can be addressed through improved data collection or
improved model development; other uncertainties cannot be reduced, such
as variability in weather events. In any case, the modeler must characterize
uncertainties in terms that are understandable to the various audiences and
must provide an evaluation of the significance of the uncertainties to the
decision to be made. For the ideal case, there may be many sources of uncertainty, but the model outputs are so rigorous that the uncertainties would
not alter the conclusions. In other cases, the uncertainties are very significant and may result in an incorrect decision; this result is common and is a
risk all decision makers face. Finally, how uncertainty is handled within the
model is an issue. For example, will the model use a Bayesian statistical
approach, use Monte Carlo simulations, or bound the range of potential
parameter values? This topic is much-debated and one that deserves equal
attention to how we explain the uncertainties. Therefore, the uncertainty in
risk assessments needs to be made explicit to the decision maker, and
models must have a truth-in-packaging aspect, in terms of both the uncertainties and the analytical approaches used to quantify them. If both of
these elements are made explicit, then the decision makers can have far
greater confidence in the results.
5.2.8 Data and Extrapolation
What are the sources, reliability, density over time and space, and applicability to the specific problem at hand of the data used in the model? A
universal issue is the question of the appropriateness of the data used to
5. Overcoming Barriers to the Use of Models
99