alternatives or reduce uncertainty (see Sidebar 3.1). Particular challenges
exist when trying to isolate causal roles for observed or predicted changes
that may be simultaneously influenced by fishing, pollution, or climatic
changes as well as by losses from entrainment and impingement. Here, too,
no fixed rules apply. Rigorous and systematic confrontations between
critical questions, data, and models frequently are not possible (Foster and
Huber 1997; Hilborn and Mangel 1997; Suter 1993).
4. What data are/will be available to parameterize, calibrate, and validate
different types of models? Here again, the scientist has to decide whether
a possible increase in model complexity is consistent with the data available to guide that increase.
5. What resources (money, time, technical expertise, and software) are
available? These resources bound the tradeoffs between model complexity
and the effort to develop and apply a model. It may be more cost effective
to make a decision with considerable uncertainty when the cost of implementing that decision compares favorably with that of reducing uncertainty
by additional study (Clemen 1991). That is, it can be advantageous simply
to act rather than to gather more information, even with considerable
uncertainty about the consequences of the decision. Such may be the situation when a CWIS has no fish protection devices for impingement but the
relative cost of determining the impact of impingement is high, and thus,
the decision is made to install fish protection devices.
Fish population modeling does not involve especially difficult mathematics, and developing a model per se is not necessarily costly. However,
substantial resources may be needed to estimate parameter values (and
their variability), calibrate the model, perform sensitivity and uncertainty
analyses, document the model [including quality assurance/quality control
(QA/QC)], and interpret and present the model results in presentations,
reports, and other publications (Ambrose et al. 1996). Hence, site-specific
modeling as a response to the weakness of 316(b) regulatory language is
balanced against a considerable commitment of resources for industries or
agencies with little experience in assessing the impacts of entrainment and
impingement. There is a natural desire, then, on the part of regulators to
look for technological criteria, such as CWIS technologies [e.g., fine-mesh
traveling screens, cylindrical-wedge wire screens, and Ristroph screens; see
Taft (2000)] and operational performance standards (e.g., intake approach
velocity no greater than 0.5 ft/sec) or other less complex and superficially
less uncertain proxies, as a substitute in assisting decision making. This
predicament is characteristic of many complex decisions, not just decisions
in environmental policy or 316(b) (Payne et al. 1993).
The above questions reflect a top-down view of model selection. The
modeling approach selected needs to reflect what is viewed as constructive
by decision makers and stakeholders as well as by scientists. As a result,
modeling is not an ancillary part of a decision process that “knows” which
3. Modeling Fish Entrainment and Impingement Impacts
55
exist when trying to isolate causal roles for observed or predicted changes
that may be simultaneously influenced by fishing, pollution, or climatic
changes as well as by losses from entrainment and impingement. Here, too,
no fixed rules apply. Rigorous and systematic confrontations between
critical questions, data, and models frequently are not possible (Foster and
Huber 1997; Hilborn and Mangel 1997; Suter 1993).
4. What data are/will be available to parameterize, calibrate, and validate
different types of models? Here again, the scientist has to decide whether
a possible increase in model complexity is consistent with the data available to guide that increase.
5. What resources (money, time, technical expertise, and software) are
available? These resources bound the tradeoffs between model complexity
and the effort to develop and apply a model. It may be more cost effective
to make a decision with considerable uncertainty when the cost of implementing that decision compares favorably with that of reducing uncertainty
by additional study (Clemen 1991). That is, it can be advantageous simply
to act rather than to gather more information, even with considerable
uncertainty about the consequences of the decision. Such may be the situation when a CWIS has no fish protection devices for impingement but the
relative cost of determining the impact of impingement is high, and thus,
the decision is made to install fish protection devices.
Fish population modeling does not involve especially difficult mathematics, and developing a model per se is not necessarily costly. However,
substantial resources may be needed to estimate parameter values (and
their variability), calibrate the model, perform sensitivity and uncertainty
analyses, document the model [including quality assurance/quality control
(QA/QC)], and interpret and present the model results in presentations,
reports, and other publications (Ambrose et al. 1996). Hence, site-specific
modeling as a response to the weakness of 316(b) regulatory language is
balanced against a considerable commitment of resources for industries or
agencies with little experience in assessing the impacts of entrainment and
impingement. There is a natural desire, then, on the part of regulators to
look for technological criteria, such as CWIS technologies [e.g., fine-mesh
traveling screens, cylindrical-wedge wire screens, and Ristroph screens; see
Taft (2000)] and operational performance standards (e.g., intake approach
velocity no greater than 0.5 ft/sec) or other less complex and superficially
less uncertain proxies, as a substitute in assisting decision making. This
predicament is characteristic of many complex decisions, not just decisions
in environmental policy or 316(b) (Payne et al. 1993).
The above questions reflect a top-down view of model selection. The
modeling approach selected needs to reflect what is viewed as constructive
by decision makers and stakeholders as well as by scientists. As a result,
modeling is not an ancillary part of a decision process that “knows” which
3. Modeling Fish Entrainment and Impingement Impacts
55