Page 759 - The Toxicology of Fishes

Aquatic Ecosystems for Ecotoxicological Research 739

have evaluated single chemicals at multiple concentrations with or without treatment replication. Designs
range from small recirculating streams (Crossland and La Point, 1992) to large, in-ground, flow-through
streams 520 m in length (Hermanutz et al., 1992). Most constructed streams are 3 or 4 m in length and
about 50 cm wide. Volume flows range considerably and usually are selected to approximate the regional
conditions. Artificial stream endpoints selected for study are almost always functional or structural
endpoints of algae, benthic invertebrates, or fish (Table 17.2). The size and scale of most artificial streams
preclude the use of predator fish, except for very large systems. For short-term studies, pools may be
constructed downstream to place herbivorous minnows or larval predators, such as bluegill or bass.

Analysis of Mesocosm Studies

Regression designs are common and suggested for use in risk assessment when experimental units are
scarce (Dyer and Belanger, 1999; Shaw and Manning, 1996). Despite problems associated with pseu-
doreplication (Hurlbert, 1984), lack of replication may be justified because within-unit (e.g., within one
replicate unit) variability due to treatments can be substantially more important than among-unit variability
(Belanger, 1997). Fewer experimental studies have used factorial designs or addressed issues of multiple
stressors (Carder and Hoagland, 1998; La Point and Perry, 1989). Factorial designs that use analysis of
variance (ANOVA) (requires replication) are efficient and allow investigation of multiple factor interac-
tions (multiple stressors) (Groten et al., 1996; Underwood, 1997). Table 17.1 and Table 17.2 provide
representative examples of experimental designs and endpoints used in outdoor stream mesocosms.

Design Considerations

Several factors must be considered when designing and implementing studies using model ecosystems.
Considerations range from the pragmatic (funding, time constraints, etc.) to the heuristic (What are the
study goals? What levels of realism are desired?). The physicochemical and biotic features of model
systems determine to what extent, if any, the systems represent natural ones. These features also influence
contaminant fate and effects. System design is therefore important in defining what inferences may be
drawn from the results of tests conducted with surrogate systems and how closely they may be extrap-
olated to fish populations in natural aquatic ecosystems. Using results from the scientific literature on
model ecosystems, the following sections seek to provide a synthesis of some key experimental design
considerations.

Scaling Effects in Artificial System Research
The question of whether artificial aquatic systems are reliable surrogates for natural ones is strongly
linked to system scale. Scale includes not only the size and physical dimensions of a microcosm or
mesocosm but also its spatial heterogeneity and attendant biotic components. Crucial physical and
chemical processes behave differently as both a function of, and contributor to, scale; thus, scaling effects
can have implications for community structure and the resultant functional attributes of the system.
Obviously, long-term studies with fish cannot be conducted in systems that are too small. Careful
consideration must be given to not only system complexity but also fish population size (and, thus, its
potential to affect the biota). Vital to the research methodology is the choice of spatial and temporal
scales in an experiment which may determine whether changes in selected endpoints (i.e., fish community
survival) can be detected during a study. Frost et al. (1988) stated that “typically, scale has not been
incorporated explicitly into sampling protocols or experimental designs.” Appropriate time scales, for
example, in model aquatic-system research must be considered when deciding on overall study duration
and sampling frequency. The decision regarding how often to sample the fish may have serious ramifi-
cations on the ultimate numbers of fish! Temporal sampling should consider life-cycle duration and
periodicities of important prey species. When making a decision concerning sampling intervals, one
should also consider the temporal behavior of key physicochemical processes, which are often related
to pesticide fates and half-lives.