9598 - Laserfiche WebLink

Home Browse Search

272 M. J. JENNINGS, L. S. FORE AND J. R. KARR How well does the current RFAI meet these three tests? The first challenge is grounded in the observation that human actions, from the construction of dams to the introduction of toxic chemicals or exotic species, cause changes in the local biota. That is, biological systems are responsive-humans do something and the system changes in return. The success of the general multimetric framework in a variety of types of environ- ments, with various taxa and in diverse geographical areas give us confidence in the RFAI concept and in the metrics selected for this project. Expectations for biomonitoring applications must be tempered with the reality that the development of ecologically meaningful, statistically sound indexes is an iterative process. Further, TVA biologists are currently testing these hypotheses and tuning the index based on their knowledge of biological patterns and new data. The challenge is to formalize and refine that knowledge in the context of the unique aspects of reservoir environments. The second challenge is more troublesome for the RFAI as it stands today. The observation that RFAI scores were not consistent over time within a year, and that no consistent seasonal trend accounted for the differences, suggests that the index as presently constructed is subject to considerable sampling error. Neither natural nor anthropogenic variation can be invoked to explain the pattern. Further, we do not accept the premise that fish assemblage health changed in any meaningful way during a single year, as did index scores from the 1992 supplemental data set. Therefore, we doubt that the annual variation in reservoir ranks observed over the four years of the study reflects real changes among reservoirs in fish assemblage health. This concern is reinforced by the pattern of RFAI variation derived from bootstrap analysis. Sampling error occurs when the entire population (e.g. all fish in the reservoir) is not sampled. For electro- fishing in reservoirs, the sampling error is caused by differences in sampling locations-that is, placement of electrofishing runs within the zone. A comparison of sampling error and measurement error for sites eval- uated as part of the 1992 supplemental studies reveals that variability among run locations (sampling error) contributes more to the overall variance of RFAI scores than measurement error (Figure 4) for 13 of 16 reservoir months. We conclude that sampling variability is the cause of the high variability of RFAI scores. Three lines of evidence support this conclusion: (1) the high sampling variability revealed by resampling analysis; (2) the failure of species accumulation curves to level off, and (3) the inconsistent ranking of reser- voirs using TAXA and ABUN. These results indicate that sampling effort at a site needs to be improved to ensure a representative sample. An ancillary problem stems from the need for more comprehensive tests of the individual metrics. Some metrics included in our RFAI are probably not appropriate. The FHAI may fail because it was originally developed to reflect the effects of bacterial and fungal infections and other stresses associated with hatchery fish, populations that are widely known for carrying such problems (Goede, 1986). The method may be less appropriate in largely healthy wild fish, especially in view of the sample bias associated with small sample sizes. Much of the decision-making with respect to biological goals in reservoirs may be directed towards improvement in fisheries. In that situation, we suggest the direct but selective inclusion of metrics such as the age structure of sport fish species in the next round of RFAI development. Such efforts should avoid the temptation to only use metrics based on harvestable fishery concepts, and should emphasize the value of naturally reproducing, self-sustaining fisheries-that is, RFAI may include fishery concepts, but should always be ecologically broader. The third challenge has not yet been adequately explored. Our use of best values across the reservoirs to establish scoring criteria is defensible, but not very satisfying. The multimetric approach relies on an ability to sample systematically sites that span a broad range of conditions, especially a range of human influences. The linear connection among the TVA reservoirs presents an important problem in extrapolating our analysis. Further, we consider our efforts to document the associations between human actions and reser- voir biology to be preliminary. A more systematic effort at defining these patterns must come in concert with a refinement of the sampling protocols (e.g. the appropriate number of electrofishing runs). In addition to defining sampling protocols, more careful study design is necessary to document the influ- ence of habitat type (e.g. sand, rip-rap, emergent vegetation) and variation among sites (within a reservoir zone), zones and reservoirs. These sources of biological pattern need to be integrated with the effect of land use and other human influences, such as the management of hydrological conditions (e.g. water-level fluctuations within a reservoir and the level of water release from upstream reservoirs).