![[logo] US EPA](http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
Development and Application of Multimetric Indexes (continued)
For the trend (regression) model, the statistical power of the fish index was highest and the invertebrate index lowest (Figure 7). The differences were small after the first few years of monitoring, however, and the indexes had very similar statistical power over the long term. USEPA's performance objectives for trend detection specify that an indicator should be capable of detecting a 2% change per year after approximately five years of sampling 30-50 sites assuming a type I error rate of 0.1 and a type II error rate of 0.2 (McCormick and Peck, 2000). On the basis of five years of monitoring at 40 sites, the diatom index could detect a 1.5% change per year, the fish index 2.1%, and the invertebrate index 2.5%.
Thus, the two statistical models ranked the three indexes differently in
terms of their statistical power; these differences are entirely an artifact
of the statistical model used to calculate power. Because the year-to-year
component of variability is relatively more important for detecting trends
through time, higher annual variability of the invertebrate index caused it
to have the lowest power for the trend (regression) model (Larsen et al.,
1995; Urquhart et al., 1998). These results illustrate that when comparing
statistical power, identical statistical models must be used because results
depend entirely on the statistical model selected and its underlying assumptions
(Blocksom, 2003).
Figure 7. The percentage change per year that would represent
a statistically significant change in multimetric index values decreases as
the number of years of sampling increases. A smaller percentage change indicates
a more precise index and greater statistical power to detect change. Thus,
the invertebrate index was the least precise although all three indexes were
very similar. Based on repeat visits to the same 40 sites each year.
Invertebrate and diatom index values were comparable for pool and riffle samples
Stream resources are naturally varied and complex but index values used for assessment must be independent of natural features and mean the same thing whether the assessment is from a low or high elevation site or a wide or narrow reach. In this way, an index must be applicable to all types of streams in a region.
The tendency when developing multimetric indexes is to select similar sites for metric testing, and eliminate data from small or unique ecoregions, unusual sites, or different habitat types. The purpose is to identify a set of homogeneous sites with similar underlying physical and geographic features so that metric response will be stronger in the absence of other confounding factors.
In the Mid-Atlantic region, a homogeneous set of sites could not be easily defined because there were so many choices for criteria including watershed size, ecoregion, and lowlands vs. uplands. A key lesson learned from the MAIA pilot was that including as much data as possible in the metric testing process was more efficient and simpler than testing multiple sets of sites. In addition, unique sites were not 'orphaned' from the assessment process.
Riffles and pools within a reach represent different habitats to invertebrates and diatoms and sampling in these areas typically yield different taxa. The MAIA sampling protocol kept the samples separate when both habitats occurred at a sample site. For invertebrates, pool and riffles were different: some metrics calculated for pools tended to indicate poorer conditions. Fortunately, similar metrics were correlated with disturbance in each habitat. To compensate for these natural differences, metrics were simply adjusted when they were converted to unit-less scores. In this way, the final index was comparable for samples from a pool or a riffle (Klemm et al., 2003). For diatoms, similar adjustments were not necessary for pool and riffle habitats because index values were similar from each habitat and contributed no measurable source of variability to the index (Fore, 2002b).
Assemblages differed in their sensitivity to disturbance types
When multiple assemblages are monitored, potentially conflicting assessments must be reconciled. What if a stream reach falls below the "impaired" threshold according to its fish index value but not according to its invertebrate index value? Several studies have found that biological indexes based on different assemblages generally agree in terms of assessment condition, that is, indexes for different assemblages are highly correlated (Lammert and Allan, 1999; O'Connor et al., 2000). Nonetheless, assemblages may disagree at specific sites and the conflict must be resolved to determine whether the site is impaired. Some of the disagreement may be associated with differential sensitivities of assemblages to specific types of disturbance (Bryce and Hughes, 2002; Norton et al., 2002).
The lesson learned from Mid-Atlantic streams was that any of the three assemblages
could be used to monitor stream condition because multimetric indexes for
all three assemblages could reliably distinguish degraded sites from sites
with little or no human influence. However, different assemblages showed different
sensitivities associated with their natural history and assessments based
on different assemblages provide more information about the type and source
of disturbance.
Multimetric indexes for all three assemblages could distinguish reference
sites from sites disturbed by nutrient enrichment, mixed impacts of development
and agriculture, and acid mine deposition (see Figure
4 on page 17). Overall, the indexes agreed, with lower values for mine
drainage than for other types of disturbance. The indexes disagreed somewhat
on the relative magnitude of degradation associated with nutrient and acid
deposition, in that diatoms indicated greater degradation associated with
nutrients, probably because several diatom metrics were related to different
aspects of nutrient enrichment (e.g., organic vs. inorganic sources; Table
5). In contrast, acid deposition sites looked more like reference sites
from the diatom point of view. Because acid deposition sites were typically
nutrient poor and steeply sloped, they lacked diatom taxa that were associated
with alkalinity, nutrients, or sediment.
The different assemblages tended to respond differently to specific stressors. Invertebrate and diatom indexes were more highly correlated with specific stressors measured at the reach scale, such as nitrogen or riparian condition than the fish index (see Table 1 on page 18). In contrast, the fish index was less significantly correlated with chemical measures and measures at the reach scale, possibly because fish are not limited to the smaller spatial scale of a reach, as are invertebrates and diatoms.
Table 5. Biological metrics included in the fish, invertebrate, and diatom indexes. Percent sign (%) denotes the percentage of individuals belonging to a given group out of the total number of sampled individuals.
| Metric type | Diatom | Invertebrate | Fish |
| Taxonomic composition | No. Ephemeroptera taxa No. Plecoptera taxa No. Trichoptera taxa |
No. cyprinid taxa % cottid |
|
| Tolerance & intolerance | % intolerant % very tolerant % salt tolerant % tolerant of low oxygen |
Tolerance index | No. sensitive taxa % tolerant |
| Assemblage structure | % non-insects % dominance |
No. benthic taxa % exotics |
|
| Autecological guild | % eutrophic % N heterotrophs % polysaprobic % alkaliphilic |
||
| Trophic guild | No. collector/filterer taxa | % piscivore/invertivore % macro-omnivore |
|
| Morphometric guild | % very motile | ||
| Reproductive guild | % gravel-spawning taxa |
<< previous page
next page >>